World Bank Document · 2016. 8. 30. · Miguel Kiguel and Stephen A. O'Connell 21 Incentives and...

TH E WORLD BANK

RESEARC SIOBSERVE19

1 7522Unitary versus Collective Models of the Household:

Is It Time to Shift the Burden of Proof?

Harold Alderman, Pierre-Andre Chiappori, Lawrence Haddad,John Hoddinott, and Ravi Kanbur

Parallel Exchange Rates in Developing Countries

Miguel Kiguel and Stephen A. O'Connell

Incentives and the Resolution of Bank DistressThomas Glaessner and Ignacio Mas

Financing Infrastructure in Developing Countries:Lessons from the Railway Age

Barry Eichengreen

The Economics of Natural Resource Extraction:A Primer for Development Economists

Stephen W Salant

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

THE WORLD BANK

RESEARCHOBSERVER

EDITOR Moshe Syrquin

COEDITORS Shantayanan Devarajan, Shahid Yusuf

CONSULTING EDITOR Elinor Berg

EDITORIAL BOARD Claire Liuksila (International Monetary Fund); Willem Buiter (Cambridge Universirv); AngusDeaton (Princeton University); Howard Pack (University of Pennsylvania); Gregory K. Ingram, Peter Muncie, GeorgePsacharopoulos, Joanne Salop, William Tyler (World Bank)

The W'orld Bank Research Observer is intended for anyone who has a professional interest in development. Observerarticles are written to be accessible to nonspecialist readers; contributors examine key issues in development economies,survey the literature and the latest World Bank research, and debate issues of development policy. Articles are reviewedby an editorial board drawn from across the Bank and the international community of economists. Inconsistency withBank policy is not grounds for rejection.

The journal welcomes editorial comments and responses, which will be considered for publication to the extent thatspace permits. On occasion the Observer considers unsolicited contributions. Any reader interested in preparing suchan article is invited to submit a proposal of not more than two pages to the editor. Please direct all editorial correspon-dence to the Editor, The World Bank Research Observer, at the address in the copyright notice below.

The views and interpretations expressed in this journal are those of the authors and do not necessarily represent theviews and policies of the World Bank or of its Executive Directors or the countries they represent. The World Bank doesnot guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any con-sequences of their use. When maps are used, the boundaries, denominations, and other information do not imply on thepart of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance ofsuch boundaries.

The World Bank Research Observer is published twice a year in February and August. Single copies may be purchasedfor US$10.95. Subscription rates are as follows:

Individuals Institutions

1-year subscription US$20 US$352-year subscription US$36 US$663-year subscription US$50 US$95

Orders should be sent to: World Bank Publications, Box 7247-7956, Philadelphia, Pennsylvania 19170-7956 U.S.A.Readers with mailing addresses in developing countries and in socialist economies in transition are eligible to receivecomplimentary subscriptions on request. Written confirmation is required every two years to renew such subscriptions.

C 1994 The International Bankfor Reconstruction and Development / THE WORLD BANK

1818 H Street, N.W, Washington, D.C. 20433, U.S.A.All rights reserved

Manufactured in the United States of AmericaISSN 0257-3032

Material in this journal is copyrighted. Requests for permission to reproduce articles should be addressed to the Officeof the Publisher at the address in the copyright notice above. The World Bank encourages dissemination of its svork andwill normally give permission promptly and, when the intended reproduction is for noncommercial purposes, withoutasking a fee. Permission is not required to make photocopies for classroom and similar immediate use.

The World Bank Research Observer is indexed by the Journal of Economic Literature, the Standard PeriodicalDirectory, ABt/INFORM, the Index to International Statistics, the Public Affairs Information Service, Inc., the SocialSciences Citation Index, and, online, by the Economic Literature Index and DIALOG. It is available in microform throughUniversity Microfilms, Inc., 300 North Zeeb Road, Ann Arbor, Michigan 48106, U.S.A.

THE WORLD BANK

RESEARCHOBSERVER

VOLUME 10 NUMBER 1 FEBRUARY 1995

Unitary versus Collective Models of the Household:Is It Time to Shift the Burden of Proof?

Harold Alderman, Pierre-Andre Chiappori, Lawrence Haddad,John Hoddinott, and Ravi Kanbur 1

Parallel Exchange Rates in Developing CountriesMiguel Kiguel and Stephen A. O'Connell 21

Incentives and the Resolution of Bank DistressThomas Glaessner and Ignacio Mas 53

Financing Infrastructure in Developing Countries:Lessons from the Railway Age

Barry Eichengreen 75

The Economics of Natural Resource Extraction:A Primer for Development Economists

Stephen W Salant 93

NOTE TO READERS

With this issue, The World Bank Research Observerwill be published in February and August.

UNITARY VERSUS COLLECTIVEMODELS OF THE HOUSEHOLD:IS IT TIME TO SHIFT THE BURDENOF PROOF?

Harold AldermanPierre-Andre ChiapporiLawrence HaddadJohn HoddinottRavi Kanbur

Most development objectives focus on the well-being of individuals. Poli-cies are targeted to increase the percentage of individuals who avoid poverty,who can read, who are free from hunger and illness, or who can find gainfulemployment. Individual welfare, however, is based in large part on a com-plex set of interactions among family members.

Until recently most policy analyses implicitly viewed the household ashaving only one set of preferences. This assumption has been a powerful toolfor understanding household behavior, such as the distribution of tasks andgoods. But a growing body of evidence suggests that this view is an expe-dience that comes at considerable, and possibly avoidable, cost. The articleargues that more effective policy instruments will emerge from analyzing theprocesses by which households balance the diverse interests of theirmembers.

xperience in several sectors shows that, when policymakers neglect pat-E terns of distribution within households, they do so at their peril. Con-sider government attempts to target programs to individuals in certain

age groups, rather than to households: nutritionists, for example, recognizingthe vulnerability of preschool children, often target supplementary feeding tothis age group. International experience, however, indicates that such interven-tions will not succeed unless the actions of other household members are takeninto account; households often reduce the amount of food given to the targetchild at home and distribute it among the child's siblings (Beaton and Ghassemi1982; Kennedy and Alderman 1987). At the other end of the age spectrum, the

The World Bank Research Observer, vol. 10, no. 1 (February 1995), pp. 1-19

© 1995 The International Bank for Reconstruction and Development/THE WORLD BANK

full impact of targeting programs to the elderly can only be effectively assessed ifthe responses of other family members are taken into consideration (Cox andJimenez 1992).

Similarly, many attempts to introduce new crops or agricultural technologieshave not fared as well as expected because policymakers did not give adequateconsideration to different household members' responsibilities for crops. For ex-ample, in her study of rice production in Cameroon, Jones (1986) found that ricewas considered to be a "male" crop, that is, men controlled any income generatedfrom rice, even if the crop was produced by women. Despite recommendations toconcentrate on rice cultivation, few women planted rice; instead, they grew sor-ghum, which they controlled, despite its lower returns.

As a result of these and similar experiences, there is increasing recognitionthat distribution of tasks and goods within households is important in projectdesign. In this paper we provide an overview of the burgeoning literature on thissubject. Our principal goal is to demonstrate that understanding the process bywhich resources are distributed within households has important implicationsfor policy.

We call the prevailing model of distribution within households the unitary model.This model implies that what matters for certain policy initiatives-such as publicworks schemes or transfer programs-is the amount of income the household re-ceives, not the identity of the individual within the household who is the target ofthe public program. Conversely, under some alternative models, the efficacy of pro-grams depends on the member of the household targeted. The guide to policy-making implied in the unitary model is simpler if it is correct, inefficient if it is not.

Models that assume that households behave as if they had a single decision-maker can lead to a failure to understand the long reach of some public interven-tions. We provide examples in which understanding how resources are distrib-uted within a household can strengthen policy design. We also review the theoryand evidence accumulated by a range of studies that indicate weaknesses in theunitary model.

Although this evidence still requires some bolstering, we suggest a shift inemphasis; what we refer to as the collective models of household behaviorshould be regarded as the standard approach, with the unitary model regardedas an important but special case. We do not counsel abandonment of the unitarymodel; it has proved to be a powerful and pliable tool for household studies. Butin many circumstances, using a unitary model of the household in inappropriatesituations has more serious policy consequences than using a collective modelwhen a unitary model would have been appropriate (see Chiappori 1992b;Haddad, Hoddinott, and Alderman 1994; and McElroy 1992).

The Unitary Model of Household Behavior

Until fairly recently, most economists viewed the household as a collection ofindividuals who behave as if they agreed on how best to combine their time,

2 The World Bank Research Observer, vol. 10, no. 1 (February, 1995)

goods purchased in the market, and goods produced at home. This approachoriginates in standard demand analysis and has been extended to include house-hold decisions about child care, crop adoption, education, fertility, health, homeproduction, labor supply, land tenure, and migration. Indeed, this view evenoffers a perspective on the formation and dissolution of the household, that is,on marriage and divorce (Becker 1973).

This approach is appealing because it allows us to analyze the impact ofchanges in policy and other relevant variables on individual behavior with rela-tive ease and it can address diverse issues. It is sometimes called the commonpreferences model, the altruism model, or the benevolent dictator model. We callit the unitary model because this label describes how the household is assumed toact (as one). Other labels tend to reflect the means by which the household is hy-pothesized to act as one. Common preferences are only one way in which thehousehold can act as one; violence or the threat of violence is another. Altruismhas also been used to explain why households might behave as one individual,but it is altruism under very restrictive conditions, as we shall show later.

The unitary model has some important limitations. It can allow prices todiffer for various household members (the wife's and husband's wages, for ex-ample), but it assumes that all household resources (capital, labor, and land) arepooled. This assumption requires that at least one member of the household isable to monitor the other members and to sanction those who fall foul of itsrules, an issue both of information flows and control.

We are critical of the unitary model because it fails to incorporate the process bywhich resources are distributed within households. The model is able to explaindifferences in individual welfare within a household, however, even when thesedifferences are exhibited systematically by gender, age, or relation to householdhead.1 These distributional inequalities could be generated by preferences for in-equality shared by household members. Moreover, unequal distribution of re-sources may be considered efficient for households. For example, resources maybe distributed on the basis of differences in individuals' ability to earn higher in-comes, and the higher income would then be shared by all members. In such amanner individual differences are treated as different prices and wages.

Pitt, Rosenzweig, and Hassan (1990), who extend the agricultural householdmodel of Singh, Squire, and Strauss (1986), illustrate the adaptability of unitarymodels. They suggest that, if some individuals can earn or produce more for thehousehold when they are healthy and better fed, then it makes sense for the house-hold to provide extra calories to those individuals. They also find that, in someseasons, individuals with the best health in a family do not receive enough caloriesto compensate them fully for their effort. Thus, within households, resources maybe distributed so that consumption is more equitable than work effort.

If, as it is likely, individual household members have different preferences,then each individual's preferences would have to be considered in assessing thetotal well-being of the household. A vast literature on social choice illustrates thetheoretical difficulties associated with aggregating individuals' preferences. It is

Harold Alderman, Pierre-Andre Chiappori, Lawrence Haddad, John Hoddinott, and Ravi Kanbur 3

difficult to reflect the preferences of all household members and not just those ofa single member, even if a single member were to act as a so-called benevolentdictator.

Various approaches to solving the problem of aggregation of preferences havebeen offered. Samuelson (1956) suggested that the aggregation of preferencesand the pooling of household resources could be achieved by consensus amonghousehold members, but he did not indicate how such a consensus is reached.Other proposed solutions include the assumption that individuals tend to seekspouses with similar preferences (assortative mating) and the treatment ofhouseholds as markets in which bartering or trade occurs (Becker 1973). Thesesolutions are not satisfactory because assortative mating does not resolve thepotential conflict in preferences across generations and a model of households asmarkets fails to address the problems of monitoring and incentives. An alterna-tive approach is based on Sen's (1966) model of cooperatives. Here, familywelfare is the weighted sum of the net utility of all members, but the model begsthe question of how these weights are determined.

Another attempt to resolve the problems of aggregation and enforcement isBecker's "rotten kid theorem" (1974, 1981). Becker considers the case of ahousehold with two members, a benefactor and a recipient. The benefactor,who is an altruist, transfers consumption to the recipient, a selfish individualwho cares only about personal consumption. Now suppose the recipient under-takes some action that raises his or her consumption but lowers that of thebenefactor (hence the "rotten kid" sobriquet). The benefactor could respond bylowering transfers so that the recipient's new level of consumption is below theoriginal level. Consequently, the recipient is not likely to behave rottenly in thefirst place. Thus, the preferences of the altruist and the preferences of the house-hold converge.

Unfortunately, the rotten kid theorem only holds under restrictive circum-stances (Bergstrom 1989; Haddad, Hoddinott, and Alderman 1994). It hasproved to be important, however, because it provides testable assumptions andbecause the underlying altruism has strong policy implications regarding theextent to which government policies are mitigated by private response.

Collective Models of Household BehaviorSeveral formulations of the unitary model contain an assumption that inequi-

table distribution of resources or leisure within a household represents a willingact on the part of all household members. Although models of the distribution ofresources within households are as much about sharing among generations asbetween genders, this assumption is viewed as particularly restrictive when it isapplied to decisionmaking between spouses. As one of the most noted critics ofthe unitary model, Folbre (1986, p. 251), comments:

The suggestion that women and female children "voluntarily" relin-quish leisure, education, and food would be somewhat more persua-

4 The Wbrld Bank Researvb Observer, vol. 10, no. I (February, 1995)

sive if they were in a position to demand their fair share. It is thejuxtaposition of women's lack of economic power with the unequalallocation of household resources that lends the bargaining power [col-lective model] approach much of its persuasive appeal.

Similar concerns have given impetus to several collective models that focus onthe individuality of household members. These models explicitly address thequestion of how individual preferences lead to a collective choice. These aresometimes referred to as bargaining models, but we prefer the more generic labelof "collective" models, partly because some important collective models do notexplicitly address bargaining. Moreover, the phrase can be neatly juxtaposedwith the term "unitary" models.

Common among the various collective models is their interest in directlyaddressing how individual household members reconcile different preferences.These approaches can be subdivided into two broad categories: those that relyon noncooperative relations; and those that rely on cooperative solutions.

In common with the unitary model, the cooperative approach begins by not-ing that individuals form a household when it is more beneficial to them thanremaining alone. Higher benefits could occur because forming a household is amore efficient way to produce household goods or because some goods can beproduced and shared by married couples but not by single individuals. Forexample, household formation may generate benefits such as "love" or "com-panionship." In any case, gains accrue from household formation, and theseneed to be distributed across the members. Where collective models depart fromunitary models is in the rule goveming this distribution.

The noncooperative approach (Ulph 1988; Kanbur 1991; Carter and Katz1992; Lundberg and Pollak 1993) relies on the assumption that individualscannot enter into binding and enforceable contracts with each other. Instead,individuals' actions are conditional on the actions of others. For example, Carterand Katz's fairly polar "reciprocal claims" model depicts the household as con-sisting of largely separate, gender-specific economies linked by reciprocal daimson members' income, land, goods, and labor. A wife's budget is separate fromher husband's; she responds to changes in her husband's allocation of laborsolely according to her own needs. The transfer of income between them estab-lishes the only link between the wife and husband. Similarly, in Lundberg andPollak's model, "each spouse makes decisions within his or her own sphere" (p.994) and responds to the other's decisions by altering the level of voluntarycontribution to shared goods.

In efficient cooperative models, it is assumed that household decisions arealways efficient in the sense that no one can be made better off without someonebeing made worse off. The models make no assumptions about how resourcesare distributed within households. A simplified version of this approach is asfollows. Suppose a household consists of two individuals. Once a decision hasbeen reached regarding expenditures on public goods, remaining household

Harald Alderman, Pierre-Andrd Chiappori, Lawrence Haddad, John Hoddinott, and Ravi Kanbur 5

income is allocated among the private goods according to a sharing rule. Thesharing rule in turn is affected by the incomes of the two household members.This model can illustrate how individual incomes affect the household's con-sumption of different goods (see Bourguignon, Browning, and Chiappori 1994).Further, it is possible to identify the household's sharing rule even if no individ-ual consumption is observed.2

A key feature of the efficient cooperative approach is that the rules regardingdistribution within households come from the data and are not assumed. This isespecially convenient for assessing the relevance of the alternative models(Chiappori 1992b).

Some cooperative models impose structure by representing household deci-sions as the outcome of some specific bargaining process and applying the toolsof game theory to this framework. Then the division of the gains from marriagedepend on the "fallback'" or "threat point," position of each member. Thesefallback positions are a function of extra-environmental parameters, that is,demographic, legal, and other macroeconomic conditions external to the house-hold. These include sex ratios in marriage markets, laws concerning alimonyand child support, changes in tax status associated with different marital states,and, in developing countries, the ability of women to return to their natal homesand prohibitions on women working outside the home (McElroy 1990, 1992).

Collective decisionmaking can be enforced in two ways. The first is throughthe threat of household dissolution. McElroy (1992) notes, however, that in thecontext of small daily decisions, it is not credible for either spouse to threatendivorce. She suggests that a second way to analyze decisions regarding short-runissues is to use differences in impatience to reach an agreement, with the nonco-operative solution acting as the threat point.

Policy Implications of Alternative Models of Distribution withinHouseholds

Is the distinction between unitary and collective models merely an arcaneacademic curiosity? Or do differences in how resources are distributed withinhouseholds, as the various theories imply, reflect appreciable differences in theoutcomes of policy measures?

Clearly, how resources are distributed within households affects the measure-ment of poverty and inequality. Consider a country in which the central govern-ment makes transfer payments to provincial or state authorities. The size ofthese transfers is determined by estimated levels of poverty. Does it matter ifpoverty is measured with reference to households or to individuals? If resourcesare equally distributed among household members, either measure will yield thesame estimate of the degree of poverty. As Haddad and Kanbur (1990) demon-strate, however, this no longer holds if resources are unequally distributedwithin the household. Drawing on individual- and household-level data on calo-ric availability in the Philippines, they estimate the incidence of poverty using the


income-gap poverty measures proposed by Foster, Greer, and Thorbecke (1984)and find that ignoring unequal distribution within households understates pov-erty by 18 to 23 percent.

Haddad and Kanbur's illustration is based on a poverty measure that paysparticular attention to food consumption, but the general result also holds whenincome is used instead of food. For example, Apps and Savage (1989) find thatthe welfare rankings of households in Australia depend critically on transfersbetween spouses. To the extent that there are empirical difficulties in identifyinghow resources are distributed within households, however, determining thechanges in rankings caused by these spousal transfers is difficult.

Does the analytical complexity associated with collective models offer anyadditional insights for policy interventions? We illustrate below four areas inwhich the choice of model is important.

Public Transfers to Individual Household Members

The claim that household decisions are not affected by the identity of theindividual earning income has been refuted in a number of settings. This hasobvious implications for policy, as the following quotations illustrate.

Many participants in the public debate concerning actual govern-ment transfers take it for granted that intrafamily distribution will varysystematically with the control of resources. When the British childallowance system was changed in the mid-i 970s to make child benefitspayable in cash to the mother, it was widely regarded as a redistribu-tion of family income from men to women and was expected to bepopular with women. (Lundberg and Pollak 1993, p. 989)

Indeed, so convinced did some Ministers become that a transfer ofincome "from the wallet to the purse" at a time of wage restraint wouldbe resented by male workers, that they decided at one point in 1977 todefer the whole child benefit scheme. (Brown 1984, quoted inLundberg and Pollak 1993, p. 989)

Most examples of income pooling revolve around the fact that women spendmore of their income on food and child care. Thomas (1990, 1992) finds that inBrazil, for example, the identity of the household member controlling incomeaffects nutrient intake, fertility, child survival, and young children's weight forheight. The results for child survival are particularly dramatic; increases in themother's unearned income raise child survival by twenty times that resultingfrom a comparable increase in the father's unearned income. We discuss suchevidence in greater detail in the next section.

The importance of potential policy failures arising from neglecting the identityof the transfer recipient is likely to grow as social safety nets are implemented toameliorate the short-run negative effects of economic adjustment. Newman,

Harold Aldermnan, Pierre-Andre Chiappori, Lawrence Haddad, John Hoddinott, and Ravi Kanbur 7

Jorgensen, and Pradhan (1991) found that in Bolivia the urban infrastructureconstruction project of the Social Emergency Fund bolstered the incomes of thepoor in a cost-effective manner. But only 2 percent of the participants in the fundwere women. The untested assumption seems to have been that fund incomewould trickle down to wives, mothers, and children or that they would be betterserved through credit and other programs in which female participation wassubstantial.

Public Transfers and Interactions between Household Members

The nature of interactions among household members determines whetherchanges in household behavior mitigate or enhance the effectiveness of publictransfers.

The potential for changes in household behavior to offset the effectiveness ofpublic transfers has been recognized as important since Barro's (1974) seminalpaper on family obligations and tax policy. Barro noted, for example, thathouseholds may respond to the introduction of a social security system byeliminating completely any private transfers from the young to the old.

If intergenerational sharing within a household is not purely altruistic, how-ever, households may respond differently. For example, Cox and Jimenez (1990)consider a hypothetical family with young members residing in towns and oldmembers living in rural areas. They consider the case in which transfers aremade from the young to the old, and individual consumption depends on thefamily's total income. Suppose a social security program is introduced that taxesthe young and subsidizes the old, leaving total family income unchanged. Ifindividuals in the household acted altruistically, this might well lead to a reduc-tion in urban-rural remittances (although consumption is unchanged). But itmay be that transfers from the young to the old were undertaken in exchange forservices (such as home production). Once social security payments are provided,rural residents might be less willing to provide services to urban residents. As aresult, the urban household members must transfer higher amounts to theirelders to retain the same services. This is the opposite result of that predicted bythe altruistic unitary model.

Policy Initiatives Directed to Individual Household Members

The unitary model implies that it does not matter how policy initiatives aredirected; the household will respond to that policy independently of the recipientof information or services. This assumption gives rise to two potential policyfailures: a resistance to particular policies that appear beneficial, and the unin-tended costs of policies that are adopted. Consider the consequences of thesetwo policy failures in terms of the adoption of new technology in developingcountries.3

The first example (mentioned in the introduction) described women's opposi-tion to recommendations to plant rice in Cameroon. In another instance, rural

8 The Wbrld Bank Research Observer, vol. 10, no. 1 (February, 1995)

households in Zambia were encouraged to intercrop maize, a male-controlledcrop, with beans, a female-controlled crop (Poats 1991). Diets would haveimproved as a result of the well-known complementary benefits from consumingthese two crops, and less work would have been required because the interplant-ing scheme reduced weeding time. Women refused to adopt this change, how-ever, because if they planted beans on land normally allocated to maize, theywould lose ownership of the beans.

By contrast, a project in Togo to encourage soybean production to supplementthe household diet with much-needed protein succeeded precisely because ittook into account the collective nature of household behavior (Dankelman andDavidson 1988). At the outset, the project was targeted toward women,through exchange visits and workshops organized in women's homes. Also,soybeans were not introduced as a cash crop, which would have changedwomen's status within the household. Instead, they were promoted as legumesthat could be used to make sauces. The result was that the crop remained in thehands of women, who in some cases were allocated small plots of land forcultivation.

Even when the neediest group is correctly identified, however, the policymight have adverse unintended impacts. Von Braun and Webb (1989) reportthat in the early 1980s in the Gambia, rice irrigation was introduced to an areaof swamp rice production to raise yields and commercialize the product. Al-though this initiative was designed to raise women's share of household income,it reduced that income because yield increases transformed the status of ricefrom a private crop under women's control into a communal crop under men'scontrol. Prior knowledge of the relative bargaining positions of men and womenmight have helped predict the outcome and enabled policymakers to redesign theprogram to meet the original goals.4

Long Reach of Policy Measures

The unitary model depicts as impotent several policy initiatives that neitherdirectly affect the technology of production nor affect household preferences butthat may in fact have a major impact on household allocation decisions. Anexample particularly relevant to developing countries is that of managing com-mon property resources, such as access to common grazing land.

Haddad and Kanbur (1992) outline the following model. A household hastwo individuals, each of whom produces output as the result of two tasks. Eachindividual is better at one or the other task, so it pays to specialize and cooperatein tasks. But how should the two individuals divide the benefits of cooperation?Suppose that the fallback option for each individual is to work alone. Nowsuppose that the government introduces a scheme that guarantees better accessfor all to common property resources. How will the government scheme affectinequality within the household? The income generated from improved access tocommon property resources might be higher than the income from working

Harold Alderman, Pierre-Andre Chiappori. Lawrence Haddad, John Hoddinott, and Ravi Kanbur 9

alone, but still less than the income from cooperation. Then, even if it is notactually used, more equitable access to common property resources actuallyimproves equality within the household. The scheme has a remarkably longreach-it equalizes distribution within the household by altering outsideoptions.

Several other features of this example are worth noting. The credibility of thescheme is at the heart of some of the policy debates on the extent of access tocommon property resources. Access might be rationed in a manner that imposeslimits and does not effect distribution within households. Moreover, the resultsof the model hold for several other policy interventions, such as Maharastra'semployment guarantee program in India, if the guarantee of employment acts asan inalienable property right.

Most important, Haddad and Kanbur's work illustrates the importance ofdistinguishing among different classes of collective models. For example, in acooperative model based on bargaining, if improved access to common propertyresources is guaranteed only to married women, distribution within householdswill be unaffected because women's position outside of marriage is unaffected.Here, only changes in access to common property resources for women outsideas well as inside marriage would alter the distribution of resources within house-holds. By contrast, if households are operating in a noncooperative setup,changing married women's access to common property resources would be suffi-cient to affect the position of women within the household.

In a similar manner, a policy might aim to change the distribution of transferswithin a household without influencing the distribution in the event the house-hold dissolves. Such a policy would be ineffective in the context of a cooperativebargaining model, but it might lead to a redistribution within the household if anoncooperative model holds. Lundberg and Pollak (1993) modeled a shift in thedistribution of public child support supplements from fathers to mothers but leftintact the distribution of support payments to mothers in the event of a divorce.In this example, the entitlement influenced the woman's position within mar-riage in a manner similar to the increased access to common property resources.Because the shift did not affect the situation in the event of household breakup(by assumption), it had no influence in the cooperative model.

Casting Doubts on the Unitary Model-Evidence

We have argued that the unitary household model faces serious theoreticalchallenges and that using this approach entails important policy implications. Inthis section, we review evidence that supports these challenges. We begin withsome informal evidence. This material is not necessarily nested within a formaltest procedure; nevertheless, it casts doubt upon certain aspects of the unitarymodel.

We then present some more formal evidence although we recognize that notall the studies are ideal. Particularly problematic is the fact that income reflects


past and current household choices and that measurement is complicated by theassumption-also held by many applications of the unitary model as well-thathousehold formation and sometimes composition can be regarded as predeter-mined. Other studies pertain only to specific regions or are based on relativelysmall samples. In the examples referred to here, aspects of the unitary model areseriously questioned or rejected using data drawn from twelve different coun-tries.5 It is this steady accretion of results that legitimates concerns about theunitary model.

Few researchers defend the unitary model on the basis of the validity of itsassumptions; these "do violence to reality" (Rosenzweig 1986, p. 233). Echoinga previous debate in economics, however, one could argue that realism is unim-portant. As Samuelson (1963, p. 232) put it: "a theory is vindicable if itsconsequences are empirically valid to a useful degree of approximation; the(empirical) unrealism of the theory 'itself; or its 'assumptions, is quite irrelevantto its validity and worth." Ultimately, then, the accumulation of the type ofevidence discussed here shifts the starting point in household studies. Again,echoing the earlier debate: "if the abstract models contain empirical falsities, wemust jettison the models, not gloss over their inadequacies" (Samuelson 1963,p. 236).

Informal Evidence

Many studies-from several disciplines and from both industrial and develop-ing countries-indicate that income is not pooled within a household. Otherarrangements that households adopt include systems where one person managesall finances and expenditures except for personal spending money; a "spheres ofresponsibility" system where, for example, a husband gives his wife a set amountfor purchasing specified commodities; and an "independent management" sys-tem, whereby each individual has income and is responsible for certain expendi-tures and no one has access to all household funds (Pahl 1983). Not surprisingly,the different ways in which households control income translate into differentpatterns of expenditures. Case study material from anthropological and socio-logical studies indicates that men spend more of the income they control for theirown consumption than do women. Alcohol, cigarettes, status consumer goods,even "female companionship" are noted in these studies. By contrast, women aremore likely than men to purchase goods for children and for general householdconsumption.

There is considerable evidence that domestic violence is prevalent in bothindustrial and developing countries and that it affects income distribution withinthe household. Domestic violence clearly refutes justification for the unitaryhousehold model based on altruism. Violence may underlie a dictatorial versionof a unitary household model. Jones (1986), for example, relates that respon-dents to a survey said that the threat of a beating influenced their decision towork. Rao (1994) finds that food purchases in India are influenced by domestic

Harold Alderman, Pierre-Andre Chiappori, Lawrence Haddad, John Hoddinott, and Ravi Kanbur 1 1

violence. The study also indicates that extra-environmental parameters mayaffect domestic violence and attempts to incorporate domestic violence into acollective model of household behavior. Tauchen, Witte, and Long (1991) pre-sent further evidence that community factors, such as access to public assistance,affect the probability of domestic violence.

Extra-environmental parameters have an explicit role in some collectivemodels and thereby provide additional indirect support for such models(Lundberg and Pollak 1993). A few recent studies use these conditions to sup-port bargaining models. An illustration is Rao and Greene's (1993) detailedanalysis of the impact of bargaining on fertility in Brazil. They find that fertilityis lower than average when the ratio of males aged 25 to 29 to females aged 15to 19 in the region is higher than average. It is reasonable to interpret this ratioas a measure of the availability of alternative spouses. As the ratio increases,women have a greater chance of remarrying if they leave their current house-hold, hence a greater ability to bargain for the smaller families they prefer.6

Formal Evidence

Several of the assumptions (or restrictions) of the unitary model do not holdwhen tested empirically. Three challenges to restrictions of the unitary model areconsidered here: nonpooling of labor income; strategic behavior in the contextof intergenerational relations; and the impact of one family member's laborchoices on those of another family member. Tests of the nonpooling of laborincome challenge an underlying assumption of the unitary model. Such tests alsoshow that policy measures might differ depending on the various methods bywhich household members control resources. The other tests of formal restric-tions are not phrased in terms of policy. The tests might seem to state obviouspoints, but they contribute evidence that challenges the unitary model.

THE INCOME-POOLING RESTRICTION. A key assumption of the unitary model isthe pooling of household income. Income pooling implies that the identity of theindividual earning the income has no effect on the household demand for goodsand leisure, except through the earning individual's choice between leisure andwork. Direct tests of the pooling of labor income have econometric problems.Some studies therefore focus on unearned income. In his study, which concludesthat not all households pool income, Schultz (1990, pp. 601-02) notes,

If non-earned income (or ownership of the underlying asset) influencesfamily demand behavior differently, depending on who in the familycontrols the income (or owns the asset), then the preferences for thatdemand must differ across individuals and such families must not com-pletely pool unearned income.

Similarly, Thomas (1990, 1992) finds that increased (nonlabor) income held bywomen leads to a greater share of the household budget devoted to expenditureson education, health care, and food.7


The use of unearned income to test the income-pooling hypothesis is subjectto econometric criticisms because it likely reflects past choices (Haddad, Hod-dinott, and Alderman 1994). Although none of the existing tests are definitive(the strongest proof may require an experimental design that randomly assignstransfers to males and females), many of the more recent studies have addressedthe possibility that the results are an econometric artifact of an unmeasuredfactor or reverse causality. Thomas observed (1992, forthcoming) that fathersand mothers behave differently toward daughters and sons. One explana-tion-albeit somewhat strained-may be that mothers with daughters choose towork or invest differently from those with sons. In his household fixed-effectsmodel, however, Thomas also includes several controls for unobserved individ-ual factors by taking differences across children and across parents. His resultsdo not disappear with these controls.

Taking a different tack, Hoddinott and Haddad (forthcoming) use traditionalcropping patterns in Cote d'Ivoire to model household behavior based on in-come sources. They also find that income is not pooled. When the share of cashincome received by wives in C6te d'Ivoire is increased, expenditures for foodincrease and expenditures on alcohol and cigarettes decrease. In their study ofinformal credit programs in Bangladesh, Pitt and Khandker (1994) also find thatcredit affects household education and consumption choices differently if it isgiven to women rather than to men. (This study also employs a methodologythat treats the availability of credit much as an experiment and uses that tocontrol for the fact that credit choices reflect household preferences.)

Collective models provide additional tests of income pooling. Moreover, theefficiency assumption within collective models strongly restricts the way inwhich different income sources may influence consumption patterns. Thus, theefficiency assumption provides additional tests of how changes in household andindividual income affect household consumption (Bourguignon, Browning, andChiappori 1994). Bourguignon and others (1992, 1993) construct a generalmodel that encompasses both the unitary and the collective frameworks asspecial cases. Using data from France and Canada, they find that householdincome is not pooled in either country. The restrictions of the collective modelsdo hold up; that is, the efficiency assumption is valid. Even more interesting isthe comparison, in the second paper, between a sample of couples and twosubsamples of singles. The unitary model fails for the couples, but not for thesingles. This would be the case if the unitary model failed because of a sharingprocess negotiated between family members.

INTERGENERATIONAL TRANSFERS. The unitary model implies that benefactorshave no incentive to behave strategically. Children, even rotten ones, do notattempt to raise their consumption at the expense of others because if they did,an altruistic benefactor would automatically reduce the size of the transfersmade to the children. Correspondingly, it is possible to test the hypothesis thataltruistic benefactors will not intentionally manipulate the behavior of the recip-


ient. If Becker's (1981) model holds, we should not find evidence of benefactorsbehaving strategically by using bequests to obtain attention or monetary trans-fers from their offspring. Lucas and Stark (1985) and Hoddinott (1992), how-ever, find that parents do behave strategically; in Botswana increased holdings ofinheritable assets lead to higher monetary transfers from nonresident familymembers (Lucas and Stark). The same pattern occurs in western Kenya in thecase of sons who anticipate receiving an inheritance (Hoddinott). Similarly, theresults in Cox and Jimenez's (1992, p. 167) study of Peru were found to be"inconsistent with the strict Barro-Becker altruism motive."

In a related vein, Altonji, Hayashi, and Kotlikoff (1992) formally test altruismby modeling the movement of children's expenditures when parents' incomechanges. They interpret their rejection of altruism as supporting the presence ofbargaining within households and as a direct challenge to the central assumptionin Barro's (1974) model.'

LABOR SUPPLY DECISIONS. Restrictions on decisions about labor supply (timespent earning income) provide some very specific tests of the effect of wages onthe labor supply of spouses. The unitary model implies that, for a given increasein total income, an increase in the husband's wage will affect the amount of timehis wife works to earn income exactly the same as an increase in the wife's wagewill affect the husband's time spent earning income.9 Evidence from the UnitedStates (Ashenfelter and Heckman 1974) rejects the equality of these effects (seealso Killingsworth 1983). Further, using panel data to control for unobservedfixed effects, Lundberg (1988) rejects the hypothesis that the labor supply of thehusband and wife is jointly determined, as predicted by the unitary model.

Similarly, recent work has focused on empirical tests characterizing the "col-lective approach. Chiappori (1988, 1992a) derives restrictions on labor supplyin a model where consumption and leisure are private goods and extends theanalysis to collective models. Fortin and Lacroix (1993) estimate a generalmodel of labor supply in which both the unitary and the collective frameworkscan be tested as special cases. Using data from Canada, they find that therestrictions of the unitary model are strongly rejected but that the analogousones from collective models are not. Browning and Chiappori (1994) analyzedata on consumption and find that the restrictions are rejected for couples butnot for singles (who do not have to share consumption). Moreover, the collectivegeneralization is not rejected for couples.

ConclusionsDespite the accumulated evidence that income is not pooled, the unitary

model, bolstered by ad hoc assumptions, retains an impressive ability to explainthe new body of evidence on inequality within the household. Moreover, inmany cases the choice of models will not affect either policy or research; Oc-cam's razor argues that in these cases the simplest approach be taken.

14 The Wbrld Bank Research Obsewer, vol. 10, no. I (February, 1995)

That said, we maintain that the burden of proof should be shifted onto thosewho would claim that the unitary model should be the rule and the collectivemodels the exception. Our intention is neither to discard the unitary model in itsentirety nor to ignore legitimate concerns about interpreting the evidence criticalof it. Although the rejection of the restrictions in some of the models discussed hasfew direct qualitative implications for policymakers, collective models themselvesdo have policy ramifications. Under many circumstances, acceptance of a unitarymodel of the household, when it is inappropriate, has more serious consequencesfor policy than does rejecting a unitary model when it is appropriate.

To reiterate an example, rejection of the collective model implies (erroneously)that targeting transfers to women is pointless; if the model is rejected when, infact, it is valid, an efficient means of directing resources to women and childrenmay be forgone. If the unitary model is rejected when it is sound, additionalcosts of targeting may be incurred. But most collective models imply equal orgreater investment in children from income using resources controlled bywomen than the unitary model implies. Thus, unless the costs of targetingprograms to women in poor households are significantly higher than targetingprograms to poor households as a unit, the available evidence may be consideredadequate to indicate that false rejection of the collective model is the moreserious error. I0

Equally important, a shift of theoretical focus will emphasize the need toquestion how resources and activities are allocated and monitored in programdesign and implication. In so doing, we surmise that more implications of house-hold allocation processes will emerge.

Implicit in our arguments is a view that household economics has not takenBecker (196S) seriously enough. "A household,' he wrote, "is truly a 'smallfactory': it combines capital goods, raw materials, and labor to clean, feed,procreate, and otherwise produce useful commodities." (p. 496) We, too, per-ceive the household as a factory, but like all factories, it consists of individualswho-motivated at times by altruism, at times by self-interest, and often byboth-cajole, cooperate, threaten, help, argue, support, and, indeed, occa-sionally walk out on each other. In fields such as labor economics, policy hasbeen informed by research that goes inside the "black box" of the factory anddiscusses individual incentives within a corporation. Those interested in thewelfare of individuals, especially in developing countries, may benefit from asimilar approach to the household.

NotesHarold Alderman is with the Policy Research Department at the World Bank; Pierre-Andr6

Chiappori is with the Departement et Laboratoire d'Economie Theorique et Applique(DELTA), Paris; Lawrence Haddad is with the International Food Policy Research Institute,Washington, D.C.; John Hoddinott is with the Lady Margaret Hall and Centre for the Studyof African Economies at the University of Oxford; and Ravi Kanbur is with the AfricaRegional Office of the Chief Economist at the World Bank.


1. Becker (1988) considers several policy issues, including growth and the intergenera-tional transmission of inequality, where an understanding of the distribution of resourceswithin households is important.

2. Apps and Rees (1988, forthcoming) offer an alternative cooperative model.

3. Other areas, such as targeting and environmental degradation, are discussed in Haddad,Hoddinott, and Alderman (1994).

4. Dey (1992) reports that more recent attempts by donors funding this project to safe-guard women's access to land were frustrated by the managers of the project, who sided withmale household heads in disputes over access to land. Thus, although attention to decision-making within households is necessary to avoid such unforeseen policy failures, this exampleindicates it may not be sufficient.

5. Additional examples can be found in Haddad, Hoddinott, and Alderman (1994), Blum-berg (1991), and Guyer (1980).

6. The relationship of sex ratios to allocation of output within marriage was also suggestedby Becker (1973). Alternative interpretations of Rao and Greene's results are offered inHaddad, Hoddinott, and Alderman (1994).

7. Thomas (1992) also explores the sensitivity of results to identification assumptions, forexample, by excluding pensions from nonlabor income.

8. For other evidence from the United States, see Cox (1987).

9. This statement is the labor analogue of the standard Slutsky restrictions of basic con-sumer demand theory.

10. Targeting to poor women should impose few additional information requirements oradministrative costs over targeting on the basis of poverty alone (and, if targeting to poorwomen is a first-stage filter, it may reduce costs). Thus, the most likely cost from suchtargeting might be the imposition of extra time burdens on women, which could reduce thewelfare of the women and, possibly, of their children. Most studies indicate, however, thatincreased labor income for women offsets any negative effects of reduced time for child care(Leslie 1988). Receiving transfers should require less time than laboring for increased income.Thus, the risks of presuming nonpooling of income in the absence of strong evidence to refuteit seem low.

ReferencesThe word "processed" describes informally reproduced works that may not be commonly

available through library systems.

Altonji, Joseph G., Fumio Hayashi, and Laurence J. Kotlikoff. 1992. "Is the Extended FamilyAltruistically Linked? Direct Tests Using Micro Data." American Economic Review 82(5,December):1177-98.

Apps, Patricia F., and Ray Rees. 1988. "Taxation and the Household." Journal of PublicEconomics 35(3, April):355-69.

. Forthcoming. "Collective Labor Supply and Household Production." Journal of Polit-ical Economy.

Apps, Patricia F., and Elizabeth Savage. 1989. "Labour Supply, Welfare Rankings and theMeasurement of Inequality." Journal of Public Economics 39(3, August): 335-64.

Ashenfelter, Orley, and James J. Heckman. 1974. "The Estimation of Income and Substitu-tion Effects in a Model of Family Labor Supply." Econometrica 42(1, January):73-85.

Barro, Robert J. 1974. "Are Government Bonds Net Wealth?" Journal of Political Economy82(6, November-December): 1095-1117.

Beaton, George, and Hussein Ghassemi. 1982. "Supplementary Feeding Programs for YoungChildren in Developing Countries:" American Journal for Clinical Nutrition 34(4, supple-ment):863-916.


Becker, Gary S. 1965. "A Theory of the Allocation of Time?' Economic Journal 75:493-517.

. 1973. "A Theory of Marriage: Part I." Journal of Political Economy 81(4, July-August):813-46.

. 1974. "A Theory of Social Interactions." Journal of Political Economy 82(6,November-December):1063-93.

1981. A Treatise on the Family. Cambridge, Mass.: Harvard University Press.1988. "Family Economics and Macro Behavior." American Economic Review 78(1,

March): 1-13.

Bergstrom, Theodore C. 1989. "A Fresh Look at the Rotten Kid Theorem-and OtherHousehold Mysteries.' Journal of Political Economy 97(5, October): 1138-59.

Blumberg, R. L. 1991. "Income under Female Versus Male Control: Hypotheses from aTheory of Gender Stratification and Data from the Third World." In Rae Lesser Blumberg,ed., Gender, Family, and Economy: The Triple Overlap. Newbury Park, Calif: Sage.

Bourguignon, Francois, Martin Browning, and Pierre-Andre Chiappori. 1994. "IdentifyingIntrahousehold Decision Processes?" DELTA, Paris. Processed.

Bourguignon, Francois, Martin Browning, Pierre-Andre Chiappori, and Valerie Lechene.1992. "Incomes and Outcomes: A Structural Model of Intrahousehold Allocation." DELTA,

Paris. Processed.

1993. "Intra Household Allocation of Consumption: A Model and Some Evidencefrom French Data.' Annales d'Economie et de Statistique/lInstitut National de la statistiqueet des &udes economiques 29(January-March): 137-56.

Brown, Joan. 1984. Family Income Support 2: Children in Social Security. Studies of theSocial Security System, No. 3. London, U.K.: Policy Studies Institute.

Browning, Martin, and Pierre-Andre Chiappori. 1994. "Efficient Intrahousehold Allocations:A General Characterization.' DELTA, Paris. Processed.

Carter, Michael, and Elizabeth Katz. 1992. "Household Resource Allocation When Income IsNot Pooled: A Reciprocal Claims Model of the Household Economy." Paper presented atthe IFPRi-World Bank Conference on Intrahousehold Resource Allocation: Policy Issuesand Research Methods. February. International Food Policy Research Institute, Washing-ton, D.C., Processed.

Chiappori, Pierre-Andre. 1988. "Rational Household Labor Supply." Econometrica 56(1,January):63-90.

. 1992a. "Collective Labor Supply and Welfare?' Journal of Political Economy 100(3,June):437-67.

1992b. "Income Sharing within the Household?' Paper presented at the IFPRi-WorldBank Conference on Intrahousehold Resource Allocation: Policy Issues and ResearchMethods. February. International Food Policy Research Institute, Washington, D.C.Processed.

Cox, Donald. 1987. "Motives for Private Income Transfers?' Journal of Political Economy95:508-46.

Cox, Donald, and Emmanuel Jimenez. 1990. "Achieving Social Objectives through PrivateTransfers: A Review?' The World Bank Research Observer 5(2):205-18.

. 1992. "Social Security and Private Transfers in Developing Countries: The Case ofPeru" World Bank Economic Review 6(1):155-69.

Dankelman, Irene, and Joan Davidson. 1988. "Women and Environment in the Third World:Alliance for the Future." London: Earthscan Publications in association with the Interna-tional Union for the Conservation of Nature (IUCN).

Dey, Jennie. 1992. "Gender Asymmetries in Intrahousehold Resource Allocation of Land andLabor: Some Policy Implications." Paper presented at the IFPRI-World Bank Conference onIntrahousehold Resource Allocation: Policy Issues and Research Methods. February. Inter-national Food Policy Research Institute, Washington, D.C. Processed.

Harold Alderman, Pierre-Andre Chiappori, Lawrence Haddad, John Hoddinott, and Ravi Kanbur 1 7

Folbre, Nancy. 1986. "Hearts and Spades: Paradigms of Household Economics." WorldDevelopment 14(2, February, special issue):245-55.

Fortin, Bernard, and Guy Lacroix. 1993. "Test of Noncooperative and Cooperative Modelsof Labor Supply." Working Paper. Universite Laval, Quebec. Processed.

Foster, James, Joel Greer, and Erik Thorbecke. 1984. "A Class of Decomposable PovertyMeasures." Econometrica 52(3, May):761-66.

Guyer, Jane I. 1980. "Household Budgets and Women's Incomes." Working Paper 28. BostonUniversity, African Studies Center, Boston, Mass. Processed.

Haddad, Lawrence James, John Hoddinott, and Harold Alderman. 1994. "IntrahouseholdResource Allocation. An Overview:' Policy Research Working Paper 1255. World Bank,Policy Research Department, Washington, D.C. Processed.

Haddad, Lawrence, and Ravi Kanbur. 1990. "How Serious is the Neglect of Intra-HouseholdInequality?" Economic Journal 100(402, September):866-81.

1992. "Intrahousehold Inequality and the Theory of Targeting.' European EconomicReview 36(2-3, April):372-78.

Hoddinott, John. 1992. "Rotten Kids or Manipulative Parents: Are Children Old Age Secu-rity in Western Kenya?" Economic Development and Cultural Change 40(3,April):545-65.

Hoddinott, John, and Lawrence Haddad. Forthcoming. "Does Female Income Share Influ-ence Household Expenditures?: Evidence from the C6te D'Ivoire." Oxford Bulletin ofEconomics and Statistics.

Jones, Christine W. 1986. "Intra-Household Bargaining in Response to the Introduction ofNew Crops: A Case Study from North Cameroon." In Joyce Lewinger Moock, ed., Under-standing Africa's Rural Households and Farming Systems. Boulder, Colo.: Westview Press.

Kanbur, Ravi. 1991. "Linear Expenditure Systems, Children as Public Goods and Intra-household Inequality." Discussion Paper 104. University of Warwick, Development Eco-nomics Research Centre, Coventry, U.K. Processed.

Kennedy, Eileen T., and Harold Alderman. 1987. "Comparative Analyses of NutritionalEffectiveness of Food Subsidies and Other Food-Related Interventions?' Washington, D.C.:International Food Policy Research Institute.

Killingsworth, Mark R. 1983. Labor Supply. Cambridge, U.K.: Cambridge University Press.Leslie, Joanne. 1988. "Women's Work and Child Nutrition in the Third World.' World Devel-

opment 16(11, November): 1341-62.Lucas, Robert E. B., and Oded Stark. 1985. "Motivations to Remit: Evidence from Bot-

swana." Journal of Political Economy 93(5, October):901-18.

Lundberg, Shelly J. 1988. "Labor Supply of Husbands and Wives: A Simultaneous EquationsApproach.' Review of Economics and Statistics 70(2, May):224-35.

Lundberg, Shelly, and Robert A. Pollak. 1993. "Separate Spheres Bargaining and the Mar-riage Market." Journal of Political Economy 101(6, December):988-1010.

McElroy, Marjorie B. 1990. "The Empirical Content of Nash-bargained Household Behav-ior." Journal of Human Resources 25(4, Fall):559-83.

1992. "The Policy Implications of Family Bargaining and Marriage Markets." Paperpresented at the IFPRL-World Bank Conference on Intrahousehold Resource Allocation:Policy Issues and Research Methods. Washington, D.C. Processed.

Newman, John, Steen Jorgensen, and Menno Pradhan. 1991. "How Did Workers Benefitfrom Bolivia's Emergency Social Fund?" The World Bank Economic Review 5(2,May):367-93.

Pahl, Jan. 1983. "The Allocation of Money within Marriage." Sociological Review32:237-64.

Pitt, Mark M., and Shahidur Khandker. 1994. "Household and Intrahousehold Impacts ofthe Grameen Bank and Similar Targeted Credit Programs in Bangladesh." World Bank,Education and Social Policy Department, Washington, D.C. Processed.

1 8 The World Bank Research Observer, vol. 10, no. 1 (February, 1995)

Pitt, Mark M., Mark R. Rosenzweig, and Md. Nazmul Hassan. 1990. "Productivity, Health,and Inequality in the Intrahousehold Distribution of Food in Low-Income Countries."American Economic Review 80(5, December): 1139-56.

Poats, Susan V. 1991. The Role of Gender in Agricultural Development. Issues in AgricultureDiscussion Paper 3. Washington, D.C.: Consultative Group on International AgriculturalResearch.

Rao, Vijayendra. 1994. "Wife-Beating, Its Causes and Its Implication for Nutrition Alloca-tion in Children: An Economic and Anthropological Case-Study of a Rural South IndianCommunity." Williams College, Department of Economics, Williamstown, Mass.Processed.

Rao, Vijayendra, and Margaret Greene. 1993. "Marital Instability, Intra-Household Bargain-ing, and Their Implication for Fertility in Brazil." University of Chicago, Population Re-search Center, Chicago, Ill. Processed.

Rosenzweig, Mark R. 1986. "Program Interventions, Intrahousehold Distribution and theWelfare of Individuals: Modeling Household Behavior." World Development 14(2, Febru-ary, special issue):233-43.

Samuelson, Paul A. 1956. "Social Indifference Curves." Quarterly Journal of Economics70(1, February):1-22.

. 1963. "Discussion." American Economic Review Papers and Proceedings 53:231-36.

Schultz, T. Paul. 1990. "Testing the Neoclassical Model of Family Labor Supply and Fertility."Journal of Human Resources 25(4, Fall):599-634.

Sen, Amartya K. 1966. "Labour Allocation in a Cooperative Enterprise." Review of Eco-nomic Studies 33(4, October):361-71.

Singh, Inderjit, Lyn Squire, and John Strauss. 1986. Agricultural Household Models: Exten-sions, Applications, and Policy. Baltimore, Md.: Johns Hopkins University Press.

Tauchen, Helen V., Ann Dryden Witte, and Sharon K. Long. 1991. "Domestic Violence: ANonrandom Affair." International Economic Review 32(2, May):491-51 1.

Thomas, Duncan. 1990. "Intra-Household Resource Allocation: An Inferential Approach."Journal of Human Resources 25(4, Fall):635-64.

. 1992. "The Distribution of Income and Expenditure Within the Household." Paperpresented at the IFPRI-World Bank Conference on Intrahousehold Resource Allocation:Policy Issues and Research Methods. February. International Food Policy Research Insti-tute, Washington, D.C. Processed.

Forthcoming. "Like Father, Like Son, or Like Mother, Like Daughter: ParentalEducation and Child Health." Journal of Human Resources.

Ulph, David. 1988. "General Noncooperative Nash Model of Household Consumption Be-havior." Discussion Paper 88-205. University of Bristol, Department of Economics, Bristol,U.K. Processed.

Von Braun, Joachim, and Patrick J. R. Webb. 1989. "The Impact of New Crop Technologyon the Agricultural Division of Labor in a West African Setting." Economic Developmentand Cultural Change 37(3, April): 513-34.


PARALLEL EXCHANGE RATES INDEVELOPING COUNTRIES

Miguel KiguelStephen A. O'Connell

Dual exchange rates and black markets for foreign exchange are commonin developing countries, and a body of evidence is beginning to emerge onthe effects that such parallel foreign exchange systems have on macro-economic performance. This article presents a simple typology of parallelsystems, discusses their emergence, and looks at why countries prefer thesearrangements to the main alternatives. The article examines the ability ofparallel markets to insulate international reserves and domestic prices fromshocks to the balance of payments. Drawing on the findings from eightdetailed case studies, the authors discuss the determination of the parallelpremium in the short and long terms, the relationship between the premiumand illegal transactions, and the fiscal effects of parallel rates. They comparethe experiences of countries that have attempted to unify their foreign ex-change markets and discuss the implications for policy alternatives.

P a arallel foreign exchange systems, in which an exchange rate determinedby the market coexists with one or more pegged or managed exchangerates, are common in developing countries. In some cases governments

respond to a balance of payments crisis by creating a legal parallel (or dual)foreign exchange market for financial transactions. The objective is to limit theshort-term effects of a depreciation of the exchange rate on domestic priceswhile maintaining some degree of control over capital outflows and interna-tional reserves. In other cases, extensive controls on foreign exchange restrictaccess to official markets and lead to the emergence of an illegal parallel market.The illegal market then grows in importance as the authorities respond to adeteriorating balance of payments by tightening and extending controls ratherthan reducing aggregate spending or devaluing the official exchange rate, orboth. (See box 1 for a glossary of terms used in this article.)

The importance of parallel markets and their effect on overall economic per-formance generally depends on the size of the parallel premium. Table 1

The World Bank Research Observer, vol. 10, no. 1 (February 1995), pp. 21-52( 1995 The International Bank for Reconstruction and Development/THE WORLD BANK 21

Box 1: Glossary of Terms

A nominal exchange rate is the price of a unit of foreign exchange in domestic currency. Anincrease in a nominal exchange rate is a depreciation of the domestic currency and a decreaseis an appreciation. When an exchange rate is pegged or managed, a discrete change in its offi-cial value is referred to as a devaluation or revaluation (a devaluation if the rate goes up and arevaluation when the rate goes down). A maxi-devaluation is a large, one-time devaluation ofa pegged or managed exchange rate.

A real exchange rate is a nominal exchange rate that is corrected for relative purchasingpower to yield a measure of external competitiveness. A real appreciation (depreciation)means an increase (decrease) in the purchasing power of domestic currency in foreign mar-kets relative to domestic markets (or in domestic traded-goods markets relative to domesticnontraded-goods markets).

A parallelforeign exchange system is one in which transactions take place at more than oneexchange rate and at least one of the prevailing rates is a legal or illegal, freely floating,market-determined rate (the parallel exchange rate). Two common examples are:

* a dual exchange rate system, in which the government assigns an important share of cur-rent account transactions to a commercial exchange rate (which we will sometimes callthe official rate; see note) and all remaining legal transactions, including capital accounttransactions, to an officially floating financial exchange rate (the parallel rate). The com-mercial rate is usually pegged or managed.

* a black market system, in which restrictions on transactions at the official exchange ratelead to the creation of an illegal market, in which transactions take place at a parallelrate. Note that a black market system can exist along with a dual exchange rate system ifsome transactions are rationed out of legal markets or excluded altogether.

The parallel premium is the percentage by which the parallel exchange rate exceeds theofficial exchange rate.

The parallel current account (and similarly parallel trade balance) is the difference be-tween the overall current account (which may be unobservable because of illegal trade) andthe current account balance that is reported to take place at the official or commercial ex-change rate(s). It is therefore the balance of private sector current account transactions thattake place (implicitly or explicitly) at the parallel exchange rate. The parallel current ac-count is primarily driven by illegal transactions, but it also includes any current accounttransactions that take place legally at an assigned parallel rate.

Note: The term official exchange rate refers to the most important legal rate in a black market systemor to the most important commercial exchange rate in a dual system. The term parallel exchange raterefers to the financial rate in a dual system or to the black market rate in a black market system. Where ablack market coexists with a dual system, one of the two parallel markets will typically be much larger,and the rate in this market is referred to as "the" parallel rate.

shows the black market premium for a sample of countries from 1970 to 1989.The premiums in these markets-which have been more important in Africa andLatin America than in Asia-increased as countries responded to macro-economic imbalances and severe balance of payments problems by tighteningcontrols on foreign exchange transactions. Even countries with low or moderatepremiums in the 1970s experienced episodes of relatively high premiums in the1980s, but in some countries, such as Ghana (1980-86) and Tanzania(1973-86), the premium remained high for five years or more.

22 The World Bank Research Observer, vol. 1 0, no. 1 (February, 1995)

Table 1. The Black Market Premium, Selected Countries(median values of annual end-of-year figures)

Largest premiumCountry 1970-79 1980-89 over total period

Low PremiumThailand -0.2 -1.5 5.1Belgium 1.0 1.3 9.9France 0.7 2.8 12.6Italy 2.8 1.4 13.4Indonesia 2.2 3.4 15.5Colombia 4.4 12.7 28.3Turkey 9.1 8.4 52.8Mexico 0.0 17.7 66.0Venezuela 0.4 75.2 213.0

Moderate premiumKenya 16.8 15.2 44.9Brazil 11.1 43.1 173.0Dominican Republic 26.5 36.0 213.0Bolivia 5.5 17.6 293.1

Large premiumPeru 51.2 27.0 278.9Sudan 85.4 78.8 344.4Zambia 102.5 40.8 361.9Tanzania 95.5 214.3 809.1Ghana 66.3 142.0 4,263.7

Note: The premium is the difference between the parallel (P) and official (E) exchange rates, definedhere as 100 x [(P - E)/E].

Source: Parallel exchange rates from International Currency Analysis, Inc. (various years) and Kauf-mann and O'Connell (1991) for Tanzania. Official exchange rates from IMF (various issues).

What explains differences in the size of the premium across countries and overtime? How effective have parallel foreign exchange markets been in balancingcapital inflows and outflows while controlling inflation? Why do countries unifythe foreign exchange market, and what policies are required to sustain unifica-tion? How does the premium affect the fiscal position? While there is a richtheoretical literature analyzing these issues (Lizondo 1990 and Agenor 1992provide an overview), less is known about how parallel foreign exchange mar-kets work in practice. This article summarizes the empirical evidence on thesequestions based on a World Bank study of eight countries-Argentina, Ghana,Mexico, Sudan, Tanzania, Turkey, Venezuela, and Zambia (Kiguel, Lizondo,and O'Connell, forthcoming).1

Economists advocated dual exchange rates for industrial countries early in thepostwar period (Triffin 1947) and again in the transition to floating rates in theearly 1970s as a way to protect international reserves and insulate the prices oftraded goods from external shocks. More recently, Dornbusch (1986a) andDonbusch and Kuenzler (1993) advocated use of dual exchange rates in develop-ing countries as a way to prevent transitory shocks to the capital account from

Miguel Kiguel and Stephen A. O'Connell 23

affecting prices and wages. The theoretical advantages of dual rates are straight-forward. A dual system is more effective than a single pegged or managed rate atinsulating international reserves from capital outflows, because these lead to adepreciation of the parallel rate rather than to a loss in reserves. It is moreeffective than a single floating rate in limiting the impact of capital outflows ondomestic prices, because current account transactions are conducted at the(pegged or managed) commercial exchange rate. In principle, similar propertiesapply to black market systems; exchange controls protect international reserves,while some current account transactions take place at the official rate.

In practice, parallel markets provide temporary relief at best from the tradeoffbetween price adjustment and reserve adjustment under unified exchange rates.Severe exchange controls are required if parallel markets are to meaningfullyinsulate reserves from capital flows, and such controls become virtually impossi-ble to enforce when the incentives for capital movements are strong and persis-tent. Insulation of domestic prices is also partial and temporary, and attempts toenhance the insulation of prices-for example, by maintaining convertibility fora wide range of current account transactions-undercut the insulation of re-serves. Parallel foreign exchange markets are emphatically not an effective wayto maintain low inflation in the longer term.

Although the restrictions that underlie parallel foreign exchange markets canprovide a useful safety valve against transitory external shocks, they often endup supporting highly distortionary policy regimes. Thus capital controls featureprominently in systems of financial repression, the costs of which have beenemphasized by Fry (1988), McKinnon (1973), Shaw (1973), and others. Ex-change controls on the capital and the current accounts are often used to but-tress chronically overvalued official exchange rates. Persistent overvaluationdiscourages exports and generates both opportunity (through the rationing offoreign exchange) and political pressure for selective protection of inefficientimport-substituting industry (Bhagwati 1978). Recent cross-country studies thatuse the premium as a summary measure of government-induced economic dis-tortions find a relation between large and persistent premiums and substantiallyslower economic growth. Barro and Lee (1993), for example, associate a 10percent premium with a reduction of nearly half a percentage point in the annualgrowth rate of gross domestic product (GDP) (see also Easterly 1994 and Fischer1993).

How is the premium determined in the short run and over time? Changes inthe parallel exchange rate-and therefore the premium-play two roles in theparallel foreign exchange market. First, they change the relative value in domes-tic currency of domestic and foreign assets held by the private sector, helping tobring about a temporary equilibrium between private demand for these assetsand existing supplies. This revaluation implies that shifts in the relative demandfor foreign and domestic assets are the primary determinants of the premium inthe short run. Second, changes in the premium alter the incentives for illegaltransactions. Because illegal trade is an important channel for accumulating or


decumulating foreign assets, this effect on illegal transactions feeds back into theshort-run determination of the premium in future periods.

The "portfolio-balancing" role of the parallel premium in the short run helpsexplain its volatility and rapid reaction to macroeconomic developments, in-cluding new information about the likely path of macroeconomic variables.Overexpansionary monetary policy, an increase in the budget deficit, or anappreciation of the real official exchange rate all cause an increase in the pre-mium. News of an impending devaluation accounts for some of the most spec-tacular spikes in the premium. Official devaluations typically reduce the pre-mium when they go into effect, but the duration of the reduction depends a greatdeal on the accompanying macroeconomic policies. Taken in isolation, a one-shot devaluation of the official exchange rate has virtually no long-run effect onthe premium.

How do parallel exchange rates feed back into the economy? The first channelis through illegal trade. Sustained increases in the premium encourage the diver-sion of exports from official to unofficial channels and the reverse for imports.Premium increases therefore tend to worsen the official trade balance (undercut-ting the insulation of reserves) and produce an accumulation of private netforeign assets through the parallel current account balance. The fiscal balancemay also be affected, altering the growth of government liabilities over time.Trade tax revenues, for example, typically fall with sustained increases in thepremium, as officially remitted export revenues fall and the authorities furthercompress imports to avoid excessive reserve losses.

The second channel through which parallel rates feed back into the economyis through prices, because the parallel rate has a direct effect on the domesticprices of goods that enter or leave the economy through parallel channels. Incases of extreme import rationing, for example, the domestic prices of importedgoods tend to be determined solely by world prices and the parallel rate. Buteven when no current account transactions take place at the parallel rate,changes in that rate alter total financial wealth and feed through to domesticprices if nominal spending is related to financial wealth.

Given the self-limiting effectiveness of parallel exchange rates-with highpremiums tending to generate responses that undercut the system-and theirmore adverse, distortionary consequences in many cases, it is not surprising thatthe controls underlying parallel foreign exchange markets are often liberalized atsome point in favor of a unified foreign exchange market. The speed with whichmarkets are unified and the type of exchange rate regime ultimately adoptedvary considerably. Experience shows that unification often occurs quickly, dur-ing a macroeconomic crisis, when parallel exchange rates no longer protectinternational reserves. This was certainly the case in Argentina in 1989 and inVenezuela in 1988. There are also cases of successful gradual unification, how-ever, especially in African countries, where it has often moved in tandem withprice deregulation and trade liberalization. Ghana and Tanzania followed thisapproach quite successfully.


The key to successful unification is choosing an exchange rate regime con-sistent with underlying fiscal and monetary policies. Countries that maintainlarge money-financed budget deficits cannot unify into a fixed exchange rate,because inflation would quickly lead to an overvalued real exchange rate.Whether countries unify to a flexible exchange rate system or a crawling peg isof secondary importance.

The theoretical literature on parallel exchange rate systems has recently em-phasized a potentially adverse link between exchange rate unification and infla-tion operating through the fiscal accounts (Pinto 1989). This view holds that thepublic sector is often a net buyer of foreign exchange from the private sector andthat these transactions form an important component of the public sector'soverall domestic borrowing requirement. In a parallel system, most of thesetransactions take place at the official exchange rate, and unification tends todepreciate this rate in real terms (the depreciation will be sharper the larger thepremium and the greater the influence of the parallel rate on domestic prices).Under these conditions, unification affects money growth and inflation by exert-ing a direct valuation effect on the public sector's real borrowing requirement.Very little evidence is available on the magnitude of this effect in practice. On thebasis of simple calculations for a small group of countries, we find that the fiscaleffect-generalized in one case to include other budgetary effects of unifica-tion-can be large and is often more favorable than the literature has suggested.

The Development of Parallel Foreign Exchange Markets

Parallel foreign exchange markets develop in one of two ways. In the first, theeconomy starts from a unified foreign exchange market, and the authoritiesadopt an official dual exchange rate system in response to a balance of paymentscrisis. In the second, a parallel (typically unofficial) market emerges gradually asthe authorities impose restrictions on access to foreign exchange in an effort tomaintain an overvalued exchange rate. As pressures on the official exchange ratemount, controls on foreign exchange are tightened, and eventually the illegalmarket becomes macroeconomically important.

Varieties of Parallel Markets

The details vary from case to case, but the array of parallel systems in devel-oping countries can be reduced to a simple classification based on the coverageand legality of the parallel rate. All the regimes examined in the World Bankstudy imposed restrictions on the capital account (mainly on outflows); theprimary distinction is the degree to which controls permeate the current account(table 2). In some episodes capital account transactions dominate the parallel


Table 2. Classification of Parallel Exchange Rate Episodes in the World Bank CaseStudy, 1970-90

Coverage of the parallel rate

Mainly capital account Capital and current accountsTransactions Country Years Country Years

Legal or toleratedArgentina 1981-89 Ghana 1983-87Europea Various Tanzania 1984-90Ghana 1987-90 Venezuela 1983-89Mexico 1982-88 Zambia 1987-88Turkey 1980-84

IllegalGhana Before 1983Sudan 1970-90Tanzania Before 1984Turkey 1970-79Zambia Except 1987-88

a. Includes Belgium, 1957-90; France, 1971-74, and Italy, 1973-74.Source: Kiguel, Lizondo, and O'Connell (forthcoming).

foreign exchange market; in others both types of transaction are conducted inthe parallel market. Cases in which controls apply solely to the current accountare not observed in practice; such controls would be virtually unenforceable inthe absence of convertibility restrictions on the capital account because foreignexchange legally traded or held for portfolio purposes can easily be channeled tocurrent transactions. The available evidence-from Indonesia, Uruguay, and theCFA countries in Africa, all with open capital accounts-indicates that, in theabsence of capital controls, the parallel market remains thin and the premium nil(or small), even in the presence of substantial tariffs, quantitative restrictions,and illegal trade.

A secondary distinction concerns the legality of transactions at the parallel rate.Formally, such transactions are either legal or illegal. Because of the high costs ofenforcement, however, governments typically tolerate a substantial amount of il-legal parallel market activity. Attempts to suppress parallel markets are notunusual, but success in such efforts-and the commitment to continue them-istypically short-lived. We therefore distinguish between systems in which transac-tions at the parallel rate are either legal or illegal but largely tolerated, and those inwhich a substantial threat of enforcement is present most of the time (which wasthe case, for example, in Ghana before 1983). (See table 2.)

In practice, coverage and legality are matters of degree rather than discretecategories. Moreover, parallel systems can and do evolve as the degree of cover-age or the legality of transactions at the parallel rate change. The classificationin table 2 is therefore imprecise, and even relatively dramatic sub-episodes, suchas Tanzania's crackdown on illegal activity in 1983, may not show up as qualita-tive shifts in the table. But major changes are relatively easy to identify. Tan-


zania's adoption of an own-funds scheme in 1984, for example, effectivelylegalized existing private holdings of foreign exchange and signaled a new toler-ance for transactions in the parallel market.

Official Parallel Exchange Rates

Governments adopt dual exchange rate systems not only to deal with balanceof payments crises, but also to increase the effectiveness of monetary policy or,somewhat surprisingly, to help unify the foreign exchange market (in thosecountries where the parallel market is illegal). Nevertheless, a balance of pay-ments crisis is the most common reason. Generally, adoption of a dual exchangerate is seen as a transitional effort to limit the inflationary effect of a devaluation(Flood 1978, Lizondo 1987, Kiguel and Lizondo 1990). The theoretical advan-tages of dual rates in such a situation are straightforward; typically, currentaccount transactions take place at the official exchange rate, and capital accounttransactions at a market-determined exchange rate. Thus international reservesare unaffected by capital outflows (which lead instead to a depreciation of theparallel rate). Relative to a unified float, the effect of a dual system on domesticprices is limited because all current account transactions take place at the official(pegged or managed) exchange rate.

Examples of dual systems in developing countries are numerous. Most are inLatin America, including those in Argentina, Mexico, and Venezuela in the1980s. Venezuela's action in 1983 was modeled in part on a successful interven-tion in 1960, when the government phased in a three-year program to devaluethe currency by 35 percent. The aim in 1983 was similar; a three-year transitionto a unified fixed rate was envisioned, with a cumulative devaluation of 39.5percent. (In the end, the program lasted six years, with a cumulative devaluationof over 700 percent.) Among industrial countries, Italy and France shifted tem-porarily to dual systems in the early 1970s to address external problems associ-ated with the collapse of the Bretton Woods fixed exchange rate system; seeMarion (in Kiguel, Lizondo, and O'Connell, forthcoming).

In most cases the adoption of a dual system was prompted by fears that thenominal depreciation required to restore external balance in the short run wouldresult in an unacceptable and potentially permanent bout of inflation. Thiscould occur through a dramatic fall in real wages that would force the authori-ties to choose between accommodating increased wage demands and inducing asevere recession. In Brazil, for example, the devaluations of the mid- and late1970s were associated with a permanent increase in inflation (Kiguel and Liv-iatan 1988). In Mexico and Argentina such concerns led the authorities to usethe commercial (official) rate as a nominal anchor whose fluctuations were to beminimized or at least smoothed out over time. Figure 1 shows the Mexicanfinancial (parallel) real exchange rate overshooting its medium-run level whenthe dual regime was adopted; when the authorities devalued the commercial ratefour months later, the magnitude of the devaluation was considerably smaller.


Figure 1. Real Excbange Rates, Mexico, 1982-1983(increase = depreciation)

Pesos per U.S. dollar700 -

600

500 \_

400-

300

200

1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12

Real official exchange rate ------- Real parallel exchange rate

Note: Real exchange rates are calculated as nominal exchange rates deflated byMexico's consumer price index.

Source: Official exchange rates and Mexico's consumer price index from JMF (variousissues); parallel rates from International Currency Analysis, Inc. (various years).

Some countries have adopted a dual exchange rate system on a long-termbasis, using the parallel market as a safety valve to deal with short-term capitalflows. This policy has permitted the authorities to exert more control overmonetary policy and has reduced the volatility in interest rates (Flood andMarion 1982). Belgium, for example, kept a dual system introduced in 1957 inplace until 1990. The Syrian Arab Republic and the Dominican Republic alsomaintained dual exchange rate systems for prolonged periods, and some coun-tries, such as Colombia, adopted quasi-official, widely tolerated parallel foreignexchange markets to support conservative monetary policy.

Increasingly, an official parallel market is introduced as a transitional deviceto unify the foreign exchange market, particularly in countries eager to phaseout exchange controls on current account transactions. Most of these countrieshave a three-tier foreign exchange market: selected current account transactionstake place at the official pegged exchange rate, others are assigned to an officialparallel rate, and the rest take place at an unofficial exchange rate. Ghana and


Zambia, for example, held official foreign exchange auctions for a short time inthe 1980s that applied to a substantial share of external transactions.

Unofficial Parallel Markets

Unofficial, or black, markets exist as a matter of course in most developingcountries because of restrictions on capital account transactions in the officialforeign exchange market. Although capital controls may be in place for the samereasons as dual exchange rates-to insulate the economy from temporary shocksto the capital account-they are also intended to influence the allocation ofprivate wealth between domestic and foreign assets. The underlying motivationis often fiscal: domestic assets can be taxed more easily than foreign ones. This isparticularly true of money balances, which can be taxed through inflation atminimal administrative and (when inflation is moderate) political cost. Thusgovernments hope to increase the demand for domestic assets by preventing orslowing down the accumulation of foreign assets. (See Fry 1988, McKinnon1973, and Shaw 1973 for studies on the costs of financial repression.) Currencyblack markets frustrate this hope to some degree by providing a channel foraccumulating foreign assets. But evidence suggests that, even in the long run,capital controls retain some effectiveness in constraining the liquidity of foreignbalances and shifting demand toward domestic assets (Adam, Ndulu, and Sowa1993; Giovannini and de Melo 1993).

In contrast to a dual system installed temporarily as part of an overall policyadjustment, the emergence of a black market often reflects a systematic biasagainst devaluation of the official exchange rate. When expansive monetaryand fiscal policies raise the rate of inflation and lead to an overvalued exchangerate, the balance of payments gradually worsens. If the government fails tocorrect this imbalance by tightening macroeconomic policies or devaluing theofficial exchange rate, it is forced to restrict access to foreign exchange at theofficial rate. Popular expectations that the authorities will impose a maxi-devaluation or tighten foreign exchange controls add to the demand for foreignexchange by encouraging importers to accumulate inventory and promotingthe substitution of domestic assets for foreign exchange. The supply is pro-vided by exporters, tourists, and workers abroad (through remittances), all ofwhom may find it profitable to divert foreign exchange from the official to theillegal market.

Many developing countries fit this pattern. In the Dominican Republic thepremium remained low during the early 1970s, when sound macroeconomicpolicies were in place, but then rose to more than 100 percent as fiscal deficitsmounted and capital flight became pervasive. In Turkey the black market forforeign exchange that emerged in the early 1940s expanded significantly in the1970s, driven by mounting macroeconomic imbalances and an overvalued offi-cial exchange rate. In Ghana inflation drove an increase in illegal foreign ex-change transactions in the 1970s that increased the premium on the black mar-


ket from 50 percent in 1974 to more than 2,000 percent in 1982. A similarsequence was recorded in Tanzania. Figure 2A shows the decline in Tanzania'sinternational reserves following the collapse of the coffee boom in 1978 and theill-fated trade liberalization of the same year. When the authorities tightenedexchange controls (rather than devaluing), the rising premium and increasingovervaluation exacerbated the decline in official exports and produced furtherdeclines in import allocations (figure 2B).

Why did the authorities prefer foreign exchange controls rather than across-the-board devaluations or tighter macroeconomic policies to deal with balanceof payments problems? In Tanzania reluctance to devalue was rooted in argu-ments that a devaluation would generate only a weak supply response in exportsof agricultural goods and would impose a contractionary and inflationary shockon the import-dependent industrial sector. By the early 1980s these argumentsgave way to a lingering ideological opposition to market-determined prices andto concerns that devaluation would trigger political instability (Loxley 1989;Hyden and Karlstrom 1993). In Ghana the maxi-devaluation of 1971 was fol-lowed immediately by a military coup. It is safe to say that the association ofthese two events, whether causal or not, exercised a strong influence againstsubsequent devaluations in Ghana.

Insulation in Theory and Practice

As noted earlier, the use of parallel markets has been linked to their ability toinsulate the domestic economy from external shocks. How well do these mar-kets insulate international reserves and prevent transitory shocks to the capitalaccount from affecting prices and wages?

Insulating International Reserves

To support the official exchange rate in a unified system, the central bankmust be willing to meet the excess demand (or supply) for foreign exchange atthe official rate. The outflow of reserves is then the sum of the excess demandfor trade transactions (the current-account deficit) and the excess demand forasset transactions (the capital-account deficit). By assigning private capital-account transactions to a market-clearing parallel rate, a parallel system allowsthe authorities to limit their net intervention, particularly with respect to short-term capital movements, which are typically the most volatile transactions.Reserves are used only to finance current transactions, so shocks to the privatecapital account are absorbed by the parallel exchange rate rather than by officialreserves. The assignment to the parallel rate can be explicit-as when financialtransactions are channeled through a legal dual rate-or implicit, as when thegovernment prohibits foreign assets from being held or traded altogether.

The effectiveness of this technique is limited, however. First, the authoritiesmay be tempted to adjust the official rate to prevent an excessive rise in the pre-


Figure 2. External Balance, Trade Flows, and Parallel Premium,Tanzania, 1970-88

A. International Reserves and Parallel Premium(percent)

Reserves Premium9 ~~~~~~~~~~~~~~600

0 500

6 / 400

5 ~~~~~~~~~~-I300

3 /2002 o

1~~~~~~~~~~~~~~~~~~~0

0- l l l I l I I I I -

1970 1972 1974 1976 1978 1980 1982 1984 1986 1988

Interational reserves as a share of GDP -Parallel premium

B. Trade Shares

Percentage of GDP50 -

40-

30 -

20 -J

10

1970 1972 1974 1976 1978 1980 1982 1984 1986 1988

Exports of goods and services-----Imports of goods and services

Source. Parallel prernium, Kaufmnann and O'Connell (1991); reserves, GDP, and tradeflows, World Bank data.

32 Tfhe World Bank Research Observer, vol. 1 0, no. 1 (February, 1995)

mium. In Mexico intervention was intended to limit the incentives for illegaltransactions and avoid a vicious circle driven by the private sector's expectationsthat devaluation was likely (Dornbusch 1986a). Partly because of this official in-tervention, capital flight was actually larger after the dual system was introducedthan before; see Kaminsky (in Kiguel, Lizondo, and O'Connell, forthcoming).

Second, because a spread between the parallel and official exchange ratesprovides incentives for illegal transactions (Fleming 1971, Lanyi 1975, Bhag-wati 1978), traders may overinvoice imports and underinvoice exports, causinga loss of reserves. Through smuggling, the prices of tradable goods reflect theparallel exchange rate, limiting the insulation of domestic prices. Outflows ofreserves may also occur through legal channels. Kamin (in Kiguel, Lizondo, andO'Connell, forthcoming) reports that exporters in Argentina aggressively usedspecial export financing facilities during the early 1980s, and Marion (1994)reviews the use of trade credits in France and Italy in the early 1970s.

Strict rationing of foreign exchange can, of course, ensure that internationalreserves are protected. In the early 1980s reserves in Ghana and Tanzaniadropped nearly to zero, and, although the exchange rate was nominally pegged,the central bank was unable to intervene to support the domestic currency. Theexchange control regime in such cases does not imply that the central bank iscommitted to finance a payments imbalance; however, reserves tend to be insu-lated at the expense of prices.

Insulating Prices

Another reason for using parallel exchange rates is to anchor domestic prices.During a balance of payments crisis, the exchange rate tends to be extremelyvolatile and is likely to exceed its long-run equilibrium level. A dual exchangerate system can limit the inflationary effect of depreciation by allowing thefinancial (parallel) rate to absorb most of the pressure. Devaluation eventuallyoccurs, but the exchange rate adjustment is smaller and smoother. Such systems,however, are most effective during the first six to nine months of the crisis; thedegree of insulation decreases significantly after that. Complete separation ofthe two foreign exchange markets becomes difficult to enforce, and the parallelrate becomes more important in determining prices.

The degree of influence of the parallel rate depends in large part on how muchcontrol the authorities exert over the current account. When foreign exchange isrationed, the official rate becomes increasingly irrelevant. Chhibber and Shafik(1991) and Younger (1993), for example, report that in Ghana in the early1980s the official exchange rate played virtually no role in domestic price forma-tion in the last few years before macroeconomic reforms. This lack of influencehad two striking implications. First, maxi-devaluations associated with thosereforms had minimal cost-push effects on domestic inflation and instead simplyundercut the rents being accrued by the recipients of official foreign exchange.Second, because the cumulative effects of overvaluation and high premiums had

MiguelKiguelandStephen A. O'Connell 33

reduced official exports to near zero, the amount of foreign exchange availableat the official rate had shrunk dramatically. Thus political resistance to devalua-tion-both from the formal wage sector and from the recipients of officialforeign exchange-was almost certainly much weaker than it had been whenexchange controls were introduced.

Ghana is an extreme example; in less dramatic instances domestic prices ontraded goods reflect a mixture of official and parallel exchange rates. Even so,the role of the parallel rate is typically out of proportion to the share of goodstraded in the parallel market for two reasons. First, the marginal source offoreign exchange for imports is the parallel-rather than the official-market.As a result exchange controls amount to an import quota, driving the domesticprice of imports up to the cost of obtaining the item through alternative chan-nels. Second, movements in the parallel rate can feed into the prices of non-traded goods through their effect on wealth and aggregate spending; seeGuidotti (1988) and Ghei and Kiguel (in Kiguel, Lizondo, and O'Connell,forthcoming).

Parallel regimes are unable to anchor the nominal exchange rates in the face ofinflationary pressures caused by loose fiscal and monetary policies. In countriesthat fail to control monetary growth, the parallel rate depreciates and domesticinflation continues despite the presence of a fixed official exchange rate. Theresulting overvaluation increases pressure on the balance of payments, and theleakages and distortions associated with increasingly tight controls eventuallyforce a devaluation. In Venezuela, for example, inflation did not increase signifi-cantly during the early phase of the dual system, but it eventually rose from 10percent in 1982 to roughly 30 percent in 1988 in the absence of policies tocontrol external imbalances. Similarly, inflation in Argentina rose from 150percent in 1982 to more than 600 percent in 1985 as a result of large budgetdeficits and protracted problems in the balance of payments. In the longer term,domestic policies determine inflation, and little is gained by having a parallelforeign exchange market.

In practice then, the degree of insulation of both reserves and prices envi-sioned in the classic dual exchange rate arrangement does not prevail in develop-ing countries. The insulation provided is partial at best and becomes less effec-tive over time, declining as the average premium rises.

Determinants of the Parallel Rate and the Premium

The stock-flow model that forms the core of the theoretical literature on dualexchange rates and black markets (surveyed in Agenor 1992 and Lizondo 1990)borrows its two central elements from an earlier literature on flexible exchangerates. The stock element comes from a view of the parallel exchange rate as anasset price, determined in the short run by the requirement that existing stocksof domestic and foreign financial assets be willingly held. Because asset demandsdepend crucially on anticipated yields, the parallel exchange rate is a forward-


looking variable, responding to news about the likely time path of any of themacroeconomic variables capable of affecting its evolution.

The flow element describes the interaction between the parallel exchange rate(or premium) and the evolution of private asset stocks over time. Two key flowsare the parallel current account, which affects holdings of private net foreignassets, and the fiscal deficit, which is a primary source of growth in domesticmoney balances. The main factors that affect these flows are, first, trade taxes,foreign exchange rationing, and the real exchange rate, for example, in the caseof foreign assets; and, second, determinants of the real fiscal deficit in the case ofdomestic assets. These become the primary determinants of the premium in thelong run.

The most striking evidence of the forward-looking behavior of the parallelexchange rate comes from movements in the premium in advance of devalua-tions. Because one-shot devaluations have little long-term effect on the pre-mium, news of an impending devaluation is associated with expectations thatthe parallel exchange rate will depreciate. Foreign assets therefore look moreattractive in the short run, inducing capital flight and raising the premiumimmediately. The premium then falls when the devaluation occurs; indeed, itmay fall by nearly the full amount of the official depreciation, because much ofthe adjustment of the parallel rate has already taken place.

Figure 3 shows this effect at work in the run-up of the Tanzanian parallelpremium in anticipation of the devaluation of April 1986. This move markedthe end of a protracted struggle over exchange rate policy that had produced twofailed International Monetary Fund programs in the late 1970s and early 1980sand culminated in the resignation of President Nyerere in 1985. The governmentimplemented macroeconomic reforms in 1984, including a minor devaluationand the introduction of an own-funds scheme for imports-a mechanism thatpermitted individuals to receive import licenses without revealing the source oftheir foreign exchange (O'Connell 1991). But it was clear that a major agree-ment with the International Monetary Fund was in the cards and that it wouldbe accompanied by a maxi-devaluation. At its predevaluation peak, the pre-mium reached nearly 800 percent.

The analysis of such episodes is complicated because causality is likely to runfrom the premium to the official exchange rate as well as from expectations of theofficial exchange rate to the premium. The rise in the premium in 1985 and 1986prompted the diversion of official foreign exchange into the parallel market andmay have pressured the Tanzanian authorities into altering the size-and perhapseven the timing-of the 1986 devaluation (see Kamin 1993, Edwards 1989). Theempirical literature, however, unanimously finds that an expected depreciationaffects the premium in the short run, even after controlling for other possible de-terminants and for feedback from the premium to the devaluation. Thus inGhana, Sudan, Tanzania, Turkey, and Zambia, increases in the expected yield dif-ferential in favor of foreign assets-a key component of which is the expected rateof official depreciation-unambiguously raised the premium.


Figure 3. Official Excbange Rate and Parallel Premium, Tanzania, 1981-90(percent)

Log of official exchange rate Premium2.4 - - 8

2.2 - k 700

2.0 - 600

1.8 - 500

1.6 - 400

1.4 - A ~~~~~~~~~~~~~~~300

1.2 ~~~~~~~~~~~~~~~~~~200

1.0 100

0.8 - 0

1981 1982 1983 1984 1985 1986 1987 1988 1989 1ggo

Log of official exchange rate ------- Parallel premium

Source: Parallel premium from Kaufmann and O'Connell (1991); official exchange rate(Tanzanian shillings per U.S. dollar) from LMP (various issues).

Devaluation and the Premium

Without support from macroeconomic policies, a devaluation of the officialrate has a negative but transitory effect on the premium, often lasting less than aquarter. Figure 4 provides two dramatic examples of the cycle of official deval-uations and increases in the premium that can emerge if policymakers fail eitherto address underlying sources of high monetary growth or to accommodatethese sources through continuous exchange rate adjustments. Each of the fourmaxi-devaluations implemented in the Sudan in the 1980s reduced the premiumsignificantly-but briefly-and each time the premium rose again, climbing be-yond 100 percent before the authorities brought it down through a new devalua-tion (see figure 4A).

Venezuela's experience was similar. In 1983 the premium on the dual systemestablished earlier in the year increased by more than 200 percent. A devaluationearly in 1984 cut the premium nearly in half. When changes in the underlyingpolicies were not forthcoming, however, the premium resumed its rise until asecond maxi-devaluation was implemented late in 1986. The devaluation-


Figure 4. Devaluation and Parallel Premium(percent)

A. Sudan, 1970-89

Devaluation Premium100 -40090 80 -

70 - 30060-

500- 20040-

10s- -' \ _____Ws - 10

0 0

1970 1972 1974 1976 1978 1980 1982 1984 1986 1988

- ~ Devaluation ------ Premium

B. Turkey, 1970-89

Devaluation PreC,ium200 - 60

150 50

-40100 If

-30

50-20

0- - 10

1970 1972 1974 1976 1978 1980 1982 1984 1986 1988

~~Devaluation - --- Premiium

Source. Official exchange rates from iMF (various issues); parallel exchange rates

from Intemnational Currency Analysis, Inc. (various years).


premium cycle was finally eliminated when the foreign exchange market wasunified in 1989. The recent (1994) reintroduction of a dual regime is consistentwith our earlier explanation that authorities tend to adopt such regimes duringbalance of payments crises.

In Turkey and Ghana, macroeconomic policies achieved a sustained reductionin the premium. In Turkey (see figure 4B), officials applied a broad package ofreforms that included devaluation, a reduction in the budget deficit, and theadoption of a crawling peg. Ghana achieved a long-lasting reduction in thepremium through reforms that combined devaluation and fiscal austerity withthe adoption of a more flexible exchange rate and substantial new aid inflows.

Fiscal Deficits and the Premium

Evidence of a positive relationship between the average parallel premium andthe average government deficit is corroborated by Ghei and Kiguel (in Kiguel,Lizondo, and O'Connell, forthcoming) for a large sample of developing coun-tries. The stock-flow model captures the essence of this link: high deficits pro-duce rapid money growth, which produces high premiums.

Domestic demand pressures on exchange markets may be associated withtemporary, intermittent episodes of high premiums, but high premiums may alsobe evidence of a long-run inconsistency among the rate of official depreciation,the growth of the domestic money supply, and the maintenance of free convert-ibility (that is, an absence of restrictions on the amount of foreign currency theprivate sector can buy or sell at the official rate of exchange). The tight linkbetween fiscal deficits and money growth in many developing countries suggeststhat overexpansionary fiscal policy is often at the heart of parallel markets withpersistently high premiums.

But the correlation between fiscal deficits and the premium may also reflectfeedback from the premium to the fiscal deficit, operating through the effect ofillegal trade on tax revenues or through the quasi-fiscal gains and losses associ-ated with intervention by the central bank (discussed below). Illegal trade canhave a significant effect when controls extend to the current account and tradetaxes are a large share of revenue. The collapse of revenues in Ghana in the early1980s, for example, was directly related to the smuggling that diverted themajority of the cocoa crop to neighboring C6te d'Ivoire (May 1985). The intro-duction of the own-funds window in Tanzania in 1984 was associated with anincrease in customs revenues (from 1983 to 1987) of 4.5 percent of GDP, eventhough underinvoicing was widely thought to be a serious problem in thatwindow.

Illegal Trade and the Premium

Although portfolio equilibrium (the "stock" element in stock-flow models)drives the premium in the very short run, changes in the extent and characteris-


tics of illegal trade flows, in the availability of foreign aid, and in the terms oftrade have an important influence in the medium to long run. Because theparallel rate is forward-looking, changes in these factors also affect the premiumin the short run.

Illegal trade primarily affects the premium by changing the stock of privatelyheld net foreign assets. As noted earlier, parallel systems are put in place toinsulate reserves (or more generally to frustrate the private accumulation ordisposal of foreign assets) by limiting central bank intervention in private capitalaccount transactions. In a parallel system the private sector can accumulateforeign assets either through a surplus in illegal trade or through more directleakages from official reserves.

The flow supply of foreign exchange into the parallel market is generated bysmuggled and underinvoiced exports, overinvoiced imports, and central banksales of reserves for capital outflows (whether legal or illegal). The flow demandis generated by underinvoiced and smuggled imports and by the "replacement"demand for foreign assets (which is aimed at keeping asset stocks at desiredlevels, given ongoing growth in income and population).

The literature on the relation between illegal trade and the parallel premium isextensive. Consider export smuggling, for example. Smuggled exports escapetaxation and offer the opportunity to convert the proceeds at the parallel raterather than at the official rate. The amount smuggled therefore increases as theexport tax rate and the parallel premium rise (Macedo 1987). In some models,aggregate exports are correlated with the official real exchange rate, while theshare of exports smuggled is correlated with the premium (Kamin 1993). Inthese models the total supply of foreign exchange from export smuggling goesup with a rise in the premium and down when the real exchange rate appreci-ates. The effect of a change in export taxation is ambiguous; it decreases aggre-gate exports but increases the share that is smuggled.

Changes in the premium have the reverse effect on the demand for foreignexchange to be used to smuggle or underinvoice imports, either to avoid thepayment of tariffs (Macedo 1987) or to take advantage of high domestic pricesassociated with import rationing (May 1985, O'Connell 1991). The demand forillegal foreign exchange arising from these imports increases with the differencebetween their border price at the official exchange rate and their price in thedomestic market.

For individuals with access to foreign exchange at the official exchange rate, arise in the parallel premium increases the profit from diverting funds from theofficial market to the parallel market. An increase in the premium is thereforeassociated with an increase in the supply of official reserves for private capitaloutflows. This effect may also operate legally, if the authorities intervene at theparallel rate to prevent excessive increases in the premium; see Kaminsky (inKiguel, Lizondo, and O'Connell, forthcoming).

The net flow supply of foreign exchange in the parallel market therefore goesup with the parallel premium, given the values of other variables that determine


trade flows and the incentives for illegal activity. Static flow theories of the blackmarket treat private capital outflows as predetermined, ignoring asset marketequilibrium and assuming that the premium is determined at each point in timeby the variables affecting the flow market for foreign exchange (May 1985,Sheikh 1976, Pitt 1974). These models are inappropriate as descriptions of theshort run, in which asset market conditions dominate; but they come into theirown as descriptions of medium- to long-run equilibrium, because the net privatecapital outflow is eventually tied down by the requirement that private netforeign assets not grow (or fall) without bound relative to income. Thus a rise inexport taxes lowers the premium in the long run by diverting export revenuesinto the parallel market; a rise in import tariffs raises the premium by increasingunderinvoicing.

Although illegal transactions are by their nature difficult to measure precisely,two kinds of evidence suggest a strong link between these activities and theparallel premium. First, trade data comparisons find that increases in the paral-lel premium generate greater underinvoicing of exports and overinvoicing ofimports (see, for instance, McDonald 1985). Second, studies based on exportsupply functions find that a rise in the parallel premium tends to reduce exportsas domestic companies resort to misinvoicing or smuggling.

Interactions between illegal exports and the parallel premium are particularlydramatic during episodes involving maxi-devaluation. In a sample of forty epi-sodes, Kamin (1993) finds that increases in the parallel premium-in anticipa-tion of the official devaluation-help explain the decline in official exportspreceding devaluation and that the immediate decline in the premium afterdevaluation helps to account for their surprisingly rapid recovery. In fact,changes in the premium have a greater effect on exports than changes in the realofficial exchange rate, suggesting that what appear to be movements of aggre-gate exports are instead largely shifts of exports between official and unofficialmarkets. When overvaluation is chronic, the cumulative effect on official ex-ports of declines in overall export volume and shifts from official to unofficialchannels can be overwhelming. In Ghana and Tanzania total recorded shipmentsfell from 23.9 and 28.4 percent of GDP, respectively, in 1970 to 11.6 and 15.6percent in 1985.

Additional "flow" determinants of the premium include aid, the terms oftrade, and other macroeconomic variables affecting the supply of official foreignexchange or the demand for quota-constrained imports. The net effect of thesevariables is often theoretically ambiguous; an increase in aid, for example,increases the supply of (divertable) official foreign exchange but may simul-taneously increase demand for foreign exchange in the parallel market, raisingthe domestic price of rationed imports and increasing the demand for (smuggled)imports. The increase in the supply of foreign exchange tends to lower thepremium, while the increase in demand raises it.

Evidence from the case studies and elsewhere (see Dornbusch and others1983) shows that a real appreciation of the official exchange rate is associated


with increases in the premium in both the short and the long run, as the theorysuggests. Because exchange rate expectations are captured by a variable measur-ing the difference in expected yields between domestic and foreign assets, thisresult suggests that overvaluation has a powerful effect on trade flows.

UnificationExchange rate unification means different things in different contexts, but

two interpretations are important in practice. Full unification refers to the adop-tion of a single exchange rate for all external transactions, with full convert-ibility if the exchange rate is managed. Partial unification refers to the adoptionof a single exchange rate for all current account transactions, while maintainingconvertibility restrictions and therefore a parallel market for portfolio and capi-tal account operations.

Full Unification in a Crisis

Most of the countries that originally created an official parallel market to dealwith a balance of payments crisis eventually decided to unify the market. Para-doxically, the decision was not part of a well-planned strategy but instead oc-curred during a second crisis when inflation was high and the premium was onthe rise. Thus Venezuela, after six years of operating a multiple system, unifiedits foreign exchange market in February 1989 by floating the exchange rate.This move was prompted by severe balance of payments problems when thepremium was close to 200 percent. In Mexico, unification was part of a packageaimed at stabilizing prices and restoring external balance. The decision wasmade following the stock market crash of October 1987, in the face of accelerat-ing inflation and a rapidly depreciating parallel exchange rate. Argentina unifiedunder a floating exchange rate to control an explosive hyperinflation in 1989.These experiences suggest that multiple systems are typically abandoned notbecause they are no longer "needed" but because they are no longer useful inprotecting reserves and maintaining low inflation. Moreover, crises may offerpolicymakers an opportunity to institute reforms that would have been politi-cally costly at other times. In other words, a dual regime may have long sinceserved its purpose, but unification was postponed because of politicalopposition.

The former socialist economies recently have eliminated dual exchange mar-kets as part of broader efforts to bolster market forces. As in earlier cases,unification occurred during a period of crisis, although in these countries theprocess was part of major systemic reforms. In Poland, for example, unificationwas essential to ensure the credibility of the fixed exchange rate (Lipton andSachs 1990). Moreover, the measure was a natural instrument for rationalizingprices after years of price controls and poorly functioning markets. Similararguments motivated the rapid unification of the foreign exchange market inRussia.


Gradual Unification in Highly Distorted Economies

In countries with widespread exchange controls, unification tends to be a longprocess. In Turkey it took nearly a decade, beginning in 1980 with a maxi-devaluation and a schedule for phasing out multiple exchange rates. In subse-quent stages the government adjusted exchange rates, liberalized imports, andrelaxed controls on the capital account. The process was completed in 1989,when residents were permitted to purchase foreign securities; since then theblack market has essentially disappeared.

In Africa most economies that regulated foreign exchange have opted forpartial-and gradual-unification. In Ghana the reform process began in 1983and included monetary and fiscal restraint, increases in producer prices, relax-ation of import controls, and more flexible management of the official exchangerate. As a result, the system-which initially consisted of an official, fixed ratemarket and a thriving black market-has been transformed into two legal mar-kets with floating rates and a negligible spread between them and a small illegalmarket. The black market premium declined from more than 2,000 percent atthe beginning of 1983 to 24 percent in April 1988, when the second legal marketbecame operational and practically absorbed the black market.

Tanzania was well on the way to more market-oriented exchange rates by theearly 1990s. The premium had declined from more than 700 percent in 1986 toroughly 50 percent in 1990. Comprehensive reforms have gradually succeededin liberalizing markets and restoring macroeconomic stability. In 1984 the au-thorities devalued and introduced an own-funds scheme. By 1986 this windowwas financing a third of total imports. The government devalued again in 1986and adopted a crawling peg as part of a macroeconomic reform package heavilysupported by external assistance. More recently, private foreign exchange bu-reaus have been authorized to deal in trade-related transactions at market-determined exchange rates.

A common element in all of these episodes has been the ability to sustainunification. Argentina, Mexico, and Turkey have been successful; Ghana andTanzania have moved gradually but steadily and remain on a course that is likelyto lead to full unification.

Ingredients for Successful Unification

Successful unification requires two ingredients. First, the price must be accept-able to those wishing to purchase or sell foreign currency for portfolio purposes;this is the main determinant of the exchange rate in the short run. Second, theexchange rate system must be consistent with underlying credit and fiscal poli-cies. In practice, this means that, if monetized fiscal deficits create inflationarypressures, the authorities need to adopt some form of crawling peg to keep aparallel market from reemerging.

The theoretical literature offers only limited guidance on choosing a unifiedexchange rate. Lizondo (1987) and Kiguel and Lizondo (1990) note that much


depends on whether unification is anticipated and on the exchange rate regimeadopted, such as a floating rate or a crawling peg. If the change is anticipated,the unified rate will coincide with the parallel rate; otherwise there would beopportunities for capital gains. In other cases the theory is more ambiguous. Anunanticipated unification to a crawling peg occurs at the parallel exchange rate ifthe central bank maintains the stock of international reserves but is likely tooccur at some rate between the official and the parallel exchange rates if thecentral bank is willing to lose reserves. The results of unifying to a floatingexchange rate are also unclear.

In practice, however, the unified exchange rate generally coincides with or isclose to the parallel rate. This was the case in Argentina and Venezuela in 1989and in Mexico in 1987. When unification is anticipated, this choice probablyreflects the need to maintain asset market equilibrium in the short run (which ina parallel regime takes place at the parallel rate). Experiences vary in countriesthat choose partial unification and maintain controls on the capital accountbecause in those cases it is only necessary for the exchange rate to balancecurrent account transactions.

The exchange rate system adopted after unification varies according to cir-cumstances. Ghana and Turkey chose a crawling peg, Mexico and Venezuela (inthe 1960s) opted for fixed exchange rates, and Argentina and Venezuela (in the1980s) used floating rates. A fixed exchange rate is possible only if fiscal andmonetary policies support stable prices. If the economy faces inflationary pres-sures, unification can be maintained only if the authorities adopt a flexibleexchange rate system.

Failure to unify successfully can often be traced to inconsistency between thenew exchange rate regime and the stance of fiscal and monetary policy. InArgentina, for example, two attempts at unification failed because the govern-ment funded large budget deficits by printing money while trying to use theofficial exchange rate as an anchor for inflation. Macroeconomic balances werestill large when unification was finally accomplished, but this time the govern-ment devalued the exchange rate rapidly enough to avoid a real appreciation.

Countries that attempt to bring down inflation and improve the externalbalance at the time of unification must adopt fiscal and monetary policies thatsupport these objectives. Ghana, Mexico, Turkey, and Venezuela all cut theirbudget deficits and tightened domestic credit to support the removal of foreignexchange controls. Although inflation did not always fall, the exchange rate wasallowed to depreciate sufficiently to avoid severe overvaluation, and the macro-economic situation usually improved.

Attempts at partial unification in Sudan and Zambia were unsuccessful. In1979 Sudan tried to unify its foreign exchange market as part of a liberalizationand stabilization program. The government shifted a growing number of trans-actions from the official market to a legal parallel market, in an attempt toreduce the importance of the illegal parallel market. Lax domestic policies,however, led to the reappearance of a large black market premium and an


expansion in the number of black market transactions. In Zambia two attemptsat unification failed. The first effort (1983-85), based on a crawling peg, wasabandoned after large adverse changes in the terms of trade led to increases inthe premium. The second episode (1985-87) was an attempt to reduce thevolume of transactions in the black market by using an auction system in theofficial market. But without the support of compatible monetary and fiscalpolicies, the premium rose and the black market continued to thrive.

Experience in Latin America suggests that rapid unification is desirable, par-ticularly when inflation is rising. The Latin American economies were less dis-torted than others in the World Bank project, however, and their capital marketswere better integrated with world financial markets. Thus parallel foreign ex-change markets (and relatively large premiums) were difficult to sustain becauseagents could easily find ways to beat the system. In economies with extensiveprice controls, barriers to trade, and thin financial markets, a gradual approachcould well be appropriate. In Turkey, Ghana, and, to some extent, Tanzania,unification has moved in tandem with structural reforms to expand the role ofthe market in determining resource allocation-an approach that has beenlargely successful.

Last, but not least, success depends on the government's commitment; it mustbe strong enough to outlast the short-term adverse consequences, such as anincrease in inflation or a drop in real wages, that unification may bring. Wheninflation increased after unification in Venezuela, the authorities were deter-mined to rely on prudent monetary and fiscal policies to control it rather thanresorting to a parallel market. In contrast, the government of Zambia wasunwilling to accept the sharp depreciation required and abandoned efforts tounify the market.

Unification and Some Pleasant (Shadow) Fiscal Arithmetic

Unification has potentially significant implications for the fiscal deficit andconsequently for money growth and inflation. For example, when the govern-ment depends heavily on trade taxes for revenue, a unified exchange rate canprovide a substantial fiscal bonus through an increase in aggregate trade and thelegalization of illegal trade. This change in the trade tax base forms part of theoverall "shadow" fiscal effect of the premium, defined as the change in thegrowth of the real domestic liabilities of the public sector associated with unifi-cation. Because money creation is typically an important means of domesticfinance in developing countries, the shadow fiscal effect represents a potentiallyimportant link between exchange rate unification and inflation.2

The literature on the fiscal effects of parallel exchange rates has tended toignore general equilibrium effects on trade flows, emphasizing instead that, evenif unification fails to change these flows, it alters their valuation by depreciatingthe real exchange rate. Because transactions between the government and thecentral bank leave the public sector's overall borrowing requirement unchanged,


the base for this valuation effect is net foreign exchange transactions between thegovernment and the private sector. Thus Pinto (1989, 1991) warns that, if thegovernment is a net buyer of foreign exchange from the private sector, unifica-tion will tend to raise inflation unless offsetting reductions are made in the fiscaldeficit (see also Lizondo 1987, 1991; Kharas and Pinto 1989).

Although the literature has focused almost exclusively on the valuation effecton foreign exchange transactions, the shadow effect of the premium also in-cludes any effect of unification on the governments' domestic currency receiptsand payments. These effects can operate either through changes in the underly-ing real flows (as in the case of changes in the trade tax base) or-holding theflows constant-through partial or full indexation of the prices at which theseflows take place. The latter would include the rise in real tariff revenue per unitof imports associated with real depreciation of the official exchange rate (assum-ing ad valorem rather than specific tariffs) and the effects that operate throughindexation of government wages or subsidies on imported foods. Changes inthese components of the real public sector deficit can feed through to moneygrowth and alter the rate of inflation. (See Easterly and Fischer 1990 for atreatment of the inflation tax in government finance.)

A full assessment of the shadow fiscal effect of the premium would require acomplete description of government budgetary flows and central bank transac-tions under the counterfactual assumption of unified exchange rates. Such anassessment has not been attempted in the literature, and little is known about theactual magnitude of the fiscal effect of parallel exchange rates (Morris, forth-coming). One problem is calculating the shadow unified real exchange rate.Another is assessing the effect of unification on illegal trade flows, which areimpossible to measure precisely. Including indirect effects on the budget ordemand for the monetary base, such as those arising from changes in income orreal wealth associated with unification, leads to further complications.

Table 3 summarizes two components of the shadow fiscal effect for five coun-tries in the World Bank study. The first is the shadow gain or loss associated withthe pricing of foreign exchange transactions at the overvalued official exchangerate-the valuation effect referred to earlier. We call this the "central bankprofits effect" to emphasize the analogy with the familiar quasi-fiscal gain or lossfrom central bank intervention in a multiple exchange rate system (see Sherwood1956, Dornbusch 1986a, 1986b, and Lizondo 1991).3 We use the parallel ex-change rate as the hypothetical unified rate; this is an imperfect proxy becausethe real exchange rate consistent with external balance in a unified regime maydiffer from the market-clearing rate under the parallel regime (Kiguel and Liz-ondo 1990). Experience suggests, however, that the market-clearing unified rateis typically much closer to the parallel rate than to the official rate, particularlywhen the system has been in place for some time. The second element summa-rized in table 3 is the change in the domestic currency component of the budgetassociated with unification. The estimates are rough and partial, given the diffi-culties mentioned above and the well-known problems in measuring public


Table 3. Fiscal Effects of Parallel Exchange Rates

Effects on shadow central bank profitsa Effects on shadow domestic

(percentage of GDP) currency budgetb

Number Direction

Country of years Maximum Minimum Average Channel of effect

Argentina- 7 3.5 1.3 2.1 Reduction in export Negativetax revenues fromunderinvoicing

Sudan - - - - Reduction in reve- Negative

nues from tradetaxes

Tanzania 14 0.4 -9.8 -4.5 Reduction in import Positive(customs and and neg-sales) tax reve- ativenues; increasedmarketing boardrevenues as resultof lower exportprices

Venezuela 4 -14.7 -25.4 -20.5Zambia 2 -12.2 -16.2 -14.2 Increase in revenues Positive

from imports (cus-toms and salestaxes)

- Not available.a. The shadow gain or loss associated with pricing foreign exchange transactions at the overvalued

exchange rate.b. The change in the domestic currency component of the budget that is associated with unification.c. The estimate of net foreign exchange sales to the private sector by the central bank includes only

merchandise trade transactions; the exchange rate for purchases includes export taxes and subsidies.Source: Aron and Elbadawi, Elbadawi, Hausmann, Kamin, Kaufmann and O'Connell (in Kiguel,

Lizondo, and O'Connell, forthcoming).

sector deficits. Except for Tanzania, the effect on the domestic currency budgetcould be assessed only in qualitative terms. Our results suggest, however, thatthe outcome of unification in practice is often more pleasant (and the implica-tions of parallel rates correspondingly more unpleasant) than the literatureindicates.

It is clear from the table that the shadow fiscal effect of the premium is notuniform across countries. In three of the four countries for which we have data,the central bank ran shadow losses. The effect on the domestic currency budgetwas positive in one country, negative in two, and ambiguous in the fourth.Moreover, the two components in table 3 do not necessarily work in the samedirection.

The qualitative effect on central bank profits depends primarily on whetherthe central bank is a net buyer or a net seller of foreign exchange to the privatesector. In Zambia, for example, the central bank was a net seller to the privatesector and thus it suffered shadow losses.4 In Tanzania, too, the central bankwas a net seller to the private sector in every year but one. In Venezuela, the


buying and selling rates were different, but the central bank was a net seller bysuch large amounts that it suffered shadow losses. We do not report calculationsfor Mexico, but the presumption again is that the dual system produced shadowlosses, as oil export revenues (which are in the public sector) far exceeded thecost of external service. In Argentina, the central bank was a net buyer of foreignexchange from the private sector and thus benefited from the dual exchange ratesystem.

The magnitude of the effect on the profits of the central bank can be quitesignificant. The absolute size of these estimates is biased upward if the parallelreal exchange rate overestimates the true shadow value of foreign exchange, butthe effects remain significant even after correcting for a possible bias. For exam-ple, a shadow exchange rate equal to 60 percent of the parallel exchange ratereduces the estimated average loss in Venezuela from 20.5 percent of GDP to 4.7percent of GDP. Moreover, the effects on inflation would be uniformly morefavorable (particularly for African countries) if the calculations included theincreases in external aid that have been triggered by moves toward unified,market-determined exchange rates in the past decade. Aid flows received by thepublic sector reduce the domestic financing required to cover any given patternof domestic currency receipts and expenditures to the extent that those flows arenot offset by increased public sector imports.

The effect of the parallel system on the domestic currency budget was esti-mated only for Tanzania; see Kaufmann and O'Connell (in Kiguel, Lizondo, andO'Connell, forthcoming). The authors assume that some components of thedomestic currency budget (custom duties, sales taxes on imports, and producerprices for exports paid to farmers by parastatals) are fully indexed to the officialexchange rate, while other components (other tax revenues and public sectorwages) are indexed to domestic prices. Under these assumptions, multiple rateshave both positive and negative effects on the domestic currency budget. On theone hand, a large premium allows the government to keep producer prices low,thereby reducing expenditures. On the other hand, it reduces declared imports,thereby reducing revenues from import taxes. From 1976 to 1989 the multiplesystem in Tanzania generated an estimated net positive average annual effectequivalent to 2.1 percent of GDP. For other countries the limited available evi-dence suggests a mixed picture. For example, an increase in the premium seemedto reduce revenues from trade taxes in Argentina and Sudan, but the oppositeappears to have occurred in Zambia.

More often than not, parallel foreign exchange markets tended to generatefiscal losses in the countries studied. Most of these losses did not appear directlyin the budgetary accounts, but were instead generated by net sales of foreignexchange to the private sector at overvalued exchange rates. In most of thecountries in the sample, the public sector was a net producer of foreign ex-change, either because public sector enterprises were the main exporters in theeconomy (as in Mexico, Venezuela, and Zambia) or because the public sectorreceived large external transfers (as in Ghana and Tanzania). In these cases, an


earlier or faster unification would have reduced the public sector's real domesticfinancing requirement, which, other things being equal, would have reducedinflationary pressures.5

Final Thoughts

Experience with parallel exchange rates in developing countries has, on thewhole, been disappointing. Most of the countries in the World Bank studytolerated high premiums for long periods, with damaging effects on the alloca-tion of resources and growth and no clear gains from maintaining a dual system.

Legal dual systems were misused more often than not, both because they wereoverextended and because the premium was excessive. Venezuela maintained itsdual system for eight years, Mexico for five, and Argentina for eight (countingofficial and quasi-official parallel exchange rates). Average premiums duringthese periods were 30 percent in Mexico, 44 percent in Argentina, and 120percent in Venezuela. In Argentina and Venezuela, governments made no clearefforts during this "temporary" period to restore external balance by alteringmonetary and fiscal policies. It is unlikely that the macroeconomic gains fromprotecting reserves and avoiding inflation in these countries were larger than thecosts resulting from the misallocation of resources. These experiences weakenthe case for recommending the adoption of dual exchange rates, even in circum-stances where, theoretically, such a recommendation would be appealing.

In other cases, the parallel market was a quasi-permanent arrangement, theresult of prolonged periods of overvalued exchange rates and expansionarymacroeconomic policies. In Ghana and Tanzania, for example, the authoritieshad to rely on extensive foreign exchange controls to avoid a full depletion ofreserves. The large premiums in these economies (exceeding 700 percent attimes) were clear evidence of a dramatic inconsistency between exchange ratepolicy and monetary and fiscal policies.

Although examples of macroeconomic mismanagement associated with thecoexistence of official and parallel foreign exchange markets are numerous, insome cases parallel systems were used judiciously. Belgium operated a dualsystem for more than three decades without producing major distortions. Col-ombia has maintained a substantial unofficial parallel foreign exchange marketfor years, while preserving macroeconomic balance. In these cases, however, thepremium was kept low on average (roughly 2 percent in Belgium and 6 percentin Colombia); larger premiums were tolerated only as a short-term safety valveduring crises. Serious distortions were avoided in these countries because thegovernments followed sound macroeconomic policies. What is more difficult todetermine is whether the parallel regime delivered greater macroeconomic bene-fits than a unified rate would have.

How important is unification? Large and persistent parallel premiums createnumerous microeconomic distortions and induce rent-seeking and corruption.Recent empirical studies on long-term growth find the premium has a significant


negative effect on investment productivity and growth. The evidence suggestsstrongly that a determined transition to unified exchange rates for current ac-count purposes should be a high priority in countries with large and persistentpremiums. Convertibility for capital account purposes is less urgent, as long asthe premium remains low on average (say, below 10 percent). Eventually,though, full unification makes sense, given the leakages between markets.

With respect to the speed of unification, we find two distinct successful pat-terns. In countries such as Argentina, Mexico, and Venezuela, where the parallelmarket was introduced to deal with capital flight, unification proceeded rapidly,generally as part of a comprehensive stabilization-liberalization package. Unifi-cation meant convertibility for current and capital account transactions.

In contrast, unification in Ghana, Tanzania, and Turkey occurred in phases,with reductions in the premium accompanying a gradual shift of transactions toa more market-determined official exchange rate. These economies were moreheavily controlled in most respects at the outset, and to some degree the liberal-ization of exchange controls was constrained by the feasible pace of reform inother areas (for instance, trade and price controls). Experience indicates thatlegalization of the existing parallel foreign exchange market is a good first steptoward full unification.

Last, but not least, a puzzling question arose from the study. In most cases,the exchange control system generated large parallel premiums and importantfiscal losses. The large premiums had detrimental effects on exports and growthwhile providing only limited insulation from external shocks. Surely a "rational"government would have pushed for unification. Why, in the face of this evi-dence, was unification typically delayed, relative to initial intentions, and half-hearted (and therefore unsuccessful)? The answer may lie in the realm of politi-cal economy. Even when parallel rates are adopted on an explicitly transitionalbasis, interest groups with enough political clout to deter immediate devaluationof macroeconomic contraction in the first place are likely to resist the policyadjustments necessary for early unification. Moreover, parallel systems generatesubstantial rents for those with access to official foreign exchange. These rentscreate strong vested interests in favor of continued controls.

NotesMiguel Kiguel is principal economist (on leave) at the World Bank and deputy general

manager for economics and finance at the Central Bank of Argentina. Stephen O'Connell isassociate professor of economics at Swarthmore College. The authors are grateful to SaulLizondo for his contribution to managing the research project and his help with this article.They would also like to thank David Bevan, Vittorio Corbo, Robert Flood, Steven Kamin,Nancy Marion, Stephen Morris, and seminar participants at Swarthmore College, OxfordUniversity, and the World Bank for helpful comments. Nita Ghei provided excellent researchassistance.

1. Kiguel, Lizondo, and O'Connell (forthcoming) includes the following papers: YawAnsu, "Macroeconomic Aspects of Multiple Exchange Rate Regimes: The Case of Ghana";Janine Aron and Ibrahim Elbadawi, "Parallel Markets, the Foreign Exchange Auction and


Exchange Rate Unification in Zambia"; Ibrahim Elbadawi, "The Black Market for ForeignExchange and Macroeconomic Management in Sudan"; Nita Ghei and Miguel Kiguel, "Dualand Multiple Exchange Rate Systems in Developing Countries: Some Empirical Evidence";Ricardo Hausmann, "Adoption, Management, and Abandonment of Multiple Exchange RateRegimes with Import Controls: The Case of Venezuela"; Steven B. Kamin, "Argentina'sExperience with Parallel Exchange Rate Markets: 1981-1990"; Graciela Kaminski, "DualExchange Rates: The Mexican Experience 1982-1987"; Dani Kaufmann and Stephen A.O'Connell, "Exchange Controls and the Parallel Premium in Tanzania, 1965-1990"; NancyP. Marion, "European Dual Exchange Rates"; and Sule Ozler, "Black Markets for ForeignExchange: The Case of Turkey."

2. Our definition of the fiscal effect emphasizes the link with inflation. To capture the fulleffect on the public sector's overall financial position, we would have to incorporate thedomestic currency value of any change in the government's net foreign assets (and domesticassets, if any) associated with unification. The evidence we present is relevant for eithercalculation.

3. The analogy is not perfect because the shadow effects we calculate do not manifestthemselves as actual cash flows when the parallel regime is in operation. Nonetheless, thegains or losses under a multiple exchange rate system must be calculated relative to a "refer-ence" exchange rate-typically the commercial rate or some other official rate-and thusimplicitly involve a counterfactual. Our shadow central bank profits effect embeds the calcu-lation in a full specification of the counterfactual. What emerges looks exactly like the quasi-fiscal effect but uses the unified rate as the reference rate. Thus, our calculation for Venezuela,for example, allows for official foreign exchange transactions occurring at more than oneexchange rate.

4. The estimates for Venezuela and Zambia differ from those presented by the respectiveauthors in their individual case studies. Although the Venezuelan oil company (PDVSA) ispublicly owned and the Zambian copper company (zccMc) is largely publicly owned, thecalculation of the fiscal effect on central bank accounts in the individual case studies treatsthese companies as part of the private sector. In our calculations, however, these companiesare included in the public sector.

5. This pleasant arithmetic must be balanced against the possibility of a decline in thedemand for domestic assets if unification includes a liberalization of capital controls. Adam,Ndulu, and Sowa (1993) argue that liberalization enhanced the liquidity and risk of foreignassets relative to domestic assets in Ghana and Kenya, producing declines in money demandand, everything else being equal, increases in inflation. Giovannini and de Melo (1993) arguemore generally that capital controls increase seigniorage revenue by supporting the demandfor domestic government liabilities.



Adam, Christopher, Benno Ndulu, and Nii Sowa. 1993. "Efficiency Gains versus RevenueLosses: Liberalization and Seigniorage Revenue in Kenya, Ghana, and Tanzania:' OxfordUniversity, Centre for the Study of African Economies, Institute of Economics and Statis-tics, Oxford, U.K. Processed.

Agenor, Pierre Richard. 1992. "Parallel Currency Markets in Developing Countries: Theory,Evidence, and Policy Implications.' Princeton Essays in International Finance 188. Prince-ton, N.J.: Princeton University.

Barro, Robert, and Jong-Wha Lee. 1993. "Losers and Winners in Economic Growth:' NBERWorking Paper 4341:1-57. National Bureau of Economic Research, Cambridge, Mass.Processed.

Bhagwati, Jagdish. 1978. Anatomy and Consequences of Exchange Control Regimes. Cam-bridge, Mass.: Ballinger.


Chhibber, Ajay, and Nemat Shafik. 1991. "Exchange Reform, Parallel Markets and Inflationin Africa: The Case of Ghana.' In Ajay Chhibber and Stanley Fischer, eds., EconomicReform in Sub-Saharan Africa. Washington D.C.: World Bank.

Dornbusch, Rudiger. 1986a. "Special Exchange Rates for Capital Account Transactions." TheWorld Bank Economic Review 1(1, September):1-33.

1986b. "Multiple Exchange Rates for Commercial Transactions." In Sebastian Ed-wards and Liaquat Ahamed, eds., Economic Adjustment and Exchange Rates in Develop-ing Countries. Chicago, Ill.: University of Chicago Press.

Dornbusch, Rudiger, and Luis Tellez Kuenzler (1993). "Exchange Rate Policies: Options andIssues:' In Rudiger Dornbusch, ed., Policymaking in the Open Economy. New York: Ox-ford University Press.

Dornbusch, Rudiger, D. V. Dantas, C. Pechman, R. R. Rocha, and S. Simoes. 1983. "TheBlack Market for Dollars in Brazil." Quarterly Journal of Economics 98:25-40.

Easterly, William. 1994. "Economic Stagnation, Fixed Factors and Policy Thresholds." Jour-nal of Monetary Economics 33:525-57.

Easterly, William, and Stanley Fischer. 1990. "The Macroeconomics of the GovernmentBudget Constraint." World Bank Research Observer 5(2, July): 127-42.

Edwards, Sebastian. 1989. Real Exchange Rates, Devaluation, and Adjustment: ExchangeRate Policies in Developing Countries. Cambridge, Mass.: MIT Press.

Fischer, Stanley. 1993. "The Role of Macroeconomic Factors in Growth." Journal of Mone-tary Economics 32(3, December):485-512.

Fleming, J. Marcus. 1971. Essays in International Economics. Cambridge, Mass.: HarvardUniversity Press.

Flood, Robert P. 1978. "Exchange Rate Expectations in Dual Exchange Markets.' Journal ofInternational Economics 8:65-77.

Flood, Robert P., and Nancy Marion. 1982. "The Transmission of Disturbances under Alter-native Exchange-Rate Regimes with Optimal Indexation.' Quarterly Journal of Economics97:43-66.

Fry, Maxwell. 1988. Money, Interest, and Banking in Economic Development. Baltimore,Md.: Johns Hopkins University Press.

Giovannini, Alberto, and Martha de Melo. 1993. "Government Revenue from FinancialRepression." American Economic Review 83(4, September):953-63.

Guidotti, Pablo. 1988. "Insulation Properties Under Dual Exchange Rates.' Canadian Jour-nal of Economics 21(4, November):799-813.

Hyden, Goran, and Bo Karlstrom. 1993. "Structural Adjustment as a Policy Process." WorldDevelopment 21 (9, September):1395-1404.

IMF (International Monetary Fund). Various issues. International Financial Statistics. Wash-ington, D.C.

International Currency Analysis, Inc. Various years. World Currency Yearbook. Brooklyn,N.Y. (Formerly Pick's Currency Yearbook,)

Kamin, Steven B. 1993. "Devaluation, Exchange Controls, and Black Markets for ForeignExchange in Developing Countries?' Journal of Development Economics 40(1,February):151-69.

Kaufmann, Dani, and Stephen A. O'Connell. 1991. "The Macroeconomics of the UnofficialForeign Exchange Market in Tanzania?' In Ajay Chhibber and Stanley Fischer, eds., Eco-nomic Reform in Sub-Saharan Africa. Washington, D.C.: World Bank.

Kharas, Homi, and Brian Pinto. 1989. "Exchange Rate Rules, Black Market Premia, andFiscal Deficits: The Bolivian Hyperinflation." Review of Economic Studies 56(July):435-47.

Kiguel, Miguel, and Nissan Liviatan. 1988. "Inflationary Rigidities and Orthodox Stabiliza-tion Programs?' The World Bank Economic Review 2(3, September): 273-98.


Kiguel, Miguel, and J. Saul Lizondo. 1990. "Adoption and Abandonment of Dual ExchangeRate Systems." Revista de Analisis Econ6mico 5(1, June):3-23.

Kiguel, Miguel, J. Saul Lizondo, and Stephen A. O'Connell, eds. Forthcoming. ParallelExchange Rates in Developing Countries. London: MacMillan; New York: St. Martin'sPress.

Lanyi, Anthony. 1975. "Separate Exchange Rates for Capital and Current Transactions." IMFStaff Papers 22:714-49.

Lipton, David, and Jeffrey D. Sachs. 1990. "Creating a Market Economy in Eastern Europe:The Case of Poland." Brookings Paper on Economic Activity 1:75-147.

Lizondo, J. Saul. 1987. "Exchange Rate Unification and Balance of Payments Under DualExchange Markets." Journal of Development Economics 26:37-53.

. 1990. "Multiple Exchange Rates and Black Market Exchange Rates: A NontechnicalSurvey." International Monetary Fund (IMF), Washington, D.C. Processed.

. 1991. "Alternative Dual Exchange Rate Regimes: Some Steady-State Comparisons."IMF Staff Papers 38(3, September): 560-81.

Loxley, John. 1989. "The Devaluation Debate in Tanzania." In Bonnie K. Campbell and JohnLoxley, eds., Structural Adjustment in Africa. New York: St. Martin's Press.

Macedo, Jorge Braga de. 1987. "Currency Inconvertibility, Trade Taxes, and Smuggling."Journal of Development Economics 27(October):109-25.

Marion, Nancy P. 1994. "Dual Exchange Rates in Europe and Latin America." World BankEconomic Review 8(2, May):213-45.

May, Ernesto. 1985. "Exchange Controls and Parallel Market Economies in Sub-SaharanAfrica: Focus on Ghana." World Bank Staff Working Paper 711. World Bank, Washington,D.C. Processed.

McDonald, Donough C. 1985. "Trade Data Discrepancies and the Incentive to Smuggle: AnEmpirical Analysis." IMF Staff Papers 32 (4, December):668-92.

McKinnon, Ronald. 1973. Money and Capital in Economic Development. Washington,D.C.: Brookings Institution.

Morris, Stephen. Forthcoming. "Inflation Dynamics and the Parallel Market for ForeignExchange." Journal of Development Economics.

O'Connell, Stephen A. 1991. "Short- and Long-Run Effects of an Own-Funds Scheme."Journal of African Economies 1(1, March):131-50.

Pinto, Brian. 1989. "Black Market Premia, Exchange Rate Unification, and Inflation in Sub-Saharan Africa." World Bank Economic Review 3(3):321-38.

1. 1991. "Black Markets for Foreign Exchange, Real Exchange Rates, and Inflation."Journal of International Economics 30 (February): 121-36.

Pitt, Mark. 1974. "Smuggling and the Black Market for Foreign Exchange." Journal ofInternational Economics 16 (June): 243-57.

Sheikh, Munir A. 1976. "Black Market for Foreign Exchange, Capital Flows, and Smug-gling."Journal of Development Economics 3 (February): 9-26.

Shaw, Edward. 1973. Financial Deepening in Economic Development. New York: OxfordUniversity Press.

Sherwood, Joyce. 1956. "Revenue Features of Multiple Exchange Systems: Some CaseStudies." IMF Staff Papers 5:74-107.

Triffin, Robert. 1947. "National Central Banking and the International Economy." PostwarEconomic Studies No. 7, Board of Governors of the Federal Reserve System, Washington,D.C. Processed.

Younger, Stephen. 1993. "Testing the Link between Devaluation and Inflation: Time SeriesEvidence from Ghana." Journal of African Economies 1(3, November):369-94.


INCENTIVES AND THE RESOLUTIONOF BANK DISTRESS

Thomas GlaessnerIgnacio Mas

Unlike prudential regulations that are put in place prospectively to developbanks, procedures for dealing with banks in distress are generally deter-mined on an ad hoc basis. Often the lack of clarity in the policy frameworkcreates incentives for bank managers, shareholders, depositors, and regula-tors that undercut prompt resolution of financial distress. The result is ofteninaction, the accumulation of bad debts, and ultimately the assumption oflosses by the state. This article argues that government intervention to relievefinancial distress should be institutionalized in a set of regulations that forcesthe authorities to comply with reporting and decisionmaking processes.Only in this way can inherent disincentives for dealing with distress becurtailed.

nsolvent banks have precipitated recurring problems in many developingcountries. In Latin America these problems have often been protracted andsystemic; some countries, such as Argentina, Chile, and Uruguay, have expe-

rienced systemwide crises, while others, such as Bolivia, Brazil, Ecuador, Peru,and Venezuela, have managed to defer or contain the problem, often at a high po-tential cost or in an unsustainable fashion. In many cases, policymakers have im-plemented preventive measures designed to avert distress rather than remedialmeasures intended to resolve crises once they occur. In others, institutional ar-rangements and legal processes have had a negative effect on efforts to resolve fi-nancial distress, reducing the incentives of regulators, managers, shareholders,depositors, employees, and borrowers to take the necessary actions. Although re-medial measures are designed to take effect once the bank is in distress, the frame-work for their execution must be adopted beforehand.

To be efficient, reliable, and credible, a policy framework for resolving bankdistress must establish incentives for all concerned parties; incentives that pre-serve financial discipline, induce cooperative solutions, and protect the rights ofclaimants (by differentiating liability holders according to their seniority and the

The World Bank Research Observer, vol. 10, no. 1 (February 1995), pp. 53-73© 1995 The International Bank for Reconstruction and Development/THE WORLD BANK 53

date of their claims). The framework for resolving bank failures should be insu-lated from political influence and vested in an appropriate regulatory group withprofessional trained staff who can identify problems before they get out of hand.Policy decisions should be based on the principle of minimum long-term costs,and financial support from the states should be both transparent and clearly tar-geted. The legal framework should force regulators to act in a timely fashion byassigning them responsibilities and giving them strong legal backing and protec-tion against pressures from political or other special interest groups. It shouldspecify the policies and procedures to be pursued under a range of possiblescenarios.

In sum, bank failures can be resolved much more effectively and in ways thatstrengthen, rather than weaken, financial discipline if governments reduce dis-cretionary supervision and require authorities to enforce laws concerning collat-eral and foreclosure. Governments must also address the underlying causes ofbank insolvency. Mechanisms for resolving bank failure constitute only one partof a well-coordinated strategy for the financial sector and will not succeed unlessstructural adjustments are made in other government financial policies, includ-ing banking-sector competition and confiscatory reserve requirements or otherforms of explicit or implicit taxation.

The discussion that follows focuses first on the incentive structure to show why li-quidating or restructuring insolvent banks is so difficult. It then goes on to showhow countries typically handle bank failure and reviews the processes for reha-bilitating and liquidating troubled banks. The third section examines the causes ofregulatory failure and shows why the discretionary approach leads to inaction onthe part of regulators and to a bias favoring rehabilitation rather than liquidation.The final sections address the design of new mechanisms to resolve bank failure.

This article focuses on private commercial banks. Although the role and size ofgovernment-controlled banks are likely to induce fiscal and banking system insta-bility in many developing countries, public banks present special issues that are be-yond the scope of this study. Because we are examining bank insolvency, rather thantemporary liquidity crises, we are not concerned with the central bank's lender-of-last-resort policy nor with deposit insurance per se, but rather with the restructuringand exit mechanisms frequently used to implement deposit insurance. (For a goodtreatment of the lender-of-last-resort policy, see Todd and Thomson 1990; for a dis-cussion of deposit insurance, see Mas and Talley 1992.) Instead, the objective is todevelop principles for reforming remedial processes and to recommend specific in-stitutional and legal reforms that can help to avert or resolve bank distress. Al-though the specific illustrations tend to be drawn from Latin America, the analysishere is generally applicable to all developing countries.

Defining and Measuring SolvencyDetermining whether a financial institution is solvent is an obstacle to prompt

action. Financial distress may not be apparent. A financial institution is consid-

54 The World Bank Research Observer, vol. 10,no. 1 (February, 1995)

ered to be viable as long as it can meet all of its obligations out of income overthe long run. Calculating the net present value of expected cash flows providesan economic measure of solvency. In the case of banks, economic value is relatedto overall economic conditions in the private sector because substantial propor-tions of banks' portfolios are concentrated in commercial loans. An institutionmay have a temporary shortfall but still have a positive net present value, inwhich case it is economically solvent but illiquid. Illiquidity, if unassisted bymonetary authorities, can turn into insolvency if the institution has to sell itsassets at distressed prices or pay above-market rates on deposits in a desperatescramble for liquidity.

Measuring economic solvency requires information about the stream of cashflows under every possible circumstance; estimating economic solvency is there-fore extremely difficult and subjective at best. Alternatively, one can use market-value solvency, or the market value of the institution (or of its underlying assets),as a proxy for net present value. Although this measure requires certain strongassumptions to hold at a theoretical level (namely, that the market value of assetsand liabilities reflects the net present value of their expected cash flows), itsadvantage is that solvency is defined in terms of current stocks rather thananticipated cash flows. But the definition is problematic for other reasons. First,unless there is a market for the shares of the bank (or for its assets), market valueis not a meaningful measure. This is true of financial assets, whose value de-pends intrinsically on the creditworthiness of the issuer (loans, for example),and with real assets such as property, whose replacement cost is hard to deter-mine. Second, commercial banks, like other businesses, have intrinsic value-orgoodwill-as going concerns, and the value of this goodwill can be particularlydifficult to estimate even when there is a market for trading bank assets.

To overcome these practical difficulties, a book-value measure of solvencybased on the nominal or historic cost of assets (net of depreciation) is sometimesused. Because of the arbitrary nature of book values, however, and the possi-bility that the bank can manipulate the way in which such statistics are pre-sented, book value does not tend to conform to market values. Thus book-valuesolvency is a very imperfect measure of economic solvency.

A better approximation can be achieved by adjusting the book value of value-impaired assets according to established but somewhat arbitrary rules. Throughprovisioning for probable losses, write-offs, revaluations, and adequate treat-ment of off-balance-sheet commitments, the net book values of certain assetcategories can more closely reflect underlying market values. In addition, rulescan be adopted that require valuing assets whose market value can be ascer-tained at whichever is lower-market or replacement cost. The resulting mea-sure of solvency, known as technical solvency, is the one bank supervisors use.Technical solvency represents a compromise between the theoretically correctconcept of economic solvency and the readily observable book-value solvency,effectively supplementing generally accepted accounting principles with specificguidelines for banks.

Thomas Glaessner and Ignacio Mas 55

Economic insolvency need not be synonymous with bank failure, however.Failure occurs only when insolvency is officially recognized and the bank isclosed. At that point the government has several options, which run the gamutfrom outright liquidation to rehabilitation. In this article, a bank is considereddistressed when it is technically insolvent. Such a bank may not fail immediatelyeither because it may still be liquid or because government actions may bekeeping it afloat.

What Is Unique about Commercial Bank Distress?In most-if not all-countries, resolving commercial bank distress involves

special procedures that do not apply to other enterprises. Three characteristicsthat distinguish commercial banks from nonfinancial corporations create orsupport disincentives that make resolution difficult:

. A liability structure characterized by deposits withdrawable on demand* Substantial leverage as measured by total liabilities (including debt and

deposit contracts) relative to the value of the institution's own capital* A smaller proportion of nonmarketable assets than most corporations, but a

larger amount than such institutions as money-market mutual funds.

The real or perceived threat of contagion across banks and the potential forhigh macroeconomic costs resulting from financial distress has often led govern-ments to adopt deposit insurance to prevent these outcomes. (See box 1 for someexamples of commercial bank distress.)

Structural Characteristics of Commercial Banks

The financial structure of commercial banks creates perverse incentives fordealing with bank distress. Liquidity problems-the institution's ability to meet

Box 1. Examples of Commercial Bank Distress in Latin America

Financial distress, defined as a situation in which a large number of financial institutionsare insolvent, has been recurrent in Latin America.

* Argentina. In the past fifteen years more than two hundred banks have had to be li-quidated, and the central bank was required to intervene in almost a hundred more.

* Uruguay. In the 1980s the government tried to prevent all bank failures by transferringproblem institutions to the public domain. A dominant public bank, which accounted fora third of the total deposits in the banking system, absorbed four ailing banks (whosecombined assets were almost 30 percent of total bank deposits).

* Chile. Following a vigorous reprivatization and liberalization of the banking system inthe mid-1970s, the government was forced to assume control of a significant number offinancial institutions in 1981 and 1983. In both years, the loan portfolio of the failingbanks was about 40 percent of the total system's loan portfolios.


demands for payment-can lead to insolvency if a bank's liabilities are moreliquid than its assets (liquid liabilities include money and money substitutes,such as savings deposits). The problem is compounded when the institution ishighly leveraged (that is, when the ratio of debt to capital is high). High leverageinduces owners and managers to engage in riskier activities (moral hazard) as thebank's financial condition deteriorates. This behavior can be aggravated by alack of financial transparency that is more likely to arise in countries wherebanking powers have been expanded or where companies are organized asfinancial conglomerates. The difficulty of interpreting accounting informationreduces the possibility of monitoring the bank's actions and inhibits the ability ofclaimants and depositors to take action to avert financial distress. Uncertaintyover the valuation of the bank's assets encourages regulatory inaction whenregulators already have incentives to avoid showing losses by actually closing thebank. Because of interbank accounts and contagion associated with bank runs,distress at one bank can infect other, fundamentally solvent, institutions.

The losses thus sustained and the lack of confidence engendered lead to realoutput costs because banks are forced to liquidate loans that, from an econ-omywide vantage point, should be continued. More broadly, such crises make itdifficult for the financial system to act as a buffer in response to real shocks.Concern over the stability of the monetary regime has served as a "public good"justification for government intervention to provide a safety net in the form ofdeposit insurance for depositors. Despite the obvious advantage of avertingbank runs, deposit insurance introduces a further set of disincentives that affectsthe behavior of bankers, depositors, and regulators.

Deposit Insurance

Governments frequently intervene in cases of bank distress to ensure thestability of the financial system. Because stability can be maintained only ifdepositors have confidence in the bank, governments typically take action toinsure the value of deposits, either explicitly through formal deposit insurance orthrough ad hoc measures to prevent bank failures. Such measures alter thedistribution of costs associated with financial distress but may also increase thetotal cost (and the probability) of distress. Thus, to minimize costs, governmentdeposit insurance must rely on, rather than substitute for, preventive measuresas well as on an effective process to ensure that insolvent financial institutionsare closed promptly (Benston and Kaufman 1988, Kane 1988).

Figure 1 shows the effect of deposit protection on the total size and distribu-tion of the losses associated with financial distress under four different assump-tions. For the sake of simplicity, this cost is distributed among three types ofbank stakeholders: depositors (D), shareholders (S), and governments (G). Atone extreme, the government offers little protection, and insolvent banks areallowed to fail systematically; the government has a stake in the bank to theextent of deferred taxes, public sector deposits, and outstanding central bank


Figure 1. The Effects of Deposit Protection on the Total Cost of Financial Distress:FourScenarios

A. Shifting costs B. Increasing confidence

Cost of Cost ofdistress distress

DD

G~~~~~~~~

Depositor Depositorprotection protection

No government Full depositintervention insurance

C. Increasing moral hazard D. A combination offactors

Optimal level ofdeposit insurance

Cost of Cost ofdistress distress

G

Depositor protection Depositorprotection

Note: Three types of stakeholders are represented: depositors (D), shareholders (S),and government (G).

credit. At the other extreme, the government provides full deposit insurance(whether through systematic bank bailouts or depositor payouts).

Panel A of figure 1 shows that as deposit protection increases, the lossesaccruing to depositors are gradually transferred to the government. The costsaccruing to shareholders depend on whether the shareholders are repaid. Thisexample shows the case in which they are not-either because depositors arepaid off upon liquidation or because old equity is written off beforerehabilitation.


In panel B the total cost of distress falls as protection increases. With increasedconfidence in the banking system, the probability of bank runs falls, and conse-quently fewer banks fail. The government's contingent liability stemming fromdeposit insurance may also create incentives for regulators to intervene in trou-bled banks early and decisively because government and taxpayer funds are atrisk.

Panel C shows the opposite case: the total cost of distress increases withprotection because of the destabilizing effects of moral hazard on each of thethree stakeholders. Here, shareholders are indifferent to risk because depositprotection can make them whole (if the form of deposit insurance is such thatthe government prevents bank failures and assumes the losses). Depositors' lackof concern about the soundness of the bank removes the discipline associatedwith the threat of bankruptcy, and regulators delay closing down insolventbanks because of the potential financial burden to the insurer.

Panel D presents a likely scenario. At high levels of protection, moral hazardis likely to be pervasive, but, at lower levels, the benefits from the sense ofconfidence imparted by protection may outweigh the problem of moral hazard.Thus the total cost is assumed to decrease (as in panel B) until the optimal levelof deposit insurance is reached and then to increase explosively when protectiongoes above the optimum level (as in panel C). At this point consumer confidenceand a reduction in the probability of depositor runs on the bank are more thanoffset by the moral hazard problem associated with high levels of protection.The implication, then, is that authorities should seek to target protection at alevel that minimizes the total costs. The exact location of this point on theprotection scale depends on the quality of prudential regulation and supervision,the nature of the threats to the banking system (micro versus systemic), thedegree of concentration in the banking industry and in the economy, the cred-ibility of the government's commitment, and the history of financial distress inthe country.

Mechanisms for Handling Bank DistressThroughout the 1980s governments emphasized the importance of mecha-

nisms and procedures to promote the stability of the financial system. Theinstitutional arrangements put into place fall into two categories: those designedto avert distress, and those designed to mitigate the consequences of distressonce it occurs.

Preventive measures include specifying the rules of the game by institutingregulations that curb risk-taking and protect the interests of depositors. A corol-lary action involves monitoring the banking system through government super-vision and external (as well as internal) auditing procedures that enforce compli-ance with the regulations, reveal the financial condition of the banks, improvemanagement, and prevent fraud. Remedial measures include deposit insurance;government intervention in troubled banks, either through conservatorship or


assumption of ownership; use of the central bank's function as lender of lastresort to improve liquidity; and liquidation or recapitalization procedures.

By its very nature, the framework for preventive measures must be establishedbefore financial crises occur and must be clear, consistent, and credible. (For agood review of preventive measures, see Polizatto 1990). Accordingly, preven-tive measures tend to be embodied (with various degrees of efficacy) in existinglegal and institutional structures, while remedial measures are often ad hoc,both in the nature of the actions taken and in the timing. There are severalreasons for the ad hoc nature of remedial measures. First, the authorities mayview the fiscal implications of recognizing and allocating huge losses as politi-cally unpalatable. Moreover, these remedial measures are implemented only inthe event of distress, which, it is hoped, may actually not occur. And finally, theevent that triggers these measures-insolvency (threatened or actual)-is, asnoted earlier, hard to define and measure precisely because of the lack of timelyand accurate accounting information and the difficulty of valuing certain assets.Thus developing countries generally lack the legal and institutional mechanismsfor dealing with distress once it occurs.

Often the path of least resistance is to do nothing-to hope for spontaneousimprovement and avoid taking action in the meantime. This results in politicalinterference in decisionmaking and a loss of the banking authorities' credibility.Inaction almost inevitably increases the cost of the cleanup. Even if the authori-ties do take decisive action, the absence of clear policy procedures may lead torestructuring rather than liquidation. If restructuring involves compensating theshareholders of failed banks, perverse incentives can be created that can lead to aloss of financial discipline and to the taxpayers' assumption of the losses.

The importance of establishing well-defined procedures for intervening incases of financial distress in developing countries is heightened by the charac-teristics of the financial systems of commercial banks. In many countries the lackof substantial debt and equity markets or credit-rating agencies makes it difficultto ascertain the market value of banks. In some countries the existence offinancial groups, in which there is little separation among banking, commerce,and other types of financial services, can further complicate the valuation ofassets and liabilities. Volatile asset prices such as interest rates tend to exacerbatefinancial distress and further complicate the rehabilitation or liquidation pro-cess. Finally, the scarcity of motivated, well-paid, and technically competentcivil servants in the agencies charged with managing bank distress can under-mine regulatory intervention.

Bank Rehabilitation

Rehabilitating insolvent banks entails three types of actions: restoring sol-vency through a recapitalization scheme that covers all existing losses and pro-vides the institution with an adequate level of capital; restoring profitability byrestructuring the institution's staff, operations, cost structure, and physical in-


frastructure; and upgrading management in the hope that new staff will improvedecisionmaking, risk management, and control systems and procedures.

Quick and successful bank rehabilitation requires that these actions occursimultaneously. Of the three, recapitalization is probably the most sensitive andhence should be handled particularly carefully. The authorities must decide howto allocate the financial losses among the government and new and old deposi-tors and shareholders. Because rehabilitation usually takes the form of explicitor implicit deposit protection, depositors are rarely forced to incur losses. Exist-ing shareholders will lose the value of their investments to the extent that "old"capital is reduced (through write-offs or provisions for bad loans). Until existinglosses are allocated, an insolvent bank is unlikely to attract new capital. There-fore, the government typically absorbs any residual losses (after writing off oldcapital) and then solicits new capital.

The primary difference from one recapitalization scheme to the next is thesource of new capital, which can come from former shareholders, new share-holders (say, other banks), the government, or a combination of these. The mostcommon schemes involve bailouts, nationalizations, assisted mergers, and pur-chases and assumptions.

* In a bailout the bank is kept afloat, with its operations and ownershipintact. The government covers all the losses.

* If the government decides to nationalize the bank, the losses again are borneentirely by the government, which provides enough capital to cover not onlythe losses but also additional outlays to keep the institution viable and toallow it to meet legal capital requirements.

* In an assisted merger the government covers all losses, restores the institu-tion to solvency, and provides sufficient resources to induce other institu-tions to invest. At this point outside investors may become interested inpurchasing the bank. Assisted mergers probably entail smaller losses for thegovernment than either of the first two options because competing bids canbe sought.

- Finally, in a purchase and assumption, which operates under the same prin-ciple as an assisted merger, the government replaces bad assets with goodones, but it must put in more fresh resources than it does in an assistedmerger to compensate for the assets that were removed. Acquiring banksprefer this option to an assisted merger if they are unable to ascertain theextent of losses in the loan portfolio or in other contingent obligations of thefailed bank or if they are convinced that the government is likely to be moresuccessful in recovering the bank's nonperforming assets.

The implementation as well as the choice of a recapitalization scheme issubject to a formal process in all countries, but often the need for rapid actionresults in extrajudicial intervention. In some cases, ownership of the troubledbank is actually transferred to the government. In developing countries theacquiring institution is typically the central bank or a public bank. In Venezuela,


for example, the central bank has assumed majority ownership of three dis-tressed commercial banks. In Uruguay, between 1985 and 1987, the state-owned Banco de la Republica bailed out four of the largest banks in the system,which accounted for about 31 percent of total banking deposits. And in Peru thestate-owned Banco de la Naci6n has directly or indirectly bailed out all thefailing banks that the authorities have decided to keep operating.

Once the government has taken over, however, it is not likely to have thetechnical expertise to turn around a commercial bank. Moreover, by drawing itinto the public sector, the government politicizes the process and makes re-privatization difficult, which effectively leads to a de facto nationalization.This could occur if the public agency that acquires the bank finds that it isprofitable but is even more likely where the troubled bank becomes a financialand administrative burden and the government has to shoulder additionallosses.

In some countries ownership of the insolvent bank is transferred to a spe-cialized institution such as a deposit insurance agency or a special-purpose cor-poration or trust. This concentration of government activities in a single agencymay help to turn around the bank and to shield the process from politicalintervention, especially if the agency is required by law to reprivatize or liquidatethe institution within a specified time. In the United States the Federal DepositInsurance Corporation (FDIC) can acquire a problem bank and operate it as asubsidiary "bridge bank." By law the FDIC must divest itself of the bridge bankwithin two years.

In other countries government intervention in the troubled institution entails ashift in the control, but not the ownership, of the bank. As in corporate reorga-nizations, legal control over the bank's assets may be vested in a conservator(usually the bank supervisory agency, such as the FDIC in the United States, orthe central bank in Argentina). The conditions that prompt the appointment of aconservator can include technical insolvency, actions by management that vio-late the bank's charter, inability to cover maturing obligations, or the presump-tion of fraud or mismanagement on the part of bank officials or directors. Theconservator is charged with operating the bank as a going concern and withinvestigating its financial condition. Once the investigation is complete, theconservator evaluates the various options for recapitalization and decides whichone to pursue. In this role the conservator makes an offer to each class ofcreditor and typically has discretion to issue new equity and deposits or toinitiate liquidation procedures.

Bank Liquidation

Liquidation involves the forced sale of bank assets once operations have beenpermanently terminated. A bank regulator must decree the suspension of opera-tions, although a judge may also have to issue a cease-and-desist order. Liquida-tion typically involves the appointment of a receiver, which in most countries


will be the central bank or a deposit insurance agency, but may also be a judge.All operations of the bank are suspended, and management and board membersare displaced. The receiver takes over the bank's assets and disposes of them,paying each class of creditor according to a hierarchy defined by the laws gov-erning extrajudicial liquidations or by the commercial or bankruptcy code.

Assets can be liquidated through market sales (if a ready market exists for theasset and the sale is not so large that it adversely affects prices), public auctions,and private placements. When a particular asset does not have the benefit ofa deep market or when the government, for political reasons, does not wish tosell assets publicly, private placements are preferred because they are lesstransparent.'

Causes of Regulatory Failure

Despite the need to resolve financial distress quickly, the authorities tend todelay the decision. A growing body of literature stresses the importance ofincentives that encourage the parties to bank distress to ignore the problems aslong as they can, delaying recovery, and leading to four types of regulatoryfailures: inaction, which permits losses of insolvent banks to mount; a biasfavoring rehabilitation over liquidation; the adoption of policies that are de-signed to prevent actual failure but that undermine financial discipline; and aloss in the authority of regulators. Two groups of incentive problems must besolved: those of regulators and those of bank managers and various claimants.

Incentives of Regulators

In analyzing the savings and loan crisis in the United States, Demirguc-Kunt(1989), Kane (1988), and Silverberg (1990) have each examined a model ofregulatory behavior and decisionmaking in which the relationship of regulatorsto political interests (see also Buchanan 1967) and the potential capture ofregulators by the banks (see Peltzman 1989; Stigler 1971) create incentives thatmake regulators unwilling or unable to let banks fail. These incentives can begrouped under three headings: political, personal, and bureaucratic.

* Political obstacles. Regulators are reluctant to alienate the politicians whoappoint and oversee them. Closing down a commercial bank is likely tohave costly budgetary consequences for taxpayers, with attendant implica-tions for the reelection of politicians. This argument assumes that a problembetween the principal and the agent exists in which the principals (tax-payers) find it difficult to monitor the actions of the agent (regulators andgovernment) who represents their interests in resolving financial distressquickly.

* Personal relationships. Regulators also must satisfy another clientele: thebanks they regulate. A regulator's interests tend to be increasingly coinci-


dent with those of the banks it regulates, particularly if salary differentialsbetween the government and the private sector are large. In developingcountries, where bankers and regulators are drawn from a small set of well-educated, economically powerful, and politically connected people, per-sonal relationships can be very important. Regulators may see their role asproviding comfort to the banks they are charged with regulating. Underthese conditions they are unlikely to move quickly to resolve financial dis-tress, whether by closing, rehabilitating, or liquidating a bank. In fact,distress can increase the incentives to compensate bank managers and evenshareholders in return for future benefits.Bureaucratic difficulties. A third factor relates to intragovernmental juris-diction, particularly in countries that have regulatory institutions with over-lapping functions and responsibilities. Instead of creating checks and bal-ances on regulators through a form of competition, such a structure oftenleads to bureaucratic inertia. If each agency acts to protect its mandate andhusband scarce information, the outcome is unlikely to advance the resolu-tion of financial distress.

When regulatory and supervisory functions are decentralized, the com-patibility of the objectives and authority of each agency must be carefullybalanced. Each agency should be permitted to issue those threats (fines,liquidation, intervention, and so on) that it is able to implement and enforceunilaterally. If the agency is unable to enforce its own threats, both itscredibility and the morale of its officials will deteriorate, underminingeffectiveness.

Examples of the institutional structure distorting the incentives of regulatorsto resolve cases of financial distress are numerous. In the United States regula-tory arbitrage among competing agencies acts to weaken the enforcement ofregulations. In Brazil and Mexico, where financial conglomerates are important,bank brokerage and investment banking operations can fall under the scrutinyof several regulatory agencies, including the central bank, the securities ex-change commission, and even agencies that regulate insurance and pensionfunds. In Brazil the confusion has resulted in numerous conflicts. The liquida-tion of several brokerage firms in 1989 in the aftermath of that country's stockexchange crisis was coupled with reports that regulatory agencies had not actedpromptly.

In Venezuela bank insolvency problems fall under the jurisdiction of the super-intendency of banks (through its power of conservatorship) or of the depositinsurance corporation (through its recapitalization instruments). The laws andprocedures do not clearly spell out who should initiate actions or how eachagency can support the other. In Nicaragua the national comptroller is chargedwith supervising banks' compliance with reserve requirements and assessing thecorresponding fines, which it does only after significant delays. Collecting thefines is the responsibility of the central bank, although it rarely does so. Because

64 The World Bank Research Observer, vol. 10, no. 1 (Febmary, 1995)

the responsibility is shared, each agency sees the other's inaction as a reason fordelaying its own efforts.

The lack of coordination among regulatory agencies means that decisions torehabilitate or liquidate troubled banks may not be based on the full budgetarycost to the government, either because the authorities fail to take relevant costsinto account or because they apply a distorted opportunity cost of funds. Inmost cases the agency in charge of rehabilitation is responsible for paying off thedepositors (if it happens to be a deposit insurance corporation), injecting freshcapital, and covering the bank's losses by purchasing the value-impaired assetsat less than market value. But other government agencies also incur costs. Thecentral bank generally gives rediscounts or credit to insolvent banks to provideliquidity, but this support may contain elements of recapitalization (if it isgranted at below-market rates or is continually rolled over and effectively cap-italized) or deposit guarantees (if it funds deposit withdrawals). Central banksupport is often indistinguishable from the rehabilitation support advanced bythe supervising agency. In fact, the central bank may subsidize the insolventbank's funds (implicitly capitalizing the bank) and may permit continual, unse-cured intraday or overnight overdrafts without limit.

Regulatory failure may also stem from the regulator's lack of informationabout the bank's financial condition. The authorities may be hard pressed toprove beyond reasonable doubt that intervention is justified and not confisca-tory. Timely and reliable accounting information is essential to shield againstlawsuits brought by owners. Moreover, uncertainty about the potential costs ofthe various alternatives leaves the authorities vulnerable to charges of mis-management, particularly when information supplied at a later date casts theappropriateness of the decision in a new light.

When information is deficient, the status quo will always look more attractivethan other alternatives. This has been the case in Argentina, where bank inter-ventions are invariably decreed only after the prospects for recovery are all butgone, and Bolivia, where a fundamentally insolvent bank continued to functionfor five years.

Incentives of Bank Stakeholders

Two strands of theoretical research, contingent claims analysis (Altman andSubrahmanyam 1985) and agency theory (Copeland and Weston 1988), shedlight on the incentives stakeholders have when a bank is in distress. All forms ofclaims on the bank can be viewed as combinations of options contracts, whosevalue depends on the volatility of the return on the bank's assets.2 When a bankis in distress, owner-managers have an incentive to expropriate the wealth of thebondholders by engaging in highly leveraged and risky operations. To preventthis behavior, bondholders try to monitor the actions of bank officials by incor-porating covenants in their contracts that restrict dividend payments and thedisposition of the bank's assets or that limit the amount of leverage the bank


may assume. These so-called agency costs (of monitoring and enforcement) tendto rise as the proportion of debt to equity rises. In addition, conflicts can alsooccur between managers and shareholders. Those who hold equity in the bankbut who do not participate in its daily management or control also have tomonitor managers or devise a compensation package that aligns their interestswith those of the shareholders (see Jensen and Meckling 1976).

Sheng (1990) suggests that the conflicts among different claimants upon thefuture cash flows of the bank and the contractual arrangements adopted forprotecting the value of these claims affect the bank's decisions on the allocationof assets and capital structure.3 Each of the three types of claimants-equityholders (or managers), depositors, and senior and subordinated debt hold-ers-faces different incentives. Equity holders have an incentive to encouragemanagers to undertake risky asset-allocation decisions to maximize their profits.Depositors, if not subject to some form of credible de jure or de facto depositinsurance, have an incentive to monitor management and move their funds to"safe" banks-imposing a certain degree of market discipline. Finally, variousclasses of subordinated debt holders cannot withdraw funds on demand, are notprovided with ex post insurance, and do not share in the profits derived fromrisk-taking. This group of claimants has an incentive to limit the riskiness ofcommercial bank activities by incorporating explicit limitations on risk-taking inlegal bond covenants associated with subordinated debt offerings.

Incentive Compatibility

All this suggests that two aspects of the rehabilitation (or liquidation) processwill also introduce important incentive problems: ill-defined property rights,and the lack of segregation of new and old claims in distressed banks.

Ill-defined property rights. When the legal rights and standing of bankclaimants are unclear, this can slow the efforts of officials to resolve anexisting crisis. If the rehabilitation process permits existing shareholders torecover their losses, shareholders have an incentive to press for risky assetselection but little incentive to avert a crisis by replacing management. Atthe same time, if depositors, bondholders, and employees perceive thatexisting shareholders will not incur losses or that the loss-allocation processwill be arbitrary, any incentive they might have to monitor bank manage-ment and avert crises evaporates. The very process of crisis resolution isslowed if claimants bring court actions to dispute the decisions of the con-servator or receiver.

The incentive for employees (as well as for potential new equity holders)to cooperate in a bank rehabilitation or liquidation depends in large part onwhether pension, severance payments, and other employee benefits areclearly delineated relative to the claims of other creditors. When employ-ment contracts permit profit sharing, or when other benefits such as pen-sions or severance payments are linked to the continued existence of the


bank, employees have a greater incentive to police management. If thepension plan is not company-related, employees, like other shareholders,may encourage management to take greater risks to maximize the value ofthe bank. Borrowers, by contrast, will have an incentive to slow down theprocess of rehabilitation or liquidation-particularly when it is clear thatthe bank will be foreclosed.Segregation of new and old claims. In rehabilitations that permit the conser-vator to take net new deposits or issue new equity, new funds might be usedto pay off old (pre-intervention) equity holders or other creditors. This typeof intermingling does not occur in standard corporate reorganizations, inwhich existing creditors of the bank are blocked from access to their assetsand pre-intervention equity holders suffer losses. Intermingling will obstructrehabilitation plans and increase the costs to the government. If new share-holders must share future returns with old creditors or shareholders, theywill not invest in the bank.

To avoid these problems, existing shareholders should be the first to assumelosses in rehabilitations and liquidations. The particular risk-reward structurebuilt into equity gives it a vital role in all corporate structures: because payoffsare linked to company performance, equity holders should have a strong incen-tive to monitor management's actions. Rehabilitation operations that do notpreserve this incentive for equity holders will weaken market discipline.

There are several ways to make equity holders shoulder the losses of banks tothe extent that limited liability permits. The cleanest option is to charge lossesimmediately against bank capital by provisioning or writing off the bank's assetsbefore restructuring. Once this has been done, government agencies can injectcapital to replace the lost capital and cover any additional losses. This is themethod that the Spanish deposit insurance fund applied. Chile followed analternative approach in 1984 that preserved old equity but modified the payoffstructure to extract the upside potential until the government was compensatedfor the cost of rescuing the bank. This process, which requires strict segregationof old and new capital, called for the central bank to purchase (in cash) the riskyportfolios of commercial banks. Each portfolio was priced at face value andcould not exceed 150 percent of the institution's capital and reserves. The valueof the portfolio was capitalized into a loan, was indexed to the price level, andaccumulated 5 percent annual interest. Commercial banks continued to managetheir loan portfolios on behalf of the central bank. Dividends were paid to newequity, but old equity did not receive dividends until the entire portfolio hadbeen repurchased. The share of dividend payments that would have normallyaccrued to old equity went toward repaying the loan.

A related point is that using capital assistance to cover negative net worth isbetter than using income enhancement assistance. Insolvency is a problem ofstocks: it occurs when the stock value of liabilities exceeds the stock value ofassets. The insolvency may result from a lack of profitability that gradually


erodes capital, but whatever the source of the insolvency, at a particular point itbecomes a stock problem.

REPLENISHING CAPITAL. An insolvent bank can be returned to solvency eitherby replenishing its capital (the stock solution) or by providing a steady subsidythat can be capitalized over time (the flow solution). The flow solution is fre-quently adopted because it allows the government to spread out the costs ofrecapitalization. The same result can be achieved under the stock solution if thecapital injection provided by the government is in the form of governmentbonds. The government acknowledges the total future cost at the time it issuesthe bonds; the cost is equal to the cumulative annual debt service on the bonds.Even though the cost is spread out, it represents a stock assistance to the bank ifit can sell the bonds in the secondary market. Politically, flow assistance may stillbe superior because it defers not only actual government payments but also therecognition of the costs. Although the bank continues to operate, it remainsinsolvent for some period. This situation is problematic for two reasons. First,the bank is especially vulnerable to any economic or financial shocks that mightoccur. Without the robustness conferred by capital, the probability that the bankwill not recover is higher, and, ultimately, the government's efforts and resourcesmay be wasted. Second, the moral hazard problem is exacerbated because exist-ing shareholders continue to run the bank without having any financial stake init.

THE INCOME SOLUTION. Income enhancement is often used to prevent theactual failure of financial institutions. Although countries have found creativeways to package and justify such assistance, the end result in terms of fiscal costsand incentives is the same. In 1989 Venezuela's deposit insurance corporation(FOGADE) made loans at 6 percent to insolvent banks under its auspices topurchase Treasury bonds at face value paying around 15 percent. In effect, thetroubled institutions were given a 9 percent annual flow subsidy for about sevenyears. In 1982 Chile's central bank purchased the risky or nonperforming port-folios of eligible banks by issuing a ten-year note to the institution, effectivelyguaranteeing a return on the portfolio equal to the interest rate on the note. Inthe United States in the early 1980s, the Federal Savings and Loan InsuranceCorporation routinely compensated ailing S&Ls for the negative spread be-tween their cost of funds and the yield on fixed-rate mortgages, essentiallyassuming the interest rate risk on the S&Ls' long-term mortgage obligations.

Allocating Losses

The principle of preserving shareholder discipline throughout the rehabilita-tion or liquidation process can be extended to other stakeholders. For instance,the blame for an institution's distress usually does not lie exclusively with itsowners or managers. Several types of depositors could be held responsible in


varying degrees: bank insiders and related parties who may have benefitedfrom excessive lending or preferential treatment; clients with deposit balancesand overdue loans; official or institutional depositors that influenced lendingdecisions (because they had a stake in the activities in which the distressedinstitution concentrated its portfolio); recent depositors attracted by high inter-est rates, who should accept the risks implicit in those returns; or very large,informed depositors who should have exercised some market discipline.

These depositors should be forced to assume their losses unless they areexplicitly covered by deposit insurance. Short of canceling the bank's obliga-tions to them, their stake could be converted into equity. In a liquidation, theirpositions in the hierarchy for distribution of proceeds would be reduced, andin the case of a rehabilitation, their payoff would be contingent on the bank'srecovery.

Wholesale debt rescheduling and other such measures that undermine mar-ket discipline are counterproductive. Government-decreed reprogrammingshave been common in Latin America; they have been instituted for social orpolitical reasons (Nicaragua), to promote investment and growth (Bolivia), orto buy time for the resolution of generalized banking crises (Argentina, Chile,and Uruguay). Typically, the government specifies the range of eligible debt-ors-by economic sector, geographic location, or income level-and compen-sates financial institutions for the income forgone. Reschedulings and debtworkouts are also performed in the context of corporate reorganizations.Filing for reorganization often allows firms to avoid repaying their debts,particularly if no penalties (in terms of future access to credit) are involved.

From the bank's point of view, it is always more profitable to reschedulenonperforming loans that are unlikely to be repaid. Lower interest rates willnot hurt the bank's income (as these loans may never be collected), andrescheduling makes overdue loans current, thereby allowing the bank to returnfunds put aside to cover the anticipated loss. This incentive to reschedulemakes the exercise a cosmetic one that simply defers problems into the future.Furthermore, to the extent that debtors understand this bias, they have anincentive to let their loans become overdue so that they become prime candi-dates for reprogramming. As a result, portfolios may worsen in the short run,and credit discipline is certain to be undermined in the medium term. Inaddition such schemes become a source of uncontrolled quasi-fiscal losses forthe central bank and confirm the public's view of the government as theresidual absorber of all losses and risks.4

Resolving Bank Crises: More Rules, Less Discretion

Many laws relating to the procedures for handling bank crises grant regula-tors and supervisors as much discretion and as many policy instruments aspossible so that they can pick the appropriate procedure in each case. This


approach has not been satisfactory, especially in developing countries, forseveral reasons. The lack of clear procedures clouds the "rules of the game,'and regulatory discretion often translates into unpredictability. Although dis-cretion may be appropriate in a particular case, the deterrent effect on otherbanks may be lost. Moreover, unless regulators have the technical expertiseand managerial skills essential in financial transactions, the benefits of discre-tion may not accrue. The use of rules may preclude better outcomes in somecases, but the downside risks are more limited. And finally, too much discre-tion opens policymakers to charges of abuse of power and permits inaction ateach stage in the remedial process. The law, initially conceived as a tool at thedisposal of authorities, can be turned into a weapon against them. Resolutionof bank crises is often paralyzed when discretion leads to charges of financialimpropriety or legal challenges to the adopted procedures.

In a crisis, regulatory discretion often deteriorates into forbearance. Regula-tors and insurers are anxious to avoid. responsibility for closing down thebank. If the financial condition of the (de jure or de facto) insurer is weak, itwill try to defer the closure to protect itself. Because rehabilitation is a costlyventure, the path of least resistance may be to turn a blind eye to insolventinstitutions and provide implicit subsidies through exemptions fromregulations.

Forbearance may be justified when it does not seek to cover up problems butrather is intended as a respite for institutions that face a financial crisis as wellas tough new regulatory standards; if accompanied by more stringent supervi-sion, forbearance may add a touch of pragmatism. But as a rule, regulatoryand accounting standards should be tougher, not weaker, during bank crises,when adverse incentives may add to the costs of distress. Strict regulatorystandards based on rules can mitigate the extent of regulatory forbearance.

For these reasons, an effective legal framework is needed to protect authori-ties, to provide clear signals to the private sector, and to force policymakers toact promptly. The legal system should clearly specify the circumstances thatwarrant liquidation, conservatorship, or rehabilitation; the range of the re-ceiver's or conservator's actions, powers, and rights; and the timing of theseactions.

The laws regulating banking activity should be tough, but realistic. Whenrequirements are too demanding or costly, the laws are frequently transgressed,undermining confidence in the laws as well as in the regulators who aresupposed to enforce them. Because regulators\would almost always prefer torehabilitate a bank than to liquidate it, the legal system should spell out theconditions that would require liquidation. For example, the law can bindauthorities to initiate the liquidation process before a bank becomes technicallyinsolvent. Such a rule would apply only to banks with a large proportion ofassets that are tradable in active markets. Where these conditions are satisfied,in countries, such as Brazil, with well-developed capital markets and a signifi-cant set of nonbanks or in countries, such as Mexico, that permit banks to buy


and sell securities, rules can lay out the mechanisms for asset disposal and thehierarchy of claims against liquidated assets.

Simply announcing the framework that has been adopted for resolving bankfailures is likely to affect the future behavior of economic agents and thus havelong-term effects on government policies. Systematically rehabilitating banks toprevent failure implies full deposit protection and compromises market disci-pline. Bailing out large banks while allowing smaller banks to fail (as theUnited States has done) confers a competitive advantage on large banks. Whenthe public perceives that rehabilitation is seldom successful, placing banksunder conservatorship will be tantamount to a public announcement that thebank is about to fail and, unless deposits are insured, is likely to trigger a bankrun that will only accelerate insolvency. The move thus becomes a self-fulfillingprophecy that entirely undermines the use of conservatorship as a policy tool.Argentina suffered from this experience in the 1980s, when interventionsoccurred only after the banks' equity and profitability positions were severelyundermined and the chances that the bank could be successfully turned aroundwere very small. It is precisely these announcement effects that rules-basedmechanisms can exploit to improve financial discipline.

Some Provisos

This article has outlined a set of principles that can guide decisionmakers inchoosing the right mechanism for resolving bank distress, but several provisosshould be borne in mind in applying these principles in a particular country.First, technological change (particularly in telecommunications) is redefiningthe very meaning of banking versus securities transactions. Systemic risks thatoriginate through the operation of interbank markets can cause contagionwhen one bank or securities subsidiary fails. As the link between regulating theform of payments system and financial distress of counterparties becomes morepronounced, the restructuring and exit processes for financial institutions needto be flexible enough to permit adequate unwinding. Complex linkages be-tween financial system policies and institutions temper the scope of rules overdiscretion.

Second, specific mechanisms to resolve bank failure are only one element ina consistent package of financial reforms and must be complemented by appro-priate financial reforms in the lender-of-last-resort function, the form of de-posit insurance, the payments system, prudential regulation, and the legalframework for delivering financial services.

Finally, in designing more effective financial policies, several criteria need tobe used to deal with tradeoffs. As noted in work by Stiglitz (1985) and Merton(1994), it is important to consider how a particular financial policy or reformmay affect the degree of financial stability, the incentives for monitoring finan-cial institutions (by stakeholders or regulators), the degree of competition in


the provision of financial services, and the set of financial contracts availableto consumers of financial services.

This article addresses only one set of criteria-incentive effects. The desir-ability of some of the mechanisms recommended in this paper might be alteredif all the above criteria were applied; integrating them provides a rich researchagenda.

Notes

Thomas Glaessner is senior financial economist in the World Bank's Country DepartmentII, Latin America and the Caribbean, and Ignacio Mas is a financial economist in the PrivateSector Development Department at the World Bank. Part of this research was done whenIgnacio Mas was a visiting professor of international business economics in the GraduateSchool of Business at the University of Chicago. The authors would like to thank CharlesBlitzer, Asli Demirguc-Kunt, Mark Dorfman, Fred Jensen, Mauricio Larrain, Ross Levine,Felipe Morris, Sarath Rajapatirana, S. Ramachandran, Ramon Rosales, David Scott, An-drew Sheng, Sam Talley, and Alfredo Thorne for helpful comments.

1. Few developing countries have set up public entities like the Resolution Trust Corpora-tion in the United States to take over the assets of troubled banks.

2. For instance, one can think of the equity owner as selling the assets of the firm tobondholders in return for cash and a call option on the value of the firm. The shareholderwill exercise the call option (that is, pay off the bondholders) if the value of the firm exceedsthe value of the bond.

3. Both high bankruptcy costs and taxes can invalidate the Modigliani-Miller propositionfor the individual firm.

4. In many Latin American countries, this phenomenon is referred to as the privatizationof gains and socialization of losses.

ReferencesThe word "processed" describes informally reproduced works that may not be available

through library systems.

Altman, Edward I., and Marti G. Subrahmanyam, eds. 1985. Recent Advances in Corpo-rate Finance. Homewood, Ill.: Richard D. Irwin.

Benston, George, and George Kaufman. 1988. "Risk and Solvency Regulation of Deposi-tory Institutions: Past Policies and Current Options.' Working Paper sM88-1. FederalReserve Bank of Chicago, Chicago, Ill. Processed.

Buchanan, James M. 1967. Public Finance in the Democratic Process. Chapel Hill, N.C.:University of North Carolina Press.

Copeland, Thomas E., and John Fred Weston. 1988. Financial Theory and CorporatePolicy. 3d ed. Reading, Mass.: Addison-Wesley Publishing Company, Inc.

Demirguic-Kunt, Asli. 1989. "Deposit-Institution Failures: A Review of Empirical Litera-ture." Economic Review 25(4):2-19.

Jensen, Michael C., and William H. Meckling. 1976. "Theory of the Firm: ManagerialBehavior, Agency Costs, and Ownership Structure." Journal of Financial Economics(October):305-60.

Kane, Edward. 1988. "Changing Incentives Facing Financial Services Regulators.' Eco-nomic Review 24(4):9-30.


Mas, Ignacio, and Samuel H. Talley. 1992. "The Role of Deposit Insurance." In DimitriVittas, ed., Financial Regulation: Changing the Rules of the Game. Washington, D.C.:World Bank.

Merton, Robert. 1994. "Operation and Regulation in Financial Intermediation: A Func-tional Perspective" Paper presented at CEPR/ESG Network in Financial Markets, Paris.June. Processed.

Peltzman, Sam. 1989. "The Economic Theory of Regulation after a Decade of Deregula-tion." Brookings Papers on Economic Activity (Special issue):1-59.

Polizatto, Vincent P. 1990. "Prudential Regulation and Banking Supervision;' Policy Re-search Working Paper 340, World Bank, World Development Report Office, Washington,D.C. Processed.

Sheng, Andrew. 1990. "The Art of Bank Restructuring: Issues and Techniques." Paperpresented at EDI Senior Policy Seminar on Financial Systems and Development in Africa,World Bank, Washington, D.C. Processed.

Silverberg, Stanley C. 1990. "The Savings and Loan Problem in the United States:' PolicyResearch Working Paper 351, World Bank, Policy Research and External Affairs Depart-ment, Washington, D.C. March. Processed.

Stigler, George S. 1971. "The Theory of Economic Regulation:' Bell Journal of Economicsand Management Science 2(Spring):3-21.

Stiglitz, Joseph E. 1985. "Credit Markets and the Control of Capital." Journal of Money,Credit and Banking 17(2, May):133-52.

Todd, Walker E, and James B. Thomson. 1990. "An Insider's View of the Political Economyof the Too Big to Let Fail Doctrine." Paper presented to the 65th Annual WesternEconomic Association International Conference, San Diego, Calif. Working Paper 9017.Federal Reserve Bank of Cleveland, Cleveland Ohio. Processed.


FINANCING INFRASTRUCTURE INDEVELOPING COUNTRIES:LESSONS FROM THE RAILWAY AGE

Barry Eichengreen

In recent years suggestions for reforming the provision and financing ofinfrastructure services in developing countries have focused on private par-ticipation. This alternative to public financing is seen as a way both tominimize the inefficiencies of public administration and to avoid the need forexternal borrowing.

In fact, for much of the nineteenth century, infrastructure projects wereprivately financed and built. This approach, however, did not obviate theneed for government intervention and foreign capital. Because of the diffi-culties of assessing projects, investors were reluctant to commit their funds,and governments turned to subsidies and loan guarantees to encourage in-vestment. Often, however, government intervention only replaced one set ofproblems with another. Investors with government-guaranteed loans had noincentive to monitor the firm's performance-a limitation that led to thediversion of funds and frustrated the public interest. This article draws outthe implications of this experience for policymakers in developing countriestoday.

F n or low-income countries, investments in infrastructure have alluring bene-fits as well as daunting costs. Where transportation, communication, andpower generation are inadequate, increased supplies can do much to boost

productivity and growth. But where income and productivity are depressed byinadequate infrastructure, the financial resources needed to underwrite invest-ments are difficult to mobilize. Because the lack of infrastructure limits invest-ment and the lack of investment limits infrastructure, low-income countries canfind themselves in a low-level equilibrium trap from which it is difficult to escape.

Two potential escape routes-government subsidies and foreign bor-rowing-are available in principle. If infrastructure is critical for raising produc-tivity and profitability elsewhere in the economy but those who finance theproject cannot capture sufficient revenues to repay their costs, the classic effi-

The World Bank Research Observer, vol. 10, no. 1 (February 1995), pp. 75-91© 1995 The International Bank for Reconstruction and Development/THE WORLD BANK 75

ciency argument for subsidies applies: that a subsidy that closes the gap betweenprivate and social benefits will prevent the relevant form of infrastructure frombeing undersupplied. And even when the returns can be appropriated, invest-ment may still not be attractive if high interest rates make domestic funds costly;investors may then seek finance abroad, where it is cheaper. Not surprisingly,government guarantees and foreign borrowing are prominent features of infra-structure finance in many developing countries.

Increasingly, these arguments for government intervention and foreign bor-rowing are regarded with skepticism. The costly "white elephants" subsidized bygovernments have underscored doubts about the efficiency of public finance,and the debt-servicing difficulties of developing countries have raised questionsabout the efficacy of foreign borrowing. Both observations encourage an interestin proposals to commercialize and privatize infrastructure projects and to fundthem by promoting the development of financial markets.

There is nothing new about these arguments or these reservations. Infrastruc-ture projects were privately financed and constructed in virtually all the overseasregions of recent European settlement in the nineteenth century (which, for thepurposes of this article, is assumed to extend to 1914). At the same time,however, government subsidies and external finance were integral to the processof developing infrastructure. Although early U.S. railways, to take a prominentexample, were private undertakings, land grants and government guaranteessubsidized their construction. Finance was raised abroad, mainly on the Londoncapital market. This history suggests that private initiative should not be viewedas obviating the need for government guarantees and foreign finance.

This article elucidates these historical patterns of public intervention andexternal finance for infrastructure investment, with a particular emphasis onrailways. Its premise is that these patterns are consequences of the structure offinancial markets in countries in the early stages of economic development.Nineteenth-century infrastructure investments included canals, docks, electricpower grids, sanitation systems, telegraph systems, tramways, and turnpikes,but railways-the most prominent and capital intensive of these invest-ments-commanded center stage. Railways forged unified national markets,linked domestic producers to the expanding world economy, facilitated the de-velopment of mass-production techniques, and incubated modern managementpractices.' The analysis therefore draws on the literature on nineteenth-centuryrailway investment, focusing most notably on recent contributions such as Bas-kin (1988) and Carlos and Lewis (1992, forthcoming).

Government intervention, external finance, and debt-servicing difficulties arecorrelates of the imperfections in financial markets that impose a heavy burdenon governments seeking to finance infrastructure projects. At the same time,government policies to overcome asymmetric information can encourage man-agement to engage in bankruptcy for profit (a problem that Akerlof and Romer1993 refer to as "looting"). This tradeoff between credit rationing and the risk ofbankruptcy for profit is at the heart of this article.


The article reviews the interaction of asymmetric information, moral hazard,and adverse selection and, in this context, describes the financial and economicenvironment in which nineteenth-century firms and governments operated. Itthen considers two arrangements-government guarantees and land grants-used to subsidize infrastructure investment and to attract foreign finance anddescribes how these arrangements worked to relax credit constraints andweaken the incentive for creditors to monitor management. The article con-cludes by drawing some implications for developing countries today.

Theoretical ConsiderationsAccording to the Modigliani-Miller theorem, investors should be indifferent

to the composition of a firm's financial structure; if a firm is highly leveraged,investors can offset the risk by adjusting the composition of their portfolios. Butin the real world there are several reasons why this strong result does not prevail.The relevant reason here is asymmetric information (see Keeton 1979 andStiglitz and Weiss 1981). That is, the entrepreneur knows more about the proba-bility of failure than do external investors. So long as all projects yield the sameexpected return and investors are risk-neutral (that is, they are willing to acceptrisk rather than pay to avoid it), entrepreneurs with riskier projects will bewilling to pay more for external funds. Because their information is poor,lenders cannot discriminate among borrowers. To cover their risk, therefore,lenders raise interest rates, prompting entrepreneurs with safer projects to dropout of the pool of potential borrowers. (This is the problem of adverse selec-tion.) Higher interest rates, in turn, encourage the borrower to take on riskierinvestments. (This is the problem of moral hazard.) Raising interest rates cantherefore reduce the lender's expected return. Under these circumstances,lenders may ration credit.

In this model, credit rationing is driven by the riskiness of the underlyingenvironment and the severity of the barriers to the dissemination of accurateinformation. The more costly it is to sort projects (and the more pervasive theinformational asymmetries), the more serious the problems of adverse selectionand moral hazard. Many developing countries fit these conditions: they are, forexample, subject to terms-of-trade shocks and lack effective regulations requir-ing financial disclosure.

What pattern of finance is likely to emerge when information is asymmetricand conditions favor adverse selection and moral hazard? Entrepreneurs orpromoters with risky but potentially profitable projects will be forced to rely ontheir own funds. The more limited the lender's information, the more capitalpromoters will have to subscribe before external finance can be obtained.De Meza and Webb (1987) show that the resulting level of investment will besocially suboptimal. Under ideal conditions (known as first-best equilibrium, inwhich all markets clear exactly, information is perfect, and there are no distor-tions), with risk-neutral investors, all projects that yield returns equal to the


world rate of return will be undertaken. But when information is distributedasymmetrically, some such projects will not be financed. In such cases, a govern-ment interest-rate subsidy or guarantee may effectively relax the credit con-straint. The problem is that such an arrangement weakens the incentive forinvestors to monitor management performance, because bondholders are guar-anteed a return. This may allow management to divert resources to nonproduc-tive uses from which it benefits (Jensen and Meckling 1976). In the extreme,promoters may resort to bankruptcy for profit. That is, they compare the re-turns from the earnings of the firm with those they would receive by taking outfunds until they exhaust the resources that are available under the interest guar-antee and are forced to declare bankruptcy. Or they inflate accounting rates ofreturn relative to economic returns to appear solvent and acquire additional debtto be invested in activities that provide a high cash flow that the owners can tap.Because the government guarantees set the process in motion by weakening theincentive for investors to monitor management performance, the taxpayer is leftholding the bag. This problem is most acute where government guarantees areunconditional, where public oversight is lax (effective surveillance and regula-tion should prevent promoters from gambling that they can get away with thisstrategy), and where promoters and their confederates attach the least value totheir reputations.

Asymmetric Information and Investment in the Railway Age

Early infrastructure projects in North America posed formidable informationproblems for investors. Three factors in particular were conducive to informa-tional asymmetries: the novelty of the technologies, the uncertain prospects forlocal market growth, and the dearth of reputable promoters.

Technology. The lack of familiarity with the technologies and the paucity ofexperience outside England hindered investors' search for information.2 Hence,so prominent and profitable a project as the Erie Canal sent a powerful signal tothe capital market. The canal was completed in 1825 at a total cost of $11million-$3 million of which came from current sources and $8 million fromlong-term loans. The project was able to meet interest payments on the debt inits first year of operation and was fully paid off within ten years. The Erie'ssuccess set off a canal-building boom that engulfed the mid-Atlantic and NewEngland coasts. And yet the costs of building the canal were of only limitedvalue to those estimating the costs of building a canal through the higher moun-tains in western Pennsylvania. It is not entirely surprising that Pennsylvania'scanals turned out to be more expensive than anticipated.

Uncertainties. In areas such as the American West that had only recentlyappeared on maps, not even geography could be taken for granted. At a dinnerthrown by the Lord Mayor of London, an English investor asked an Americanguest whether Cincinnati or Illinois was the larger city. Even when the locationwas known, potential profitability was not: the amount of traffic a railway


could generate was contingent on the economic development of adjoining re-gions, which depended on such unknowns as the fertility of the soil, the re-liability of the rainfall, and the extent of mineral reserves.3 Where the volume oftraffic ultimately depended on these reserves, uncertainty posed a considerablerisk. Construction of many North American railways was based on crude fore-casts of coal or silver deposits. Where land had to be settled and cleared before itcould be farmed, it took years before the need for transport networks caught upwith the investment in railroad infrastructure. In Canada, for example, althoughrailway construction peaked in the final decades of the nineteenth century, sig-nificant gains in wheat production and rail traffic did not occur until the seconddecade of the twentieth century (Ankli 1980).

Experience with promoters. In addition to evaluating the economic prospectsof the project, investors had to assess the reputability of the promoter. Recentlysettled, sparsely populated regions were prime locations for fly-by-night opera-tors who could strike "sweetheart" deals with construction companies thatwould permit them to siphon off resources, saddling the project with insupport-able debts. Typically, the uncertainty was reduced by drawing on the expertiseand information of local investors. Where industrial and commercial develop-ment was precocious, it was possible to finance infrastructure through limitedpartnerships of local residents. Because early canals, turnpikes, and railroadshad modest capital requirements (by the standards of the long-distance rail linesthat followed), a local partnership could raise the requisite capital.

Local Finance

Examples from New England illustrate the point. The region was the centerof American textile manufacturing and hence of American industry in the earlynineteenth century, as well as the heart of commerce, shipping, and whaling. Agrowing number of small industrial towns provided a fertile market for short-haul railways (Chandler 1954). From the trade with China, Boston merchantslearned how to use entrepreneurial and managerial techniques to overcome thelong spans of time and distance (Johnson and Supple 1967). Much of NewEngland's railway finance was raised the same way in which the region financedits textile mills, by relying on family, friends, and other personal contacts.Where contract enforcement was problematic and information was difficult toverify independently, the markets made heavy use of such links. Friends andassociates vested their confidence in individual financiers with reputations forhonest dealing who signaled their commitment by putting their own funds atrisk (Baskin 1988, Lamoreaux 1986). As Johnson and Supple (1967, p. 338) putit, "investment tended to be a cumulative social process in an environmentlacking an impersonal, national money market." Thus, the danger of looting byfly-by-night operators was correspondingly reduced. Local farmers, bankers,merchants, landowners, contractors, and manufacturers subscribed the majorityof New England's early railway shares. Not only did such individuals have


favored access to information, but they stood to benefit from the transportationlinks. At the head of many early syndicates were textile producers seeking roadsthat would serve their mills and Boston merchants looking to railroads as a linkwith the hinterland market and the Great Lakes (Platt 1984).

That these projects were relatively modest (they connected Boston withnearby Portsmouth, New Hampshire, or Providence, Rhode Island) facilitatedthe local mobilization of capital. Even a more ambitious line built in the late1830s and early 1840s, the Western Rail-Road linking Boston with Albany,raised most of its finance locally.4

Underdeveloped markets could, however, impede efforts to raise local fi-nance. The attempt to market bonds for Canada's St. Lawrence and AtlanticRailway in the 1840s illustrates these difficulties: farmers who had no cash paidtheir subscriptions in the form of pork and eggs to feed the construction gangs.Some early U.S. railways similarly took subscriptions in the form of labor andmaterials (Cleveland and Powell 1912).

External Finance

This model of local finance was difficult to generalize because the capitalrequirements of early railways were more modest than those of subsequentprojects, and the funds available in New England exceeded those of other re-gions. Elsewhere it was necessary to seek external finance.5 Such funds were nota substitute for local finance; local investors still had to subscribe to indicatetheir willingness to put their money where their mouths were. If locals put upfunds, external investors could be confident that those in the best position toassess the needs of the project and monitor its progress and the actions of itspromoters would do so.

When Boston began to invest in the railroads of the U.S. South and West inthe 1840s, personal ties played a significant role. Railroad men coming toBoston contacted merchants who had invested in earlier railways. The promo-ters invested their own money in the project as evidence of their commitmentand talked friends and business acquaintances into investing as well. Long-termrelations between entrepreneurs from the West and merchants from Boston andbetween the merchants and their contacts provided a conduit for informationabout investment projects and individual promoters.

Railway securities tended to be traded in distant markets before such tradedeveloped in manufacturing and commercial concerns. Manufacturing usedmore exotic technologies, and commercial undertakings had less tangible assets(knowledge of customer requirements, for example), so investing in industryand commerce had to surmount even higher information hurdles. The railwayswere consequently among the first enterprises to access external finance on asignificant scale (Baskin 1988). According to Adler (1970), as early as the 1830sseveral lines around Philadelphia and in Virginia and North Carolina were ableto market securities in London, attracting support from British investors famil-


iar with financing rail networks. The pattern persisted: as late as 1914 railwaysecurities accounted for perhaps half of all outstanding foreign investments inthe United States. Foreign financing entailed the intermediation of specializedinstitutions that had grown up in the principal European financial centers todeal with information problems: issue houses, private banks, bill brokers, andfinancial investment companies. An illustration of the importance of these insti-tutions in Britain is shown in table 1. British investment houses typically re-tained American agents familiar with the American economy and railway proj-ects. They specialized in recommending high-quality foreign bonds, usuallythose of railroads that were well known and long established or were backed bythe credit of a state government. To show that they had confidence in theproject, these firms often bought the same bonds for their own portfolios. MostBritish investors followed their advice, limiting their purchases to a few largeeastern companies that provided adequate information (Adler 1970).

Financial institutions were not the only conduits for surmounting informationproblems. Immigrant communities were another, as were specialized publica-tions such as The American Railroad Journal and Poor's Manual of the Rail-

Table 1. The Proportion of Overseas New Issues Introduced by the Main Types ofBritish Issuing Houses, 1870-1914

Totalamount

Official Compan- issuedand semi- Joint- Overseas ies via (millions

official Private stock banks and their Other ofYears agencies banksi banks agencies bankers mediab pounds)

1870-74 1.8 53.0 4.4 9.6 18.2 13.0 390.61875-79 14.5 36.5 0.8 24.7 13.0 10.5 149.21880-84 6.7 38.5 3.3 14.1 26.7 10.7 355.31885-89 9.9 43.7 5.3 7.5 26.1 7.5 479.21890-94 10.4 46.4 9.0 8.8 19.6 5.8 349.61895-99 8.7 25.1 11.2 20.3 25.2 9.5 359.61900-04 27.4 19.2 17.8 14.4 16.7 4.5 258.21905-09 10.3 32.7 12.2 22.4 18.7 3.7 509.91910-14 8.3 35.2 17.4 18.8 17.5 2.8 738.81870-1914

Average (percent) 9.8 37.2 10.3 15.4 20.5 6.8 100.0Total amount

issued (millionsof pounds) 355.0 1,354.0 371.0 562.0 746.0 248.0 3,636.0

Note: Figures are percentages unless otherwise specified.a. That is, merchant bankers.b. Investment trust, 23 million pounds; finance, land, and property companies, 18 million pounds;

special-purpose syndicates, 41 million pounds; issue house with stock exchange connections, 22 millionpounds; companies as their own issuers, 13 million pounds; and miscellaneous issuers, 131 millionpounds.

Source: Based on a table prepared by W. A. Brown published in The Economist (November 20, 1937)and reprinted in Balough (1947, p. 233).


roads of the United States. From the 1860s British investors organized them-selves as the Council of Foreign Bondholders and the English Association ofAmerican Bond and Shareholders to collect information on arrears and negoti-ate with debtors (Wilkins 1989 and Eichengreen and Portes 1989). Protectiveassociations were established in France, Germany, and Holland as well. Someforeign investors traded in the securities of small or obscure railways, which theyobtained from jobbers and dealers who purchased blocks of stock in the UnitedStates for sale in Europe. But the vast majority of investors concentrated on first-class securities issued by the London offices of prominent American railwaysand endorsed by British issue houses or banks.

The preferred financial instruments varied with economic and geographicdistance. Nearby lenders, such as New Englanders lending to the Midwest,purchased common stock, because personal and business contacts provided areliable flow of information. The short, inexpensive lines of central New Yorkwere able to supplement local subscriptions with sales of equity in New YorkCity. (According to Chandler 1954, trading in these shares played a central rolein the early development of the New York Stock Exchange.) A group of investorsin New York State purchased a substantial block of shares in Canada's WellandCanal, built in the late 1820s and 1830s to circumvent Niagara Falls and openMontreal to the western trade. British investors sometimes purchased commonstock as well, although the majority of their-as well as other inves-tors'-holdings were in bonds which, as primary claims, were perceived as lessrisky (Lewis 1938, Wilkins 1989). Such bonds were secured by mortgages on therailroad's property or were guaranteed by the government. Many lines issuedbonds that were convertible into stock at the holder's option, and such bondseventually became the standard instruments for financing railways and otherinfrastructure projects (Chandler 1954). A very few railways, such as certainearly southern lines, were able to issue stock, but the returns were guaranteed bycities such as Charleston and Savannah.

Normally, the regulations and surveillance of an organized stock market helpto attenuate the moral hazard and adverse selection problems caused by poorinformation. But the institutions of the London market carried out these func-tions only to a limited extent. The ability of the London Stock Exchange torestrict trading in particular securities was constrained by competition fromEuropean and provincial exchanges as well as from outside brokers and so-called bucket shops (the nineteenth-century equivalent of modern discount bro-kers) (see Adler 1970, Wilkins 1989). Paish (1951, p. 4) notes that "before 1914the [London] Stock Exchange made no attempt to restrict or control in any waythe right to deal in any security, whether British or foreign.... It was in generalmore concerned with arrangements to ensure a reasonably free market in thesecurities than with the intrinsic merits of the company or with the adequacy oraccuracy of the information provided."

This meant that the portfolio of projects to be financed grew riskier as interestrates rose. Promoters had an incentive to take on excessive debt because they


stood to make huge profits if the venture succeeded but could lose no more thantheir equity stake if it failed. Contemporaries consequently complained thatmany worthwhile investment projects were unable to raise external funds, andthis inability to obtain credit created an obvious argument for governmentintervention.

Government Subsidies and Guarantees"When great schemes of public utility are brought before the country,' the

editors of The Economist wrote in 1858, "it is natural that the Governmentshould extend its aid to such enterprises" (MacPherson 1955, p. 181). In thecase of investments in infrastructure, government aid came in three forms: inter-est guarantees on bonded debt, subsidies, and aid-in-kind (often financed by abond issue designated for the purpose or by earmarked revenues). Canadiangovernments borrowed $20 million to build canals in the 1840s (Jenks 1938). Inthe United States, state governments spent $121 million of the $195 millionallocated to canal construction between 1815 and 1860; private companiesspent only $74 million (Cranmer 1960). State and local governments were alsokey subscribers to the securities of early American railroads. Before 1840 nearlyall East-West projects-both railways and canals-were financed by publicbonds. Thus, in the 1830s the Commonwealth of Massachusetts took a one-thircl partnership in the Western Railroad Corporation of Massachusetts, whichit financed by floating state paper in London (Platt 1984, p. 156), and the stateof Ohio subscribed one share in its state's railroads for every two shares pur-chased by private investors (Chandler 1954). Only the early North-South rail-ways and the Pennsylvania coal roads were paid for largely by bonds of privatecorporations; these lines were shorter and cheaper to build and more certain ofregular traffic.

The classic efficiency argument for subsidization rests on externalities: that aproject's social returns exceed its private returns. The historical literature sup-ports the proposition that the railways were a source of positive externalities.Fogel's (1960) study of the Union Pacific Railroad in the United States, forexample, estimated that the social return averaged 30 percent a year, two-and-a-half times the private return. Yet it is not clear that private entrepreneurs werealways unable to capture these returns. In many cases the promoter of an infra-structure project purchased adjoining lands whose productivity and value wereenhanced by construction of the turnpike, canal, or railway. In others, textilemills and mercantile enterprises whose profits were boosted by infrastructureinvestments that moved a steady supply of raw materials to the factory andfinished products to the market could be-and often were-owned by those whoorganized the infrastructure projects. This ability to capitalize on their invest-ment clearly weakened the case for subsidization.6

A further justification for government intervention was the need to offset theimperfections in capital markets that resulted from asymmetric information.


Even if investors could otherwise capture the social returns, incomplete informa-tion that led to credit rationing sometimes prevented them from doing so.

A very different explanation for government intervention is rent-seeking bythose who stood to benefit from government subsidies and guarantees. Lewisand McKinnon (1987) argue that the Canadian Northern Railway may havebeen socially as well as privately unprofitable but those who stood to benefitfrom its construction succeeded in enlisting government subsidies for the proj-ect. Rent-seeking was certainly prevalent, but one need not dismiss it as unim-portant in order to acknowledge that at least some government intervention wasjustified on other grounds.

Government Guarantees

Capital market constraints explain one form of subsidy commonly used forcanal and railway construction in the nineteenth century-interest guarantees ongovernment bonds. In India, for example, if a railway company did not attain aminimum rate of return of, say, 5 percent, the government made up the differ-ence. The interest clause in the bond covenant was backed by the government'sfull powers of taxation. All of India's early railways were built under the termsof the guarantee.

Government guarantees were particularly important in attracting foreigninvestors, who found it difficult to obtain accurate information on railwayprojects in India. Without guarantees, infrastructure projects were consideredimpossible to finance. MacPherson (1955, p. 180) reports that because theNorth Bengal Company was refused a guarantee, it was unable to begin con-struction and was forced to return all deposits to shareholders. Once the guaran-tee was provided, however, India's railways had no difficulty raising fundsabroad. "The motives of the British investors can be explained almost entirely interms of the 5 percent guarantee of interest offered by the Indian Government.Indian bonds were regarded as perfectly safe; investors included widows, barris-ters, clergymen, bankers, and retired army officers."

Canal projects in Canada in the first half of the nineteenth century receivedgovernment guarantees under the aegis of the British Colonial Office. Before1849 attempts to build railways in Canada had foundered on the difficulty ofraising capital. That year legislation was passed guaranteeing interest at no morethan 6 percent on half of the bonds of any railway more than seventy-five mileslong, provided that half of the line was already built (Easterbrook and Aitken1956).7 The guarantee, which covered the principal as well as the interest,enabled Canadian railways to attract significant amounts of foreign finance (seetables 2 and 3).8 Funds for the Grand Trunk line were raised from individuals,municipalities, and contractors, but roughly half of its bonds were guaranteedand were heavily subscribed by British investors. The Canadian Pacific and theGrand Trunk Pacific also enjoyed public support. Glazebrook (1938) concludedthat not one of these lines could have been built without government guarantees.

84 The World Bank Research Observeer, vol. 10, no. 1 (February, 1995)

Table 2. Distribution of Total Flow of Capital to Canada, 1900-14(millions of dollars)

All Great United OtherRecipient countries Britain States countries

Dominion and provincial governments 179 175 4Municipal governments 260 200 60Railroads 767 670 50 47Industrial 630 420 180 30Land and timber 305 80 145 80Mining 125 65 60Insurance 82 32 50Other 198 111 81 6

Total 2,546 1,753 630 163

. Negligible.Source: Buckley (1955, p. 66).

Although the guarantees helped railway promoters to surmount credit ration-ing, they also weakened the incentive for investors to hold management account-able. Investors no longer stood to lose-or to lose as much-if promoters andtheir confederates diverted resources from productive uses, because the govern-ment promised to bail them out. In the extreme, this might encourage theconstruction of railway lines where there was no hope of generating sufficienttraffic to service the debt that was incurred. More generally, it gave promotersan incentive to negotiate sweetheart deals with contractors that made it possibleto channel cash into their own accounts. Such practices were difficult to detectbecause reasonable costs for idiosyncratic projects such as railways and canalsare intrinsically difficult to ascertain. And because construction generated an

Table 3. Gross Construction Outlays in Major Transport Fields, Canada, 1901-30

Canals andYears Railways Highways harbors Total

A. Values (millions of dollars)1901-05 124.3 3.3 32.1 159.71906-10 380.7 11.7 48.0 440.41911-15 537.4 38.5 93.7 669.61916-20 252.5 39.4 59.7 351.61921-25 253.2 100.4 109.8 463.41926-30 389.4 172.4 138.2 700.0B. Percentage distribution1901-05 77.8 2.1 20.1 100.01906-10 86.4 2.7 10.1 100.01911-15 80.3 5.7 14.0 100.01916-20 71.8 11.2 16.9 100.01921-25 54.6 21.7 23.7 100.01926-30 55.5 24.6 19.7 100.0

Source: Buckley (1955, p. 32).


abundant cash flow, the diversion of resources into the pockets of the promoterswas relatively easy to arrange. Moreover, many partnerships were temporary, sopromoters had little reason to be deterred by considerations of reputation. Onlythose ultimately responsible for the financial liability-or more precisely, theirelected representatives-had an incentive to monitor the accounts.

Thus the potential for looting was created. Bondholders, whose rate of returnwas guaranteed by the government, had little incentive to expend resources todetermine whether promoters had identified a project capable of generating anadequate net revenue stream or whether contractors were siphoning off theproject's resources. Only if government authorities monitored the actions ofpromoters and contractors and threatened them with legal sanctions did thelatter have reason to be deterred. Although the problem of quantifying mini-mum construction costs makes the prevalence of looting difficult to establish,the qualitative evidence is suggestive. Of Canada's Great Western Railway, Jenks(1938) wrote that its directors, who enjoyed a government guarantee, soughtnot to minimize construction costs, but to finance the contractor and share in hisprofits. The history of Canada's Grand Trunk Railway provides additional gorydetails. Almost immediately upon floating government-guaranteed bonds, asEasterbrook and Aitken (1956, p. 309) noted, the company found itself unableto pay interest. To a large degree, its problems reflected "unanticipated costs ofconstruction,' as contractors pressed for new links to the railways of New Yorkand Michigan rather than using existing lines. In 1851 Gzowski and Company, acontracting firm run by former directors of railways with connections to theGrand Trunk, was awarded the contract for the construction of these lines. Thecontractors were paid in cash, "and the individual members of the firm realizedsizable fortunes" (p. 310). The British group of Peto, Brassey, Betts, and Jack-son, itself deeply entangled in Canadian politics, was "helped . . . over everydifficulty" by "a complaisant legislature and a winning governor-general" (Jenks1938, p. 204). Existing lines were added to the network for "inflated" purchaseprices. Operating expenses in the first ten years ran between 58 and 85 percentof gross receipts-far above the 40 percent that had been forecast and wastypical of other railways.

This type of fraud is consistent with the predictions of the Akerlof-Romermodel-that government guarantees extended to relax credit-rationing con-straints weaken the effectiveness of corporate control if they are not accom-panied by effective public sector oversight and regulation. Currie (1957, p. 9)summarized the situation: "As the Government would guarantee bonds up toone-half the cost of the road, hard-pressed promoters were tempted to inflatetheir costs, to effectively force the Government to assume responsibility for morethan its proper share of the actual expenditure, and to reduce the real value ofthe assets against which, under the Guarantee Act, the Government held a firstmortgage."

In India there were layers of principal-agent problems: the taxpayers under-wrote the guarantee, but the government that extended it was responsible not to


them, but to the British Crown; Jenks (1938) attributes failed projects such asthe Madras Canal, which was completed but could not be filled with water,to the lack of monitoring encouraged by this weak structure. Of the railways,he reports that an Indian finance minister testified, "All the money came fromthe English capitalist, and so long as he was guaranteed 5 percent on therevenues of India, it was immaterial to him whether the funds that he lent werethrown into the Hooghly [River] or converted into brick and mortar"(pp. 221-22).

Land Grants

Land grants, which served to correct capital market imperfections by provid-ing collateral, were another prevalent form of government subsidy. Approxi-mately 150 million acres of land were granted to western U.S. railways between1850 and 1870. Legislation authorizing the use of land grants was adopted inCanada in 1852, although it was not used until the 1870s. Land grants wereattractive on several counts. First, because the prairies in both countries wereunsettled and the land remained in government hands, such grants obviatedproblems of assembling parcels for right-of-ways. Second, a land grant tended toconfront less political resistance than did government financial subsidies andinterest guarantees, which implied the imposition of distortionary taxes. Argua-bly, ceding land adjoining railways and canals to the promoters of those projectsalso allowed promoters to recover at least some of the externalities that werethrown off by their investments. Third, compared with other bonds, thosebacked by mortgages on land had minimal bankruptcy costs; the interest andprincipal due to the primary creditors could be paid off, at least in part, throughthe sale of the land if the project failed. The loan was collateralized, thusmitigating the moral hazard and adverse selection problems that otherwise con-strain credit.9 Adler (1970) notes that the importance of this support is reflectedin the fact that only American railroads receiving land grants were able to issueregular bonds (as opposed to convertible issues). Alternatively, receipts from thesales of land, known as "land income" bonds, could be mortgaged. In the case ofthe Atchison, Topeka and Santa Fe Railroad, this backing was attractive toforeign investors, who had limited opportunities for monitoring the project.

In principle, these land grants collateralized only a portion of an enterprise'sbonded debt. Compared with an unlimited guarantee, this should have encour-aged closer monitoring by outside investors. But in fact, railways that ran intodifficulties were frequently offered additional guarantees and subsidies. In prac-tice, it is questionable that the negative side effects of land grants were lesspronounced than those of bond guarantees.

Fishlow (1965) estimates that the land subsidy amounted to roughly 5 percentof total railroad investment between 1850 and 1880; Mercer (1969, 1972)arrives at smaller numbers. The small size of these estimates suggests that landgrants only partially collateralized the railways' liabilities. Land grants were not


uniformly distributed, however; they were concentrated in the period 1865 to1870 and were awarded disproportionately to certain risky investments, such asthe first transcontinental lines. Mercer concludes that many land grants werewasted on railroads that would have been built in any case. He bases his view onthe finding that the private rate of return exceeded the return on alternative usesof funds. In other words, the railways still would have wished to borrow at theprevailing rate. An asymmetric information perspective casts doubt on this con-clusion, however, because it implies that, without land grants, some railwaysmight not have been able to obtain external finance at any price.

Policy ImplicationsRecent suggestions for reforming the ways infrastructure in developing coun-

tries is provided and financed include encouraging private provision as a way toavoid the inefficiencies of public administration and tapping local savings as away to avoid excessive reliance on external borrowing. These suggestions have aback-to-the-future quality: private provision and local finance were characteris-tic of infrastructure investments in many countries-notably the North Ameri-can case considered here-for much of the nineteenth century. Consequently, thehistorical record is a potentially rich source of information on the circumstancesunder which these approaches are workable and on their limitations.

What the record reveals is that government intervention continued to beimportant. The ability of domestic financial markets to underwrite the construc-tion of ports, canals, and railways was constrained, in part because of informa-tional asymmetries characteristic of markets in the early stages of development.To help with these problems and to attract private investment, lenders turned tofinancial institutions that specialized in assessing projects and monitoring man-agement. These were typically foreign institutions with foreign clienteles whoseexperience with privately financed projects had given them a head start in raisingcapital and judging risk. This approach relieved-but did not eliminate-con-cerns about inadequate information. Nor did private investment and local capi-tal reduce the government's involvement or the need for foreign borrowing.

All too often, however, government intervention simply replaced one set ofproblems with another. Investors, assured of a guaranteed return, had less incen-tive to hold management accountable. Management, freed of investor scrutinyand provided with access to capital markets, courtesy of the government, ar-ranged deals with construction companies that left taxpayers holding the bag.Guaranteed loans encouraged investors to finance infrastructure projects, butwithout built-in mechanisms to monitor spending and protect the public inter-est, it was impossible to ensure that resources were allocated efficiently.

These failings imply that exploiting nontraditional approaches to financinginfrastructure requires two further policy initiatives. First, efforts should bemade to enhance the effectiveness of public administration. Government agen-cies or departments should be responsible for monitoring the efficiency and


performance of the enterprise and should be backed by a credible threat ofsanctions against managers who are tempted to enrich themselves. Second, poli-cymakers need to encourage the development of financial institutions and instru-ments such as banks, mutual funds, and bond-rating agencies that can surmountinformation problems and relieve the government of the need to provide sub-sidies and interest guarantees.10

These are essential tasks for governments in any setting. Idealistic strategiesfor privatizing the provision of infrastructure notwithstanding, it seems likelythat government's traditional role-and the traditional problems associated withgovernment intervention-will necessarily remain.

Notes

Barry Eichengreen is John L. Simpson professor of economics and professor of politicalscience at the University of California, Berkeley. He would like to thank Ashoka Mody forguidance, Ann Carlos, Michael Edelstein, Frank Lewis, Richard Sylla, and Mira Wilkins forcomments, and Lisa Ortiz and Andrea Cu for research assistance.

1. Fogel (1964), however, concludes that the social savings attributable to railway construc-tion in the United States were small. Given geographical and topographical differences,subsequent studies have yielded larger estimates for other parts of the world. Nor do calcula-tions of social savings attempt to quantify the dynamic effects emphasized by authors asdiverse as Chandler (1990), Jenks (1944), and Williamson (1974).

2. Foreign investors were put off by political uncertainty as well. The disruption caused bythe Civil War, for example, lingered into the 1870s.

3. The same was true elsewhere, of course. Thus, in 1852 the chairman of the MadrasCompany cited "want of local knowledge" as an obstacle to attracting external finance forrailroads in India (MacPherson 1955, p. 180).

4. Although the Western Rail-Road had more than 2,000 shareholders, most of these werelocated in Boston. Only 17 percent had one hundred or more shares as of 1841. To continuethe road from the state line to Albany, the city of Albany subscribed the entire capital, payingfor it with city bonds (Johnson and Supple 1967).

5. 1 refer to "external" rather than "foreign" finance for two reasons. First, entities such asCanada and India, which were not fully independent, relied on external finance from theimperial center, Great Britain. Second, regions that were late in developing, such as thewestern United States, relied on external finance from regions that had developed earlier, suchas New England.

6. Because many different landowners and merchants typically benefited from the construc-tion of a single railway line, the problems of organizing collective actions by a large number ofvaguely interested parties might impede efforts to internalize the externalities, leaving arationale for subsidization.

7. In 1851 the guarantee was restricted to railroads that formed part of a main, or trunk,line. This legislation was passed partly in response to pressure from the Canadian govern-ment's British bankers, Baring Brothers and Glyn, Mills and Company, who worried that anunlimited guarantee would encourage excessive building and result in an unsupportable debt.Another act, in 1852, liberalized this condition somewhat, allowing individual municipalitiesto borrow from a provincial fund to help establish branch and feeder lines (Currie 1957, p. 9).

8. An exception was the Great Western Railway, which obtained initial capital from mer-chants in Detroit and from New York investors who had financed the New York Central.Completion, however, required floating a bond in London in 1852. Other major railways


initiated after the passage of the Guarantee Act relied almost entirely on British finance(Carlos and Lewis, forthcoming).

9. This was true at least of early mortgage bonds. Subsequently, some promoters issued"collateral trust mortgage bonds" that were secured not by real property, but by the stocksand bonds of other companies. See Bryant (1971) for details.

10. The case for subsidization may only be reduced, rather than eliminated, insofar asinfrastructure investments continue to throw off positive externalities that private agentscannot internalize fully. This distinguishes the problem of financing infrastructure invest-ments from general problems of financing enterprises in developing countries.

ReferencesAdler, Dorothy. 1970. British Investment in American Railways. 1834-1898. Charlottesville,

Va.: University of Virginia Press.Akerlof, George, and Paul Romer. 1993. "Looting: The Economic Underworld of Bank-

ruptcy for Profit." Brookings Papers on Economic Activity 2:1-73.Ankli, Robert E. 1980. "The Growth of the Canadian Economy, 1896-1920." Explorations

in Economic History 17:251-74.Balough, Thomas. 1947. Studies in Financial Organization. Cambridge, U.K.: Cambridge

University Press.Baskin, Jonathan B. 1988. "The Development of Corporate Financial Markets in Britain and

the United States, 1600-1914: Overcoming Asymmetric Information." Business HistoryReview 62:199-237.

Bryant, Keith L. 1971. Arthur E. Stilwell. Nashville, Tenn.: Vanderbilt University Press.Buckley, K. A. H. 1955. Capital Formation in Canada 1896-1930. Toronto: University of

Toronto Press.Carlos, Ann M., and Frank D. Lewis. 1992. "The Profitability of Early Canadian Railroads:

Evidence from the Grand Trunk and Great Western Railway Companies." In ClaudiaGoldin and Hugh Rockoff, eds., Strategic Factors in Nineteenth Century American Eco-nomic History. Chicago, Ill.: University of Chicago Press.

. Forthcoming. "Foreign Financing of Canadian Railroads: The Role of Information."In Michael D. Bordo and Richard Sylla, eds., Anglo-American Financial Systems: Institu-tions and Markets in the Twentieth-Century. Burr Ridge, Ill.: Irwin.

Chandler, Alfred. 1954. "Patterns of American Railroad Finance." Business History Review28:248-63.

.1990. Scale and Scope. Cambridge, Mass.: Harvard University Press.Cleveland, Frederick A., and Fred Wilbur Powell. 1912. Railroad Finance. New York: D.

Appleton.Cranmer, H. Jerome. 1960. "Canal Investment, 1815-1860." In National Bureau of Eco-

nomic Research, Trends in the American Economy in the 19th Century. Princeton, N.J.:Princeton University Press for the National Bureau of Economic Research.

Currie, A. W. 1957. The Grand Trunk Railway of Canada. Toronto: University of TorontoPress.

De Meza, David, and David C. Webb. 1987. "Too Much Investment: A Problem of Asymmet-ric Information." Quarterly Journal of Economics 102:281-91.

Easterbrook, W. T., and G. J. Aitken. 1956. Canadian Economic History. Toronto: Univer-sity of Toronto Press.

Eichengreen, Barry, and Richard Portes. 1989. "After the Deluge: Default, Negotiation, andReadjustment on Defaulted Foreign Bonds During the Interwar Years." In BarryEichengreen and Peter Lindert, eds., The International Debt Crisis in Historical Perspec-tive. Cambridge, Mass.: MIT Press.


Fishlow, Albert. 1965. American Railroads and the Transformation of the Antebellum Econ-omy. Cambridge, Mass.: Harvard University Press.

.1986. "Lessons from the Past: Capital Markets and Foreign Lending During and 19thCentury and the Interwar Period:' In Miles Kahler, ed., The Politics of International Debt.Ithaca, N.Y.: Cornell University Press.

Fogel, Robert. 1960. The Union Pacific Railroad: A Case of Premature Enterprise. Balti-more, Md: Johns Hopkins University Press.

. 1964. Railroads and American Economic Growth. Baltimore, Md.: Johns HopkinsUniversity Press.

Glazebrook, G. P. 1938. A History of Transportation in Canada. Toronto: McClelland andStewart.

Jenks, Leland J. 1938. The Migration of British Capital to 1875. New York: Knopf.1944. "Railroads as an Economic Force in American Development." Journal of

Economic History 4:1-20.

Jensen, Michael C., and William H. Meckling. 1976. "Theory of the Firm: ManagerialBehavior, Agency Costs and Ownership Structure." Journal of Financial Economics3:305-60.

Johnson, Arthur H., and Barry E. Supple. 1967. Boston Capitalists and Western Railways.Cambridge, Mass.: Harvard University Press.

Keeton, William. 1979. Equilibrium Credit Rationing. New York: Garland Press.Lamoreaux, Naomi. 1986. "Banks, Kinship, and Economic Development: The New England

Case " Journal of Economic History 46:647-68.

Lewis, Cleona. 1938. America's Stake in International Investments. Washington, D.C.: TheBrookings Institution.

Lewis, Frank, and Mary McKinnon. 1987. "Government Loan Guarantees and the Failure ofthe Canadian Northern Railway." Journal of Economic History 47:175-96.

MacPherson, W. J. 1955. "Investment In Indian Railways, 1845-1875.' Economic HistoryReview 8:177-86.

Mercer, Lloyd. 1969. "Land Grants to American Railroads: Social Cost or Social Benefit?"Business History Review 43:134-5 1.

. 1972. "Taxpayers or Investors: Who Paid for the Land Grant Railroads?" BusinessHistory Review 46:279-94.

Paish, F. W. 1951. "The London New Issue Market.' Economica 18:117.Platt, D. C. M. 1984. Foreign Finance in Continental Europe and the United States

1815-1870. London: Allen & Unwin.

Stiglitz, Joseph, and Andrew Weiss. 1981. "Credit Rationing in Markets with ImperfectInformation?' American Economic Review 71:393-410.

Wilkins, Mira. 1989. The History of Foreign Investment in the United States to 1914.Cambridge, Mass.: Harvard University Press.

Williamson, Jeffrey G. 1974. Late Nineteenth-Century Economic Development: A GeneralEquilibrium Approach. New York: Cambridge University Press.


THE ECONOMICS OF NATURALRESOURCE EXTRACTION: A PRIMERFOR DEVELOPMENT ECONOMISTS

Stephen W Salant

As developing countries become major consumers of the global supply ofcommercial energy, it is essential to understand the determinants of futureenergy prices. At the same time, many developing countries are relying onexports of their own natural resources-tropical hardwoods, oil, tin, gold,and other minerals-to generate badly needed foreign exchange. Govern-ment policies influence how much of a resource is extracted today and howmuch is saved for the future. Flawed policies needlessly waste precious na-tional wealth.

etween 1970 and 1990 consumption of commercial energy in developingcountries tripled, reaching 27 percent of the world total. Accordingto projections (World Bank 1992), energy use in developing countries

will dominate energy markets worldwide in a matter of decades. At thesame time many developing countries rely on exports of their own naturalresources-tropical hardwoods, oil, tin, gold, and other minerals-to generateforeign exchange. Thus, greater domestic demand for energy is likely to be offsetby increasing the pace of resource extraction rather than by shipping fewerresources abroad.

In some countries, state-run enterprises extract and sell natural resources. Inothers, resources are privately owned but subject to government regulation. Ineither case development economists asked to advise on resource policy are dis-covering that the theory of the firm requires substantial revision when the firmextracts a natural resource. The following questions are among many that arise:

Suppose a state-owned enterprise is charged with selling a natural resourceon the world market so that discounted profits are maximized (discountedprofits refer to the equivalent value today of all expected profits in thefuture).' For a small country-that is, one that sells its natural resources onworld markets but that does not have a large enough market share to be able

The World Bank Research Observer, vol. 10, no. 1 (February 1995), pp. 93-111.©) 1995 The International Bank for Reconstruction and Development/THE WORLD BANK 93

to influence world prices (a so-called price-taker)-can it be a mistake toextract to the point where the cost of extracting an additional unit (themarginal cost) equals the market price? Is it ever sensible to refrain alto-gether from extracting a resource, even though the current price exceeds themarginal cost of extracting the first unit-that is, even though it is profitableto begin extracting?

* How does the fact that credit may be more tightly rationed (and interestrates higher) in one country than another affect the rationed country's rela-tive rate of natural resource extraction? Assume that the goal of each coun-try is to maximize national wealth and that each has approximately thesame extraction costs, reserves, and expectations about future prices.

* Suppose a government decides to restrict the extraction of a resource that isprivately owned and extracted because the resource is deemed exceptionallyvaluable (such as tropical hardwoods in Cameroon) or because the extrac-tion process generates pollution (such as gold mining in Brazil, which usesmercury, and in Ghana, which uses arsenic). Why is it a mistake in suchcases to give extractors a grace period before the restrictions go into effect tomitigate the dislocations that such policies will cause?

The theory of natural resource extraction can illuminate these and otherquestions. What follows is a self-contained introduction with suggestions forfurther readings. (Space limitations prevent consideration of misallocations thatresult when a natural resource is common property and how such problems canbe attenuated. A valuable discussion of this topic is contained in Dasgupta1982.)

The Basic Model

Natural resource economics traces its origins to classic articles by L. C. Gray(1914) and Harold Hotelling (1931) (for a comparison of Gray's and Hotelling'scontributions, see Crabbe 1983). Gray examined the supply behavior over timeof an individual extractor who anticipates a sequence of real prices and attemptsto maximize discounted profits. This is the relevant case for a "small" country.Hotelling extended Gray's theory by predicting the sequence of market pricesthat Gray took as given.

Supply Behavior of a Small Extractor Selling on World Markets

The supply behavior of the owner of a depletable resource such as oil differsfrom that of the firm portrayed in intermediate microeconomics textbooks.According to these texts, a price-taking firm with no fixed costs and increasingmarginal costs will adjust production in all periods so that its marginal cost ofproduction in each period either equals the market price in that period or, ifproduction is zero and can be contracted no further, exceeds it. This result can


be unclerstood intuitively. If the cost of producing an additional unit is greaterthan the price that will be paid for that unit, then the supplier will lose byincreasing production; indeed, he will benefit from contracting output becausehe will save in costs more than he will lose in revenues. If the marginal cost is lessthan the price, the supplier can raise current profits by increasing production.When the firm is the extractor of a nonrenewable resource, however, suchbehavior might require the extraction of more reserves than the firm hasavailable.

In such cases, some modification of the standard theory is clearly required.But what modification is appropriate? It might at first seem that an extractorshould equate the marginal cost of production to the market price in everyperiod before reserves run out, after which there are no choices left. But thatstrategy would not maximize discounted profits. A barrel of oil saved until afterthe proposed date of exhaustion could in some cases be sold later at a largerdiscounted profit.

The important element missing in the above discussion is the principle ofopportunity cost. A barrel of oil extracted and sold today is a barrel unavailablein the future.2 In deciding whether to extract and sell an additional barrel today,the extractor must consider not only the cost of pumping the barrel, but also thecost of forgoing the highest profit that could have been earned if the oil hadinstead been pumped and sold in the future. It is optimal to adjust extraction inany given period until this more inclusive definition of marginal cost, or "aug-mented marginal cost' equals the market price or exceeds it when extraction iscut to zero.

When marginal cost is redefined in this way, the old maxim once again isvalid: a firm produces in each period to the point where its (augmented) margi-nal cost equals the market price. But beneath the apparent simplicity of thismaxim lies a wealth of subtlety. The maxim implies, for example, that thecurrent extraction rate of a private owner of a natural resource depends not onlyon the current price, as in the standard theory, but on expectations about futureprices. These price expectations determine the opportunity cost of additionalcurrent extraction. A competitive supplier that expects future prices to be suffi-ciently low compared with the current price may extract and sell intensively inthe current period, judging the opportunity cost of additional extraction to besmall. But if future prices are expected to be sufficiently high, the same currentprice may elicit no extraction whatsoever today.

What sales path will maximize the discounted profit of a small (price-taking)oil extractor or state-owned oil company selling on world markets? Suppose thatafter determining a sequence of extraction and sales that uses all the under-ground reserves in the most profitable way, the company finds an additionalbarre]L of oil. Assume that extracting and selling this barrel at the most pro-pitious time would increase net discounted profits by a certain amount, to bedenoted by x. Because x would also be the loss in net discounted profits if theextractor had one less barrel of initial reserves, it represents what the extractor

Stephen W Salant 95

would lose if he was forced to cut back extraction by a barrel in one period tooffset a barrel of additional extraction in another period. That is, x can beinterpreted as the opportunity cost of additional extraction expressed in initialperiod dollars.

If the value of x were known, the optimal extraction path could be determinedby computing the firm's augmented marginal cost in each period and then ad-justing extraction so that the augmented marginal cost equaled the price in thatperiod (or exceeded it when extraction was cut to zero). But how is x deter-mined? Clearly, this opportunity cost cannot be so small that it encouragesaggregate extraction that exceeds underground reserves. Nor can inducedextraction be so large that reserves are left in the ground indefinitely, since theopportunity cost of additional pumping would be zero rather than x. The op-portunity cost must be such that planned cumulative extraction exactly matchesunderground reserves: this condition uniquely determines x and, consequently,the optimal path of extraction.

Even this elementary analysis illuminates questions of practical policy thatmight otherwise appear baffling. Each of the problems mentioned in the intro-duction ceases to be puzzling once attention is properly refocused on the oppor-tunity cost of current extraction and on how that cost changes when a newpolicy is anticipated.

Expanding state-run oil production so that the marginal cost of pumpingequals the market price sounds sensible but is not. Such pumping is excessivebecause it fails to account for the profit forgone when an additional barrel ispumped. Indeed, if this opportunity cost is sufficiently high, no amount of theresource should be extracted today even though the current price exceeds thecurrent marginal cost of extracting the first unit.

It is often thought that a rise in the real rate of interest will not affect naturalresource extraction if it has no effect on market prices. An increase in the realrate of interest will, however, cause the extractor to pump more in early periodsand less in later ones. When the interest rate rises, the opportunity cost ofextracting another barrel declines because the future profit (from the barrel thatmust be forgone) is worth less today. Hence, even if the marginal cost of pump-ing does not shift, the augmented marginal cost in early periods will fall. As aresult, the same sequence of prices would trigger an initial surge in extraction onproperties that are privately owned. However, the opportunity cost componentwould increase at a faster rate than it would have if the real interest rate hadremained unchanged. The augmented marginal cost beyond some date would behigher, which would eventually reduce extraction below what it otherwisewould have been.

This example can be reinterpreted to explain differences in the behaviors of twocountries selling the same natural resource on the world market. Assume thatboth countries have approximately the same underground reserves and costs ofextraction. If credit is rationed more tightly in one country, that country shouldextract more rapidly in the near term in order to maximize national wealth.


Officials who propose shutting down a polluting gold mine or a leakinglandfill but who permit a grace period to mitigate the dislocations caused by theclosure may aggravate the very problems they seek to solve. Premature closinglowers the opportunity cost of extracting an additional unit before the shut-down. Consequently, the augmented marginal cost in each period before thedate of closing is lower than it was before the policy was announced. If thesequence of world prices is unchanged, the polluter will find it profitable tointensify mining throughout the grace period. A similar policy response wouldoccur if the extracted resource were the remaining space in a polluting landfill.Indeed, in the U.S. solid waste industry, the frenzied dumping in landfills duringthe grace periods preceding mandated closings is quaintly termed "stuffing"(Ley, Macauley, and Salant 1994).

Imposing an export tax on tropical hardwoods may have a similar effect ifpolicymakers give loggers time to react. Because loggers anticipate that they willreceive less net profit from exports after the tax is imposed, cutting a tree has alower opportunity cost than it did before they learned of the future export tax. Ifthe price sequence does not change, more logging will occur before the tax isimposed and less afterwards. The anticipatory logging may be massive if it doesnot unduly increase marginal costs. As a result, the very trees the policy isintended to protect may be harvested before the tax is imposed.

A similar perverse response is possible at the global level. To combat globalwarming caused by carbon dioxide emissions from fossil fuels, a tax on thecarbon content of these fuels has been proposed. If the tax is phased in gradu-ally, carbon dioxide emissions may increase in the short run.

Efforts to mitigate painful economic dislocations resulting from policy inter-ventions need not backfire. Financial compensation, for example, would havenone of the allocative effects discussed above. What the preceding examplessuggest, however, is that more than heartfelt generosity is required when decid-ing how best to mitigate such dislocations.

Determining Prices for Resources

Hotelling (1931) extended Gray's analysis of the extraction of price-takingfirms by determining the sequence of prices that balances supply and demand inevery year. Hotelling made the useful simplification that all firms have the sameexpectations about future prices and that these expectations are subsequentlyborne out. It is simplest at the outset to assume that each unit can be extracted atconstant marginal cost.

In this case, a firm can determine the present value of the revenue it wouldexpect to receive, net of extraction costs, from producing and selling a unit ofthe resource in any year, given its expectations about future prices.3 Leaving anyreserves unexploited forever is foolish as long as the net price (the excess of themarket price over the constant per-unit cost of extraction) is strictly positive insome year. Yet selling any of the resource when the net price in a later year is

Stephen W. Salant 97

expected to have a larger present value is also foolish. Hence, the supply behav-ior of each firm is to sell its entire reserve during years when the present value ofthe net price is highest. (Under the previous analyses, the present value of thisnet price is the value x.)

Given this supply behavior, there turns out to be only one series of prices at orbelow the choke price (the lowest price that generates zero demand) that balancessupply and demand in every year. Because prices below the choke price inducepositive demand, market equilibrium requires that demand be matched by anequal supply. One condition must be satisfied for producers to supply the resourcein every period: net prices must grow at the rate of interest. A numerical examplemay help to illustrate this fundamental condition. Suppose the interest rate is 10percent and the net price per barrel of oil is $20. If the net price is not expected togrow to $22 next year, it pays to extract more oil in the current year, because theincome will earn 10 percent interest. If the net price is expected to go above$22-that is, to grow faster than the rate of interest-the producer will have noincentive at all to pump oil in the current period, because any barrel pumped to-day will (even after interest receipts) be worth less in a year than a barrel pumpedand sold a year from now.

Many price paths involve the net price rising at the rate of interest, but onlyone path generates cumulative demand equal to aggregate reserves. Any lowerprice path would induce cumulative demand that exceeded industry reserves;consequently, the industry would be unable to satisfy demand in some year. Anyhigher price path would induce cumulative demand that fell short of industryreserves; consequently, consumers would be unwilling to purchase all the oiloffered for sale in some year. For supply to match demand in every year, cumula-tive demand must match industry reserves.

If extraction costs are zero, the dictum that the net price must rise in succes-sive years by the interest rate implies that the price itself rises by that factor. Inreality, of course, extraction costs are never zero. Whenever they are positive, aprice increase equal to the interest rate would cause the net price to rise by morethan the interest rate. For example, suppose the price per barrel of oil is $30 andthe net price is $25. A 10 percent price rise (the rate of interest) would boost thenet price from $25 to $28, or by more than 10 percent. That difference wouldgive every extractor an incentive to postpone extraction rather than satisfycurrent demand. If such imbalances are to be avoided, the price in successiveyears must rise by less than the rate of interest.

The price path is determined by the size of aggregate reserves.4 The largerthese underground stocks, the smaller the difference between the initial priceand the marginal extraction cost, the smaller the percentage increase in the priceinitially, and the longer the time required for the reserves to be depleted.

Market Power in the Extraction IndustryBecause natural resources often tend to be geographically concentrated and

because firms without access to a particular resource cannot enter an extraction


industry, such industries tend to have noncompetitive elements (see Hotelling1931). If a single extractor owns all underground reserves and sells the resourceat the same price per unit, which may vary over time, the monopolist will planextraction so that the marginal profit (the excess of marginal revenue overmarginal cost) received from selling another unit of the resource in another yearis equal, in discounted terms, to the marginal profit received from selling it thisyear. In other words, the monopolist will plan extraction so that there is no gainfrom extracting more in one period and less in another.

How does such behavior affect market prices over time? Even the simplestcases are ambiguous. Assume a fixed real rate of interest and costless extraction.Undeir competition, the price is predicted to rise each year by the interest factorso long as all shifts in demand are foreseen. Under the same circumstances, amonopolist would extract enough so that marginal revenues would rise eachyear by the interest factor. Unfortunately, when shifts in demand are fully antici-pated, the implied path of extraction might yield virtually any path of prices.

If the demand in future years does not shift, the monopolist always finds itoptimal to raise prices over time-by less than the interest factor in all butexceptional cases, where the insensitivity of consumer demand to price mightlead a monopolist to raise the resource price faster (Stiglitz 1976).5

The surge in oil prices in 1973 stimulated research into the strategic interplaybetween resource cartels such as the Organization for Petroleum ExportingCountries (OPEC) and independent extractors. Unfortunately, these models werebuilt before economists gained their current level of understanding of the subtle-ties of multistage strategic games so even the most sophisticated models containconceptual weaknesses. (See Cremer and Salehi-Isfahani 1991 on various at-tempts to model recent developments in the world oil market.) In addition, mostmodels (including Cremer and Weitzman 1976 and Salant 1982) treat OPEC as amonolith that maximizes the joint profits of its members, although, as Bain(1948) pointed out, joint profit maximization is unlikely in the absence of sidepayments.

Technologies for Energy

The analysis has to be modified when it departs from the assumptions under-lying the Hotelling model. In the scenario discussed in this section, the analysisagain revolves around the fundamental principle that for supply to occur invarious time periods, the net price must increase over time by the rate of interest.

Extracting Oil Reserves at Different Marginal Costs

If the products of different firms are indistinguishable in the eyes of con-sumers, they will be purchased at the same time only if they sell at the sameprice. If firms have different but constant marginal costs of extraction, firmswith the lowest marginal cost will operate first while the others wait. Only after


firms with the lowest-cost reserves exhaust their stocks will firms with higherextraction costs operate.

To understand why, suppose there are only two cost levels, high and low.High-cost and low-cost extractors cannot operate simultaneously in successiveyears because any price increase that makes the high-cost extractor indifferentto selling in one year or the next will prompt the low-cost extractor to sell onlyin the earlier year. A high-cost extractor will be indifferent about selling inadjacent years only if the net price is expected to rise by the interest rate, butthat would necessarily cause the net price of the low-cost firm to rise by lessthan the interest rate, making it more profitable for the low-cost firm to sellonly in the earlier year. Nor can the high-cost firm extract for a succession ofyears before the low-cost firm operates. During the entire phase when the netprice of the high-cost firm rises at the rate of interest, the net price of the low-cost firm rises by a smaller factor, giving it an incentive to extract its entirereserve at the outset of the phase. Instead, the low-cost firm extracts during aninitial period and then, when its reserves are depleted, the high-cost firmextracts. During the first phase, the high-cost producer always finds postpon-ing extraction until the second phase to be profitable. At the same time, thelow-cost firm never regrets having extracted all of its reserves during the firstphase.

If, on the contrary, every firm owned some reserves at each cost level, neitherthe competitive price path nor the total supply in any year would change. Eachfirm would now extract in every year, but all low-cost oil would still be extractedbefore any high-cost oil was. The extraction technology in this case is said toexhibit "depletion effects"; each firm's marginal cost of extraction rises as itsremaining reserves decline.

The most efficient way to extract reserves is in the order of their constantmarginal costs.6 As long as the government does not intervene, a competitive ora monopolized industry would extract reserves in this order.

Governments, of course, do intervene in resource markets and may pervertthis sensible pattern of resource exploitation. Gold provides an important exam-ple. Gold costs approximately $300 an ounce to extract from beneath theground but can be removed at virtually no cost from the stockpiles of govern-ments and international financial institutions. Official stockpiles are vast, equal-ing twenty years' worth of world extraction at the 1992 rate of 64 millionounces. Official stockpiles were accumulated during an era when gold played acentral role in international monetary arrangements. But it no longer plays thatrole. Only inertia can account for the retention of such vast stockpiles. As a1993 article in the Economist remarked: "Strip away history and it is ludicrousthat central banks in industrial countries still hold, on average, 40 percent oftheir foreign reserves in this metal.... If central banks were to construct theirportfolio from scratch, they would hold little (if any) gold."7

Yet, society extracts high-cost gold reserves while zero-cost official stocks gounutilized. Henderson, Irons, and Salant (1993) have shown that governments


could earn approximately $75 billion without adversely affecting the privatesector if they corrected this production inefficiency.8

Cost-Increasing Depletion Effects versus Cost-Reducing TechnologicalChange

In principle, extraction costs may rise over time as lower-cost reserves aredepleted and higher-cost reserves remain to be exploited. If technological changecauses the marginal cost of extraction to fall as time passes, the price need nolonger strictly increase over time when demand is unchanged. The price need notchange at all if the rate of decline in the marginal cost increases the net price byexactly the interest factor. If the marginal cost is expected to decline by a largerpercentage, the price may fall.

Norgaard (1975) investigated empirically the extent to which improvementsin the technology of extraction offset depletion effects. He found that, althoughcosts of oil extraction tended to increase as depletion occurred, cost increaseswould have been much larger in the absence of technological change: "Overall,the decline in resource quality out-weighed new technology, and the real cost ofsuccessful wells increased 64 percent. This study shows, however, that the costwould have risen by 233 percent if there had been no improvement in knowledgeand inputs" (p. 266). Norgaard's finding may explain in part the absence of anupward trend that Barnett and Morse (1963) observed in the prices of a widevariety of other natural resources from 1870 to 1957. (For a discussion ofBarnett and Morse, see Fisher 1981, ch. 4.; for an attempt to reconcile thesefindings with exhaustible resource theory, see Dasgupta and Heal 1979, section15.8.)

Diminishing Returns in Resource Extraction

If, as Gray assumed, increases in a firm's rate of extraction cause its marginalcost to increase because of diminishing returns (that is, the faster the rate ofextraction, the more effort is required and the larger the marginal cost),Hotelling's analysis must be modified. A firm choosing an extraction path hasmaximized its discounted profits only if the difference between price and margi-nal cost in the first year is the same as the discounted value of this difference inevery subsequent year. When the demand curve does not shift over time, theprice of the resource will rise in successive years (although more slowly than theinterest factor). If demand for the resource falls overtime, however, and the firmanticipates these shifts, the direction of the change in prices will depend onwhether marginal costs are constant or increasing. To understand this result,recall that the excess of the price over the marginal cost of extraction mustincrease over time by the interest factor. If the marginal cost is constant, theprice will increase. But if marginal cost instead depends on how much is extrac-ted, the excess of price over marginal cost can increase over time by the interestfactor even though the market price is declining. For this to occur, extraction


must decline. This phenomenon might occur in equilibrium if extractors antici-pated that the demand curve for the resource would fall over time. Firms mightthen extract more initially and less in successive periods and yet, because of adecline in demand, the amount people would want to purchase at the lowerprice would match the declining supply in each period.

Producing a Perfect Substitute for Oil: the Deus ex Machina Called theBackstop

Forecasts about the future prices of oil (and other natural resources) areaffected by one's assumptions about new sources of energy that might serve as abackstop. From Nordhaus (1973) it is standard to assume that the next energysource will be competitively held, inexhaustible, and operated at a constantmarginal cost. From this assumption, it is almost invariably argued that themarginal cost of producing the so-called backstop acts as a future ceiling on theprice of oil and will therefore induce a drop in the current price of oil. Forexample, Nordhaus (1979) estimated that, in a world without OPEC and othernoncompetitive producers, the price of oil under competition would have beenas low as $3 a barrel in 1975 (in 1975 dollars) because of extractors' anticipa-tions of future backstop production.

Such conclusions, however, do not follow from the explicit assumptions madeabove but from additional assumptions that, if made explicit, might be regardedby some as wishful thinking. Whether the backstop represents fusion or someother inexhaustible energy source, the capacity to deliver it on a scale that canreplace oil is currently unavailable.9 It is unrealistic to assume that this capacitycould be created costlessly. If it cannot be, then no backstop producer wouldundertake the costly expansion of capacity merely to receive a price per barrel-equivalent equal to the marginal operating cost.

If backstop capacity could be expanded at a constant marginal cost, thisexpansion would be deferred until the moment the backstop is needed becausethere would be no cost-saving in spreading the buildup of capacity over a longperiod, and it is always beneficial to defer costs. Because efforts to expandcapacity are undertaken in advance of production from the backstop, it seemsplausible to assume that backstop expansion involves increasing marginal costs.This assumption implies that compressing a task into a short period is morecostly than building ahead of time.

For simplicity, assume that the initial capacity of the backstop is zero, andconsider alternative cost scenarios under which this capacity can be expanded. Ifcapacity could be expanded costlessly, as the literature assumes, then the margi-nal operating cost of the backstop would serve as a ceiling on the oil price, andthe initial oil price would be comparatively low. At the other extreme, if expand-ing backstop capacity is prohibitively expensive, even though the backstopwould be cheap to operate once expanded, it cannot serve as a ceiling on the oilprice because for all practical purposes it is not a feasible option. In fact, the


current oil price would be the same as if no backstop existed. Thus, in bothcases, the price path depends on a factor not typically addressed: the marginalcost of expanding backstop capacity.

The intermediate case, where the marginal cost of capacity expansion is nei-ther zero nor infinite and increases with the rate of expansion, seems the mostplausible. In such a case, backstop expansion would begin before the pricereached the marginal operating cost so that the needed capacity would be builtin the most economical way. A given level of capacity would be provided atlowest cost when the discounted marginal cost of expansion is the same in allperiods. Therefore the marginal cost of expanding capacity before the backstopis in operation would itself grow by the interest factor, which implies thatexpansion would proceed at an increasing rate until the backstop begins to besold. At this point, the price must overshoot the backstop's marginal operatingcost; otherwise, the expansion would be unprofitable because the investment tobuild capacity would not be recovered. Once the price overshoots the marginaloperating cost, however, the backstop must operate at full capacity. The netprice would continue to rise by the interest factor, inducing continued extrac-tion. The price would rise because the expansion in the backstop is less than thereduction in extraction of the oil it is replacing. Eventually the price would peak,extraction of oil would cease, and backstop production would remain at capac-ity. The continued expansion of the backstop would then result in a price thatdeclines to the marginal operating cost (for further analysis, see Switzer andSalant 1986).

Although such overshooting will occur only in the distant future, its conse-quences will be felt immediately. Oil withheld for use during the overshootingphase is unavailable earlier. As a result, even if its marginal operating cost isconstant and the underlying technology is licensed competitively, the backstopmay not serve as a potent force in depressing the current price.

A boveground Storage

Because natural resources are sometimes stored above ground, it is importantto understand how such additional opportunities affect resource markets. Inprinciple, three kinds of costs are associated with either aboveground or under-ground storage: costs of acquisition, costs of maintenance, and costs of extrac-tion. In the case of underground storage, oil is never purchased from the marketto augment reserves because the cost of forcing it below ground is prohibitive.Maintaining reserves under ground or above ground is typically costless, how-ever. Hence, the focus is on the costs of extraction. In the case of abovegroundstorage, the costs of augmenting or depleting a stockpile are typically negligible,and the focus is on the costs of protecting and preserving stocks.

For stocks to be held above ground, the price must rise by more than theinterest factor to compensate for the additional costs incurred in maintaining thestockpile. Recall that the price would rise by less than the interest factor if all

Stephen W Salant 103

stocks were instead held below ground and a positive cost had to be incurred toextract them. Although the price would rise in both cases by the interest factor ifextraction and storage costs were zero, even then extractors and abovegroundstockpilers would sometimes behave in fundamentally different ways. Even ifstorage and extraction costs are zero, there remains one essential differencebetween aboveground and belowground storage: while aboveground stockpilesare often augmented by purchases from the market, underground reservesare never augmented by such purchases. Owners of aboveground storagecapacity-unlike extractors-are, therefore, always searching for opportunitiesto buy low, store their acquisitions, and sell high later.

Tests of the ModelWhat distinguishes models of natural resource extraction from static models is

the role of expectations and anticipatory behavior. Agents are assumed to haveforesight and to base current behavior in part on expectations about futureprices. These expectations greatly complicate the analysis of resource modelsand influence their predictions. Pragmatists will certainly want to know whethersuch anticipatory behavior is ever observed and, if so, whether it is sufficientlyimportant to be taken into account when formulating policy.

In my view, such anticipatory behavior is frequently observed; indeed, itconstitutes qualitative evidence in support of the Hotelling model. Failure totake such behavior into account has contributed to important policy errors.

Suppose, for example, the government uses a buffer stock to peg the price of anatural resource above its initial market level, as happened earlier in this centurywhen gold prices were pegged at $35 an ounce. Suppose the government com-mits to purchase any amount of the resource offered for sale and to sell anyamount (up to its entire accumulated inventory) demanded at this official price.What would be the consequences of this policy?

In the new equilibrium, the price could not remain permanently at the peggedlevel because cumulative demand would eventually exceed the government'sreserves (or, indeed, any finite amount). Nor could the price remain at thepegged level and then jump to the choke price when government reserves weredepleted, because such a jump would be anticipated and would provoke unlim-ited stockpiling. Instead, the Hotelling model predicts the following sequence ofevents: producers will extract and sell their entire reserves to the government atthe outset; the government will then begin reselling to consumers at the officialprice; at a predictable moment, private speculators will "attack" the governmentstockpile and purchase all the remaining reserves; and the speculators will there-after resell the acquired official reserves to consumers at prices that rise gradu-ally to compensate for interest and storage costs. The speculative attack ispredicted to occur when the government reserves decline to the point at whichthe official price would exactly equal the market price if the remaining officialstocks were held entirely by private stockpilers. Under this scenario each private


agent is maximizing, the government is following its policy rule, and the marketclears in each period.

If speculators have foresight, their anticipatory behavior is predicted to resultin an attack on the government scheme to peg the official price (Salant andHenderson 1978). This prediction applies equally when uncertainty is intro-duced (Salant 1983). Attacks are likewise predicted to occur in response tobuffer stock schemes to defend ceilings or bands on the prices of commodities ornatural resources.1 0 If speculators lacked foresight, however, no such specula-tive attacks would occur.

Actual government attempts to defend price ceilings by sales from officialstockpiles provide many opportunities to assess this qualitative prediction. Thesudden speculative attacks provoked by official schemes to impose ceilings onthe prices of various commodities (such as gold, tin, and coffee) and to fix manyexchange rates constitute qualitative evidence of anticipatory behavior too im-portant to ignore.

Similar points can be made about stockpiling in anticipation of government-imposed reductions in availability. If quotas, sanctions, tariffs, or regulationsare anticipated to limit the availability of a storable good from some knownfuture date onward, the Hotelling model predicts that inventories will be accu-mulated as that date approaches. If, however, that date remains uncertain, themodel predicts that inventories will be maintained continually. In that case, themodel also predicts that speculators will increase their stockpiles if the prospectsof reduced availability increase. None of these actions would occur if stockpilerslacked foresight.

Government policies that limit the availability of storable goods providecountless opportunities to assess these qualitative predictions. Observed stock-piling behavior commonly corresponds to the model's predictions. For example,as the day approached when Prohibition was about to go into effect, anticipa-tory stockpiling surged. Similarly, although no one in the United States knowswhen additional regulations will reduce the availability of firearms, the percep-tion that such regulations are now more likely to be enacted than they once werehas set off a surge in anticipatory stockpiling. Examples of such behavior byowners of stocks above or below ground are abundant. Policymakers would befoolish to ignore them.

Do Declining Prices Refute the Model?

As Barnett and Morse (1963) emphasized, real prices in many resource mar-kets have declined over long intervals. These declines have sometimes beentaken as evidence against the Hotelling model. Although declining real pricesmay be inconsistent with particular assumptions about extraction technologyand the nature of the resource, when these assumptions are suitably modified,the model does in fact predict declining prices. Four alternative reasons whyresource prices might decline are considered.


* Cost-reducing technical change. Technical change, as noted earlier, mightreduce extraction costs and cause prices to decline.

* Expansion of extraction capacity. Resource prices might also decline-evenin the absence of technical change-if extraction capacity were limited ini-tially and the marginal cost of expanding it increased with the rate ofcapacity expansion. In such circumstances, the prospect of a lower real pricenext year might induce the extractor to sell up to current capacity con-straints, and the expansion of capacity between one year and the next wouldresult in a price reduction (Switzer and Salant 1986).

* Revised estimates of old reserves and discoveries of new reserves. Discov-eries of new oil or revisions in estimates of existing oil reserves may reduceprices. Although the Hotelling model assumes that all reserves are knownfrom the outset, reserves are rarely estimated precisely, and owners are oftenrelieved to learn that their reserves are larger than their most pessimisticestimates (Gilbert 1979). Moreover, exploration results in the discovery ofadditional reserves at unknown times (Arrow and Chang 1982).

* Durables that are costly to extract rapidly. Prices of depletable resources,such as gold, that some consumers enjoy as possessions, might declineinitially as the stock held above ground increased. For a price decline tooccur, there must be diminishing returns to additional extraction or capacitylimits so that resource owners will not extract the entire stock at the outset.As underground reserves are extracted and the aboveground stock ex-panded, the rentals that these stocks yield decline, and so too may theirprice (Levhari and Pindyck 1981).

A Formal Test

Although econometric tests have been used to question the predictions of theHotelling model, the data used in these tests render their results questionable.Miller and Upton (1985) devised a clever cross-sectional test of the Hotellingtheory. Using the market price of the firm's shares as an estimate of the marketvalue of the extractor, the theory predicts that the market value per unit ofreserves of any operating extractor ought to equal that firm's current net pricewhen marginal costs are constant. Miller and Upton deduced how uncertaintyor rising marginal costs could modify this relationship because of diminishingreturns or depletion effects and concluded that the Hotelling model could ac-count for a substantial portion of the variation in market values of the sample offirms. They also found that their regression provided a method of predicting themarket value of extraction firms superior to Herold's Oil Industry ComparativeAppraisals, a commercial service widely quoted in the financial press that usesmore extensive information and more complicated methods.

This is not to say, on the one hand, that complexity inevitably leads to poorerpredictions or, on the other, that stylized Hotelling models of the kind reviewedhere are the best that can be devised to understand real-world resource markets.


The merit of these stylized models is that they capture essential elements of aresource market, they are easily understood and manipulated, and they can beused as a framework for more detailed analysis. Thus Nordhaus (1979) has builta complex Hotelling-style competitive model that uses detailed informationabout alternative sources of energy, reserves, and costs. Salant (1982) has devel-oped a model for the U.S. Department of Energy that extends the Hotellingframework to encompass the noncompetitive market structure of the world oilmarket. Both of these are complicated simulation models involving hundreds ofequations; neither can be solved without a computer. But because both of thesecomplex models adhere rigorously to the Hotelling framework, their structureand predictions can be understood by anyone familiar with the stylized modelsreviewed here. In both of these models, for example, one can have confidencethat each competitive extractor currently operating will behave as predicted bythe Hotelling model, a behavior that Upton and Miller found at work in real-world data. Thus, although Hotelling's stylized model is intuitively appealing toacademic economists, the framework also has something very useful (relative toavailable alternatives) to offer market analysts and policymakers.

Applications to Environmental Planning

Most environmental problems arise when production or consumption of agood generates a flow of pollution that eventually accumulates in the air orwater. Damage typically arises not from the current flow of pollution but fromits long-term accumulation. Economists contend that the competitive marketoften fails to maximize social welfare in such circumstances because agentsmaking the production or consumption decisions do not take into account thecosts that their actions impose on others.

The profession has not been as quick to recognize the central technologicalfeature of these pollution problems. Keeler, Spence, and Zeckhauser (1972)were the first to emphasize the resemblance between the dynamic aspects ofpollution and extraction problems.

To illustrate, suppose that a good (for simplicity, not a natural resource) canbe produced domestically at constant marginal cost and that consuming it gener-ates benefits that are subject to diminishing returns. Suppose, further, that aflow of pollution is also generated in fixed proportion to the rate of productionand that any damages from this pollution affect third parties. Suppose that thispollution generates no social costs until it accumulates beyond a known thresh-old-at which point the cost imposed on the third parties is unconscionablyhigh. The competitive market, left to function freely, will not produce the levelof output that is socially optimal. The socially optimal production path wouldcoincide with the equilibrium extraction path in the simplest Hotelling model.

Once more detailed considerations are introduced, the resemblance betweendynamic extraction and dynamic pollution problems becomes less striking. Forexample, as Dasgupta (1982) emphasizes, the environment has a capacity to

Stephen W Salant 107

recover from many forms of pollution, and society has the opportunity, at a realcost, either to abate the flow of new pollution or to clean up the stock of pastaccumulations. Although these features of environmental problems may have nocounterparts in extraction, the social planning problem associated with theHotelling model can easily be modified to accommodate these additionalconsiderations.

The literature summarized in this article is therefore not only useful in predict-ing prices and behavior in well-functioning resource markets, it can also identifysocially optimal allocations when environmental problems make the marketwork poorly.

Further Reading

The literature elaborating on Hotelling's classic article has been surveyedrepeatedly in the last two decades. Readable introductions are provided in arti-cles by Solow (1974) and Devarajan and Fisher (1981). Short monographs byBohi and Toman (1984) and Hartwick (1989) comprehensively cover Hotelling'smodel and its extensions. Textbooks include Hartwick and Olewiler (1986) andTietenberg (1992) at the undergraduate level and Dasgupta and Heal (1979) andConrad and Clark (1987) at the graduate level. Course outlines, exams, andproblems are collected and published by Tower (1990). Careful evaluations ofthe application of the Hotelling model to the world oil market are contained in amonograph by Cremer and Salehi-Isfahani (1991) as well as in Epple and Lon-dregan's (1993) contribution to the three-volume handbook edited by Kneeseand Sweeney (1993).

Dasgupta (1982) draws the connections between resource and environmentaleconomics. A list of journal articles in the field through 1992 is provided inHoagland and Stavins (1992).

Notes

Stephen W. Salant is a professor of economics at the University of Michigan.1. Recall that one dollar received in period t is equivalent to 1 / (1 + r)t- 1 dollars in period 1

where r is the (constant) interest rate, or that one dollar in the initial period is equivalent to(1 +r)t-1 dollars in period t. We refer to (1 +r) as the interest factor. At an interest rate of 10percent, a firm will be equally well off if it earns $100 today (which can generate $10 ofinterest income) or if it earns $110 one year from now.

2. Assume for now that storing extracted oil above ground and selling it later is prohib-itively expensive. As will become clear, the extra costs of premature extraction and above-ground storage make such behavior optimal only when prices are expected to rise rapidly.

3. This case is based on the following assumptions: The marginal cost of extraction isconstant and independent of the rate of extraction (that is, the first barrel of oil is just as costlyto extract as the next one). Neither the marginal cost nor the demand for the resource shifts inreal terms over time. The demand for oil is zero at sufficiently high prices. The real interestrate does not vary over time, and each firm's oil reserve is finite.


4. Any initial price has a corresponding price path. Because the net price on this path riseseach year at the interest rate, the present value of the net price in every year is constant.

5. If aboveground stockpilers anticipate rapid price increases, however, they are likely topurchase as much of the resource as is available, and in so doing prevent such price increases.

6. Herfindahl (1967) discovered this result for the case where resource pools can be extrac-ted at different but constant marginal costs. Weitzman (1976) discovered a generalization ofHerfindahl's result when demand for one unit is continual and the marginal cost of extractingan additional barrel at a given pool of oil can vary in any specified way as cumulativeextraction changes.

7. "Fool's Gold: When Central Banks Lose Their Lust for Gold, Gold Bugs Should Be-ware," Economist, January 23, 1993, p.17.

8. Suppose, for example, that global gold mining ceased and that whenever an extractorwanted to sell an ounce of gold, the government gave it to him in exchange for $300 and titleto one ounce of his underground gold. The extractor would not be affected by this schemeand should sell exactly as before and at the same prices. The extractor would still pay $300 anounce, but payment would go to the government instead of the miners; the extractor wouldcease to own each ounce of gold sold (having relinquished title to it). Governments wouldreceive $300 an ounce for the gold they would sell during the next twenty years and wouldthen pay the cost of extracting the gold (to which they had gained title) to restock theirreserves. Even at a 4 percent interest rate, governments would earn approximately $75 billionfrom this scheme.

9. It is important to distinguish two dimensions of costs: the cost of building or adding tocapacity, and the cost of operations. If, for example, the backstop represents nuclear energy,the first category would be the cost of building the reactor, and the second would be the costof producing nuclear energy. (The two categories are assumed to be independent.)

10. Krugman (1979) recognized that Salant and Henderson's reasoning could explainspeculative runs on fixed exchange-rate regimes. The subsequent literature on speculativeattacks in the foreign exchange market is nicely surveyed in Krugman and Obstfeld (1991).



Arrow, Kenneth J., and Sheldon Chang. 1982. "Optimal Pricing, Use, and Exploration ofUncertain Natural Resource Stocks' 7Journal of Environmental Economics and Manage-ment9(1, March):1-10.

Bain, Joe S. 1948. "Output Quotas in Imperfect Cartels." Quarterly Journal of Economics62:617-22.

Barnett, Harold J., and Chandler Morse. 1963. Scarcity and Growth: The Economics ofNatural Resource Activity. Baltimore, Md.: Johns Hopkins University Press.

Bohi, Douglas R., and Michael A. Toman. 1984. Analyzing Nonrenewable Resource Supply.Washington, D.C.: Resources for the Future.

Conrad, Jon M., and Colin W. Clark. 1987. Natural Resource Economics: Notes and Prob-lems. Cambridge, U.K.: Cambridge University Press.

Crabbe, Philippe J. 1983. "The Contribution of L. C. Gray to the Economic Theory ofExhaustible Natural Resources and Its Roots in the History of Economic Thought' "Jour-nal of Environmental Economics and Management 10(3, September):195-220.

Cremer, Jacques, and Djavad Salehi-Isfahani. 1991. Models of the Oil Market. London:Harwood Academic Publishers.

Cremer, Jacques, and Martin L. Weitzman. 1976. "opEc and the Monopoly Price of WorldOil." European Economic Review 8(2, August):155-64.

Stephen Wv Salant 109

Dasgupta, Partha. 1982. The Control of Resources. Cambridge, Mass.: Harvard UniversityPress.

Dasgupta, Partha, and G. M. Heal. 1979. Economic Theory and Exhaustible Resources.Cambridge, U.K.: Cambridge University Press.

Devarajan, Shantayanan, and Anthony C. Fisher. 1981. "Hotelling's 'Economics of Exhaust-ible Resources': Fifty Years Later." Journal of Economic Literature 19(1, March):65-73.

Epple, Dennis, and John Londregan. 1993. "Strategies for Modelling Exhaustible ResourceSupply." In A. Kneese and J. L. Sweeney, eds., Handbook of Natural Resource and EnergyEconomics. Vol. 3. Amsterdam: North Holland.

Fisher, Anthony C. 1981. Resource and Environmental Economics. Cambridge, U.K.: Cam-bridge University Press.

Gilbert, Richard J. 1979. "Optimal Depletion of an Uncertain Stock." Review of EconomicStudies 46(1, January):47-57.

Gray, L. C. 1914. "Rent under the Assumptions of Exhaustibility." Quarterly Journal ofEconomics 28:466-89.

Hartwick, John M. 1989. Nonrenewable Resources, Extraction Programs and Markets.London: Harwood Academic Publishers.

Hartwick, John M., and Nancy D. Olewiler. 1986. The Economics of Natural Resource Use.New York: Harper and Row.

Henderson, Dale W., John S. Irons, and Stephen W. Salant. 1993. "Making Better Use ofGovernment Gold: Qualitative and Quantitative Effects of Alternative Policies." Resourcesfor the Future, Washington, D.C. Processed.

Herfindahl, Orris C. 1967. "Depletion and Economic Theory." In Mason Gaffney, ed., Ex-tractive Resources and Taxation: Proceedings. Madison: University of Wisconsin Press.

Hoagland, P., and R. N. Stavins. 1992. "Readings in the Field of Natural Resource andEnvironmental Economics." John F. Kennedy School of Government, Harvard University.Processed.

Hotelling, Harold. 1931. "The Economics of Exhaustible Resources." Journal of PoliticalEconomics. 39(2, April):137-75.

Keeler, Emmett, Michael Spence, and Richard J. Zeckhauser. 1972. "The Optimal Control ofPollution." Journal of Economic Theory 4(1, February): 19-34.

Kneese, A., and J. L. Sweeney, eds. 1993. Handbook of Natural Resource and EnergyEconomics. Vols. 1, 2, 3. Amsterdam: North Holland.

Krugman, Paul. 1979. "A Model of Balance-of-Payments Crises." Journal of Money, Credit,and Banking 11(3, August):311-25.

Krugman, Paul R., and Maurice Obstfeld. 1991. International Economics Theory and Policy.2d ed. New York: HarperCollins.

Levhari, David, and Robert S. Pindyck. 1981. "The Pricing of Durable Exhaustible Re-sources." Quarterly Journal of Economics 96(3, August):365-77.

Ley, Eduardo, Molly Macauley, and Stephen W. Salant. 1994. "Spatially and IntertemporallyEfficient Solid Waste Management." Report prepared for the Environmental ProtectionAgency (EPA). Washington, D.C. Processed.

Miller, Merton H., and Charles W. Upton. 1985. "A Test of the Hotelling Valuation Princi-ple." Journal of Political Economy 93(1, February): 1-25.

Nordhaus, William D. 1973. "The Allocation of Energy Resources" Brookings Papers onEconomic Activity 3:529-70.

. 1979. The Efficient Use of Energy Resources. New Haven, Conn.: Yale UniversityPress.

Norgaard, Richard B. 1975. "Resource Scarcity and New Technology in U.S. PetroleumDevelopment." Natural Resources Journal 15(2, April):265-82.

Salant, Stephen W. 1982. Imperfect Competition in the World Oil Market: A ComputerizedNash-Cournot Model. Lexington, Mass.: Lexington Books.

1 10 The World Bank Research Observer, vol. 10, no. 1 (February, 1995)

. 1983. "The Vulnerability of Price Stabilization Programs to Speculative Attack."Journal of Political Economy 91(1, February):1-38.

Salant, Stephen W., and Dale W. Henderson. 1978. "Market Anticipations of GovernmentPolicies and the Price of Gold." Journal of Political Economy 86(4, August):627-48.

Solow, Robert M. 1974. "The Economics of Resources or the Resources of Economics.:American Economic Review 64(2, May):1-14.

Stiglitz, Joseph E. 1976. "Monopoly and the Rate of Extraction of Exhaustible Resources."American Economic Review 66(4):655-61.

Switzer, Sheldon, and Stephen W. Salant. 1986. "Expansion optimale de capacite par desexploitants prevoyants d'une ressource renouvelable." In Gerard Gaudet and Pierre Las-serre, eds., Ressources Naturelles et Theorie Economique. Quebec: Presses de l'UniversiteLaval.

Tietenberg, Thomas H. 1992. Environmental and Natural Resource Economics. 3d ed.Glenview, Ill.: ScottForesman.

Tower, Edward. 1990. Environmental and Natural Resource Economics: Economics ReadingLists, Course Outlines, Exams, Puzzles and Problems. Durham, N.C.: Eno River Press.

Weitzman, Martin L. 1976. "The Optimal Development of Resource Pools:' Journal ofEconomic Theory 12(3, June):351-64.

World Bank. 1992. World Development Report: Development and the Environment. NewYork: Oxford University Press.

Stephen W Salant 1ll

World Bank 3New Publications

World Debt Tables 1994-95: ExternalFinance for Developing Countries

A complete and up-to-date presentation of World Population Projections 1994-95:data and a review of development on the Estimates and Projections with Relatedexternal debt of and financial flows to develop- Demographic Statisticsing countries. Volume I--Analysis and Summary Contains the most recent demographicTables. 248 pages. Stock No. 12912. $21.95. statistics on the world's population, with esti-Volume II--Country Tables. 560 pages. Not mates for 206 countries and territories projectedavailable separately. Two-volume set--Stock as far as the year 2150. This edition incorporatesNo. 12914. $150.00. census and fertility-mortality survey results that

were unavailable for the previous edition. ItForthcoming: contains a new subregion, Central Asia, toWorld Debt Tables 1994-95: Data on Dis- categorize the southernmost former Sovietkette. Stock No. 12915. $95.00. republics. 532 pages. Stock No. 44947. $34.95.

Averting the Old Age Crisis: Policies toProtect the Old and Promote Growth

The first comprehensive and global examina- Voting for Reform: Democracy, Politicaltion of old age security. Systems providing Liberalization, and Economic Adjustmentfinancial security for the old are under growing Drawing on case studies from 8 countries,strain, and changing demographics have led to this volume assesses the interaction between twoan increase in the number of old people. global trends: political liberalization and market-Economists, policymakers, and government oriented reform. The study shows that althoughofficials will find valuable evaluations of policy there is no necessary contradiction between theoptions in this detailed report, along with two trends, the success of efforts to consolidatecritical analyses of the issues affecting income both democracy and reform depends on appropri-sucurity for the elderly in both developing and ate poldcal institudons and strategies. 536pagesindustrial nations. 432 pages. Stock No. 60996. Stoc a No s 60987 $24 95.$19.95. SokN 08 2.5

World Bank Atlas 1995An easy-to-use collection of color maps,

charts, and graphs representing key social,economic, and environmental indicators for 209countries and territories. Organized under three l r pmaedevelopment themes: The People, TheEconomy, and The Environment. 36 pages. *........

Stock No. 12922. $7.95.

World BankOn Film

Challenges and Change:The World Bank TodayFrancis Dobbs, producer

This thought-provoking new video on the World Bank recounts the institution's history, fromthe ruins of World War II and the Bretton Woods meetings through the present, and explains itsmission of global development and poverty reduction.

The World Bank is committed to reducing poverty and improving the standard of living of itspoorer member countries by providing development loans and technical assistance. This videoexplains the history, purpose, and practical workings of the Bank and explores projects in sectorssuch as health, education, nutrition, infrastructure, sustainable development, environmentalprotection, and many more.

Footage from sites around the world shows World Bank projects in progress and the results ofcompleted projects. Scenes depict programs for educating girls in Bangladesh, Mexico, and Togo;reforestation projects in Burkina Faso; urban development through community and governmentaction in Brazil; and the auction of state-owned enterprises in the Russian Federation.

The film concludes with the Bank's view of the future-that despite a 50 percent increase inlife expectancy and a doubling of per capita incomes over the past few decades, many challengesremain for global development.

June 1994 125 minutes I VHS and Betacam. NTSCIPALISECAM versions. EnglishlFrenchISpanish/ $10.95

Order: English edition: Stock No. 12931 (ISBN 0-8213-2931-6) NTSCI Stock No. 12932 (ISBN 0-8213-29324) PAL I Stock No. 12933 (ISBNO-8213-2933-2) SECAM.

French edition: Stock No. 12934 (ISBN 0-8213-2934-0) NTSC I Stock No. 12935 (ISBN 0-8213-2935-9) PAL / Stock No. 12936 (ISBN 0-8213-2936-7) SECAM.

Spanish edition: Stock No. 12937 (ISBN 0-8213-2937-5) NTSC I Stock No. 12938 (ISBN 0-8213-2938-3) PAL / Stock No. 12939 (ISBN 0-8213-2939-1) SECAM.

1W ~~~~~~ .. ...

World Bank m

CD-ROM WORLD,DATA 1994:World Bank Indicators on CD-ROM

Introducing the most comprehensive resource on the global economy the World Bank hasever produced, with the convenience of CD-ROM.

Over 700 socioeconomic time-series indicators. WORLD*DATA includes information from1960 through 1992, with complete data sets from our print publications World Tables, World DebtTables, and Social Indicators of Development, plus the complete text of Trends in DevelopingEconomies.

Data for over 200 economies. The CD-ROM format allows you to select regions (e.g. SoutheastAsia) or create your own groups of countries for easy comparison of indicators.

Information you can't get from any other World Bank source. Statistics in WORLD*DATAgo as far back as 1960 wherever available, compared with 1970 for our print publications. GDPfigures are restated in U.S. dollars for easy country-to-country comparison.

A powerful tool. WORLD*DATA contains the complete version of Javelin Plus 3.1 software.More than a spreadsheet, Javelin lets you visualize and analyze data in multidimensional ways.

Compatibility and ease of use. DOS-based WORLD*DATA uses the World Bank's popularand user-friendly Socioeconomic Time-Series Access and Retrieval System, (STARS) and workswith any IBM-compatible computer that has a CD-ROM reader and 512K of available memory. A24-page user's guide helps you get up to speed fast.

"The volume of data was incredible on WORLD*DATA"-Michael R. Yurkiv, Analyst, Dugg &Phelps Credit Rating Co.

"A very exciting product"-Laura Guy, Data and Program Library Service, University ofWisconsin.

Stock No. 13106 / $275.00

L World Bank Publications Order Coupon

CUSTOMERS IN THE UNITED STATES: CUSTOMERS OUTSIDE THE UNITED STATES:Complete this coupon and return to Contact your local World Bank Publications distributor

The World Bank for information on prices in local currency and paymentBox 7247-8619 terms. (See next page for a complete list of distributors.)Philadelphia, PA 19170-8619 If no distributor is listed for your country, use this orderU.S.A. form and return it to the U.S. address. Orders that are

Charge by credit card by calling (202) 473-1155 or sent to the U.S. address from countries with distribu-fax this completed order coupon to (202) 676-0581. tors will be returned to the customer.

Quantity Title Stock Number Price Total

Subtotal US$* If a purchase order is used, actual postage will Postage and handling* US$

be charged. If payment is by check or creditcard, postage and handling charges are Total US$US$5.00 per order. For air mail deliveryoutside the U.S., include US$8.00 for the firstitem and US$6.00 for each additional item.

CHECK METHOD OF PAYMENT

O Enclosed is my check payable to The World Bank.

O Charge my [I VISA O MasterCard [ American Express

Credit card account number

Expiration date Signature

O Bill me. (Institutional customers only. Purchase order must be included.)

PLEASE PRINT CLEARLY

Name

Firm

Address

City State Postal code

Country Telephone

1102

Distributors of World Bank PublicationsPrices and terms vary by country

ARGENTINA COLOMBIA GREECECarlos Hirsch, SRL Infoenlace Ltda. Papasotiriou S.A.Galeria Guemes Apartado Aereo 34270 35, Stournara Str.Florida 165,4th Floor-Ofc. 453/ Bogota D.E. 106 82 Athens

4651333 Buenos Aires COTE D'IVOIRE HONG KONG, MACAO

Centre d'Edition et de Asia 2000 LtdOficina del Libro Intemacional Diffusion 4648 Wyndham StreetAlberti 40 Africaines (CEDA) Wmning Centre1082 Buenos Aires 04 B.P. 541 7th Floor

Abidjan 04 Plateau Central Hong KongAUSTRALIA,PAPUA NEW GUINEA, CYPRUS HUNGARYFIJI, SOLOMON ISLANDS, Center of Applied Research Foundation for MarketVANUATU, AND Cyprus College EconomyWESTERN SAMOA 6, Diogenes Street, Engomi Dombovari Ut 17-19D.A. Information Services P.O. Box 2006 H-1117 Budapest648 Whitehorse Road NicosiaMitcham 3132 INDIAVictoria CZECH REPUBLIC Allied Publishers Private Ltd.

National Information Center 751 Mount RoadAUSTRIA PO Box 668 Madras - 600 002Gerold and Co. CS-113 57 Prague IGraben 31 INDONESIAA- 011 Wien DENMARK Pt. Indira Limited

SamfundsLitteratur Jalan Borobudur 20BANGLADESH Rosenoerns A116 11 P.O. Box 1181Micro Industries Development DK-1 970 Frederiksberg C Jakarta 10320

Assistance Society (MIDAS)House 5, Road 16 DOMINICAN REPUBLIC IRANDhanmondi R/Area Editora Taller, C. por A. Kowkab PublishersDhaka 1209 Restauraci6n e Isabel la P.O. Box 19575-511

Cat6lica 309 TehranBELGIUM Apartado de Correos 2190 Z-1Jean De Lannoy Santo Domingo IRELANDAv. du Roi 202 Government Supplies Agency1060 Brussels EGYPT, ARAB REPUBIJC OF 4-5 Harcourt Road

Al Ahram Dublin 2BRAZIL Al Galaa StreetPublicacoes Tecnical Cairo ISRAEL

Internacionais Ltda. Yozmot Literature Ltd.Rua Pexoto Gomide, 209 The Middle East Observer P.O. Box 5605501409 Sao Paulo, SP 41, Sherif Street Tel Aviv 61560

CairoCANADA R.O.Y. InternationalLo Diffuseur FINLAND P.O. Box 56055151 A Boul. de Mortagne Akateeminen Kiirjakauppa Tel Aviv 61560Boucherville, Quebec P.O. Box 128J413 5E6 SF-00101 Helsinki 10 ITALY

Licosa CommissionariaRenouf Publishing Co. FRANCE Sansoni SPA1294 Algoma Road World Bank Publications Via Duca Di Calabria, 1/1Ottawa, Ontario Kl B 3W8 66, avenue d'lena Casella Postale 552

75116 Paris 50125 FirenzeCHINAChina Financial & Economic GERMANY JAMAICA

Publishing House UNO-Verlag Ian Randle Publishers Ltd.8, Da Fo Si Dong Jie Poppelsdorfer Allee 55 206 Old Hope RoadBeijing 53115 Bonn Kingston 6

JAPAN PERU SRI LANKA & THE MALDIVESEastern Book Service Editorial Desarrollo SA Lake House BookshopHongo 3-Chome, Bunkyo-ku 113 Apartado 3824 P.O. Box 244Tokyo Lima 1 100, Sir Chittampalam A.

Gardiner MawathaKENYA PHILIPPINES Colombo 2Africa Book Service (E.A.) Ltd. International Book CenterQuaran House, Mfangano St. Suite 1703, Cityland 10 SWEDENP.O. Box 45245 Condominium Tower 1 Fritzes FackboksforetagetNairobi Ayala Avenue, H.V. dela Regeringsgatan 12, Box 16356

Costa Extension S-106 47 StockholmKOREA, REPUBLIC OF Makati, Metro ManilaPan Korea Book Corporation SWITZERLANDP.O. Box 101, Kwangwhamun POLAND Librairie PayotSeoul International Publishing Service Case postale 3212

Ul. Piekna 31/37 CH 1002 LausanneKorean Stock Book Centre 00-677 WarszawaP.O. Box 34 Van Dierman Editions TechniquesYeoeida PORTUGAL P.O. Box 465Seoul Livraria Portugal CH 1211 Geneva 16

Rua Do Carmo 70-74MALAYSIA 1200 Lisbon TANZANIAUniversity of Malaya Cooperative Oxford University Press

Bookshop, Limited SAUDI ARABIA, QATAR Maktaba StreetP.O. Box 1127, jalan PantaiBaru Jarir Book Store P.O. Box 529959700 Kuala Lumpur P.O. Box 3196 Dares Salaam

Riyadh 11471MEXICO THAILANDINFOTEC SINGAPORE, TAIWAN, Central Department StoreApartado Postal 22-860 MYANMAR,BRUNEI 306 Silom Road14060 Tlalpan, Mexico D.F. Gower Asia Pacific Pte Ltd. Bangkok

Golden Wheel BuildingNETHERLANDS 41, Kallang Pudding, #04-03 TRINIDAD & TOBAGO,De Lindeboom/InOr-Publikaties Singapore 1334 Systematics Studies UnitP.O. Box 202 #9 Watts Street7480 AE Haaksbergen SLOVAK REPUBLIC Curepe

Slovart G.T.G. Ltd. Trinidad, West IndiesNEW ZEALAND Krupinska 4EBSCO NZ Ltd. P.O. Box 152 UGANDAPrivate Mail Bag 99914 852 99 Bratislava 5 Gustro Ltd.New Market Ist Floor, Room 4, GeogiadisAuckland SOUTH AFRICA, BOTSWANA Chambers

For single titles: P.O. Box 9997Oxford University Press Plot (69) Kampala Road

NIGERIA Southern Africa KampalaUniversity Press Limited P.O. Box 1141Three Crowns Building Jericho Cape Town 8000 UNITED KINGDOMPrivate Mail Bag 5095 Microinfo Ltd.Ibadan For subscription orders: P.O. Box 3

International Subscription Service Alton, Hampshire GU34 2PGNORWAY P.O. Box 41095 EnglandNarvesen Information Center CraighallBook Department Johannesburg 2024 ZAMBIAP.O. Box 6125 Etterstad University BookshopN-0602 Oslo 6 SPAIN Great East Road Campus

Mundi-Prensa Libros, S.A. P.O. Box 32379PAKISTAN Castello 37 LusakaMirza Book Agency 28001 Madrid65, Shahrah-e-Quaid-e-Azam ZIMBABWEP.O. Box No. 729 Libreria Internacional AEDOS Longmian Zimbabwe (Pte.) Ltd.Lahore 54000 Consell de Cent, 391 P.O. Box ST 125

08009 Barcelona SouthertonHarare

Date post:	18-Mar-2021
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

World Bank Document · 2016. 8. 30. · Miguel Kiguel and Stephen A. O'Connell 21 Incentives and...

Documents