+ All Categories
Home > Documents > Wheat Supply Response: Some Evidence on Aggregation Issues

Wheat Supply Response: Some Evidence on Aggregation Issues

Date post: 13-Feb-2022
Category:
Upload: others
View: 7 times
Download: 0 times
Share this document with a friend
23
Wheat Supply Response: Some Evidence on Aggregation Issues Nurs ¸en Albayrak* Recent concerns about food security claim that since about 1984 world population has been growing faster than cereal production, and that world per capita cereal output is therefore now falling, in other words, the contemporary ‘neo-Malthusian’ case which disproves any Malthusian. Likewise, it has been argued that per capita cereal production has fallen in all the world’s main regions (Brown, 1997; Dyson, 1997). Indeed, with regard to Turkish wheat production, Figure 1, looking at the 5-year moving average curve, shows that the level of per capita wheat output has declined since the early 1980s. Figure 1 Per capita wheat production in Turkey (1935–96) Note: The world average figure is 363 kg/per capita (Dyson, 1997). In a recent article in this journal, Shaw (1997), reviewing a sample of an increasing number of books addressing questions of world food security, comes to a worrying view about the prospects for future generations. Dyson (1997) states that the single most important reason for the decline in world per capita * Post-Doctoral Research Fellow, Harper Adams Agricultural College, School of Management, Newport, Shropshire, UK. Development Policy Review Vol. 16 (1998), 241–263 © Overseas Development Institute 1998. Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF, UK, and 350 Main Street, Malden, MA 02148, USA.
Transcript
Page 1: Wheat Supply Response: Some Evidence on Aggregation Issues

Wheat Supply Response: Some Evidence onAggregation Issues

Nursen Albayrak*

Recent concerns about food security claim that since about 1984 worldpopulation has been growing faster than cereal production, and that world percapita cereal output is therefore now falling, in other words, the contemporary‘neo-Malthusian’ case which disproves any Malthusian. Likewise, it has beenargued that per capita cereal production has fallen inall the world’s mainregions (Brown, 1997; Dyson, 1997). Indeed, with regard to Turkish wheatproduction, Figure 1, looking at the 5-year moving average curve, shows thatthe level of per capita wheat output has declined since the early 1980s.

Figure 1Per capita wheat production in Turkey (1935–96)

Note: The world average figure is 363 kg/per capita (Dyson, 1997).

In a recent article in this journal, Shaw (1997), reviewing a sample of anincreasing number of books addressing questions of world food security, comesto a worrying view about the prospects for future generations. Dyson (1997)states that the single most important reason for the decline in world per capita

* Post-Doctoral Research Fellow, Harper Adams Agricultural College, School of Management,Newport, Shropshire, UK.

Development Policy ReviewVol. 16 (1998), 241–263

© Overseas Development Institute 1998. Published by Blackwell Publishers, 108 Cowley Road,Oxford OX4 1JF, UK, and 350 Main Street, Malden, MA 02148, USA.

Page 2: Wheat Supply Response: Some Evidence on Aggregation Issues

242 Development Policy Review

cereal production since the early 1980s is deliberate policy-induced reductionsin cereal output. The per capita production of non-cereal foods has beenincreasing as a result of farmers switching land to other food (and non-food)crops. For example, farmers in Europe have transferred cereal cropland tosunflowers and other oilseeds (p. 6). In this context, the role of providing the‘right’ price incentives to increase production has been repeatedly emphasisedin the development literature (Behrman, 1968; Thamarajakshi, 1977; Newberyand Stiglitz, 1981; Stevens and Jabbara, 1988; Tsakok, 1990; Sadoulet and deJanvry, 1995).

Early estimates of supply elasticities point to a pattern of low and positiveprice responses in agricultural supplies around the world.1 Consequently, theprevailing orthodoxy in the development literature of the 1950s and 1960s was

that agriculture was ‘not responsive to normal economic incentives’ (Schultz,1964: 8). Although encouraged by the observation that the producer’s responseto price incentives is low (and positive), some economists have expressedconcern that farmers are not given the right incentives. For instance, Krueger(1992: 126) reports that it was believed that most of agriculture represented‘backwardness’, that agricultural output was unresponsive to incentives, and thatagriculture could therefore be discriminated against in order to raise a surplusfor industry without large economic costs. Gafar (1997: 205) agrees and sees thefact that the supply response of agriculture was low and not statisticallysignificant as providing, in part, the justification for taxing agriculture andfavouring an inward-looking import-substitution industrialisation strategy. Thereason given for this is that the low measurable values of supply responses toprice changes tend to limit the damaging consequences of price distortions indeveloping agriculture. Indeed, analysis of a large sample of World BankCountry Economic Reports and Agricultural Sector Reports has revealed that theeffects of administered agricultural prices, taxes and subsidies wereunsatisfactory in many respects. The case studies concluded that the distortionswere substantial in many developing countries, the domestic terms of trade hadbeen turned heavily against agriculture as compared with the rest of theeconomy, cropping patterns had been distorted from what long-run comparativecost advantage would indicate was the optimum pattern, income distribution hadworsened and policies were internally inconsistent and not achieving their statedpurposes (Krueger et al., 1988, 1991).

1. Considerable effort has been devoted to the estimation of supply responses for agriculturalproducts in various countries around the world. Literally hundreds of estimates of some variantof supply relations at aggregate or disaggregate level have been made. See Behrman, 1968,1990; Askari and Cummings, 1976; Lin, 1977; Sobhan, 1977; Scandizzo and Bruce, 1980;Bond, 1983; Henneberry, 1986; Rao, 1989; Tsakok, 1990; Sadoulet and de Janvry, 1995;Schiff and Montenegro, 1997; Albayrak, forthcoming, as well as individual publications.

Page 3: Wheat Supply Response: Some Evidence on Aggregation Issues

Albayrak,Wheat Supply Response: Some Evidence on Aggregation Issues 243

Since 1981, many developing countries have adopted IMF-World Bankstructural adjustment programmes aimed at improving the competitiveness oftheir economies by removing the distortions caused by price controls on foodand other price rigidities, as well as monopolies, and an overvalued exchangerate, reducing public sector deficits and implementing market-friendly policies.The success of these reforms depends fundamentally on the quick growth ofagricultural output to overcome food shortages. It is essential to assess the‘price’ impact and to know to what extent the policies adopted affect theproduction of agricultural commodities, in what direction the distortions takeeffect, and what possible adjustments can be made to improve policy.Furthermore, reliable supply response estimates are particularly important whenpredicting the impacts of changing trade policies on agricultural markets. Thesereasons underlie the importance of farmer response to economic incentives forthe policy-maker. They are the focus of this article.

Motivation

The aim of this article is to investigate some alternative ways of specifying thearea response function for established wheat-growing areas in the Mediterraneanregion of Turkey as well as at the national level. The scope of the study isrestricted to analysis of the supply response at individual crop level. The wheatmarket seems reasonable to investigate for several reasons. Wheat is the mostimportant crop and food commodity in Turkey when measured in terms of thecropland under cultivation, the number of farmers involved in its production,and its importance in the diet of the population. In terms of acreage andproduction, cereals, especially wheat, are easily Turkey’s most important crop,occupying almost 9.8 million ha of arable land (about 70% of the crop area,excluding fallow). At 21 million tons, it would appear that Turkey is one of theworld’s biggest wheat producers, alongside Russia, the United States, India,Canada and France (FAO, 1988: 9; EIU, 1995: 31). In terms of number offarmers, the 1990 Agricultural Census indicates that 75% of the 4 millionhouseholds engaged in agriculture deal with wheat production. Because of theimportance of wheat as Turkey’s leading agricultural commodity, thegovernment has long intervened in the sector. Justification for this interventionis based on the need to ensure food security for the nation’s 60 million peopleand to guarantee adequate and stable incomes to its farmers. Like manydeveloping countries, Turkey has embarked on a process of market liberalisationwith the reform of most agricultural marketing boards. As government controlover agricultural prices is relaxed and new policies are developed, knowledgeabout farmer responsiveness to price adjustments is essential. Such informationcan improve the effectiveness of price policy reform and form a basis fordeveloping transitional policies.

Page 4: Wheat Supply Response: Some Evidence on Aggregation Issues

244 Development Policy Review

The article is organised as follows. A synthesis of available wheat supplyelasticities is provided in the next section. The following one deals with issuesrelating to the specification of a wheat supply model. Section three presentsbackground information on the analytical framework and the methodology, andthe fourth section examines the estimation procedure. Conclusions andrecommendations for future research are summarised in the final section.

Previous studies: a synthesis of available results

In this section, the available empirical evidence on the supply response to wheatproducer prices will be scrutinised with a view to establishing some generalconclusions on the subject. Table 1 lists country-specific estimates of wheatsupply elasticity to its own-price. It is unfortunate that the review is restrictedalmost exclusively to studies of only one commodity/country. A glance at thetable reveals a wide range of wheat supply elasticity estimates. This is largelybecause of the variety of ways that expected prices are formed, althoughdifferent time periods, levels of data aggregation, and estimation methods havealso contributed. Despite these variations, there is a pattern of significant priceresponses in wheat supplies. In other words, while some of the reportedestimates of elasticities may be biased, there seems to be little doubt thatpositive and significant supply elasticities for the crop under investigationunequivocally dispel the notation of perversity.

Table 1Wheat supply elasticities from Turkish agriculture

Author Own-price elasticity

Short-run Long-run

Kip (1972)Soral (1973)Ekmekcioglu and Kasnakoglu (1979)Somel (1979)AFC/GAP (1992)Albayrak (1997)

--

0.093–0.1290.108

-0.183

0.3760.108

0.211–0.2550.5210.3000.230

In terms of the magnitudes of elasticities reported in Table 1, the followingpoints should be made. First, the estimates of short-run elasticities of supplyresponse with respect to own-prices range between 0.093 and 0.183. Byconvention an elasticity less than 1 is considered inelastic. The short-runresponse in agriculture is very low because the main inputs — land, labour and

Page 5: Wheat Supply Response: Some Evidence on Aggregation Issues

Albayrak,Wheat Supply Response: Some Evidence on Aggregation Issues 245

capital — are fixed. However, one should bear in mind that while changes inproduct prices typically (but not always) explain a relatively small proportionof the total variation in output, short-run changes in output are often influencedby the weather and pests, while long-run changes are attributable to such factorsas improvements in technology which result in higher yields. This issue is alsoaddressed by a number of researchers, for example Mundlak (1985) andBinswanger (1990) who argue that to get a good response requires moreresources and/or better technology and infrastructural investments in areas suchas roads, markets, irrigation, and education. Secondly, the elasticity of supplyresponse increases with time as the desired factor reallocation becomes morecomplete and as factors which are fixed in the short run become variable, i.e.,the elasticity estimates in the long-run column are greater than the reportedresults in the short-run column. However, the sizes of short-run and long-runelasticities are close to each other. This could be explained by the low elasticityof supply response of the fixed factors. In other words, long-run elasticity ofsupply response can be very high (or low) depending on the elasticity of supplyresponse of the fixed factors. And thirdly, another type of contribution in termsof size of elasticity relates to the level of data aggregation. If the estimatedelasticities reasonably reflect the responsiveness of the cultivators concerned,considerable differences in elasticity exist for the same crop in the variousregions. Obviously, supply elasticities for districts which differ in resourceendowment, agro-climatic conditions, etc. will be different. There are alsodifferences in alternative crop(s) and production techniques according to regions.Several attempts within the literature already provide evidence on regionaldifferences in agricultural supply response, for example papers by Janssen andPerthel (1990), Du (1995), and Albayrak (1997) relate to the magnitudes of theelasticity estimates using aggregate and disaggregate data.

Empirical measure of supply response: specification issues

The degree of supply responsiveness is basically an empirical question. Varioustheories have been developed, adapted and applied to explain the dynamics ofsupply in agriculture. The approaches to estimating output responses varyconsiderably in the extent to which they depend upon the theory of the firm toproduce results. Colman (1983), Just (1993) and Sadoulet and de Janvry (1995)provide excellent reviews of these methods as well as an extensive review ofthe literature employing them. Since the objective of the present study is acomparison of short- and long-run forecasting of supply response, it hasemployed a ‘directly estimated single commodity supply model’ on market leveltime series data, where the behavioural parameters are obtained directly fromstatistical analysis of historical time series data. Typically, this involves

Page 6: Wheat Supply Response: Some Evidence on Aggregation Issues

246 Development Policy Review

specifying supply as a function of important current and past economicvariables.

Many steps need to be undertaken before the theory of supply response canbe translated into empirical measures of market response. The direction andmagnitude of supply elasticity depend upon many factors, such as the natureof the production function during the supply period, the elasticity of supply toinputs, the cost structure as regards fixed and variable costs, the motivatingforce behind the production response of farmers, the price expectation offarmers, and so on. The non-inclusion of important determining factors producesestimate bias. There are a number of attempts to show that the resultingestimates (elasticities) are highly sensitive to the way the theory was specifiedmathematically, the estimation procedure, and the data set used. For instance,Tsakok (1990) discusses the fact that empirical results are sensitive to (oftenarbitrary) specification decisions. Just (1993) argues that a major challenge insupply analysis is to determineprior to econometric estimation which marketand policy variables are potentially relevant to producers’ decisions. He furtheradds that another problem is missing data. When data are unavailable for someprices, inputs, or competing outputs, a common practice is simply to ignorethem.

However, the role of missing variables can change the appropriatespecification and the interpretation of the results. Kingwell (1996) echoes Just’sconcerns by illustrating the influence of specification errors in a programmingmodel of farm wheat supply response. He states that unless the farming system,with its array of production technologies, resources and enterprise alternatives,is described in some detail, it is highly likely that specification errors willsignificantly bias estimates of the supply response. He concludes that modelmisspecification often causes large changes in production response, with wheatarea and output elasticities changing by 40% or more. The same argument hadalready been pointed out by Chhibber (1982), who demonstrated that theomission of important structural variables may lead to an overstatement of thetrue magnitude of the response. For example, by adding irrigation to Peterson's(1979) model, Chhibber (1989) finds that price elasticity passes from 1.27 to0.97. Ogbu and Gbetiouo (1990) support this, stating that the empirical literatureon agricultural supply behaviour in sub-Saharan Africa is full of models that aredeficient in either their methodology or their choice of relevant explanatoryvariables. This is because the models failed to recognise the structure ofagricultural production in these countries and over-simplified the issues andconstraints facing the farmers.

Since model specification influences the values and the significance of theestimated supply elasticities, the reported results vary substantially regarding theidentification of relevant observed exogenous variables. Considerable judgment,based on local knowledge, is required to select the variables that really matter.There may be several competing crops, several inputs, several institutional

Page 7: Wheat Supply Response: Some Evidence on Aggregation Issues

Albayrak,Wheat Supply Response: Some Evidence on Aggregation Issues 247

variables (including policies), and even other environmental variables, but nottoo many as there will be a limited number of observations. It is important toconsider which variables are most relevant to the question. At least six factorscan influence crop acreage responses, such as expected price (or more generally,a vector of relative prices including the price of the crop itself, prices ofcompeting crops, and factor prices, with one of these prices chosen asnumeraire), factors of physical production of the crop concerned, a set of otherexogenous shifters, principally private and public fixed factors and trulyexogenous variables such as weather, government commodity programmes, etc.The hypothesised supply function in this article is expressed as follows:

Supply = f {expected output price and expected price of alternative crops, inputprice, technology, physical environment, policy environment} (1)

Mamingi (1997) addresses another important issue in estimating supplyelasticity. Econometric studies of agricultural supply response are based on anumber of somewhat arbitrary assumptions about the appropriate set of causalvariables and relationships. The specified supply relationship implies that thereis a unidirectional causality from right-hand-side variables to agricultural supply,in other words, price and other explanatory variables are uncorrelated with theerror term. In reality, it may well be the case that price and supply aresimultaneously determined, in which case estimates suffer from demand/supplysimultaneity bias. For instance, the price of agricultural exportables (exportcrops) in a given country is most likely exogenous, as it depends on the worldprice and production and the latter do not depend on the country's production.Nevertheless, the price of a given export crop is probably endogenous if thecountry’s share of world production is substantial. Hence, the recourse to someexogeneity tests, i.e., Hausman exogeneity tests, should become the rule ratherthan the exception for deciding on the simultaneity issue. Mamingi reports thatfailure to deal properly with the simultaneity problem gives rise to inconsistentestimates, and very few authors have dealt with the issue of simultaneity in thesupply response literature (p.23). Another point for consideration is that supplymodels are sometimes estimated by inappropriate methods, e.g., the treatmentof serial correlation in the disturbance term (Tomek and Myers, 1993).Consequently, the quality of many empirical agricultural product analyses is indoubt.

There are still a number of important omissions in the list of parameters ofthe supply function of equation (1). First, the concept of supply and supplyelasticity as defined are bothex ante, while what we observe in the real worldgenerally areex-postphenomena, which include the impact of many otherfactors on production. Secondly, equation (1) is still inadequate. This is becauseit is static, i.e., it implies that a change in an explanatory variable will inducean instantaneous and complete response in supply and that there are no delays

Page 8: Wheat Supply Response: Some Evidence on Aggregation Issues

248 Development Policy Review

in adjustment. In fact, there are a number of reasons for delayed adjustment inagricultural markets; hence we must differentiate between the immediate, orshort-run, response and the long-run response. Thus, a dynamic approach, whichrecognises the time lags in agricultural supply response, should be adopted inempirical analysis. Thirdly, there are also measurement problems involved in theestimation of supply elasticities. Measurement problems include segregating theimpact of the weather and technology from the production, generating farmers’expected prices, and isolating the effect of a change in the product price fromother factors. Regarding methodology, a careful specification of expected prices,weather, institutional constraints and technological variables should be made sothat an unbiased estimate of elasticity can be obtained.

Measurement of technology and expectations variables

The estimated supply model and the elasticities will be useful for policy analysisand forecasting if the proxies have adequately captured the relevant variable. Asneither ‘technology’ nor the ‘expectations formation’ of the producer is directlyobservable to the empirical analyst, they must be inferred from observedbehaviour. The main goal is to minimise error in the measurement of thevariables involved.

Technological change may be defined as a change in production function.Fulginiti and Perrin (1993) argue that typically the effects of technical changeare modelled in anad hoc fashion. Advances in technology that essentiallyresult in higher output per unit of input may be confused with supply responseand hence may lead to biased estimates of the price coefficients (p. 471).Similarly, improvements in the human factor because of training and outwardcontacts may increase productivity, but again there is no satisfactory way ofmeasuring it. The effects of these changes are well known, but it is difficult toidentify and measure precisely how much of a given change in output is due totechnical improvements and how much to changes in factor or product prices.Within the literature it is common to use a simple linear trend to representtechnology. Indeed, many models that analyse agricultural supply responsecontain a linear trend term as an independent regressor, implying the sameautonomous external source for technology.

The justification often given for including trend terms is their perceivedability to capture the effects of omitted or unmeasurable variables. The omittedvariable is assumed to be technology, i.e., suggesting smooth deterministicchanges in technology. But such a measure lacks validity because the use oftime implies that the technology increased at a constant rate every year. Some,however, have sought to determine the important causes of technical change inagriculture, and employed some crude measures of technological change, suchas public expenditure, the rate of adoption of improved production practices,

Page 9: Wheat Supply Response: Some Evidence on Aggregation Issues

Albayrak,Wheat Supply Response: Some Evidence on Aggregation Issues 249

output-input ratio, output-labour ratio, as it is presumed that these are the sourceof technical change, and hence the rate of change will be captured by thisvariable.

Others, e.g., Bardhan (1973); Parikh and Triverdi (1979); Krishna (1982),argue that simply an irrigation variable (the number of hectares irrigated) canbe used as a measure of the contribution of technical progress, which is thecritical precondition, and most important determinant, of the growth of areaunder HYVs, fertiliser consumption, and crop intensity. Alternatively, somehave sought to determine the important causes of technical change inagriculture, including the contributions of public expenditures for research, andto measure the rate of adoption of improved production practices. Most recently,the relationship between research and development expenditures and technology-based productivity growth in agriculture has attracted the attention of a numberof observers.2

With regard to expectational variables, there are a number of certaintyequivalents that might transform expectational variables into ones which aremeasurable.3 Yet there is no theoretical or empirical model of expectations thathas been universally embraced as the true or optimal representation. Since anarbitrary choice of expectations formation may not be a good strategy, anotherproblem in modelling supply response is the choice among alternativeformations of expectations. The question of which variable best represents theunknown prices has also been usefully addressed by a series of papers. Theliterature review on price expectations of field crops suggests the followingpoints. First, the problem of measuring expectations arises in every model ofagricultural supply. Several papers have examined the trade-offs in the choiceof a price expectations model and have aimed to determine the alternativemodels that are significantly dominant. The conclusion suggests that empiricaltests of the relative predictive power of these specifications of price expectationsshow that no one model dominates the rest. Most supply elasticity estimates arederived by incorporatingad hocassumptions about price expectations. In mostcases, naive and adaptive schemes have been used as proxies for priceexpectations in previous studies of supply response. On the other hand, therehave been some successful empirical applications of the rational expectationshypothesis in agricultural supply response models since the late 1970s.However, rational expectations models for agricultural products have not beenfound to be uniformly superior.

2. See Alston et al. (1995: Chapter 3) for a review of the methods for assessing research-induced technological changes in supply response.

3. Various postulated expectations generators have proceeded from the theories of Ezekiel,Goodwin, Koyck, Cagan, Nerlove and Muth. Each has different inherent strengths andweaknesses.

Page 10: Wheat Supply Response: Some Evidence on Aggregation Issues

250 Development Policy Review

Secondly, since most formulations of price expectations have shortcomings,the choice of any particular specification needs to be based on the performanceof alternative formulations in the supply response equations of a particularcommodity. However, this may not be a matter of simply selecting one of thealternative schemes to represent price expectations. Expectations formation islikely to be conditioned by the particular geographical location of thecommodity, and the institutional setting. This task is particularly difficult forcrops subject to farm programmes, as these programmes change and tend tocomplicate supply estimation because the relevant variables and structuralparameters may change over time. How these expectations are incorporated intoa commodity model, including the choice of an expectation theory, depends onthe information which is available to industry participants and how thatinformation is used. Moreover, agents (farmers) may not use this informationas intelligently as the model; that is, they do not know the model, or they havean incomplete understanding of the mechanism of price determination. Underthe assumption that firms exploit efficiently all available current information inmaking their forecasts, the rational expectations hypothesis should be found tobe reasonable. However, if firms fail to exploit the current information fully orsuch information is lacking or cost-prohibitive, then a lagged price responsemodel may be reasonable.

For the estimated elasticities to be meaningful, agricultural supply responsemodels must be built with a proper understanding of the realities of thecountries where these models will be applied. The estimated supply model andthe elasticities will be useful for policy analysis and forecasting if the proxieshave adequately captured the relevant variable. Of course, the choice amongalternative specifications is important as alternative models would yield differentforecasts or policy implications. In this article, price expectations formation isbased on the price determination mechanism observed in the Turkish wheatmarket. The marketing of wheat, over the sample period, was carried out understrict government control through the Turkish Grain Board (TMO). In otherwords, the government influences the position of the supply curve directly attime t by its own procurement price at time (t-1). Given the structure of thesupport provided to output prices in the Turkish wheat market, it would not betoo unreasonable to assume the expectation coefficients of prices to be equal toone, implying that the expected prices are simply determined by the previousyear’s prices.

Assuming that the supply response of each crop is homogeneous of degreezero (i.e., changing the prices of all the crops and inputs in the same proportionhas no effect on acres planted), one price can be used to normalise other prices.In this article, the nominal prices received by farmers have been deflated by thePrices Received By Farmers Index (PRBFI) in order to adjust for increases in

Page 11: Wheat Supply Response: Some Evidence on Aggregation Issues

Albayrak,Wheat Supply Response: Some Evidence on Aggregation Issues 251

inflation.4 The resulting price variable gives a ‘real’, i.e., normalised, price.Furthermore, the use of normalised rather than actual prices as regressorsreduces multicollinearity in prices. At the province level the price (orprofitability) of the second-best alternative crop can be effectively incorporated,and the aggregation problem becomes more manageable. For instance, wheatand cotton are important alternative crops over much of the Mediterranean (thespring wheat) region of Turkey. As wheat is of such primary importance interms of food security, its competitor (i.e., cotton) is one of the main cash cropscontributing significantly to foreign-exchange earnings. At the time of theproduction decision, acreage is allocated based on known input prices.Therefore, a perfect foresight assumption is employed for input (fertiliser)prices, i.e. FPt

e=FPt. Other variables included in the wheat supply equation werea set of dummy variables to capture the effects of policy parameters.

The methodology

To estimate how the quantity supplied responds to price and other variables, weneed to move from an economic model to a statistical model that we canestimate.5 With all variables in logarithmic terms for convenience ofmathematical manipulations and for direct estimation of elasticities, the proposedmodel is:

(2)

If we recall that the most important input of production is land, all factorslimiting the mobility of this input among its alternative uses will introduce awedge between what is desired and what is realised. After the usual substitutionof Nerlove's formulation between the desired area and the actual area planted:

(3)

4. The Prices Received By Farmers Index consists of the prices of 29 crops produced. Thisprice index is obtained by employing the weights of the crops in the total value of cropproduction (1961=100). On average, the Prices Received By Farmers Index incorporates 71%of the total value of crop production.

5. The measure of supply is planted acreage, a fairly common practice in agricultural supplystudies. By using area sown, not quantity, as the dependent variable, it is aimed to eliminateyield uncertainty and concentrate on the main decision variable.

Page 12: Wheat Supply Response: Some Evidence on Aggregation Issues

252 Development Policy Review

(4)

However, the model in equation (4) still cannot be directly estimated becauseof the unobservable variables. Hence, the next step is to use an appropriateexpectation variable. The specific form of the model is derived essentially froma Nerlovian partial adjustment model:

(5a)

(5b)

The simplification of the system will yield an estimation equation that containsonly observable variables:

(6)

Since equation (6) is estimated in logarithms the parameters are ready-madeestimates of the elasticities. The names and the definitions of variables in theestimated wheat supply function are presented in Table 2.

Page 13: Wheat Supply Response: Some Evidence on Aggregation Issues

Albayrak,Wheat Supply Response: Some Evidence on Aggregation Issues 253

Table 2Symbols and definitions of the variables used

Name Definition

AWtd desired area planted to wheat at time t

AWt acreage planted to wheat at time t

PRBFWt nominal prices received by farmers of wheat

PRBFCt nominal prices received by farmers of cotton

PRBFIt prices received by farmers index

PRBFWPRBFI

e

t

PRBFCPRBFI

e

t

expected wheat price received by farmers, relative toprices received by farmers index

expected cotton price received by farmers, relative toprices received by farmers index

FPte expected fertiliser price

IA t irrigated area

RFt rainfall

u1t, u2t and vt random disturbances with zero mean and constantvariance

γ area adjustment coefficient

λ1 andλ2 own and alternative crop’s price expectation coefficient

The political calendar used for creating dummyvariables

D(HYV) a policy variable representing the introduction of HYVfor wheat

D(1967) a dummy variable representing the application ofsupport prices to cotton

D(1980) a dummy variable representing the introduction ofstabilisation policies

S1 the modified effect of stabilisation policies in terms of

prices received by farmers:

Page 14: Wheat Supply Response: Some Evidence on Aggregation Issues

254 Development Policy Review

Estimation of parameters of wheat supply function

The foregoing model was applied to annual data on wheat area response inprovince-level (Mediterranean region) as well as national-level wheat data. Thedata set used are annual observations covering the period 1950–90. At the firststage, different variable specifications were considered but some of them wereexcluded because of a lack of significant contributions to specified province-level and national-level wheat supply models. For instance, rainfall wasconsidered an important supply shifter. However, the major part of wheatproduction at the province level is based on the irrigated area; hence the rainfallin the region does not play any role in its agriculture. In fact, if rainfall is apurely random variable, and if it is not included in the supply equation, itsimpact is merged in the residual term which is a random variable. Also, theintroduction of HYVs (1968) was limited to the spring wheat areas of thecoastal part of the country.

The model depicted by equation (6) provides a reasonable fit to bothprovincial and national-level data. Various measures of the extent of the degreeof fit of individual parameters or the entire empirical model are reported inTable 3. For provincial estimates the reported LM tests up to lag 12 indicatethat serial correlation is not likely. On the other hand, however, for the national-level estimates both Durbin's h-statistics and the LM test statistic at order oneindicate that first order serial correlation of the disturbance process is a potentialproblem, i.e., both statistics are significant at the 5% level, suggesting thatsignificant correlation remains in the residuals. One can take the error structureof vt explicitly into account by using the Maximum Likelihood method.6 Theresults, with a correction for first-order autocorrelation, are presented in the thirdcolumn of Table 3. Empirical analysis suggested that all the variables specifiedabove were important in explaining the area response of wheat in provincial andnational-level data. The values of the coefficient of determination, R2, indicatethat 86% and 95% of the variation in the planted area at the provincial and thestate level are explained by the model. The next section extends the analysis bydiscussing the role of the estimated elasticities under consideration.

6. If the autocorrelation was left uncorrected, it would lead to inappropriate inferences beingmade from biased t-ratios. A Cochrane-Orcutt iterative technique was used to correct forautocorrelation. The specific form of autocorrelation estimated is vt=-0.66vt-1+et with a t-ratioof (-3.78*).

Page 15: Wheat Supply Response: Some Evidence on Aggregation Issues

Albayrak,Wheat Supply Response: Some Evidence on Aggregation Issues 255

Table 3Provincial and national level wheat area response

Independentvariables

Province National

Constant -11.323(-1.50)

5.185(2.17)**

2.903(1.89)***

Own-Price 0.399(1.71)***

0.039(1.09)

0.041(1.69)***

Cross-Price -0.385(-1.87)***

--- ---

Fertilizer Price 0.058(2.52)**

0.008(1.85)***

0.005(2.05)**

Irrigated Area 1.519(2.69)**

--- ---

D(HYV) 0.374(3.108)*

0.042(2.57)**

0.027(2.63)**

D(1967) -0.158(1.36)

--- ---

D(1980) --- -0.049(-1.59)

-0.040(-2.25)**

S1 --- -0.042(-0.96)

-0.034(-1.27)

Lagged Variable 0.284(1.73)***

0.675(4.47)*

0.818(8.422)*

Diagnostic tests

Adjustment Coefficient

Durbin’s h-statistic

0.716

---

0.325

-2.81*

0.18

---

R2 0.86 0.93 0.95

R 2 0.81 0.92 0.94

Heteroskedasticity

(White’s)~ℵ12 0.04 0.62 ---

Test for Normality

(Jarque-Bera)~ℵ22 0.54 1.14 ---

Page 16: Wheat Supply Response: Some Evidence on Aggregation Issues

256 Development Policy Review

Test for FunctionalForm (Ramsey’s Reset

Test)~ℵ12 0.89 --- ---

LM Test (t-ratios)Lag 1 -0.24 -2.39**

---

2 -0.52 -0.63 ---

3 0.33 0.76 ---

4 -0.75 -0.87 ---

5 0.04 -3.06* ---

6 -0.55 -2.07** ---

7 -0.07 -1.86*** ---

8 -1.24 -0.06 ---

9 -0.16 -1.02 ---

10 -0.18 -1.35 ---

11 0.14 -1.65 ---

12 -0.54 -1.87*** ---

Notes: Asymptotic t-statistics appear in parentheses. Significance levels: *= 1%, **= 5%,***= 10%.

Short-run and long-run elasticities

The estimated models behave quite well, yielding significant elasticities. Table4 presents a summary of the estimated short- and long-run elasticities atprovincial and national level.

The first stage is to examine the plausibility of the results in terms ofa prioriexpectations of signs and magnitudes. These estimates are consistent in signwith economic theory and are generally significant. The estimated elasticitiesfurther indicate the following points.

First, the provincial-level estimates are positively associated with own-price,and are associated negatively with cross-prices. According to the results of theprovincial-level estimates the wheat area response is found to be 0.399. In otherwords, a 10% increase in the price of wheat relative to the Prices Received ByFarmers Index will bring only a 3.99% increase in the area planted to wheat inthe province. The negative sign of cross-price elasticity suggests decisioninterdependence, i.e., with competitors. The values of the cross-elasticitiesprovide evidence of the percentage change in wheat area in response to a givenchange in the expected price of cotton. It is worth noting that the estimated areaelasticity with respect to the fertiliser price is relatively small and significant.

Page 17: Wheat Supply Response: Some Evidence on Aggregation Issues

Albayrak,Wheat Supply Response: Some Evidence on Aggregation Issues 257

Table 4Short-run and long-run wheat area response elasticities

Elasticity withrespect to

Province levelestimates

National levelestimates

Short-run Long-run Short-run Long-run

Own-price 0.399 0.557 0.041 0.225

Cross-price -0.385 -0.537 - -

Fertiliser price 0.058 0.081 0.005 0.027

Irrigated area 1.519 2.121 - -

D(HYV) 0.374 0.522 0.027 0.148

D(1967) -0.158 -0.220 - -

D(1980) - - -0.040 -0.219

S1 - - -0.034 -0.186

Partial AdjustmentCoefficient

0.716 0.182

Note: The reader should check that the long-run results are in agreement with the argumentsof the relationship between partial adjustment elasticity and short-run elasticity.

Secondly, the coefficients of lagged acreage are highly significant for bothnational and province-level estimates. Consistent with the theory, thesecoefficients are positive and less than one, implying that long-run elasticitiesexceed short-run elasticities, i.e., a period of more than a year is required forwheat farmers fully to adjust their planting decisions in response to exogenousshocks. The short-run elasticities are low, because the main inputs, such as land,labour and capital, are fixed. Such results provide evidence that asset fixitieswill become less restrictive in influencing the area planted to wheat in the longrun. The partial adjustment coefficients are found to be 0.716 and 0.182 forwheat acreage at the province and state level, respectively. The coefficientsindicate that economic adjustment is faster in province-level than in national-level wheat area response.

Thirdly, it is well known that model specifications which do not take accountof regional and inter-crop dependencies may lead to misleading conclusions. InTurkey, there is a wide diversity in resource endowments, in agroclimaticconditions, and in the performance of wheat production techniques acrossdifferent regions. Hence, the wheat area responses to price vary by district andnational level data. Note that the own-price elasticities of provincial estimatesare greater than those of national-level estimates. This may be because national-level aggregation has many limitations, since price and other variables at thenational level lose their significance for the farmer who takes decisions on the

Page 18: Wheat Supply Response: Some Evidence on Aggregation Issues

258 Development Policy Review

basis of local information. Furthermore, at the national level, one encounters notonly the aggregation problem but also the difficulty of including the competingcrops’ price variable in the supply equation. In conclusion, by concentrating onregional application, it is expected that one will get a clearer understanding ofthe associated supply response.

Fourthly, the impact of the stabilisation policies started in 1980 became acentral issue in the policy debate over whether to continue government controlof the sector or to let the private market determine producer prices. While theintroduction of new wheat varieties provides a significant and positive impact,both the introduction of the price policy for cotton and the stabilisation policiesdecrease the area planted to wheat. In the light of such evidence, the case fora positive price policy for agricultural development is strong indeed.

Finally, based on the overall performance of the model, the statistical fit andthe forecast evaluation, the empirical work seems to provide some usefulinformation which could form a solid basis for policy action and analyses.Reliable area response estimates are particularly important when predicting theimpact of changing policies in agricultural markets. The next step involves theuse of elasticity estimates to calculate the responses. The following areexamples. If we suppose that the change in (PRBFW/PRBFI) is 30%, thechange in wheat area planted will be 11.97%, i.e. if the initial amount of wheatacreage is 286,000 ha (actual data for 1990), the change in acreage would be34,000 ha. Similarly, for the state-level estimates, if the price elasticity of wheatis 0.041 and the total wheat acreage is 9.45 million ha (actual data for 1990),then a price change of 10% translates into a change of 387,450 ha switched toor from a given crop in the short run. The same exercise can be applied to theother elasticities. Note that, although the response is inelastic in terms of theelasticity estimates, it may be major in terms of absolute levels of acreageinvolved.

Results and discussion

Based on the results of this study, the following conclusions can be drawn:

(i) The short-run elasticities for the province-level wheat area planted withrespect to own-price and cross-price to the Prices Received By Farmers Indexare 0.399 and -0.385, respectively. The corresponding long-run estimates are0.557 and -0.537. Overall, the results suggest that policies with regard to farmprices are crucial for the further expansion of wheat production and that recentpolicy initiatives to remove price subsidies may have an adverse effect on wheatoutput. Even within the present predominantly free market for wheat, thecontinuation of the guaranteed price scheme, at least on a limited scale, appearsto play a positive role in providing incentives to producers.

Page 19: Wheat Supply Response: Some Evidence on Aggregation Issues

Albayrak,Wheat Supply Response: Some Evidence on Aggregation Issues 259

(ii) The estimated own-price, cross-price and fertiliser price elasticities canprovide some insights to policy-makers concerning the quantitative effects ofchanges in farm prices. Price policies not only stimulate production but alsofavour the production of some crops over others. If a programme favours onecrop or another, there may be an allocation of land and other resources to thatparticular crop which is unwarranted, given market conditions. The resultsindicate that it would be hazardous to formulate a price policy on the basis ofa single crop's cost of production, and any change in the variables of one cropcould have adverse repercussion on other prices. In other words, own- andcross-price elasticities suggest that price can be an effective way to mobilisearea planted to wheat.

(iii) Perhaps the most important finding of the study relates to the magnitudesof the elasticity estimates using province and national-level data. There are largevariations in the price response of area planted at the provincial and nationallevels. The results further indicate that, without knowledge of regional acreageresponse elasticities, caution is required in assessing the impact of national-levelacreage response. The own-price elasticity for the province level is as high as0.399, i.e. ten times higher than the national-level estimates (0.041). Theimplication of different size elasticities is that the elasticity estimates developedfor the country as a whole may not be accurate for the region for predictingacreage response to changing policy environments or market conditions.

(iv) The effect of irrigation on wheat acreage was found to be significantlypositive and several times the price elasticity, i.e., 1.519 and 0.399, respectively.The results also raise a policy problem about how to increase output. ShouldTurkish wheat policy pay greater attention to the measures of technological andinfrastructure improvements or to that of ‘getting the price right’? It is wellknown that the process of achieving higher agricultural productivity and greateroutput can be accelerated by an incentive agricultural price policy. However,such a policy cannot be framed rationally unless the effects and implications ofprice changes and differentials are taken into consideration. The resultsrecommend that support for agriculture should take the form of measures toimprove the agricultural structure and encourage modernisation throughinvestment rather than using the short-term solution of price intervention totackle problems in the agricultural sector. In other words, a unit percentagechange in the important shifter variable (technology) will yield much greatergrowth than a unit percentage price shift; hence a balanced policy should stressa technology more than a price policy.

Page 20: Wheat Supply Response: Some Evidence on Aggregation Issues

260 Development Policy Review

References

AFC/GAP (1992) ‘Agricultural Marketing and Crop-Mix Planning in Turkey’.Report on joint project by AFC (Agriculature and Food InternationalConsulting, Bonn, Germany) and GAP (South Anatolia RegionalDevelopment Administration, Ankara, Turkey), Vol. V. Ankara, August.

Albayrak, N. (1997) ‘Applying Time Series Analysis to Supply Response andRisk’. Unpublished PhD Thesis, Dept. of Economics, University of Leicester,UK.

Albayrak. N. (forthcoming) ‘Time-Series Estimates of Agricultural SupplyFunctions and Systematic Determinants of It: A Synthesis of AvailableResults’,International Review of Applied Economics.

Alston, J. M., Norton, G. W. and Pardey, P. G. (1995)Science Under Scarcity:Principles and Practices for Agricultural Research Evaluation and PrioritySetting.Published in co-operation with the International Service for NationalAgricultural Research. Ithaca, NY and London: Cornell University Press.

Askari, H. and Cummings, J. T. (1976)Agricultural Supply Response: A Surveyof the Econometric Evidence.New York: Praeger Special Studies inInternational Economics and Development.

Bardhan, P. K. (1973) ‘Size, Productivity and Returns to Scale: An Analysis ofFarm-Level Data in Indian Agriculture’,Journal of Political Economy:1370–86.

Behrman, J. R. (1968)Supply Response in Underdeveloped Agriculture: A CaseStudy of Four Major Annual Crops in Thailand, 1937–1963.Amsterdam:North-Holland Publishing Company.

Behrman, J. R. (1990) ‘Agricultural Supply’ in J. Eatwell, M. Milgate and P.Newman (eds)Economic Development. The New Palgrave. Basingstoke:Macmillan.

Binswanger, H. (1990) ‘The Policy Response of Agriculture’,World BankResearch Observer4: 231–58.

Bond, M. (1983) ‘Agricultural Responses to Prices in Sub-Saharan AfricanCountries’,International Monetary Fund Staff Papers30(4): 703–26.

Brown, L. R. (1997) ‘Can We Raise Grain Yields Fast Enough?’,World Watch,July/August.

Chhibber, A. (1982)Dynamics of Price and Non-price Response of Supply inAgriculture. Stanford, CA: Stanford University Press.

Chhibber, A. (1989) ‘The Aggregate Supply Response: A Survey’ in S.Commander (ed.)Structural Adjustment and Agriculture: Theory andPractice in Africa and Latin America.London: Overseas DevelopmentInstitute.

Colman, D. R. (1983) Review of the Arts of Supply Response Analysis',Review of Marketing and Agricultural Economics51(3): 201–30.

Page 21: Wheat Supply Response: Some Evidence on Aggregation Issues

Albayrak,Wheat Supply Response: Some Evidence on Aggregation Issues 261

Du, W. (1995)Agricultural Marketed Surplus Response in China.Aldershot:Avebury.

Dyson, T. (1997) ‘Feeding The World to 2020-Prospects for Demand andSupply’. Paper Presented at the 51st Oxford Farming Conference.

EIU (1995)World Outlook, Turkey. Country Annual Report. London: EconomistIntelligence Unit.

Ekmekcioglu, C. and Kasnakoglu, H. (1979) ‘Supply Response in TurkishAgriculture: Preliminary Results on Wheat and Cotton (1955–75)’,METUStudies in Development6 (22/23): 113–43.

FAO (1988) Agricultural Review for Europe, Vol. I (31), ECE/AGRI/103.Rome: FAO.

Fulginiti, L. and Perrin, R. (1993) ‘Prices and Productivity in Agriculture’,Review of Economics and Statistics75(3): 471–782.

Gafar, J. (1997) ‘The Supply Response of Aggregate Agricultural Output inJamaica,Agricultural Economics16: 205–17.

Henneberry, S. R. (1986)Review of Agricultural Supply Response forInternational Policy Models. Department of Agricultural Economics,Oklahoma State University.

Janssen, M. and Perthel, D. (1990) Seasonal and Regional Differences inAgricultural Supply Response in Benin',European Review of AgriculturalEconomics17: 407–20.

Just, R. E. (1993) ‘Discovering Production and Supply Relationships: PresentStatus and Future Opportunities’,Review of Marketing and AgriculturalEconomics61: 11–40.

Kingwell, R. (1996) ‘Programming Models of Farm Supply Response: TheImpact of Specification Errors’,Journal of Economic Systems50: 307–24.

Kip, E. (1972) ‘Supply-Price Relationship: Some Evidence’, (in Turkish),ZiraatEkonomisi Dergisi3(9), Ankara.

Krishna, R. (1982) ‘Some Aspects of Agricultural Growth, Price Policy andEquity in Developing Countries’,Food Research Institute StudiesXVIII (3).

Krueger, A. O. (1992)The Political Economy of Agricultural Pricing Policy.Washington, DC: World Bank.

Krueger, A. O., Schiff, M. and Valdes, A. (1988) ‘Agricultural Incentives inDeveloping Countries: Measuring the Effect of Sectoral and Economy widePolicies’, The World Bank Economic Review2(3): 255–71.

Krueger, A. O., Schiff, M. and Valdes, A. (1991)The Political Economy ofAgricultural Pricing Policy.Baltimore, MD: Johns Hopkins University Press.

Lin, G. W. (1977) ‘Measuring Aggregate Supply Response under Instability’,American Journal of Agricultural Economics59: 903–07.

Mamingi, N. (1997) ‘The Impact of Prices and Macroeconomic Policies onAgricultural Supply: A Synthesis of Available Results’,AgriculturalEconomics16: 17–34.

Page 22: Wheat Supply Response: Some Evidence on Aggregation Issues

262 Development Policy Review

Mundlak, Y. (1985)The Aggregate Agricultural Supply Response, Centre forAgricultural Economic Research Working Paper 8511, Rehovot, Israel.

Nerlove, M. (1968)The Dynamics of Supply: Estimation of Farmers' Responseto Price. Baltimore, MD: Johns Hopkins University Press.

Newbery, D. G. M. and Stiglitz, J. E. (1981) The Theory of Commodity PriceStabilisation: A Study in the Economics of Risk.Oxford: Clarendon Press.

Ogbu, O. M. and Gbetiouo, M. (1990) ‘Agricultural Supply Response in Sub-Saharan Africa: A Critical Review of the Literature’,African DevelopmentReview2: 83–99.

Parikh, A. and Trivedi, P. K. (1979) ‘Estimation of Returns to Inputs in IndianAgriculture’, (unpublished paper).

Peterson, W. L. (1979) ‘International Farm Prices and the Social Costs of CheapFood Policies’,American Journal of Agricultural Economics61: 12–21.

Rao, J. M. (1989) ‘Agricultural Supply Response: A Survey’,AgriculturalEconomics3: 1–22.

Sadoulet, E. and de Janvry, A. (1995)Quantitative Development PolicyAnalysis.Baltimore, MD: Johns Hopkins University Press.

Scandizzo, P. L. and Bruce, C. (1980)Methodologies for MeasuringAgricultural Price Intervention Effects. World Bank Staff Working Paper394. Washington, DC: World Bank.

Schiff, M. and Montenegro, C. E. (1997) ‘Aggregate Agricultural SupplyResponse in Developing Countries: A Survey of Selected Issues’,EconomicDevelopment and Cultural Change45(2): 393–410.

Schultz, T. W. (1964)Transforming Traditional Agriculture. New Haven, CT:Yale University Press.

Shaw, D. J. (1997) ‘World Food Security: The Impending Crisis?’,DevelopmentPolicy Review15(4): 413–20.

Sobhan, I. (1977) ‘Agricultural Price Policy and Supply Response - A Reviewof Evidence and Interpretation for Policy’. AGREP Division Working Paper,2. Washington, DC: World Bank (mimeo).

Somel, K. (1979) ‘Agricultural Support Policies in Turkey: A Survey ofLiterature’,METU Studies in Development6 (24/25): 275–325.

Soral, E. (1973) ‘Agricultural Price Formation and Government Intervention’ (inTurkish), Eskisehir, ITIA,50th Year Special Issue: 146–76.

Stevens, R. D. and Jabara, C. L. (1988)Agricultural Development PrinciplesEconomic Theory and Empirical Evidence.Baltimore, MD: The JohnHopkins University Press.

Thamarajakshi, R. (1977) ‘Role of Price Incentives in Stimulating AgriculturalProduction’ in Douglas Ensminger (ed.),Food Enough or Starvation forMillions. New Delhi: Tata McGraw Hill.

Tomek, W. G. and Myers, R. J. (1993) ‘Empirical Analysis of AgriculturalCommodity Prices: A Viewpoint’,Working Papers in Agricultural Economics

Page 23: Wheat Supply Response: Some Evidence on Aggregation Issues

Albayrak,Wheat Supply Response: Some Evidence on Aggregation Issues 263

93(1), Dept. of Agricultural Economics, New York State College ofAgriculture and Life Sciences.

Tsakok, I. (1990)Agricultural Price Policy: a Practitioner’s Guide to PartialEquilibrium Analysis.Ithaca, NY and London: Cornell University Press.


Recommended