+ All Categories
Home > Documents > Do governance indicators predict anything? The case...

Do governance indicators predict anything? The case...

Date post: 31-Aug-2018
Category:
Upload: lamthien
View: 242 times
Download: 0 times
Share this document with a friend
49
Do governance indicators predict anything? The case of “fragile states” and civil war * James D. Fearon Department of Political Science Stanford University May 24, 2010 1 Introduction The term “fragile state” may be the most successful and influential development policy euphemism of the last 10 years. It has been embraced as an important operational concept by the World Bank, the OECD’s Development Assistance Committee, the United Kingdom’s Department for International Development, and the U.S.’s Agency for International Development, among many other governmental and non-governmental donor agencies. “Fragile state” is a euphemism because it is a delicate way of saying that a country has weak or dysfunctional institutions and/or is poorly governed. The term is also used to suggest the possi- bility or actuality of significant political violence, something that for many years the World Bank and other aid agencies viewed as none of their business. 1 “Fragile states” are thought to be at risk of becoming “failed” or “collapsed,” with terrible consequences for economic welfare and development. Implicit in the concept is a theory of economic development that has become more and more influential, after years of project lending that has often had disappointing results. The theory is that economic growth requires above all good policies and capable government institutions to * To the discussants: This is an INCOMPLETE DRAFT, prepared as a background paper for the 2011 World Development Report and including material that will form the basis for a paper for Annual Bank Conference on Development Economics. As such, it is something of an uncomfortable hybrid at present – sorry about that. The ultimate ABCDE paper will be focused on the question posed in the title about governance indicators and civil war risk. 1 There was more civil war in developing countries in the mid 1980s than there is now, but there was essentially no discussion of the problems posed for development by conflict at that time. 1
Transcript

Do governance indicators predict anything?The case of “fragile states” and civil war∗

James D. FearonDepartment of Political Science

Stanford University

May 24, 2010

1 Introduction

The term “fragile state” may be the most successful and influential development policy euphemismof the last 10 years. It has been embraced as an important operational concept by the WorldBank, the OECD’s Development Assistance Committee, the United Kingdom’s Department forInternational Development, and the U.S.’s Agency for International Development, among manyother governmental and non-governmental donor agencies.

“Fragile state” is a euphemism because it is a delicate way of saying that a country has weak ordysfunctional institutions and/or is poorly governed. The term is also used to suggest the possi-bility or actuality of significant political violence, something that for many years the World Bankand other aid agencies viewed as none of their business.1 “Fragile states” are thought to be atrisk of becoming “failed” or “collapsed,” with terrible consequences for economic welfare anddevelopment.

Implicit in the concept is a theory of economic development that has become more and moreinfluential, after years of project lending that has often had disappointing results. The theoryis that economic growth requires above all good policies and capable government institutions to

∗To the discussants: This is an INCOMPLETE DRAFT, prepared as a background paper for the 2011 World DevelopmentReport and including material that will form the basis for a paper for Annual Bank Conference on Development Economics. Assuch, it is something of an uncomfortable hybrid at present – sorry about that. The ultimate ABCDE paper will be focused on thequestion posed in the title about governance indicators and civil war risk.

1There was more civil war in developing countries in the mid 1980s than there is now, but there wasessentially no discussion of the problems posed for development by conflict at that time.

1

implement them. Violent conflict may be caused by bad policies and institutions, and violentconflict may in turn cause bad policies, the destruction of institutions, and so more poverty –fragile states are thus thought to be at risk of being stuck in a “conflict trap” (Collier et al. 2003).One of the central questions for development aid at present is whether and what sorts of aid cando anything to improve policies and institutions in a “fragile state.” Indeed, another dimension ofthe euphemism is that “fragility” suggests something that needs to be taken care of, for example,by providing more aid. But one could argue, and some have, that there is little point to providingaid if a government is riddled with corruption and led by elites who are not much interested indevelopment.

At a conceptual level, the idea of a “fragile state” remains murky: What exactly are “weak in-stitutions” and how do we recognize and measure them? Despite the conceptual or theoreticalvagueness, aid agencies have managed to produce operational criteria for identifying fragility. Inmost cases, states are considered fragile if they score below a threshold value on governance in-dicators that are produced by expert ratings. For example, the World Bank designates a state as“fragile” if its aggregate score falls in the bottom 40% of the Bank’s Country Policy and Institu-tional Assessment (CPIA), a set of governance indicators based on annual surveys completed byBank officials working in particular regions and countries.2

There is almost no research on the question of whether these expert-based governance indicatorsactually forecast a country’s performance in the next five or ten years. If aid allocation and styledecisions are going to be conditioned to a significant degree on these indicators, we would like toknow if perceptions-based measures of somewhat unclear concepts actually are picking up any-thing relevant to performance and outcomes. For example, is it actually true that “fragile states” asdesignated by governance indicators are at greater risk of violent conflict?

In this paper I consider whether low values on governance indicators such as the CPIA index, theWorld Governance Indicators (Kaufmann, Kraay and Mastruzzi 2009), and the International Coun-try Risk Guide predict an elevated risk of civil violence in subsequent years. It would not be thesurprising if there were a bivariate relationship, since it is well-known that governance indicatorsof all sorts are strongly related to per capita income levels, and it is also well documented that civilwar is much more common in poor countries. What is less obvious is whether, controlling for acountry’s level of economic development, expert-based perceptions of the quality of governancewill have value for forecasting subsequent conflict experience.

I find that they do. A country that was judged in one year to have worse governance than one wouldexpect given its income level has a significantly greater risk of civil war outbreak in the next fiveor ten years. This is true for all three sets of governance indicators considered here, and it doesnot matter much which indicator one chooses – for instance, “government effectiveness” (WGI),

2The Bank also designates states as “fragile” if they have had a peacekeeping operation in any of theprevious three years ...

2

“investment profile” (ICRG), “corruption” and “rule of law” (several of them) all work. Resultsare weakest for the Bank’s CPIA indicator.

These results may also have relevance for current debates on the causes of civil war. Are thepoorest countries more likely to have civil war (a) because of direct labor market effects, wherebyjoining an armed group is relatively attractive in a poor country; or (b) because low income proxiesfor weak governance, which raises civil war risk either because (b.1) more people are frustratedand unhappy because service provision is so poor, or (b.2) the state’s weak administrative andcoercive capabilities create better opportunities for rebel groups? The finding here is that whenyou control for governance quality, level of income has little or no predictive power for civil waronset, but when you control for income, measures of governance quality do predict future conflictexperience. This is supportive of (b) versus (a), although as I discuss below there are a few otherstudies that can be interpreted as finding evidence of a direct causal effect from low income toconflict risk.

In terms of policy implications, the results lend support to the view that aid in conflict-affectedcountries needs to do more than try to raise incomes through project lending. If capable govern-ment is indeed the root of the problem of conflict and development more than a “poverty trap,”for example, then a more integrated approach that draws from the “peace building” and “statebuilding” experience of U.N. and other peacekeeping operations may be necessary.

In the next two sections I reexamine some of the most common findings in the recent literatureon civil war onset using cross-national panel models and the coding of civil conflict that is em-ployed for the 2011 World Development Report. Section 3 introduces the conflict data. Section 4presents a series of “onset models,” and includes consideration of some factors that have not beenmuch examined in the existing literature, such as the relationship between human rights abusesby government and civil war risk. In section 5 I then discuss what these essentially correlationalfindings may or may not imply about the causes of civil war and lower levels of violent conflict.Section 6 returns to empirical analysis, examining the relationship between governance indicatorsand conflict onset.

2 Correlates and causes of civil war and lower-level violence conflict

Since the end of the Cold War, a moderately large literature has developed that uses cross-nationaldata to study the correlates of large-scale civil violence, usually for the 65 years since the end ofthe World War II. The typical design is an annual panel for roughly 160 countries (micro-states areoften omitted for lack of data or other reasons) with on average 35 to 45 observations per country.

Researchers have formulated the dependent variable in these statistical models in two main ways.Most of the literature examines correlates of civil war onset, meaning that the dependent variable

3

is coded as “1” for country years with a civil war onset, and zero for other country years.3 Otherresearchers look at the correlates of civil war incidence, meaning that the dependent variable iscoded as “1” for every country year with (at least one) civil war occurring, and zero otherwise(Montalvo and Querol 2005; Besley and Persson 2009). A problem with the latter approach isthat the estimated coefficients are complicated averages of the “effect” of a covariate on both theonset and the duration of civil conflicts. Distinguishing between onset and continuation (duration)almost invariably shows that we can reject the hypothesis that an independent variable has the samerelationship to onset as it does to continuation. A more natural alternative is to study determinantsof conflict onset and duration separately, using survival models for the latter (Balch-Lindsay andEnterline 2000; Collier, Hoeffler and Soderbom 2004; Cunningham 2006; Fearon 2004). In thispaper I consider factors related to the onset of civil conflict and war, rather than incidence orduration.4

For the most part, the cross-national statistical models of civil war onset and incidence shouldbe viewed as descriptive more than structural (or causal). That is, they have been of great valuefor making clearer which political, economic, and demographic factors are associated with highercivil war propensity in the last 60 years, which factors are not, and which are associated withonset when you control for other factors. But for many covariates found to be statistically andsubstantively significant in these models, the argument for interpreting the estimated coefficientsas causal effects is tenuous or speculative. For instance, do we really believe that if the share ofmountainous terrain in, say, Namibia, increased from 10% to 20% that its expected annual odds ofcivil war onset would increase by about 15%? Maybe, but this is not an experiment that we willever get to run. Even so, it is of some interest to learn that in the last 65 years, there has been sometendency for more mountainous countries to have more civil wars, even controlling for a range ofother country characteristics.

At least for the literature to this point, found opportunities for natural experiments to allow cleaneridentification of causal effects have been rare. Miguel, Satyanath and Sergenti (2004) cleverly used

3Some drop country years with ongoing civil war (Collier and Hoeffler 2004), others include them andintroduce a variable indicating whether war was ongoing in the previous year as a control (Fearon andLaitin 2003). The latter approach avoids dropping onsets that occur when another civil war is already inprogress.

4Duration dependence is also a bigger problem in incidence models than in onset models. To date,researchers who have taken incidence as the dependent variable have tried to fix the problem simply byclustering the errors within countries, but not including explicit dynamics in the form of, say, a laggeddependent variable. Implicitly, then, they assume that there is no direct causal effect of war in one year onthe probability of war in the next year. This is strongly counter to theory and case-based evidence, since it isclear that there are large fixed costs to generating an insurgent movement, and major commitment problemscan prevent easy dissolution of a movement once it has started (?; Fearon and Laitin 2007). Omitting alagged dependent variable when it should be there in an incidence regression tends to bias sharply upwardsthe estimated coefficients for any covariates that are positively serially correlated.

4

exogenous variation in rainfall in subSaharan Africa to estimate the effect of changes in income oncivil conflict propensity (see also Bruckner and Ciccone (2010b), who examine the same data andreach a quite different conclusion). And there are a few papers that use variation in internationalcommodity prices in a similar fashion (Besley and Persson 2009; Bruckner and Ciccone 2010a).But even these papers can be of limited value for understanding why an effect is observed – what isthe causal mechanism connecting changes in income to civil war propensity? – and thus whetherand how it might generalize. For these reasons, arguments about causes of civil conflict havegenerally taken the form of attempts to make sense of the complicated pattern of associationsobserved in the cross-national panel data, often informed by theoretical arguments and additionalevidence from cases.

In the next two sections, I review the major findings on correlates of civil war and conflict on-set, reproducing and revisiting these using the conflict codings of the Uppsala Armed ConflictDatabase. In section 5, I consider in more detail arguments about causes of civil conflict based onthese findings.

3 The conflict data and global trends

3.1 The ACD conflict measure

The UCDP/PRIO Armed Conflict Database (ACD) codes for each country and year since 1945whether a violent conflict occurred between a named, non-state armed group and governmentforces that directly killed at least 25 people.5 I work with a version of the data used in the prepara-tion of the 2011 World Development Report that has (rough) estimates of annual battle deaths foreach conflict. Following the WDR’s categorization scheme, we will distinguish between “major”conflicts, or civil wars, that are estimated to have killed at least 1000 over the whole episode ofconflict; “medium” conflicts estimated to have killed between 500 and 999; “small” conflicts esti-mated to have killed between 250 and 499; and “minor” conflicts estimated to have killed between25 and 249. “Killed” in all cases is intended to refer to battle deaths rather than “indirect” deathsdue to starvation, deprivation from medical services, and so on.

Two features of the ACD conflict data should be noted before we present some descriptive statistics.First, the ACD does not commit to any particular scheme for identifying episodes of civil war.That is, the raw data simply records whether there is enough fighting going on in the country year(and other conditions are satisfied on the nature of the fighting) to qualify for inclusion. So it isdifficult to know how to use these data to produce a list of distinct civil wars or conflicts, which isunfortunately just what we need if we want to study determinants of civil war onset or duration. For

5http://www.prio.no/CSCW/Datasets/Armed-Conflict/UCDP-PRIO/

5

that we need to know when a conflict started and when it ended. In many cases this is intuitivelyclear, but in many others it is not. For example, some low-level conflicts flit above and below the25-dead threshold for a period of many years. Is that one conflict, or many?

While this is not my preferred way of approaching the problem, here I will follow the suggestedpractice for WDR 2011 (which is also employed by many other scholars using the ACD). We willconsider a new civil war or conflict to have started if it is preceded by at least two years of peacebetween the present adversaries. Thus, if a ‘1’ represents a year with a conflict that killed at least25 in a country, then a sequence like . . . , 0, 0, 1, 1, 0, 1, 1, 1, 0, 0, . . . would be coded as one civilwar episode despite the one year break in fighting above the 25 dead threshold. By contrast, asequence like . . . 0, 0, 1, 1, 0, 0, 1, 1, 0, 0 . . . would be coded as having two civil wars onsets. Anadvantage of this approach is that it avoids potentially more subjective or difficult judgments aboutwhether a break in fighting is an “end” or a merely a pause in a continuing war. A disadvantage isthat it tends to render long-running, relatively low level conflicts as many civil wars. For example,by this coding the most onset-prone countries since 1945 are India with 14, Burma with 11, andEthiopia with 9 civil wars.6 Does it seem right to think of India as having had 14 distinct civil warssince independence?

A second consequential feature of the ACD coding rules concerns the treatment of multiple con-flicts within a country. For example, if two rebel groups are fighting the government at the sametime, is that two civil wars or one civil war? At one extreme one could argue for making the thingto be explained whether there is any war (or conflict) that starts in a country in a given year whenthe country doesn’t already have a war occurring – thus ruling out the possibility of multiple civilwars in one country. The ACD approach comes closer to the opposite extreme of coding a new civilwar every time there is a conflict with a new rebel group that meets the threshold requirements.ACD codes conflicts as being of two types, “government” where the rebel group aims to capturethe central government, and “territorial” where the rebel group aims to secede or win increasedautonomy for a specific region.7 As a result, every time a violent rebel group appears that advo-cates for a particular region or part of the country, the country gets a new conflict in the ACD data.This leads to coding of large numbers of distinct “civil wars” in countries like Burma, Ethiopia,and India, where multiple, often very small territorial rebel groups have operated, and often in thecontext of what is arguably a larger civil war.

By contrast, if one rebel group defeats the central government and then some new rebel groupimmediately arises to contest its control – think of Afghanistan or Somalia in 1991 – then for ACDthere is no new civil war since the fighting is still about taking over the government. An alternativeapproach codes a new civil war when one or both of the main combatants has changed (e.g., Fearon

6If one considers all conflict, minor and major, these numbers are 21, 19, and 13 onsets, respectively!7This leads to some odd codings for cases where there is ambiguity or variation over time in announced

objectives. For example, Anya Nya in Sudan is coded as territorial, but both the SPLM and Darfur are codedas “government.” Conflicts between Israel and Palestinian groups are coded as territorial.

6

and Laitin 2003, COW). Both approaches seem defensible.

As it happens, many of the cross-sectional patterns are fairly robust to different civil war codings.The main difference that results from using an ACD-based series versus other common alterna-tives8 is that the ACD approach increases the number of civil war onsets in some highly ethnicallyfractionalized countries that have had long-running internal conflict. And not just any ethnicallyfractionalized countries, but in particular ones where there is a predominant ethnic group that con-trols the center so there is not much chance that ethnic minorities at the periphery can take poweror play a significant role in coalition politics. As noted, the ACD approach to coding multiplecivil wars leads to large numbers of onsets for Ethiopia, Burma, and India, almost entirely due todistinguishing conflicts between the state and multiple regional minorities as distinct civil wars.9

The effect in conflict regressions is to increase the strength of association between onset and var-ious measures of ethnic fractionalization. These tend not to be significantly related in other civilwar lists (Collier and Hoeffler 2004; Fearon and Laitin 2003), but they are often significant inACD-based analyses, especially when researchers include minor-level conflicts.

To foreshadow the results on ethnic diversity and civil war risk discussed more below, the upshot isthat while more ethnically diverse countries have not been much more likely to have civil war overthe whole post-war period, if they do have it, they are more likely to have conflicts with multiplerebel groups fighting in the name of diverse regional minorities.

3.2 Global and regional trends

Figure 1 below uses the ACD data as described above to display the number of civil wars (major)and all conflicts (major, medium, small, and minor) by year from 1946. Figure 2 is the same exceptthat it shows the percentage of countries with a civil war or any conflict by year.

The basic features of these graphs track with previous studies using other civil war and conflictlists. There was a steady increase in conflicts from the end of World War II to the early 1990s.Since then there has been something of a decline, although civil war prevalence remains quitehigh, with total conflicts in the 30s in 2008, in a bit more than 12% of 193 independent countries.One interesting and possibly novel observation from the data shown here is that (major) civil warshave continued to trend down in the last five or six years, but minor conflicts have jumped back

8Such as the Correlates of War-based coding used by Collier and Hoeffler (2004), Sambanis’ (2001), orFearon and Laitin’s (2003).

9For example, in India: Nagaland, 2 war onsets, 4 of all sizes; Mizoram, 1 war onset; Tripura, 2 majorwar onsets; Manipur, 3 major war onsets; Punjab, 1 major; Kashmir, 1; Assam and Bodoland, 1 major and1 minor each. Other countries with a similar ethnic configuration and many ACD onsets include Indonesia,Iran, and Pakistan.

7

up. This could augur a return of more major conflicts (since minor conflicts become major if theycontinue over time), or it could reflect a change in the distribution of conflict sizes.

Figure 3 presents the same data but broken down by regions, which sheds some light on the sourcesof the decline in total conflicts in the last 15 years. In Eastern Europe and the former Soviet Unionthe spate of early 90s conflicts quickly subsided, or “froze” in some locales. There has also beenstriking decline in Latin America, which is plausibly related to the end of the Cold War for mostcases. Elsewhere, in subSaharan Africa, North Africa/Middle East, and Asia, the total amount ofconflict has not changed much, although there is perhaps some evidence of gradual improvement,especially in Africa.

Figure 4 shows that despite the post-Cold War decline in the number and proportion of countrieswith civil conflict, the share of the world’s population living in conflict-affected countries hasremained fairly steady or even increased, to one-third in 2008. Of course this does not mean thatone third of the world’s population has been directly affected by organized violent conflicts, sincethey are mostly localized within countries. Still, it is a measure of prevalence and impact, andreflection of the fact that conflict is much more likely in larger countries (discussed below).

Figure 5 considers where conflict-affected countries are found in the global distribution of incomeover time. Notice that close to 80% of conflicts have occurred in countries with incomes belowthe global median over the whole period, a number that shows no strong trend. However, therehas been a fairly steady increase in the share of conflicts in the second quartile on income, whilethe share occurring in the poorest 25% has declined from almost two thirds in the early 60s toabout one third. Thus there has been some tendency for conflict to become more common amongmiddle- and lower-middle income countries.

4 Correlates of civil war onset

Table 1 reports the results of logit models with civil war or conflict onset as the dependent vari-able, using as covariates things that previous studies (and in particular Fearon and Laitin (2003))found to be substantively and significantly related. The first model is for ACD civil wars, “major”conflicts that reached 1000 dead or more. The second uses an up-to-date version of Fearon andLaitin’s civil war list, and shows that the results are almost identical, with the partial exception ofethnic fractionalization (ELF) and the different sign on “prior war,” which marks whether therewas already conflict occurring in the country. In the third model the dependent variable is theonset of any level of ACD conflict, from minor to major. Results are again quite similar, althoughmost coefficients shrink a bit towards zero except for ethnic fractionalization, which has a largerapparent “effect” when we consider both minor and major conflicts.

Model 4 is Model 1 but estimated with country and five-year-period fixed effects. The factors

8

that vary a lot within countries over time have similar effect estimates; those that do not, likeincome and population, change markedly. This already suggests that these variables may appear to“matter” in the cross-sectional analysis due to omitted variables rather than a direct causal effect(to be discussed more below).

The variables are:

• log of per capita income in the previous country year, in 2005 U.S. dollars, using Penn WorldTables 6.3 data extended where possible and necessary by World Bank growth rates.10

• Log of country population in the previous country year.

• Log of the percentage of mountainous terrain in the country (plus one), as judged by geogra-pher A.G. Gerard.

• oil, which marks whether the country is a major oil producer, coded by whether one third ormore of its GDP comes from natural resources (based on World Bank data).

• new state, which marks whether the country is in its first two years of political independence.

• political instability, which marks whether, in the previous country year (t− 2 to t− 1), therewas any change in the Polity 2 score. The Polity score is a measure of democracy that runsfrom -10 (extreme autocracy) to 10 (full democracy).

• anocracy, which marks whether the country’s Polity 2 score -5 and 5 in the previous year.This is a measure of partial, or weak democracy.

• ELF is a measure of the ethnic fractionalization of the country, based on estimates of ethnicgroup populations from a Soviet ethnographic atlas complied in the early 1960s, and updatedfor some newer countries (Fearon and Laitin 2003). It can be interpreted as the probabilitythat two randomly drawn individuals from the country are from different ethnic groups.

• religious fractionalization is a similar measure of religious diversity (Fearon and Laitin2003).

10Listwise deletion due to missing income data is a big problem for civil conflict regressions, because it isnot missing at random – civil war countries are more likely to have missing income data. The data used hereis a relatively complete set of estimates, drawing primarily on PWT6.3 but using World Bank and Maddisonestimates for some missing years and countries

9

Table 1: Correlates of civil war onset, 1946-2008Model 1 2 3 4DV ACD civil wars FL civil wars all ACD conflicts ACD civil warslog(gdpt−1) −0.425∗∗ −0.404∗∗∗ −0.351∗∗∗ 0.038

(0.129) (0.111) (0.099) (0.289)log(popt−1) 0.339∗∗∗ 0.313∗∗∗ 0.238∗∗∗ 0.173

(0.072) (0.066) (0.056) (0.559)log(% mountains) 0.235∗∗ 0.186∗ 0.151∗

(0.086) (0.072) (0.062)oil producer 0.507∗ 0.698∗∗ 0.715∗∗ 0.631

(0.248) (0.242) (0.219) (0.549)new state 1.820∗∗∗ 1.913∗∗∗ 1.336∗∗∗ 1.418∗∗∗

(0.340) (0.307) (0.292) (−.403)pol instabilityt−1 0.729∗∗∗ 0.746∗∗∗ 0.466∗∗ 0.784∗∗∗

(0.190) (0.205) (0.170) (0.235)anocracyt−1 0.471∗ 0.482∗ 0.355∗ 0.608∗

(0.206) (0.199) (0.175) (0.262)democracyt−1 0.108 −0.466 0.080 0.152

(0.272) (0.329) (0.215) 0.384ELF 1.045∗∗ 0.521 1.153∗∗∗

(0.370) (0.344) (0.287)relig frac 0.046 −0.176 −0.228

(0.610) (0.499) (0.429)prior war −0.011 −0.548∗∗ 0.204 −1.520∗∗∗

(0.274) (0.185) (0.228) (0.275)constant −5.379∗∗∗ −4.795∗∗∗ −4.001∗∗∗

(1.279) (1.080) (0.982)N 7929 7985 7873 3517

Fixed effectsa No No No YesStandard errors clustered by country in parentheses† significant at p < .10; ∗p < .05; ∗∗p < .01; ∗∗∗p < .001a By country and five-year-periods.

4.1 “Effect” magnitudes

Table 2 provides estimates of the substantive magnitude of effect estimates in Table 1, Model 1.The baseline country in this example is a stable, non-oil producing autocracy with no civil warin progress, with the median income ($4,100), population (8.3 million), and “mountainousness”(9%). Such a country had about a .72% chance of civil war outbreak in any given year, whichtranslates to 3.6% over five years and 7% over ten. These small numbers should remind us that

10

new civil wars do not break out that often – with the ACD civil wars, the rate is about 1.7% of allcountry years from 1946 to 2008, and about 2.5 new civil wars per year.

The “relative risk” column shows how the indicated change affects the odds of civil war onset inthe next year. For example, moving from the 75th to the 25th percentile on per capita incomeis associated with a 2.11-fold increase in the annual odds of onset. Thus, the relative risk scoreprovides a rough way to compare the magnitude of the associations across covariates.

By this measure, the most striking pattern is that newly independent states have a much higherrisk of civil war onset than other states – more than six times greater, using the ACD data. Thefact that the effect estimate is similar in the fixed-effects model shows that this is not an artifactof states that became independent since 1945 having higher average conflict risk than that of olderstates; rather, even within former colonies, the first two years are the most dangerous. I discuss theimplications and interpretation of this observation in section 5 below.

After “new state,” several factors have roughly similar estimated associations (using the 25 to75th percentile comparison, and 0-1 for the dichotomous variables). As noted, the annual odds ofconflict outbreak slightly more than double comparing countries at the 75th percentile on incometo the 25th. Similarly, any change in governing arrangements that attracts the attention of Politycoders (whether in a more democratic or less democratic direction) in one year doubles the riskof civil war onset in the next year. Slightly smaller associations are observed for moving fromthe 25th to the 75th percentile on population size, mountainousness, and ethnic fractionalization.Anocracy (partial democracy) and having 1/3 or more of GDP from oil production is associatedwith 66% greater annual odds of civil war outbreak.

4.2 Other country characteristics of potential interest

4.2.1 Population growth, land pressure, and “youth bulges”

So-called “neo-Malthusians” fear that rapid population growth, or population per hectare of arableland, causes violent conflict by increasing competition for resources.11 A more specific argumentin this vein is that what matters is the ratio of young males to the rest of the population (or to justthe adult population) (Huntington 1996; Urdal 2006). Fearon and Laitin (2003) and Urdal (2005)found little support for a strong relationship between either population growth rate or populationdensity on arable land, but Urdal (2006) finds evidence in favor of a relationship between theshare of the adult population under 25 and conflict risk. Fearon and Laitin (2003) and Collier andHoeffler (2004) had found no relationship for the (very similar) measure of share of total populationbetween 15 and 24.

11Huntington (1996) sometimes stressed population pressures. See also Homer-Dixon (2001).

11

Table 2: Magnitude of “effect” estimates, Model 1variable level pctile % chance onset % chance onset % chance onset relative risk

in 1 year over 5 years over 10 yearsbaselinea 50 0.72 3.56 6.98

gdp/cap $1,665 25 1.06 5.17 10.07 2.11$9,612 75 0.50 2.50 4.93

population 3.3m 25 0.53 2.62 5.1721.5m 75 0.99 4.86 9.49 1.88

% mountains 1.7 25 0.53 2.60 5.1427.2 75 0.91 4.47 8.74 1.74

ELF .12 25 0.57 2.81 5.54.66 75 1.00 4.90 9.56 1.77

baselinea 50 0.72 3.56 6.98

new state 4.29 6.17instability 1.48 7.20 13.89 2.07oil 1.19 5.82 11.30 1.66anocracy 1.15 5.62 10.93 1.60aan autocracy that in the previous year was stable, at peace, a non-oil producer, with median income($4,117), population (8.3 mill.), mountains (9.4%), ELF (.35), and relig. fractionalization.

Using our ACD-based measure for onset of civil war, I find that lagged population growth rates arepositively associated with civil war onset risk, but the statistical significance varies a lot with whatother variables are in the model. When added to Model 1 in Table 1, lagged population growth rategets an estimated coefficient of about .10 with a standard error of .07 (p = .16).12 Smaller standarderrors and slightly higher estimated coefficients can be found if one drops various other variablesfrom the model. Population growth rates are moderately well correlated with several other vari-ables, such income (negative), oil producer (positive), anocracy (positive), democracy (negative),and ethnic diversity (positive). Overall, it tends to lose out in the “battle of the covariances” withthese other variables, but not by much.

12I drop observations that have a population growth rate above the 99th percentile or below the 1st per-centile, because changes in state form and some bizarre data points make for some extreme and influentialoutliers. If these are left in, results are highly erratic depending on specification.

12

Note also that results for population growth are much weaker if one uses “all civil conflict” asthe dependent variable instead of major ACD civil wars, or if one uses the Fearon/Laitin civil waronset indicator. In addition, if one constructs the variable as the average growth rate for the fiveyears previous to the current year – which gets rid of a lot of noise in the annual measure – thecoefficient estimate diminishes some and gets a much larger p value. Overall, it is hard to knowwhat to make of these correlational results – rapid population growth could increase the annualrisk of civil war onset on average, but it is clear that there is no strong and persistent associationwhen one controls for other plausible determinants.

Results for a measure of arable land per capita (from the World Development Indicators) are sim-ilar, though perhaps a bit weaker. Arable land (in hectares) per capita is negatively related to civilwar onset when added to Model 1, with a p value of .07, though this is fairly sensitive to what othervariables are in the model. Substantively, moving from the 25th to the 75th percentile associateswith about a 20% annual reduction in civil war onset odds, which is not particularly large. More-over, the measure is highly skewed – Canada, Australia, New Zealand, and Niger have extremelyhigh values – and the strength of the association weakens dramatically if we use log of arable landper capita instead. Overall, this measure of resource scarcity, if that is what it really is, shows littleconsistent association with higher civil war risk.

Urdal (2006) found that although “youth bulge” as measured by the share of 15-to-24 year oldsin the total population is not related to onset risk – which is what Fearon and Laitin (2003) andCollier and Hoeffler (2004) also found with this measure – “youth bulge” as measured by the shareof 15-to-24 year olds in the adult population was related to civil war onset in his data. He arguedthat “from a theoretical perspective” (p. 615) the latter indicator is to be preferred, but it is notclear what the theoretical argument is (to me, anyway). The two measures are highly correlated(r = .85), which increases the difficulty in saying what is different about them.

For the reanalysis undertaken here, I began by importing Urdal’s measure, which has “youth bulge”data for 1950 to 2000. Added to Model 1 the estimated coefficient is positive but substantivelysmall and not close to statistically significant (p = .22). The estimated coefficient for incomeweakens slightly but remains significant at p = .05; otherwise nothing changes much. It thusappears that “youth bulge” as Urdal formulates it performs less well as a predictor when the de-pendent variable is based on this ACD civil war list. It does slightly better, and can be “significant”at .10 under certain specifications, if we use Fearon and Laitin’s civil war onset variable.

I then updated the population estimates using the same source Urdal employed (U.N. PopulationDivision data), which allowed an extension of the data to 2009. I also used the data for males be-tween the ages of 15 and 25, rather than both males and females (although this almost surely wouldmake no difference at all). I find that neither young males as a share of adult or total population issignificantly related to civil war onset in these data. Whether added to the specification in Model1 or paired with income and other combinations of variables, it takes a positive coefficient but theestimates are not close to being statistically or substantively significant. Once again, young males

13

as a share of adult population (but not young males as a share of total population) performs muchbetter as a predictor if the dependent variable is from Fearon and Laitin’s coding of civil war.

What to conclude about “youth bulge” and civil war onset? As Fearon and Laitin (2003) noted,“youth bulge” is strongly negatively correlated with per capita income, which makes it difficultto get stable or sharp estimates of the partial correlation with civil war onset controlling for othervariables. There is a tendency for income per capita to “trump” youth bulge in these data, atendency that is very strong when youth bulge is measured as youth over total population andconsiderably weaker when it is measured as youth over adult population. Contrary to Urdal’s view,I find it difficult to come up with a plausible or clear theoretical rationale for why the results ofthese two different measures should be particularly different. So I am inclined to think that theevidence is not very good that population structure explains a big part of why poor countries beingmore civil war prone, rather than something else about poor countries explaining why countrieswith young populations happen to be more civil war prone.

4.2.2 Vertical and horizontal income inequality

The bivariate relationship between income inequality and civil war onset in these data is actuallynegative (more inequality, lower odds of conflict), although not statistically significant. The neg-ative sign persists when we add the covariates in Table 1, Model 1, or subsets of them, and isactually close to significant (p = .056) in the full model. This is not a matter of inequality pickingup regional “effects,” as the estimate gets even more negative and more significant when regionaldummies are added.13

Not too much should be made of this in the absence of better inequality data, and a more theoret-ically informed model specification. Still, it is interesting: Contrary to some long-standing claimsabout the causes of civil conflict, not only is there no apparent positive correlation between incomeinequality and conflict, but if anything across countries those with more equal income distributionshave been marginally more conflict prone.

Some have argued that a more relevant form of income inequality is “horizontal,” meaning acrossgroups within a country as defined by ethnicity or religion (Stewart 2002). Good measures ofinequality across groups are hard to find and construct, however. Early efforts to use the Minoritiesat Risk data (Gurr 199x) to examine the relationship between economic disadvantage of “minoritiesat risk” and propensity to rebel found inconsistent or no evidence (Gurr 199x, Moore 199x, Fearonand Laitin 1999). More recently there have been some efforts to use the Demographic and HealthSurveys (http://www.measuredhs.com/) to construct measures of the relative economic standing of

13I have used the WIDER inequality measure, which is based on an updated version of Deininger andSquire. The variable is constructed as the average of all observations within each country. Results aresimilar if we interpolate.

14

different ethnic groups in subSaharan Africa (Østby 2008; Condra 2009). Results are inconsistent.Østby finds a positive relationship between horizontal inequalities and conflict, while Condra doesnot.

Cederman and Girardin (2007) examine another possible interpretation of horizontal inequality,in the idea that ethnic groups whose members are excluded from political office will be morelikely to rebel. They construct a measure that takes high values when the population share of the“ethnic groups in power” (EGIP) is small and the population share of “marginalized ethnic groups”(MEG) is high. They find that this measure is associated with higher onset probabilities, using datafor Eurasia and North Africa (Latin America and Africa were not coded).

Fearon, Laitin and Kasara (2007) examined their data and found that the results are completelydriven by the four observations where the coded EGIP is a minority. Doubtful about what thecoding rules were for EGIPs, Fearon, Kasara, and Laitin coded instead the ethnicity of the rulerfor all country years in all regions. They find a positive but statistically insignificant relationshipbetween rule by a member of an ethnic minority and civil war onset.14 In their analysis (which,like Cederman and Girardin, used the Fearon and Laitin (2003) model and civil war list), ethnicminority rule has no association with civil war onset at all in subSaharan Africa or Latin American,but some signs of a positive relationship in the rest of the world. However, even in Eurasia ethnicminority rule is quite rare, so it is hard to establish any real pattern.15

More recently, Wimmer, Cederman and Min (2009) have analyzed the results of a more systematiceffort they undertook to code EGIPs and MEGs. They used some process of expert surveys andconsultations to assess first whether ethnicity was “politically relevant,” in a country year, meaningthat their coders perceived discrimination or that they said there were politicians mobilizing basedon ethnic appeals. This leads to a number of ethnically diverse countries – such as Burkina Faso,Tanzania, and Papua New Guinea – being coded as having no “ethnopolitical groups” at all, andthus no possibility of ethnic exclusion.16 Second, for each “ethnopolitical group” in each country-year their experts coded whether the group has “monopoly power” or is “dominant,” or if it is an“excluded group” that has “regional autonomy,” is “powerless,” or is ”discriminated” against. They

14Related variables, like the size of the leader’s ethnic group, or the ratio to the second largest group,perform worse than a simple dummy for ethnic minority rule.

15There are also (as usual) concerns about endogeneity. Cross-sectional analysis could understate theconflict-generating effects of ethnic minority rule, if it is more likely in precisely those places where it isviewed as tolerable. On the other hand, ethnic minorities may in some cases work especially hard to attainand hold onto power where they would be highly threatened if they lost power (e.g., Syria, Iraq). Fixedeffects models that try to control for country-specific characteristics like this yield unstable, though positive,estimates, because there are so few cases to go on.

16In all three cases politicians have mobilized along ethnic lines (in Papua New Guinea, this is all thereis), but the broader concern is whether this is coding on the dependent variable or not.

15

find that the log of the proportion of what they call “excluded” ethnic groups in all “ethnopolitical”groups is robustly associated with civil war onset, and even more strongly related to the onset ofethnic civil wars.

Importing Wimmer et al.’s country-level codings into the data set considered here, I find that the logof the lagged share of “excluded groups” is positively related to civil war onset odds when addedto Model 1.17 The coefficient is .159 with a p value of .07; substantively this implies that movingfrom a country with no excluded groups to one where 23% of the population of ethnopoliticalgroups are excluded (25th to 75th percentiles) associates with 66% greater annual odds of civilwar outbreak. The unlogged version of the variable is much more weakly related. With fixedeffects the estimated coefficient is essentially zero, which suggests that “exclusion” is picking upenduring characteristics of countries more than that variation in policies over time within countriespredicts onset.

One could argue that the more relevant measure should be the share of population that is “dis-criminated against,” as “excluded” includes groups that could be content with their regional auton-omy arrangements or are not particularly unhappy with being “powerless” (whatever exactly thismeans). Using as a predictor the share of ethnopolitical groups that are “discriminated against”by Wimmer et al.s’ codings yields similar results to “excluded,” though possibly stronger. In sum,Wimmer et al.’s study suggests that countries that raters judge to have bad ethnic relations anddiscrimination against relatively larger groups are more civil war prone.

Two caveats about these findings should be noted. The first mirrors a similar issue for the expert-ratings based governance indicators that are examined below. Namely, these measures of politicalexclusion and discrimination are based on the subjective judgements of diverse coders, trying tocode somewhat impressionistic things. Countries where there has been no ethnic conflict and whereethnic relations have been calm are for that reason judged to have a low value on “exclusion” –thus, the dependent variable determines the coding of the independent variable. More generally,one can reasonably worry that a coder’s knowledge that there was an ethnic conflict in a countryincreases the probability that he or she judges that, earlier on, groups were discriminated against orpolitically excluded. And codings of discrimination at time t may be based on earlier experiencesof conflict, again making it hard to sort out causes and effects.

The second concern is that when we include a variable that tries to measure the extent of thepopulation that is “excluded” or discriminated against by government policy, we are now runninga policy regression. That is, we have put a variable that is a direct policy choice on the right-hand-side. Income per capita, and even ethnolinguistic fractionalization, can be viewed as the resultsof policy choices, as well. But they are produced by policy choices over longer periods of time,and arguably much more indirectly, than a variable trying to measure current government policywith respect to an ethnic minority. If we have concerns about the endogeneity of income and ELF,

17I added .01 to avoid log of zero; not immediately clear to me what Wimmer et al. do.

16

which we should, then we must have them far more strongly about a direct policy.18

It is very important to understand that the endogeneity of the policy choice might lead us to over- orunderestimate the average effect that introducing a more (or less) exclusionary policy might havein a typical country. For example, if governments tend to calibrate levels of exclusion to what theycan get away with, then estimates from panel data may understate what would be the causal impactof arbitrarily switching to more exclusionary policies (Fearon and Laitin 2010). Alternatively, tothe extent that exclusionary policies are themselves driven by fear of rebellion for other reasons(such as rebel opportunity), then panel data estimates will tend to overestimate the causal impactof exclusion or discrimination.

Buhaug, Cederman and Rød (2008) use the same core data as Wimmer et al on ethnic power rela-tions to examine determinants of ethnic conflict at the group (rather than the country) level. Theyfind that Eurasian ethnic groups that are larger, live farther from the capital, and in more moun-tainous terrain are more likely to be involved in ACD minor or major level conflicts.19 Curiously,they end up interpreting this as supporting an “exclusion” versus an “opportunity” explanation,although their three main variables (size, distance, and terrain) are usually considered to be at leastas plausible as measures of capability to sustain rebellion as of motivation to rebel.20

4.2.3 Civil liberties, human rights abuses

Are countries whose governments abuse human rights and restrict civil liberties more prone to civilwar onset? This certainly seems plausible and likely on its face. The question is motivated by anintuition similar to that behind studies of horizontal inequality, though here the focus in on whetherrepressive or restrictive government policies favor war even if they are not necessarily directed atany particular ethnic or religious group.

Once again, however, we need to be careful about the interpretation of results, since abuse of humanrights and restriction of civil liberties are policy choices and thus almost surely are endogenous toother causes of conflict. Even worse, there is a major danger that these indicators may simply pickup onset of civil conflict before it happens to get coded by ACD or others. Is this governmentabuse that is causing a conflict, or government abuse that is already part of a conflict we are tryingto explain? Still, it may be interesting to know how strong is the correlation between existingmeasures and civil war onset.

18See Rodrik (2005) for a nice discussion of this issue in the context of studies of economic growth.19Condra (2009) reports similar findings for groups in Africa.20All the groups in the sample are what they call “marginalized ethnic groups,” so that there is no oppor-

tunity to estimate the effect of “marginalization” by this design.

17

Freedom House provides a 1-to-7 scale of government observance of civil liberties for a largenumber of countries since 1972. Although procedures have varied over the years, for the most partthe scale is constructed from expert responses to 15 questions grouped into four areas, concerning“Freedom of Expression and Belief, Associational and Organizational Rights, Rule of Law, andPersonal Autonomy and Individual Rights.” Higher values on the scale indicate fewer civil libertiesfor citizens of the country. One problem with this measure for our purposes is that while govern-ment behavior is clearly the focus, the measure is not in principle limited to government behavior.For example, a country may also be judged to have worse civil liberties if it has “groups opposedto the state [that] engage in political terror that undermines other freedoms.” Thus to some extentthe scale may incorporate a measure of civil conflict.

The civil liberties measure proves to be highly correlated with other measures of democracy – forinstance, r = −.85 with Polity – which we have already seen is not a significant predictor of civilwar or conflict onset. When added to Model 1, the coefficient is positive (worse civil liberties,higher conflict risk) but close to zero (p = .66). This is true whether or not we include the vari-ables for anocracy and democracy as measured by Polity, and if we look at simpler specificationsprovided income is included. Basically, observance of civil liberties behaves similar to democracyindicators. Indeed, if we add a squared term (and drop the Polity measures), we find evidence ofthe inverted U. Other things equal, civil war risk is highest for countries with a Freedom Housecivil liberties score of 5 out of 7, which they describe as countries “with a combination of high ormedium scores for some questions and low or very low scores on other questions.”

Since the Freedom House civil liberties measure may be coded in part for civil conflict, one concernmay be that including lagged civil liberties is like including an indicator for whether there wasconflict in the previous period. I also tried running Model 1 (and variants) only for cases in whichthere was no civil war in the prior period. This leads to the coefficient on civil liberties doubling(or more), but it still comes up short of statistical significance at p = .10, usually.

A potentially more focused measure of government abusiveness is the Political Terror Scale, anannual 1-to-5 index based on coding of Amnesty International and State Department human rightsreports from 1976 to 2008 (Gibney, Cornett and Wood 2008). Mark Gibney writes that “Codersare instructed not to turn a blind eye towards violence by non-state actors, but that their primarygoal is to measure levels of violence by the state.”21

21There are two indices, one based on Amnesty reports and the other based on State Department reports.They are well correlated at .8. As is common, I use the average of the two. The description of scale levelsis: “(Level 5) Terror has expanded to the whole population. The leaders of these societies place no limitson the means or thoroughness with which they pursue personal or ideological goals. (Level 4) Civil andpolitical rights violations have expanded to large numbers of the population. Murders, disappearances, andtorture are a common part of life. In spite of its generality, on this level terror affects those who interestthemselves in politics or ideas. (Level 3) There is extensive political imprisonment, or a recent historyof such imprisonment. Execution or other political murders and brutality may be common. Unlimited

18

The main difficulty that this measure poses is that it is hard to separate out government abusivenessthat is part of a civil war from government abusiveness that causes a civil war onset. Of the 774country years coded as 4 or 5 on the PTS (the highest two levels of human rights violations), fully65% are ACD civil war years, and 71% are Fearon/Laitin civil war years. In other words, most ofthe worst human rights abuses, according to these data, are carried out by governments during civilwars. The problem is then that if lagged human rights performance predicts onset in a regressionmodel, we don’t know how much this is because the onset of civil war is coded with error, and howmuch it is because government abusiveness caused rebellion. A second difficulty is that the laggedPTS variable can act as an indicator for prior civil war(s) in progress.

Table 3 reports the results of adding one- and five-year lags of the PTS measure to Model 1, bothfor the full set of observations and for only those country years with no civil war occurring in theprevious year. The scale appears to be very strongly related to subsequent civil war onset, both interms of a substantively large estimated coefficient and statistical significance. The estimate forPTS lagged five years is less than half of the estimate for PTS lagged one year (Model 2 versusModel 1), which suggests that there may indeed be a big problem with the one-year lag pickingup conflicts that have already started (this may also happen for the five-year lag, but hopefully lessso).22

Substantively, the estimated association for the one-year-lag is simply enormous: the .79 estimatein Model 1 means that each step up the Political Terror Scale associates with a 2.2 times increasein the odds of civil war onset in the next year. This implies that country years with a 5 on PTS haveabout a 50 times greater odds of onset in the next year than those with 1! If we use the smallerestimate for the five-year lag, each step associates with an 33% increase in odds, and the differencebetween 5 and 1 is a factor of 5.3, which is comparable to the effect of “new state.”

The models in Table 3 implicitly treat PTS as a linear scale, estimating an average “effect” ofmoving from 1 to 2, 2 to 3, and so on. If we dummy out each of the five levels, we find thatwe cannot estimate a logit model because there were zero ACD civil war onsets for states at PTSlevel 1.23 (Level 1 states are described by Gibney as countries “under a secure rule of law, [where]people are not imprisoned for their views, and torture is rare or exceptional.”)

An alternative then is to use a linear probability model (with robust standard errors to try to address

detention, with or without a trial, for political views is accepted. (Level 2) There is a limited amount ofimprisonment for nonviolent political activity. However, few persons are affected, torture and beatingsare exceptional. Political murder is rare. (Level 1) Countries under a secure rule of law, people are notimprisoned for their views, and torture is rare or exceptional. Political murders are extremely rare.”

22Coefficients for several of the other variables weaken considerably here, mainly because the sample isnow for 1976 (or 1981 in columns 2 and 4) to 2008, but also because the PTS variable picks up some of theimpact of anocracy, political instability, and mountains.

23This means that the odds ratio of PTS > 1 versus PTS = 1 is infinity, so the logit model doesn’t converge.

19

the heteroscedasticity issue). These estimates are shown in Table 4, and although we should be abit cautious about the standard errors, the results shed considerable light on the patterns in thedata concerning conflict and human rights performance. Since this is now ordinary least squares,the coefficients for the PTS dummies give the difference in the probability of onset for the givenlevel versus level 1. So, for example, using a one year lag of the PTS score (the first column ofresults), on average level 2 countries had only a .00019 greater probability of onset in the nextyear as compared to level 1 countries. This is hardly anything. Going to level 3 associates with anincrease above the level 1 baseline of about one third of one percent on average, a difference that isnot statistically significant. In all of these models, we see that the big jump is at level 4, for whichthe estimated probability of onset in the next year (or five years later in Models 2 and 4) rangesfrom about .02 to .05 greater than a level 1 country.

According to Gibney, level 4 refers to countries in which “Civil and political rights violations haveexpanded to large numbers of the population. Murders, disappearances, and torture are a commonpart of life. In spite of its generality, on this level terror affects those who interest themselves inpolitics or ideas.” As compared to levels 1-3, this is likely to pick up mainly countries with actualor incipient civil war.

Table 5 lists country years for which ACD codes the start of a civil war, and the PTS scale was 4or 5 for the previous year, and there was no ACD civil war occurring in the country in the previousyear. There are a few cases where it is plausible to see the conflict arising in some significant wayfrom an oppressive prior dictatorship – for example, Cambodia 1978, Uganda 1978, Yugoslavia1998, and Indonesia 1999. But there are more cases where the lagged PTS coding is picking upcivil war activity that is for different reasons missed by ACD – for example, Mozambique 1977, ElSalvador 1979, Sri Lanka 1984, Rwanda 1997, Liberia 2000, Iraq 2004, DRC 2006, and Somalia2006.

Overall, the analysis suggests that very poor human rights performance is a very bad sign for agovernment: major civil conflict is then much more likely to begin, if it has not already started.

The causal, as opposed to diagnostic, role of government human rights abuses is less clear. Itis plausible that government abuses would often encourage support for rebellion, but it is alsopossible that in some or many cases rebellion would be even more likely without the repression– else why is the government being so repressive? This is another “policy regression” problem.Cross-national data of the sort examined here cannot help us much in sorting this out.

There is, however, a growing literature on a closely related problem – what is the impact of indis-criminate counterinsurgency strategies on support and success of rebel movements? Much of thisliterature consists of case studies that suggest that indiscriminate brutality by government forcesincreased local support for (already existing) rebel groups. More recent work tries to identify nat-ural experiments or to use fine-grained incident data from specific conflicts. Results so far are abit mixed. Lyall (2009) finds, somewhat surprisingly, that random artillery shelling by Russian

20

troops in Chechnya suppressed rebel activity. Lyall and Wilson (2009), who examine the effectof mechanized counterinsurgency on government success in a large sample of guerrilla wars andin a case study of Iraq, would seem to suggest the opposite, as does Condra and Shapiro’s (2010)incident study of Iraq.

Table 3: Human rights abuses and civil war onset, 1976-2008 (or less)1-year lag 5-year lag 1-year lag 5-year lag

no prior war no prior warprior war −0.747∗ −0.106

(0.358) (0.400)log(gdpt−1) −0.420∗ −0.358† −0.500∗∗ −0.412†

(0.175) (0.206) (0.185) (0.214)log(popt−1) 0.301∗ 0.379∗∗ 0.036 0.062

(0.151) (0.142) (0.129) (0.126)log(% mountains) 0.155 0.150 0.151 0.193

(0.113) (0.120) (0.118) (0.127)oil producer 0.166 0.071 0.328 0.098

(0.384) (0.431) (0.447) (0.569)pol instabilityt−1 0.205 0.193 0.537 0.417

(0.274) (0.329) (0.345) (0.394)anocracyt−1 0.298 0.555† 0.555 0.991∗

(0.286) (0.328) (0.342) (0.429)democracyt−1 0.107 0.022 −0.357 −0.401

(0.453) (0.440) (0.482) (0.580)ELF 1.177∗ 1.374∗ 0.562 0.748

(0.491) (0.549) (0.497) (0.619)

PTSt−1 0.793∗∗∗ 0.875∗∗∗

(0.177) (0.191)PTSt−5 0.332† 0.484∗∗

(0.192) (0.181)

constant −7.152∗∗∗ −7.282∗∗∗ −4.005∗ −4.186†

(1.585) (1.820) (1.977) (2.419)

N 4597 3974 3854 3326

DV = ACD civil war onset. Standard errors clustered by country† significant at p < .10; ∗p < .05; ∗∗p < .01; ∗∗∗p < .001

21

Table 4: Human rights and civil war onset, linear probability modelPTS 1-year lag PTS 5-year lag PTS 1-year lag PTS 5-year lag

no prior war no prior warprior war −.01674 .00225

(.01025) (.01156)log(gdpt−1) −.00490∗ −.00377 −.00493∗ −.00367

(.00241) (.00274) (.00217) (.00258)log(popt−1) .00530 .00601† .00084 .00130

(.00337) (.00336) (.00164) (.00138)log(% mountains) .00089 .00099 .00106 .00138

(.00143) (.00149) (.00123) (.00134)oil producer .00304 .00253 .00476 .00177

(.00558) (.00581) (.00661) (.00641)pol instabilityt−1 .00347 .00262 .00961 .00706

(.00625) (.00725) (.00814) (.00821)anocracyt−1 .00530 .00894 .01367∗ .01911∗∗

(.00629) (.00661) (.00681) (.00728)democracyt−1 .00178 .00083 .00042 −.00042

(.00663) (.00644) (.00391) (.00409)ELF .02058† .02156† .00621 .00684

(.01111) (.01105) (.00847) (.00916)

PTS= 2 .00019 .00456 −.00002 .00367(.00336) (.00393) (.00270) (.00348)

PTS= 3 .00338 −.00341 .00281 −.00115(.00512) (.00541) (.00471) (.00471)

PTS= 4 .03665∗∗ .02553∗∗ .04565∗∗ .01950∗

(.01165) (.00805) (.01726) (.00959)PTS= 5 .05653∗∗ −.00050 .09077 .03406

(.01987) (.01694) (.05846) (.03287)

constant −.00979 −.02706 .03353 .01772(.02842) (.03095) (.02420) (.02812)

N 4597 3974 3854 3326R2 0.02759 0.02415 0.02900 0.01931

DV = ACD civil war onset. Cluster robust standard errors in parentheses† significant at p < .10; ∗p < .05; ∗∗p < .01; ∗∗∗p < .001

4.3 Political reform and civil war onset

The results in Table 1 show a strong relationship between changes in governing institutions, asmeasured by change in components of the Polity index, and an increased probability of civil war

22

Table 5: ACD onsets in countries that had PTSt−1 ≥ 4 and no ongoing warcountry onset year PTSt−1 PTSt−5

Mozambique 1977 4.0 .Cambodia 1978 5.0 .Uganda 1978 5.0 .El Salvador 1979 4.5 .Sri Lanka 1984 4.5 2.5Pakistan 1995 4.0 3.5DRC 1996 4.0 3.5Rwanda 1997 5.0 4.0Angola 1998 4.0 5.0Yugoslavia/Serbia 1998 4.0 4.0Indonesia 1999 4.0 4.0Russia 1999 4.5 4.5Liberia 2000 4.5 5.0Israel 2000 4.0 3.5Iraq 2004 5.0 5.0Pakistan 2004 4.0 4.5DRC 2006 4.5 4.5Somalia 2006 4.0 4.0

onset in subsequent years. The coefficient estimate actually increases when we add country fixedeffects, so it is not just a matter of more unstable countries having more civil wars. Rather, civil warsomewhat reliably follows institutional changes in the direction of greater democracy or greaterautocracy.

Does it matter whether the change is towards more democracy or autocracy? Previous work on civilwar onset finds no significant differences (e.g., Fearon and Laitin 2003), and that result carries overhere. If to Model 1, Table 1, we add dummy variables marking whether there was a one-or-greaterchange in Polity from year t − 2 to t − 1 in the democratic direction and the autocratic direction(so that the excluded category is no change), we find that both variables are significant at the .01level, with estimated coefficients of .82 for autocratizing change and .66 for democratizing change.The difference between them is not significant, though as a descriptive matter this indicates thatautocratizing change is slightly more prone to be followed by war than democratizing change.

The likelihood that democratic reform leads to civil conflict does not vary much across levels ofdevelopment. The risk is slightly lower as one moves up the income scale, but not significantlyso. Nor does it matter much what the initial level of democracy is. One might have expectedthat democratic reform would be more likely to end in violent conflict the more authoritarian thereforming state. But the opposite seems to hold, at least as judged by interacting the initial level of

23

democracy with whether there is Polity change in a democratic direction. It is of course possiblethat there are more complicated paths between reform efforts and civil war risk. For instance,reform might lead to authoritarian retrenchment and higher risk of rebellion for that reason (e.g.,Algeria in 1991, perhaps Thailand today).

4.4 Natural resources

other measures of oil, gems, primary commodity exports ...

5 Causes of civil war

Patterns of correlations like those presented above have been the basis for a new round of academicand policy-maker arguments about the causes of civil conflict that started after the end of the ColdWar.24 In early versions of their very influential paper “Greed and Grievance in Civil War,” Collierand Hoeffler (2004) interpreted the strong association they found between civil war onset andmeasures of dependence on primary commodity exports and low education levels – versus theweak association of onset with measures of ethnic diversity, income inequality, and democracy –as supporting the argument that rebel groups are primarily motivated by opportunities for profitrather than by a desire to right perceived wrongs. They suggested that aspiring “grievance-based”rebellions might face a more severe collective action problem than would “greed-based” or “loot-seeking” rebellions. In this initial formulation, the implicit assumption was that the causes of civilwar would be located in the motivations of rebel groups.

Later versions of the paper converged (at least on this point) with the interpretation in Fearon andLaitin (2003), who argued that “Surely ethnic antagonisms, nationalist sentiments, and grievancesoften motivate rebels and their supporters. But such broad factors are too common to distinguishthe cases where civil war breaks out” (p. 76). The idea is that grievances that could potentiallymotivate a rebellion are regrettably common (and reasonable) in much of the world, so that moreof the “action” in terms of explaining cross-national variation in civil war propensities is likely tobe found in variation in factors that affect the viability of, or opportunity for, rebellion.

Very few civil wars since 1945 have been, or have emerged out of, popular revolutions character-ized by mass protests and mass action – the French revolution model. Instead, the vast majority ofviolent civil conflicts in this period have been fought as guerrilla wars or militia-based conflicts,

24There is of course a long history of theorizing about civil war and revolutions before this, mainly on thebasis of particular cases (e.g., Moore 1966, Skocpol 1979), but including an earlier round of quantitativecross-sectional analysis (e.g., Hibbs 1973).

24

typically by small rebel groups that often number in the hundreds, especially in their early years.Fearon and Laitin suggested that “because insurgency can be successfully practiced by small num-bers of rebels under the right conditions, civil war may require only a small number with intensegrievances to get going” (p. 76).

The prevalence of small-scale guerrilla or militia-based wars does not rule out the possibility thatvariation in the level of broad social grievances across countries could be an important explanationfor civil war risk. In principle, it could be that social support from many sympathetic people isnecessary for even a small rebel band to operate successfully. In practice, however, if this is thecase it seems to apply mainly to rebellions in countries where the central state is relatively capable.Where states are less capable, rebel groups often appear to be able to operate without broad or deepsocial support, or they can coerce it.25

Collier and Hoeffler (2004) and Fearon and Laitin (2003) give mainly “opportunity” interpretationsfor the patterns of correlation described in the last section. For example, when a colony receivesindependence, its central government receives a negative shock to its capability to deter and fightrebels, if and when the colonial army leaves. Likewise, political instability may signal weaknessat the center, as may anocracy (partial democracy) (Hegre, Ellingsen, Gates and Gleditsch 2001).Large populations are suggested to make rebellion more feasible on average by making the gover-nance problem harder for the center (harder to develop reliable chains of principals and agents tomonitor what is going on at the local level, for example; India is much harder to govern than, say,the Maldives).

In their data Collier and Hoeffler found strong results for a variable measuring the share of primarycommodity exports in GDP. They interpreted this as an indicator of greater financing opportunitiesfor would-be rebel groups.

opportunity interpretation of covariates ...

possibility that we are not good at measuring grievance, or variation in grievances ...

But why is opportunity enough? Why aren’t all “opportunities” bought off by appropriate bar-gained solutions? Commitment problem explanations – governments can’t commit not to renegeon agreements after rebel groups have disbanded or demobilized. This leads to theoretical expla-nations for conflict onset that stress shocks to relative capability of state versus rebel groups: Forexample, loss of colonial power support, death or decline of an old autocrat, dramatic economicdecline, entry of third-party supporter of a rebel group.

25Cf. Weinstein (2007). Drawing on an exhaustive reading of “micro-level” accounts of particular civilwars, Kalyvas (2006) provides almost innumerable examples of how the relative local strength of rebeland government forces shapes local support for the rebels versus the government, more than the other wayaround.

25

Disposition variables like per capita income or population versus shocks. Time-invariant countrycharacteristics like low income make states more or less prone to conflict onset due to shocks (moreor less “resilient”).

Evidence on capability shocks ...

6 Governance indicators and civil conflict

As discussed above, several of the most striking cross-national patterns in civil war onset mightbe explained by an interpretation that puts “state capabilities” at the center of the story. In thisview, low per capita income is strongly related to conflict onset because it is a proxy for the centralstate’s capability to deter and suppress armed challengers, and possibly also to provide publicservices. A set of alternative interpretations argues that there is a direct effect of low income oncivil war propensity through some labor market channel. For instance, it is argued that povertymakes joining a rebel group relatively more attractive for young men.

This section uses three sets of governance indicators – the World Governance Indicators producedby Kaufmann, Kraay and Mastruzzi (2009), the International Country Risk Guide (ICRG) indica-tors, and the World Bank’s aggregate Country Policy Institutional Assessment (CPIA) – to try tobetter assess the “state capabilities” argument about civil war onset.

The ICRG series, which starts in 1984, is produced and sold by the company Political Risk Ser-vices; the variables are derived from expert surveys of business and political conditions in about140 countries. The WGI project began in the 1990s at the World Bank. Kaufmann and Kraayassembled a large set of expert-based governance ratings produced each year by think tanks, aca-demic research groups, NGOs, international organizations, and businesses, and divided them intosets that they argue correspond to six dimensions of governance: “government effectiveness,”“voice,” “political instability,” “rule of law,” “corruption,” and “regulatory quality.”26 They then

26Kaufmann, Kraay, and Mastruzzi (p. 6) describe these six areas as follows: “(1) Voice and Accountabil-ity (VA) capturing perceptions of the extent to which a country’s citizens are able to participate in selectingtheir government, as well as freedom of expression, freedom of association, and a free media. (2) PoliticalStability and Absence of Violence (PV) capturing perceptions of the likelihood that the government willbe destabilized or overthrown by unconstitutional or violent means, including politically-motivated violenceand terrorism. (3) Government Effectiveness (GE) capturing perceptions of the quality of public services,the quality of the civil service and the degree of its independence from political pressures, the quality ofpolicy formulation and implementation, and the credibility of the government’s commitment to such poli-cies. (4) Regulatory Quality (RQ) capturing perceptions of the ability of the government to formulate andimplement sound policies and regulations that permit and promote private sector development. (5) Rule ofLaw (RL) capturing perceptions of the extent to which agents have confidence in and abide by the rules

26

use techniques akin to factor analysis to extract a common dimension in each area. This hasyielded a panel for 212 countries and territories for 1996 to the present (but not including 1997,1999, and 2001).

Since 1977, each year World Bank personnel have coded Bank client countries on 16 or moredimensions concerning the quality of policies and institutions. These codings are then aggregatedto a summary measure called the Country Policy and Institutional Assessment, which is used forvarious purposes including decisions about aid allocation. The aggregate index varies from 1to 6, with higher scores indicating a better policy and governance environment from the Bank’sperspective. Unfortunately the CPIA index is produced only for aid recipient countries, so we havenearly complete series for only 85 countries.

To my knowledge this paper is the first to exploit these data for an analysis of civil war onset.An earlier version for the ICRG data, for 1982 only, has been used as a measure of “governance”or “good institutions” in a number of studies of the determinants of economic growth, includinginfluential papers by Acemoglu, Johnson and Robinson (2001), Knack and Keefer (1995), andMauro (1995). But the longer time series employed here appears not to have been used even inthat much larger literature on growth.

Whether the thing to be explained is growth or civil war onset, expert-survey based measures of“good institutions” or “good governance” face a number of problems. In the first place, it is notcompletely clear what the expert ratings are measuring. This is partly due to lack of clarity aboutwhat we are trying to measure. Just what are “good governance” and “good institutions”? Manypeople have strong intuitions here, based on experiencing the relative efficiency, competence, andcorruption of public services and officials in different countries. In more theoretical terms, thetradition associated with North and Thomas (1973) identifies “good institutions” with formal andinformal political institutions that render unlikely the expropriation of private wealth and invest-ments by political elites. In work on state capabilities and civil war, the focus tends to be on theefficiency and competence of the police, armed forces, and judiciary (“rule of law,” in part).

But neither the competence of public management, expropriation risk and contract enforcement,nor “rule of law” is easily observed and measured. Ideally, we would like to have objective in-dicators for these constructs, but even if we did, concepts like “efficiency,” “competence,” and“expropriation risk” seem inherently to be latent variables that would have to be inferred fromdiverse observations of different things.

This fact makes expert surveys a natural approach for measuring the quality of “governance” and

of society, and in particular the quality of contract enforcement, property rights, the police, and the courts,as well as the likelihood of crime and violence. (6) Control of Corruption (CC) capturing perceptionsof the extent to which public power is exercised for private gain, including both petty and grand forms ofcorruption, as well as “capture” of the state by elites and private interests.”

27

“good institutions,” but it is also makes it hard to know exactly what the experts are doing. Forexample, are they really making judgements about the quality of governance and particular insti-tutions, or implicitly are they answering the general question “How do you think things are goingthese days in country X (perhaps implicitly as compared to other countries in the region)?” An-swers to the latter might partly measure quality of governance or institutions, but could also includemany considerations that we would not associate conceptually with governance and institutions. Insum, there are reasons to be concerned about both the validity and reliability of the expert-surveybased measures of governance, but it is also not obvious what a better approach would be.27

The second major problem one faces when trying to use governance indicators to assess the causesof economic growth or civil war onset is endogeneity. If an indicator is well correlated withcontemporaneous growth or civil war onset, we cannot infer causality, because it could be that theobservation of growth is leading the experts to think that governance is good, or that the observationof civil war leads them to infer that governance or institutions are bad.

When one has only a single observation of governance quality for a set of countries, and a singleobservation of level or growth of income, or conflict performance, the only feasible solution isto find an instrument for governance – an exogenous variable that affects growth or conflict onlythrough its effect on governance. Such variables are very hard to find and the exclusion restrictionis not testable.28

With data from time t on governance and from time t+1 on growth or conflict, we can ask whetherthe former predicts the latter, controlling for other possible determinants of growth or conflict. Animportant advantage of this design is that it cannot be that observation of the outcome (growthor conflict in time t + 1) caused the experts to code better governance or institutions in time t,because of course it had not happened yet when they made those judgments. So if we have enoughyears of data on governance and growth or conflict, then we can ask whether expert assessmentsof governance actually predict subsequent conflict or growth experience.

If the answer is yes, this still does not settle the question of causality, since it could be that omittedvariables are causing both expert assessments of quality of governance at time t and conflict orgrowth performance subsequently. In particular, as we will see below, all of the WGI and ICRGindicators are highly correlated with per capita income. This is as it should be, if it is true that

27A signal advantage of Kaufmann and Kraay’s approach is that by drawing on a large number of differentexpert-based measures, their measures may have greater reliability than any one source.

28For studies of institutions and governance as causes of economic growth, Mauro (1995) used ethnicfractionalization as an instrument for corruption as measured by expert surveys; however, it is highly im-plausible that the only path through which ethnic fractionalization would be related to economic growth iscorruption. Acemoglu, Johnson and Robinson (2001) famously used settler mortality in colonies hundredsof years ago as an instrument for 1983 expropriation risk (ICRG). Knack and Keefer (1995) did not reallyaddress the endogeneity issue.

28

income is a proxy for state capabilities and that good governance and good institutions causeeconomic growth over the long term. But it raises the problem of how to separate out the causalimpact of governance on conflict or growth, versus that of other determinants of high income.

The basic approach taken below will be simply to control for prior income levels, thus asking aboutthe relationship between what we might call “surprisingly good governance” and civil war onset.A country has surprisingly good governance when experts gave it high ratings as compared to othercountries at the same level of per capita income. The attempt to identify the causal impact of gov-ernance quality on conflict then comes from seeing whether surprisingly good or bad governancein one period of time predicts conflict onset subsequently. The strategy will be effective to theextent that whatever determines surprisingly good/bad governance in one time period influencessubsequent conflict risk primarily through governance and institutions, rather than via some otherpath.

With the ICRG and CPIA indicators, we have long enough time series and enough variation overtime within countries to go further. We can consider models with country fixed effects, so control-ling for all manner of unobserved time-invariant country characteristics.

[for ABCDE paper, more will follow here on this strategy and potential threats. ABCDE paperwill probably look also at determinants of surprisingly good/bad governance.]

6.1 The WGI, ICRG, and CPIA governance indicators

As noted, the WGI project produces indicators for six dimensions of “governance,” labeled gov-ernment effectiveness, voice and accountability, political instability, rule of law, corruption, andregulatory quality. ICRG produces a large set of indicators, which have varied somewhat over theyears. In this paper I consider four ICRG indicators that have the longest history and that cor-respond most closely to the WGI categories. These are called “investment profile,” “corruption,”“rule or law” (or “law and order”), and “bureaucratic quality.” The correspondence with the WGIindicators is clear except for investment profile. ICRG intends this measure as a general indica-tor of business climate and political risks to business in a country year. It is the successor to the“expropriation risk” and “observance of contracts” variables from the 1982 ICRG data used by anumber of growth studies.

Because they are derived from a factor-analysis-like technique, the WGI indicators all have meanzero and standard deviation of 1, with higher values indicating better quality governance on thatdimension.29 For ICRG, corruption and rule of law are on 1-to-6 scales. Investment profile varies

29One problem with this approach is that a country’s rating may change from one year to the next notbecause anything changed in the country, but because other countries changed; these measures have morevalidity as a ranking within a given year than as a time series measure.

29

from 1 to 12, and bureaucratic quality from 1 to 4. Higher values are better.

The World Bank’s CPIA indicator varies from 1 to 6, with higher values indicating better gov-ernance. The scale is an average of a large number of components, which since 1997 have beengrouped into four equally weighted clusters, described as “economic management,” “structuralpolicies,” “policies for social inclusion/equity,” and “public sector management and institutions.”

For our purposes, a major liability of the CPIA index is that it is only coded for countries thatreceive IDA loans, and that countries can “graduate” out of and enter into this category dependingon economic and government performance. As a result, the CPIA sample is already truncated byhaving relatively poor countries, and, even worse, there is a built-in selection bias that will workagainst identifying the impact of governance on conflict (or growth) outcomes. Namely, countriesthat perform well are more likely to exit the CPIA sample, and countries that perform poorly mayenter it.

Figures 7 and 8 show scatterplots that illustrate how closely related are the different ICRG andWGI governance indicators to each other and to (the log of) per capita income. The data arefor 2005, but the picture of course looks very similar for other years. The correlations are given inTable 6, which also shows that associations among the WGI, ICRG, and CPIA indicators are strongas well.30 There is not much indication that correlations across the ICRG and WGI indicators arehigher within what should be the same dimension – for example, “rule of law” in the two differentdata sets.

Table 7 shows the correlations between the residuals of the WGI, ICRG, and CPIA indicators afterregressing each of them on log per capita income. They remain substantial, which is encouragingin that it suggests that raters’ perceptions of quality of governance or institutions are not com-pletely determined by level of economic development. Instead, there appears to be some level ofagreement about surprisingly good or bad governance. However, there is not much indication thatagreement is markedly higher within categories (e.g., corruption, rule of law, etc.) than acrossthem. This suggests either that these various dimensions of governance quality tend in practiceto go together very closely, or that the expert raters really have in mind some general notion of“country has its act together” rather than being able to separate out dimensions of performanceclearly.

One other descriptive statistic about these indicators is worth presenting before moving to theanalysis. Table 8 shows the percentage of variation for each indicator that is due to variation acrosscountries as opposed to over time within countries. Almost of all of the variation in the WGIindicators is across countries, which makes sense given that the time period is just over a decadeand no one could think that “state capabilities” would change a great deal from year to year. Thereis much more within country variation for the longer ICRG and CPIA series, and especially for the

30This is partly mechanical, since ICRG indicators are one of the many inputs into the WGI indicators.

30

Table 6: Income and governance indicator correlationsWGI ICRG

income ge voice pol. stab. corr. rol reg. qual. ip corr. rol bqgovt eff. 79voice 58 75pol. stab. 67 79 71corruption 74 94 72 77rule of law 77 95 79 83 94reg. qual. 75 94 79 75 88 91inv. prof. 72 81 73 73 78 83 87corruption 61 84 73 67 87 83 78 64rule of law 68 73 48 71 75 77 65 60 64bur. qual. 76 89 77 66 82 85 83 72 75 62cpia 51 78 58 47 62 67 82 73 51 39 62

Table 7: Correlation between WGI and ICRG, netting out incomeWGI

ICRG ge voice pol. stab. rol corruption reg. qual.ip 38 37 33 41 35 47corruption 60 51 41 60 65 51rol 42 13 42 55 45 33bq 67 54 27 58 53 57cpia 61 41 24 51 47 64

ICRG “investment profile” indicator. This will allow us to consider a fixed effects model with theICRG and CPIA data.

6.2 ICRG governance measures and civil war onset

Table 9 reports our baseline model (Model 1 of Table 1), adding each of the four ICRG governanceindicators in turn. The indicators are lagged two years to try to lessen the risk that their values arebeing caused by knowledge of a civil conflict already in progress.

We find that all of them have estimated coefficients that correspond to very large substantive ef-fects, with all but “bureaucratic quality” statistically significant (investment profile and corruptionstrongly so). For investment profile, moving from the 75th to the 25th percentile (8.5 to 3) is esti-mated to associate with an increase in annual civil war odds of a factor of 3.8. Going from 12 to 1,

31

Table 8: Percentage within vs. between country variation in governance indicatorsvariable between % within %log(income) 85 15ACD war onset 6 94

WGI: 1996-2008ge 96 4voice 97 3pol. stab. 91 9corruption 96 4rol 96 4reg. qual. 94 6

ICRG: 1984-2006ip 41 59corruption 71 29rol 71 29bq 81 19

CPIA: 1977-2008cpia 58 42

the full range of the scale, increases the estimated odds by a factor of 96. Moving from 6 to 1 onthe rule of law and corruption measures (the full range of these scales) associates with increases inthe odds of 5.7 and 3.7, respectively.

Model 1 of Table 9 is Model of Table 1 but restricted to the sample used when we add ICRGindicators, which is limited to 1984-2006 and countries with ICRG data. It shows that in thissubsample the standard errors for several of the other covariates increase and some of the estimatedcoefficients diminish. For these countries and years, ethnic fractionalization and population arestronger predictors, while per capita income weakens somewhat. We also see, however, that theestimated coefficients for per capita income diminish further (or even become negative) when weadd the governance indicators in Models 2-4. This is consistent with the hypothesis that income“matters” because it proxies for state capabilities, or governance.31

Table 10 repeats the exercise, but using conditional fixed effects logit and thus controlling forheterogeneity associated with unmeasured country characteristics. Remarkably, given the smallnumber of countries in the sample now, three of the ICRG indicators get negative coefficients(bureaucratic quality is essentially zero), and the estimate for investment profile is statisticallysignificant (p = .017). As in Table 1, Model 5 (fixed effects on the larger sample), per capita

31This effect is stronger if we consider models that drop other covariates, for example leaving just priorwar, income, population, and ethnic fractionalization.

32

Table 9: ICRG governance and civil war onsetModel 1 Model 2 Model 3 Model 4 Model 5

constant −8.24∗∗∗ −9.96∗∗∗ −8.38∗∗∗ −8.77∗∗∗ −9.21∗∗∗

(1.97) (1.76) (2.03) (2.23) (1.98)prior war 0.50 0.19 0.43 0.31 0.45

(0.42) (0.42) (0.42) (0.40) (0.44)log(gdpt−1) −0.21 0.13 −0.10 −0.08 −0.10

(0.20) (0.18) (0.22) (0.23) (0.23)log(popt−1) 0.45∗∗ 0.54∗∗∗ 0.43∗∗ 0.48∗∗ 0.48∗∗

(0.15) (0.15) (0.15) (0.16) (0.15)log(% mountains) 0.08 0.03 0.10 0.11 0.07

(0.16) (0.16) (0.16) (0.18) (0.15)oil producer 0.33 0.09 0.23 0.17 0.25

(0.45) (0.44) (0.47) (0.47) (0.46)pol instabilityt−1 0.42 0.42 0.41 0.33 0.40

(0.35) (0.33) (0.35) (0.37) (0.35)anocracyt−1 0.67 0.84∗ 0.67 0.70† 0.70†

(0.43) (0.38) (0.42) (0.42) (0.42)democracyt−1 −0.38 −0.15 −0.33 −0.37 −0.31

(0.52) (0.49) (0.52) (0.52) (0.49)ELF 1.58∗ 2.12∗∗∗ 1.65∗∗ 1.59∗ 1.70∗∗

(0.62) (0.59) (0.63) (0.62) (0.62)

ipt−2 −0.38∗∗∗

(0.10)corruptt−2 −0.22∗

(0.11)rolt−2 −0.29†

(0.16)bqt − 2 −0.16

(0.22)

N 2750 2750 2749 2749 2749

Standard errors clustered by country† significant at p < .10; ∗p < .05; ∗∗p < .01; ∗∗∗p < .001

income takes “the wrong sign” and is not significant. Thus, within countries over time, civilwar onset has been somewhat more likely when investment profile, corruption, and rule of lawwere judged worse in recently preceding years. This result supports a causal interpretation of therelationship between governance quality and conflict onset more than the previous models, becausehere identification is based on within-country comparisons (and because it is somewhat remarkableto find anything given the lack of within-country variation in governance indicators).

33

Table 10: ICRG governance and civil war onset, fixed effectsModel 1 Model 2 Model 3 Model 4

prior war −1.894∗∗ −1.803∗∗ −1.835∗∗ −1.782∗∗

(0.531) (0.527) (0.534) (0.527)log(gdpt−1) 0.994 0.466 0.644 0.506

(0.742) (0.684) (0.693) (0.690)log(popt−1) −0.329 −1.022 −0.714 −1.045

(1.274) (1.192) (1.261) (1.190)oil 2.719† 2.276 2.500 2.780

(1.614) (1.706) (1.624) (1.702)pol. instab. 0.421 0.389 0.380 0.367

(0.441) (0.435) (0.434) (0.433)anocracyt−1 1.260† 1.126† 1.129† 1.192†

(0.647) (0.657) (0.653) (0.650)democracyt−1 −0.130 −0.339 −0.236 −0.250

(0.881) (0.893) (0.893) (0.897)

ipt−2 −0.276∗

(0.115)corruptiont−2 −0.172

(0.255)rule of lawt−2 −0.170

(0.202)bqt−2 0.080

(0.332)

N 536 535 535 535N (countries) 25 25 25 25Country fixed effects yes yes yes yes† significant at p < .10; ∗p < .05; ∗∗p < .01

[ICRG indicators studied decade to decade: We might still worry that a two-year lag is not enoughto rule out the possibility that expert raters are coding based on indications of incipient civil war,more than on the quality of governance or institutions. Constructed a panel with three waves,for the 80s, 90s, and 00s, asking if average ICRG ratings in one decade forecast conflict in nextdecade, controlling for prior conflict experience and lagged income levels. Again, ICRG indicatorsforecast conflict outbreak even in next decade.]

34

6.3 WGI indicators and civil war onset

The WGI series is only for 1996 to 2008, with some of the early years missing. It also has verylittle over time variation within countries, and the method of its construction raises some questionsabout whether and how best to treat it as panel data in any event.

However, because 14 years have passed since the first set of WGI indicators were constructed,we can ask whether expert-based assessments of different dimensions of governance quality in1996 or 1998 actually forecast conflict experience in the next decade, controlling for initial levelof income and prior conflict experience. Using the ACD civil war variable, 19 countries hadonsets between 1997 and 2009. Tables 11 and 12 show that the WGI estimates of “governmenteffectiveness,” “political stability,” “rule of law,” and “corruption” are indeed significantly relatedto subsequent conflict experience. This is perhaps not so surprising for “political stability,” whichis based on expert surveys intended to capture “perceptions of the likelihood that the governmentwill be destabilized or overthrown by unconstitutional or violent means” (Kaufmann, Kraay andMastruzzi 2009). But the results are also strong for “government effectiveness” and “rule of law.”

In Table 11, we control for the number of ACD civil war onsets before 1997, log of income in 1996,and ELF. Oil production in 1996, and log of population in 1996 can be added without changingresults (both take positive coefficients but are not significant). Table 12 differs only by the additionof dummies for each of six regions (coefficients not shown). Notice that the “government effec-tiveness” and “rule of law” remain significantly negatively related to future onsets even controllingunmeasured region-specific influences. Ethnic fractionalization weakens considerably for this timeperiod when regional dummies are included, which is due to the fact that most onsets occurred inethnically diverse subSaharan Africa.

Model 1 in Table 11 shows that per capita income is negatively related to subsequent onset if we donot control for governance; the coefficient is similar to that in our baseline model although it is notquite statistically significant. Adding the governance indicators (or regional dummies, in Table 12)greatly reduces or even flips the sign of income, which again supports the hypothesis low incomecountries are more civil war prone because of low state capability rather than due to direct labormarket effects.

I have run the same models using the WGI indicators from 1998, and the dependent variableas onsets after 1998. The results are quite similar, though marginally weaker. This lowers thelikelihood that the results are a fluke from one year of WGI data (which as we have seen are highlystable over time anyway).

Inspection of added variable plots shows that Liberia is a highly influential data point. Probablybecause of the 1996 election that brought Charles Taylor to power and appeared to be a promisingend to the civil war, Liberia does extremely well on the WGI indicators in 1996 despite its very lowincome. A new war began in 2000, however. If Liberia is dropped from the sample, the coefficients

35

for all of the WGI indicators in Tables 11 and 12 except for regulatory quality become markedlymore negative and more statistically significant.

There is not much evidence that different dimensions of governance as measured by WGI shownotably stronger or weaker relationships to subsequent conflict risk. “Political stability,” whichis supposed to be an expert appraisal of conflict risk, is indeed the most strongly related, while“voice and accountability,” which is based on assessments of democracy, is the weakest. (Thisis consistent with our earlier findings of little link between democracy and conflict risk in poorcountries.) Rule of law and “government effectiveness” – which KKZ describe as “capturingperceptions of the quality of public services, the quality of the civil service and the degree of itsindependence from political pressures, the quality of policy formulation and implementation, andthe credibility of the government’s commitment to such policies” – appear to be most predictiveafter political stability, followed by corruption. But overall, just as we saw that there are quite highcorrelations among these different dimensions, no strong conclusions can be drawn about whatdimension of governance is most important for increasing the odds of civil peace.

To recall, the identification strategy here is plausible in so far as the following argument is plau-sible: once we control for 1996 income, prior conflict experience, and other factors, variation incountries’ quality of governance as measured by expert ratings in 1996 is essentially random withrespect to unmeasured other determinants of subsequent civil war risk. I find it difficult to thinkof omitted variables distinct from “governance” or “institutions” that would plausible cause bothrater perceptions of governance quality and conflict performance over the subsequent ten years.But of course the possibility still exists.

The next step in this part of the investigation should be to examine the determinants of surprisinglygood/bad governance – what explains variation in expert perceptions once we have netted outincome level? In preliminary work, I find that, interestingly, ethnic fractionalization is almostcompletely unrelated to surprisingly good governance. This is surprising in light of the literaturein economics finding and arguing that ethnic diversity directly causes low public good provisionand corruption. Second, I find that oil producers and autocracies are consistently judged to haveworse governance, even controlling for income.32

6.4 CPIA indicators and civil war onset

Table 13 is essentially our basic model using the set of countries and year with CPIA estimates(Model 1), adding one- and two-year lags of the CPIA index (Models 2 and 3), and then Model 2with country fixed effects. The estimated coefficients for the lagged CPIA index are quite similarin magnitude to those for the other governance indicators, despite the sample truncated sample and

32As noted, the results in Tables 11 and 12 are not changed if we control also for oil or democracy.

36

Tabl

e11

:WG

Igov

erna

nce

in19

96pr

edic

tsci

vilw

aron

set,

1997

-200

8M

odel

1M

odel

2M

odel

3M

odel

4M

odel

5M

odel

6M

odel

7on

sets

pre

1997

0.37∗∗

0.36∗

0.32∗

0.18

0.34∗∗

0.32∗

0.33∗

(0.14)

(0.14)

(0.14)

(0.12)

(0.13)

(0.14)

(0.14)

log(

inco

me)

1996

−0.44

0.03

−0.26

0.07

0.07

−0.12

−0.16

(0.29)

(0.37)

(0.32)

(0.35)

(0.36)

(0.35)

(0.34)

EL

F1.89

1.96†

2.18†

2.26†

2.06†

1.92

2.03†

(1.18)

(1.18)

(1.21)

(1.23)

(1.21)

(1.17)

(1.19)

gov

effe

ct.1

996

−1.03∗

(0.51)

voic

e19

96−0.58

(0.38)

pol.

stab

.199

6−1.25∗∗

(0.37)

rule

ofla

w19

96−1.38∗∗

(0.51)

corr

uptio

n19

96−0.98†

(0.56)

reg.

qual

1996

−0.61

(0.41)

cons

tant

0.20

−4.17

−1.69

−4.84

−4.96

−2.88

−2.38

(2.62)

(3.37)

(2.92)

(3.23)

(3.30)

(3.17)

(3.13)

N156

156

156

156

156

156

156

DV

=A

CD

civi

lwar

onse

taft

er19

96.L

ogit,

with

stan

dard

erro

rsin

pare

nthe

ses

†si

gnifi

cant

atp<

.10

;∗p<

.05

;∗∗p<

.01

;∗∗∗p<

.001

37

Tabl

e12

:WG

Igov

erna

nce

in19

96an

dci

vilw

aron

set,

regi

onfix

edef

fect

sM

odel

1M

odel

2M

odel

3M

odel

4M

odel

5M

odel

6M

odel

7on

sets

pre

1997

0.46∗∗

0.43∗∗

0.43∗∗

0.30∗

0.39∗∗

0.40∗

0.42∗

(0.17)

(0.16)

(0.17)

(0.15)

(0.15)

(0.16)

(0.17)

log(

inco

me)

1996

−0.06

0.28

0.03

0.60

0.44

0.18

0.13

(0.34)

(0.40)

(0.35)

(0.45)

(0.41)

(0.38)

(0.38)

EL

F0.75

0.80

0.91

0.80

0.78

0.62

0.87

(1.29)

(1.32)

(1.31)

(1.44)

(1.37)

(1.31)

(1.30)

gov

effe

ct.1

996

−0.84†

(0.50)

voic

e19

96−0.40

(0.39)

pol.

stab

1996

−1.41∗∗

(0.41)

rule

ofla

w19

96−1.39∗∗

(0.52)

corr

uptio

n19

96−0.88

(0.56)

reg.

qual

1996

−0.49

(0.42)

cons

tant

−2.96

−5.93†

−4.00

−9.48∗

−7.59∗

−5.26

−4.65

(3.02)

(3.52)

(3.20)

(4.04)

(3.59)

(3.37)

(3.33)

N156

156

156

156

156

156

156

DV

=A

CD

civi

lwar

onse

taft

er19

96.L

ogit

with

stan

dard

erro

rsin

pare

nthe

ses

†si

gnifi

cant

atp<

.10

;∗p<

.05

;∗∗p<

.01

;∗∗∗p<

.001

38

the selection bias issue. However, they are barely statistically significant, and not at all with fixedeffects. The estimate hardly changes when we use the two- instead of the one-year lag, whichsuggests that there probably isn’t very much “coding of CPIA on civil war in progress” going on.

Table 13: CPIA governance and civil war onsetModel 1 Model 2 Model 3 Model 4

prior war −0.17 −0.30 −0.19 −2.96(0.35) (0.32) (0.36) (0.60)

log(gdpt−1) −0.54∗∗ −0.40† −0.40† 0.32(0.19) (0.22) (0.22) (0.59)

log(popt−1) 0.47∗∗∗ 0.54∗∗∗ 0.53∗∗∗ −0.22(0.14) (0.13) (0.12) (1.03)

log(% mountains) 0.24∗ 0.25∗ 0.23†

(0.12) (0.12) (0.13)oil producer 0.25 0.11 0.18 2.23†

(0.48) (0.49) (0.50) (1.34)pol instabilityt−1 0.24 0.19 0.18 0.15

(0.31) (0.31) (0.35) (0.42)anocracyt−1 0.60† 0.61† 0.69† 1.18∗

(0.36) (0.35) (0.37) (0.59)democracyt−1 0.12 0.18 0.15 −0.09

(0.47) (0.47) (0.46) (0.81)ELF 1.21∗ 1.21∗ 1.20∗

(0.59) (0.58) (0.57)

cpiat−1 −0.39† −.20(0.21) (0.25)

cpiat−2 −0.35(0.24)

constant −5.94∗∗∗ −6.34∗∗∗ −6.47∗∗∗

(1.80) (1.76) (1.80)

N 3120 3120 3063 851(33)Country fixed effects? No No No YesRobust standard errors in parentheses† significant at p < .10; ∗p < .05; ∗∗p < .01; ∗∗∗p < .001

7 Conclusion

39

1950 1960 1970 1980 1990 2000 2010

010

2030

4050

year

# of

war

s/co

nflic

ts

Figure 1. # of civil conflicts and wars, 1946−2008

all conflicts (kia > 25/year)major conflicts (kia > 1000 total)war onsetswar ends

40

1950 1960 1970 1980 1990 2000 2010

0.00

0.05

0.10

0.15

0.20

0.25

year

% o

f cou

ntrie

s w

ith c

onfli

ct

all conflicts (kia > 25/year)major conflicts (kia > 1000 total)

% countries with conflict or war, 1946−2008

41

1950 1970 1990 2010

05

15

year

# co

nflic

ts

SSA

1950 1970 1990 2010

05

15

year

# co

nflic

ts

Asia

1950 1970 1990 2010

05

15

year

# co

nflic

ts

NA/ME

1950 1970 1990 2010

05

15

year

# co

nflic

ts

LA/Ca

1950 1970 1990 2010

05

15

year

# co

nflic

ts

EEur

1950 1970 1990 2010

05

15

year

# co

nflic

ts

West

42

1950 1960 1970 1980 1990 2000 2010

0.0

0.1

0.2

0.3

0.4

0.5

year

shar

e of

wor

ld p

op in

con

flict

cou

ntrie

s

all conflictscivil wars

Figure 4. % of world population in conflict countries

43

1960 1970 1980 1990 2000

5 year periods

0.0

0.2

0.4

0.6

0.8

1.0

botto

m q

uart

ile2n

d3r

d

top

Figure 5. Distribution of conflicts by income quartiles and years

44

lgdp

−2 0 2 −2 0 −2 0 2

68

10

−2

02

ge

voice

−2

0

−2

0 polstab

corrupt

−1

1

−2

02

rol

6 8 10 −2 0 −1 1 −2 0 2

−2

02

regqual

WGI indicators and income, 2005

45

lgdp

2 6 10 1 3 5

68

10

26

10

ip

corrupt.1

02

46

13

5

rol.1

6 8 10 0 2 4 6 0 1 2 3 4

01

23

4bq.1

ICRG indicators and income, 2005

46

References

Acemoglu, Daron, Simon Johnson and James A. Robinson. 2001. “The Colonial Originsof Comparative Development: An Empirical Investigation.” American Economic Review91(5):1369–1401.

Balch-Lindsay, Dylan and Andrew J. Enterline. 2000. “Killing Time: The World Politics of CivilWar Duration, 1820-1992.” International Studies Quarterly 4:615–642.

Besley, Timothy and Torsten Persson. 2009. “The Logic of Political Violence.” Ms., LSE and IIES,Stockholm University.

Bruckner, Markus and Antonio Ciccone. 2010a. “International Commodity Prices, Growth, andthe Outbreak of Civil War in subSaharan Africa.” The Economic Journal .

Bruckner, Markus and Antonio Ciccone. 2010b. “Transitory Economic Shocks and Civil Conflict.”Unpublished paper, Universitat Pompeu Fabra.

Buhaug, Halvard, Lars-Erik Cederman and Jan Ketil Rød. 2008. “Disaggregating Ethno-Nationalist Civil Wars: A Dyadic Test of Exclusion Theory.” International Organization62(3):531–551.

Cederman, Lars-Erik and Luc Girardin. 2007. “Beyond Fractionalization: Mapping Ethnicity ontoNationalist Insurgencies.” American Political Science Review 101(1):173–185.

Collier, Paul and Anke Hoeffler. 2004. “Greed and Grievance in Civil War.” Oxford EconomicPapers 56:563–595.

Collier, Paul, Anke Hoeffler and Mans Soderbom. 2004. “On the Duration of Civil War.” Journalof Peace Research 41:253–273.

Collier, Paul, V. L. Elliott, Havard Hegre, Anke Hoeffler, Marta Reynal-Querol and Nicholas Sam-banis. 2003. Breaking the Conflict Trap: Civil War and Development Policy. Washington,DC: World Bank and Oxford University Press.

Condra, Luke and Jacob Shapiro. 2010. “Who Takes the Blame? The Strategic Effects of CollateralDamage.” Unpublished ms., Stanford and Princeton Universities.

Condra, Luke N. 2009. “Ethnic Rebellion against the State: Perils of the Periphery.” AnnualMeetings of the APSA, Toronto, 2009.

Cunningham, David E. 2006. “Veto Players and Civil War Duration.” American Journal of PoliticalScience 50(4):875–92.

Fearon, James D. 2004. “Why Do Some Civil Wars Last So Much Longer Than Others?” Journalof Peace Research 41(3):275–301.

47

Fearon, James D. and David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” AmericanPolitical Science Review 97(1):75–90.

Fearon, James D. and David D. Laitin. 2007. “Civil War Termination.” Unpublished paper, Stan-ford University.

Fearon, James D. and David D. Laitin. 2010. “Sons of the Soil, Migrants, and Civil War.” WorldDevelopment . Forthcoming.

Fearon, James D., David D. Laitin and Kimuli Kasara. 2007. “Ethnic Minority Rule and Civil WarOnset.” American Political Science Review 101(1):187–93.

Gibney, Mark, L. Cornett and R. Wood. 2008. “Political Terror Scale 1976-2008.” Date Retrieved,from http://www.politicalterrorscale.org/.

Hegre, Havard, Tanja Ellingsen, Scott Gates and Nils Petter Gleditsch. 2001. “Toward A Demo-cratic Civil Peace? Democracy, Political Change, and Civil War 18161992.” American Polit-ical Science Review 95(1).

Hibbs, Douglas A. 1973. Mass Political Violence. New York: Wiley.

Homer-Dixon, Thomas F. 2001. Environment, Scarcity, and Violence. Princeton, NJ: PrincetonUniversity Press.

Huntington, Samuel. 1996. The Clash of Civilizations and the Remaking of World Order. NewYork: Simon and Shuster.

Kalyvas, Stathis N. 2006. The Logic of Violence in Civil War. New York: Cambridge UniversityPress.

Kaufmann, Daniel, Aart Kraay and Massimo Mastruzzi. 2009. “Governance Matters VIII: Aggre-gate and Individual Governance Indicators, 19962008.” World Bank, Development ResearchGroup, Policy Research Working Paper 4978.

Knack, Stephen and Philip Keefer. 1995. “Institutions and Economic Performance: Cross-CountryTests Using Alternative Institutional Measures.” Economics and Politics 7(3):207–227.

Lyall, Jason. 2009. “Does Indiscriminate Violence Incite Insurgent Attacks? Evidence fromChechnya.” Journal of Conflict Resolution 53(3):331–362.

Lyall, Jason and Isaiah Wilson. 2009. “Rage against the Machines: Explaining Outcomes in Coun-terinsurgency Wars.” International Organization 63:67–106.

Mauro, Paolo. 1995. “Corruption and Growth.” Quarterly Journal of Economics 110(3):681–712.

Miguel, Edward, Shanker Satyanath and Ernest Sergenti. 2004. “Economic Growth and CivilConflict: An Instrumental Variables Approach.” Journal of Political Economy 112(4):725–753.

48

Montalvo, Jose and Marta Reynal Querol. 2005. “Ethnic Polarization, Potential Conflict, and CivilWar.” American Economic Review 95(3):796–816.

Moore, Barrington. 1966. Social Origins of Dictatorship and Democracy. Boston, MA: BeaconPress.

North, Douglass C. and Robert Paul Thomas. 1973. The Rise of the Western World: A New Eco-nomic History. New York: Cambridge University Press.

Østby, Gudrun. 2008. “Polarization, Horizontal Inequalities and Violent Civil Conflict.” Journalof Peace Research 45(2):143–162.

Rodrik, Dani. 2005. “Why We Learn Nothing from Regressing Economic Growth on Policies.”Unpublished, Kennedy School of Government.

Sambanis, Nicolas. 2001. “Do Ethnic and Non-Ethnic Civil Wars Have the Same Causes? ATheoretical and Empirical Inquiry (Part 1).” Journal of Conflict Resolution 45(3):259–82.

Skocpol, Theda. 1979. States and Social Revolution. Princeton, NJ: Princeton University Press.

Urdal, Henrik. 2005. “People vs Malthus: Population Pressure, Environmental Degradation, andArmed Conflict Revisited.” Journal of Peace Research 42(4):417–434.

Urdal, Henrik. 2006. “A Clash of Generations? Youth Bulges and Political Violence.” InternationalStudies Quarterly 50:607–629.

Weinstein, Jeremy M. 2007. Inside Rebellion: The Politics of Insurgent Violence. New York:Cambridge University Press.

Wimmer, Andreas, Lars-Erik Cederman and Brian Min. 2009. “Ethnic Politics and Armed Con-flict: A Configurational Analysis of a New Global Data Set.” American Sociological Review74:316–337.

49


Recommended