                                                                                         62009



WORLD DEVELOPMENT REPORT 2011
BACKGROUND PAPER




GOVERNANCE AND CIVIL WAR ONSET

James D. Fearon
Department of Political Science
Stanford University


August 31, 2010




The findings, interpretations, and conclusions expressed in this paper are entirely those of the
authors. They do not necessarily represent the views of the World Development Report 2011
team, the World Bank and its affiliated organizations, or those of the Executive Directors of the
World Bank or the governments they represent.
Contents
1 Introduction                                                                                                                    2

2 Correlates and causes of civil war and lower-level violence conﬂict                                                             5

3 The conﬂict data and global trends                                                                                              6
  3.1 The ACD conﬂict measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                                         6
  3.2 Global and regional trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .                                      9

4 Correlates of civil war onset                                                               10
  4.1 Baseline models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
  4.2 “Effect” magnitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

5 Other country characteristics of potential interest                                                                            14
  5.1 Population growth, land pressure, and “youth bulges”       .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   14
  5.2 Vertical and horizontal income inequality . . . . . .      .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   16
  5.3 Civil liberties, human rights abuses . . . . . . . . . .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   20
  5.4 Political reform and civil war onset . . . . . . . . . .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   25
  5.5 Rapid governance reform and conﬂict onset . . . . .        .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   25
  5.6 Natural resources . . . . . . . . . . . . . . . . . . .    .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   30
  5.7 Geography . . . . . . . . . . . . . . . . . . . . . . .    .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   32
  5.8 Gender inequality and civil conﬂict . . . . . . . . . .    .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   34

6 Causes of civil war                                                                                                            36

7 Governance indicators and civil conﬂict                                                                                        42
  7.1 The WGI, ICRG, and CPIA governance indicators .        .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   46
  7.2 ICRG governance measures and civil war onset . .       .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   48
  7.3 WGI indicators and civil war onset . . . . . . . . .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   52
  7.4 CPIA indicators and civil war onset . . . . . . . .    .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   55




                                               1
1 Introduction

Since the end of the Cold War, violent political conﬂict has increasingly been seen as a major obsta-
cle to the economic development of many low and middle-income countries. Despite a signiﬁcant
decrease since the peak in the early 1990s, the prevalence of serious civil conﬂict remains quite
high. About one third of the world’s population lives in countries affected by signiﬁcant political
violence, and about 25% of the world’s population outside of China and India. Most of these are
in low and middle income countries. In addition, very many people live in post-conﬂict countries
that are recovering from major and minor civil wars.

It is obvious that major civil conﬂict prevents economic development. Indeed, it is obvious that
major civil wars can and do any destroy economic progress that may have been made in previous
years. It is less obvious, however, what causes civil wars and lower levels of violent civil conﬂict
to begin with. Why have some countries been more prone to violent conﬂict than others, and what
explains why conﬂicts break out when they do?

In this paper I review and extend the ﬁndings of a growing, post-Cold War literature on the cor-
relates of civil war that has relied primarily on cross-national statistical analyses. Using a version
of the Uppsala Armed Conﬂict Database’s coding of conﬂicts that distinguishes between high and
low intensity conﬂict, I reproduce a number of the main ﬁndings of this literature. For the period
since World War II (or since the early 1960s), the countries most prone to major civil war have
been distinguished by low income, large populations, mountainous terrain, and possibly by oil
production, a high share of “politically excluded” ethnic minorities, and (perhaps) greater gender
inequality. Recent independence and recent changes in degree of democracy augur a higher risk of
major conﬂict onset in the next few years. For all conﬂicts including low intensity conﬂicts, these
same factors are statistically associated with conﬂict outbreak, as is, additionally, higher levels of
ethnic fractionalization. Some factors that show no very consistent relationship with a propensity
for violent civil conﬂict include income inequality (as measured by Gini coefﬁcients) and level of
democracy, although there is some indication that countries with “anocracy” (partial democracy)
are at greater risk if we consider both low and high intensity conﬂicts.

As is the case for most of the literature in this area, these ﬁndings are based on analysis of a panel
of 150 to 180 or so countries, observed anywhere from about 10 to 65 years depending on the time
since independence or 1946. Many of the factors mentioned above, such as income, ethnic diver-
sity, and mountainousness, vary little or not at all over time within countries. This means that many
of the results are based on comparisons in rates of conﬂict outbreak across countries, and we may
have substantial reason to worry that the observed associations are “spurious correlations” rather
than estimates of causal effects. It could be that, for example, that low income does not directly
cause higher conﬂict risk, but happens to be correlated with some third, unmeasured factor that
does cause conﬂict. In some cases, it could be that the mere expectation of conﬂict for unmeasured
reasons inﬂuences one of our presumed causal factors, such as income or perceived ethnic diver-


                                                  2
sity, again leading to spurious correlations. So it is important to keep in mind that much caution is
called for in interpreting many of the associations reported below.

For a few variables that vary a lot over time within countries, such as recent independence or change
in governing arrangements, we can ask whether change in these variables is reliably followed
within countries by a higher risk of conﬂict outbreak (this is called a “ﬁxed effects” analysis).
Recent independence, change in governing arrangements, and the onset of partial democracy all
predict conﬂict outbreak in this sense in the data examined here, and so may have a stronger claim
on being (or marking) causal effects.

The main novelty of this paper is an examination of the relationship between several indicators of
the quality of country governance or institutions and conﬂict onset. The correlational literature on
civil war onset has seen some debate and discussion over how to interpret the fact that low income
is one of the strongest and most signiﬁcant correlates of a country having a higher propensity for
civil war (although there is at best weak evidence that increasing income within a country lowers
its civil war risk). Some have argued for a direct causal effect via a country’s labor market. For
example, perhaps poverty makes joining a rebel band relatively attractive for underemployed young
men, whether for ﬁnancial or more ideological reasons. Others argue that the strong association is
more likely due to a spurious correlation – perhaps low income is a proxy for countries that have
central states with low administrative and coercive capabilities, and/or are ineffective at supplying
public goods.

One obstacle to assessing these arguments has been the lack of more direct measures of the quality
of a country’s “governance” or “institutions.” In this paper I employ several sets of governance
measures that are based ultimately on expert ratings, in particular, the World Bank’s Country Policy
Institutional Assessement index, the Worldwide Governance Indicators (Kaufmann, Kraay and
Mastruzzi 2009), and some of the International Country Risk Guide indicators. To a limited extent
some of these measures have been used in research on governance quality and economic growth,
but they have not been much employed in research on civil conﬂict.1

Not surprisingly, and consistent with the view that low income might stand in for low institutional
capabilities, governance indicators and per capita income tend to be highly correlated. However,
there is some variation in governance quality even conditional on income, and this is what I try to
exploit to assess the impact of the former on the latter. I ﬁnd that a country that was judged in one
year to have worse governance than one would expect given its income level has a signiﬁcantly
greater risk of civil war outbreak in subsequent years. This is true for all three sets of governance
indicators considered here, although effects are barely present for the CPIA aggregate indicator.
Remarkably, given that governance indicators change little over time, I ﬁnd some evidence that
even within countries, improvements in governance tend to reduce subsequent conﬂict risk. Both
  1
   But see Fearon (2005) and papers in the May 2010 special issue of the Journal of Peace Research, which
focuses on state capacity.


                                                   3
in the ﬁxed effect and cross-sectional models, including measures of governance tends to weaken
or in some cases eliminate the association between income and conﬂict onset.

Moreover, it may be interesting to learn that it does not matter much which governance indicator
one chooses: “government effectiveness” (WGI), “investment proﬁle” (ICRG), “corruption” and
“rule of law” (several of them) all work similarly. This could mean that different dimensions have
similar effects on conﬂict; that in practice, good governance is like a syndrome and “all good things
tend to go together”; or that expert ratings are not very good at distinguishing different dimensions
of governance/institutions.

Overall, the results tend to support the view the low income countries have been at greater risk
of violent conﬂict due to poor governance and weak institutions, more than due to direct labor
market effects. The strategy for identifying the causal effect of good governance here is admittedly
not iron-clad. It could be, for example, that some unmeasured factor, completely distinct from
governance, is causing both “surprisingly good governance” given income, and also lower conﬂict
risk. However it is not obvious what such factors might be, and there is at the same time a plausible
argument for why the strategy employed here would underestimate the total impact of governance
on conﬂict risk, if such an effect exists. Namely, since current income incorporates the effects
of past governance and current governance is surely measured with signiﬁcant error, income may
tend to pick up effects of governance.

In terms of policy implications, the results lend support to the view that aid in conﬂict-affected
countries 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 “state
building” 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 literature
on 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 4
presents a series of “onset models,” and includes consideration of some factors that have not been
much examined in the existing literature, such as the relationship between human rights abuses
by government and civil war risk. In section 5 I then discuss what these essentially correlational
ﬁndings 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 indicators
and conﬂict onset.




                                                 4
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-national
data to study the correlates of large-scale civil violence, usually for the 65 years since the end of
the World War II. The typical design is an annual panel for roughly 160 countries (micro-states are
often 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 ar onset, meaning that the dependent variable
is coded as “1” for country years with a civil war onset, and zero for other country years.2 Other
researchers look at the correlates of civil ar inci ence, meaning that the dependent variable is
coded 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 is
that the estimated coefﬁcients are complicated averages of the “effect” of a covariate on both the
onset 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 same
relationship to onset as it does to continuation. A more natural alternative is to study determinants
of conﬂict onset and duration separately, using survival models for the latter (Balch-Lindsay and
                                          o
Enterline 2000; Collier, Hoefﬂer and S¨ derbom 2004; Cunningham 2006; Fearon 2004). In this
paper I consider factors related to the onset of civil conﬂict and war, rather than incidence or
duration.3

For the most part, the cross-national statistical models of civil war onset and incidence should
be viewed as descriptive more than structural (or causal). That is, they have been of great value
for making clearer which political, economic, and demographic factors are associated with higher
civil war propensity in the last 60 years, which factors are not, and which are associated with
onset when you control for other factors. But for many covariates found to be statistically and
   2
    Some drop country years with ongoing civil war (Collier and Hoefﬂer 2004), others include them and
introduce a variable indicating whether war was ongoing in the previous year as a control (Fearon and
Laitin 2003). The latter approach avoids dropping onsets that occur when another civil war is already in
progress.
   3
    Duration 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 by
clustering the errors within countries, but not including explicit dynamics in the form of, say, a lagged
dependent variable. Implicitly, then, they assume that there is no direct causal effect of war in one year on
the probability of war in the next year. This is strongly counter to theory and case-based evidence, since it is
clear that there are large ﬁxed costs to generating an insurgent movement, and major commitment problems
can prevent easy dissolution of a movement once it has started (Walter 2002; Fearon and Laitin 2007).
Omitting a lagged dependent variable when it should be there in an incidence regression tends to bias
sharply upwards the estimated coefﬁcients for any covariates that are positively serially correlated.


                                                       5
substantively signiﬁcant in these models, the argument for interpreting the estimated coefﬁcients
as causal effects is tenuous or speculative. For instance, do we really believe that if the share of
mountainous terrain in, say, Namibia, increased from 10% to 20% that its expected annual odds of
civil war onset would increase by about 15%? Maybe, but this is not an experiment that we will
ever get to run. Even so, it is of some interest to learn that in the last 65 years, there has been some
tendency for more mountainous countries to have more civil wars, even controlling for a range of
other country characteristics.

At least for the literature to this point, found opportunities for natural experiments to allow cleaner
identiﬁcation of causal effects have been rare. Miguel, Satyanath and Sergenti (2004) cleverly used
exogenous variation in rainfall in subSaharan Africa to estimate the effect of changes in income on
                                        u
civil conﬂict propensity (see also Br¨ ckner and Ciccone (2010 ), who examine the same data and
reach a quite different conclusion). And there are a few papers that use variation in international
                                                                           u
commodity prices in a similar fashion (Besley and Persson 2009; Br¨ ckner and Ciccone 2010a).
But even these papers can be of limited value for understanding why an effect is observed – what is
the causal mechanism connecting changes in income to civil war propensity? – and thus whether
and how it might generalize. For these reasons, arguments about causes of civil conﬂict have
generally taken the form of attempts to make sense of the complicated pattern of associations
observed in the cross-national panel data, often informed by theoretical arguments and additional
evidence 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 Conﬂict
Database. In section 5, I consider in more detail arguments about causes of civil conﬂict based on
these ﬁndings.


3 The conﬂict data and global trends

3 1 The C conﬂict           easure

The UCDP/PRIO Armed Conﬂict Database (ACD) codes for each country and year since 1945
whether a violent conﬂict occurred between a named, non-state armed group and government
forces that directly killed at least 25 people.4 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 for
each 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 on average per year over the
whole episode of conﬂict; “medium” conﬂicts estimated to have killed an average per year between
  4
      http://www.prio.no/CSCW/Datasets/Armed-Conﬂict/UCDP-PRIO/


                                                   6
500 and 999; “small” conﬂicts estimated to have killed a per year average between 250 and 499;
and “minor” conﬂicts estimated to have killed between 25 and 249 on average per year. “Killed”
in all cases is intended to refer to battle deaths rather than “indirect” deaths due to starvation, de-
privation from medical services, and so on. These levels are thus measures of the intensity of the
conﬂict.

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 e iso es 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 is
difﬁcult to know how to use these data to produce a list of distinct civil wars or conﬂicts, which is
unfortunately just what we need if we want to study determinants of civil war onset or duration. For
that we need to know when a conﬂict started and when it ended. In many cases this is intuitively
clear, but in many others it is not. For example, some low-level conﬂicts ﬂit above and below the
25-dead threshold for a period of many years. Is that one conﬂict, or many?

In this paper I will follow the suggested practice for WDR 2011 (which is also employed by many
other scholars using the ACD). We will consider a new civil war or conﬂict to have started if it is
preceded by at least two years of peace between the present adversaries. Thus, if a ‘1’ represents a
year with a conﬂict that killed at least 25 in a country, then a sequence like . . . , 0, 0, 1, 1, 0, 1, 1, 1, 0, 0, . . .
would be coded as one civil war episode despite the one year break in ﬁghting above the 25 dead
threshold. By contrast, a sequence like . . . 0, 0, 1, 1, 0, 0, 1, 1, 0, 0 . . . would be coded as having two
civil wars onsets. An advantage of this approach is that it avoids potentially more subjective or dif-
ﬁcult judgments about whether a break in ﬁghting is an “end” or a merely a pause in a continuing
war. A disadvantage is that it tends to render long-running, relatively low level conﬂicts as any
civil wars. For example, if we consider all levels of conﬂict (minor through major), then by this
coding the most onset-prone countries since 1945 are India with 21, Burma with 19, and Ethiopia
with 13 conﬂict onsets. Though each country has certainly suffered a great deal of violent conﬂict,
the main reason that these numbers are so high is that they have suffered a particular form of con-
ﬂict – small rebel groups ﬁghting mainly over speciﬁc territories that often jump above or below
the 25 dead threshold. At the highest level of intensity (1000 or more dead per year on average),
this problem is much less serious. At that level, the most onset-prone countries in the sample are
Indonesia with 6, and Iraq and China with 5.

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 same
time, is that two civil wars or one civil war? At one extreme one could argue for making the thing
to be explained whether there is any war (or conﬂict) that starts in a country in a given year when
the country doesn’t already have a war occurring, thus ruling out the possibility of multiple civil
wars in one country. The ACD approach comes closer to the opposite extreme of coding a new civil
war 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 capture

                                                         7
the central government, and “territorial” where the rebel group aims to secede or win increased
autonomy for a speciﬁc region.5 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 conﬂicts in countries like Burma, Ethiopia,
and India, where multiple, often very small territorial rebel groups have operated, and often in the
context of what is arguably a larger civil war.

By contrast, if one rebel group defeats the central government and then some new rebel group
immediately arises to contest its control – think of Afghanistan or Somalia in 1991 – then for ACD
there is no new civil war since the ﬁghting is still about taking over the government. An alternative
approach codes a new civil war when one or both of the main combatants has changed (e.g., Fearon
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-
tives6 is that the ACD approach increases the number of civil war onsets in some highly ethnically
fractionalized countries that have had long-running internal conﬂict. And not just any ethnically
fractionalized 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 power
or play a signiﬁcant role in coalition politics. As noted, the ACD approach to coding multiple civil
conﬂicts leads to large numbers of conﬂict onsets for Ethiopia, Burma, and India, almost entirely
due to distinguishing conﬂicts between the state and multiple regional minorities as distinct civil
wars.7 The effect in conﬂict regressions is to increase the strength of association between onset
and various measures of ethnic fractionalization, when the dependent variable includes low-level
conﬂict. These tend not to be signiﬁcantly related in other civil war lists with higher magnitude
or intensity thresholds (Collier and Hoefﬂer 2004; Fearon and Laitin 2003), but they are often
signiﬁcant in ACD-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 is
that while more ethnically diverse countries have not been much more likely to have civil war over
the whole post-war period, they have been more likely to have lower-level conﬂicts with multiple
rebel groups ﬁghting in the name of diverse regional minorities. This appears to be particularly
  5
    This 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 coded
as “government.” Conﬂicts between Israel and Palestinian groups are coded as territorial.
  6
   Such as the Correlates of War-based coding used by Collier and Hoefﬂer (2004), Sambanis’ (2001), or
Fearon and Laitin’s (2003).
  7
   For example, in India: Nagaland, 4 conﬂict onsets; Mizoram, 1; Tripura, 2; Manipur, 3; Punjab, 1;
Kashmir, 1; Assam and Bodoland, 2 each. Other countries with a similar ethnic conﬁguration and many
ACD onsets include Indonesia, Iran, and Pakistan.


                                                    8
true for countries with many small minority groups, and relatively large “dominant” ethnic group
at the center.


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 except
that 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 conﬂict
lists. 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 quite
high, 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 wars
have continued to trend down in the last ﬁve or six years, but minor conﬂicts have jumped back
up. This could augur a return of more major conﬂicts (since minor conﬂicts become major if they
continue 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 interesting light on
the sources of the decline in total conﬂicts in the last 15 years. In Eastern Europe and the former
Soviet Union the spate of early 90s conﬂicts quickly subsided, or “froze” in some locales. There
has also been striking decline of both major and minor conﬂict in Latin America, which is plausibly
related to the end of the Cold War for most cases. Elsewhere, in subSaharan Africa and Asia there
have been some declines in major, high intensity conﬂicts, but no change in minor conﬂict in Asia
and at best a slight decline in Africa. The number of major conﬂicts has stayed fairly steady in
North Africa/Middle East, while the number of all conﬂicts there has declined slightly.

Figure 4 shows that despite the post-Cold War decline in the number and proportion of countries
with civil conﬂict, the share of the world’s population living in conﬂict-affected countries has
remained fairly steady or even increased, to one-third in 2008. Of course this does not mean that
one third of the world’s population has been directly affected by organized violent conﬂicts, since
they are mostly localized within countries. Still, it is a measure of prevalence and impact, and
reﬂection 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 income
over time. Notice that close to 80% of conﬂicts have occurred in countries with incomes below
the global median over the whole period, a number that shows no strong trend. However, there
has been a fairly steady increase in the share of conﬂicts in the second quartile on income, while
the share occurring in the poorest 25% has declined from almost two thirds in the early 60s to
about one third. Thus there has been some tendency for conﬂict to become more common among
middle- and lower-middle income countries.

                                                9
We should also describe the variation in the main outcome variables of interest in the analysis that
follows – conﬂict onset at different levels of intensity. Table 1 shows the total number of conﬂicts
at different intensity levels by region, both for the whole 1946-2008 period and for the post-Cold
War years (1990-2008). SubSaharan Africa and Asia have had the largest number of both large and
small conﬂicts over both periods. We see again that Latin America and, in terms of new conﬂicts,
Middle East/North Africa, have become rather more peaceful since the end of the Cold War.

Finally, note that while for the whole period major and minor conﬂicts were similar in frequency,
since the end of the Cold War there have been relatively fewer high intensity wars and relatively
more low-level conﬂicts. Figure 6 shows more directly. The decline in the proportion of high
intensity conﬂict is probably mainly due to the end of the Cold War, since it is known that foreign
intervention is strongly related to more intense conﬂicts and there was more foreign intervention
by major powers during the Cold War. However, improved international response through peace-
keeping and related methods may also be a factor.

                               Table 1: # conﬂicts by region and intensity
                                 1946-2008                                 1990-2008
                   major medium small minor             all major medium small minor          all
     SSA               24       22        8      40     94       8         11      3 28       50
     Asia              31       21        8      25     85       4          6      7 16       33
     MENA              20         6       3      16     45       5          1      0  9       15
     L. America        11         4       3      17     35       0          0      0  5        5
     E.Eur/FSU         12         7       2      11     32      10          5      2 10       27
     West               1         0       0       5      6       0          0      0  2        2
     total             99       60      24     114 297          27         23     12 70      132
     average kia/year: major ≥ 1000, medium 500-999, small 250-499, minor 25-249.




4 Correlates of civil war onset

41     aseline   odels

Table 2 reports the results of logit models with civil war or conﬂict onset as the dependent variable,
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 had average annual fatalities greater than 1000. The second uses an up-to-date version of




                                                 10
Fearon and Laitin’s civil war list, and shows that the results are nearly the same.8 In the third
model the dependent variable is the onset of any level of ACD conﬂict, from minor to major.
Results are again similar, although a number of coefﬁcients shrink a bit towards zero. The most
striking difference is that ethnic fractionalization has a much larger apparent “effect” when we
include minor conﬂicts; this is mainly for the reasons discussed in the previous section.

Models 4 and 5 are Models 1 and 3 but estimated with country and decade ﬁxed effects. The factors
that vary a lot within countries over time have similar effect estimates. Those that do not, like
income 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 World
         Tables 6.3 data extended where possible and necessary by World Bank growth rates.9
       • 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 John Gerrard.
       • oil, which marks whether the country is a major oil producer, coded by whether one third or
         more of its GDP comes from natural resources (based on World Bank data).
       • ne state, which marks whether the country is in its ﬁrst two years of political independence.
       • olitical insta ility, which marks whether, in the previous country year (t − 2 to t − 1), there
        was any change in the Polity 2 score. The Polity score is a measure of democracy that runs
        from -10 (extreme autocracy) to 10 (full democracy).
       • anocracy, which marks whether the country’s Polity 2 score is between -5 and 5 (inclusive)
         in the previous year. This is a measure of partial, or weak democracy.10
   8
    FL’s threshold criterion is 1000 total killed, rather than the annual intensity threshold used here. As a
result the FL is slightly more permissive in the sense of letting in more conﬂicts. It is, in effect, intermediate
between “major” and “all ACD,” though much closer to major.
   9
     Listwise deletion due to missing income data is a big problem for civil conﬂict regressions, because it is
not missing at random – civil war countries are more likely to have missing income data. The data used here
is a relatively complete set of estimates, drawing primarily on PWT6.3 but using World Bank and Maddison
estimates for some missing years and countries
  10
    As Vreeland (2009) points out, some components of the aggregate Polity measure may be coded in part
on the basis of observation of armed conﬂict, in which case the “effect” of anocracy may not be an effect of a

                                                       11
     •         is a measure of the ethnic fractionalization of the country, based on estimates of ethnic
         group populations from a Soviet ethnographic atlas compiled in the early 1960s, and updated
         for some newer countries (Fearon and Laitin 2003). It can be interpreted as the probability
         that two randomly drawn individuals from the country are from different ethnic groups.
     • religious fractionali ation is a similar measure of religious diversity (Fearon and Laitin
       2003).



42         ffect    agnitudes

Table 2 provides estimates of the substantive magnitude of “effects” in Table 1, Model 1. The
baseline country in this example is a stable, non-oil producing autocracy with no civil war in
progress, with the median income ($4,100), population (8.3 million), and “mountainousness” (9%).
Such a country had about a .57% chance of civil war outbreak in any given year, which translates
to 2.8% over ﬁve years and 5.6% over ten. These small numbers should remind us that new civil
wars do not break out that often. With the ACD major wars, the rate is only 1.14% of all country
years from 1946 to 2008, and about 1.44 new civil wars per year (using the sample with data on
the covariates, which has about 156 countries in it). Counting all conﬂicts, major through minor,
the corresponding rates are 3.3% and 4.1%.

The “relative risk” column shows how the indicated change affects the odds of civil war onset in
the next year. For example, moving from the 75th to the 25th percentile on per capita income
is associated with a 2.03-fold increase in the annual odds of onset. Thus, the relative risk score
provides 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 higher
risk of civil war onset than other states – more than six times greater, using the ACD data. The
fact that the effect estimate is similar in the ﬁxed-effects model shows that this is not an artifact
of states that became independent since 1945 having higher average conﬂict risk than that of older
states. Rather, even within former colonies, the ﬁrst two years are the most dangerous. I discuss
the implications and interpretation of this observation in section 5 below.

After “new state,” oil producers are estimated in this model to have about triple the annual risk of
major civil war onset. Any change in the Polity index in the previous year (whether in a democratic
or autocratic direction) about doubles the subsequent odds, as does moving from 75th to 25th
percentile on income or from the 25th to 75th on rough terrain. Anocracy (partial democracy) is
associated with 29% greater annual odds of civil war outbreak, although this is not statistically
signiﬁcant in this model where the dependent variable is major war).

class of political systems. Instead, it could be that “partial democracy” is at least partly picking up incipient


                                                       12
                                  Table 2: Correlates of civil war onset, 1946-2008
      Model                            1                 2             3             4             5
      DV                         ACD major war FL civil war all ACD ACD major war all ACD
      log(gdpt−1 )                      −0.404∗         −0.404∗∗∗ −0.351∗∗∗             0.012      0.123
                                         (0.177)         (0.111) (0.099)               (0.384)    (0.220)
      log(popt−1 )                         0.203∗∗        0.313∗∗∗ 0.238∗∗∗           −0.192     −0.588
                                         (0.067)         (0.066) (0.056)               (0.654)    (0.423)
      log(% mountains)                     0.360∗∗∗       0.186 ∗     0.151 ∗

                                         (0.100)         (0.072) (0.062)
      oil producer                         1.095∗∗∗       0.698∗∗ 0.715∗∗               0.978      0.326
                                         (0.297)         (0.242) (0.219)               (0.617)    (0.415)
      new state                            1.862∗∗∗       1.913∗∗∗ 1.336∗∗∗             1.084∗     1.125∗∗∗
                                         (0.377)         (0.307) (0.292)               (0.452)    (0.321)
      pol instabilityt−1                   0.706∗∗        0.746∗∗∗ 0.466∗∗              0.714∗     0.484∗∗
                                         (0.261)         (0.205) (0.170)               (0.293)    (0.184)
      anocracyt−1                          0.258          0.482 ∗     0.355 ∗           0.281      0.636∗∗
                                         (0.252)         (0.199) (0.175)               (0.313)    (0.200)
      democracyt−1                      −0.245          −0.466        0.080             0.232      0.276
                                         (0.408)         (0.329) (0.215)               (0.451)    (0.263)
      ELF                                  0.534          0.521       1.153∗∗∗
                                         (0.389)         (0.344) (0.287)
      relig frac                           0.597        −0.176 −0.228
                                         (0.756)         (0.499) (0.429)
      prior war                         −0.519          −0.548∗∗ 0.204                −1.861∗∗∗ −1.111∗∗∗
                                         (0.320)         (0.185) (0.228)               (0.400)    (0.203)
      constant                          −4.868  ∗∗      −4.795  ∗∗∗ −4.001∗∗∗

                                         (1.583)         (1.080) (0.982)
      N                              7929              7985          7873        2804 (55)     4805 (96)
      Fixed effectsa                  No                No            No            Yes          Yes
      Standard errors clustered by country in parentheses
      †
          signiﬁcant at p < .10; ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001
      a
          By country and decades.




armed conﬂict.




                                                                13
                            Table 3: Magnitude of “effect” estimates, Model 1
   variable            levelpctile % chance onset % chance onset % chance onset              relative risk
                                          in 1 year       over 5 years    over 10 years
   baselinea                   50               .57               2.84              5.6
   gdp/cap           $1,665    25              0.83               4.06             7.96              2.03
                     $9,612    75              0.41               2.03             4.01

   population          3.3m      25              0.48              2.37               4.68
                      21.5m      75               0.7              3.43               6.74           1.46

   % mountains           1.7     25              0.35              1.76               3.49
                        27.2     75              0.82              4.04               7.92           2.33

   ELF                  0.12     25              0.51              2.52               4.97           1.34
                        0.66     75              0.68              3.35               6.59

   baselinea                     50              0.57              2.84                5.6

   new state                                      3.59                                                6.44
   instability                                    1.16                5.66             10.99          2.03
   oil                                             1.7                8.21             15.75          2.99
   anocracy                                       0.74                3.66              7.18          1.29
   a an 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.



5 Other country characteristics of potential interest

51        opulation growth land pressure and youth bulges

So-called “neo-Malthusians” fear that rapid population growth, or population per hectare of arable
land, causes violent conﬂict by increasing competition for resources.11 A more speciﬁc argument
in this vein is that what matters is the ratio of young males to the rest of the population (or to just
the 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 population
density on arable land, but Urdal (2006) ﬁnds evidence in favor of a relationship between the
share of the adult population under 25 and conﬂict risk. Fearon and Laitin (2003) and Collier and
Hoefﬂer (2004) had found no relationship for the (very similar) measure of share of total population
  11
       Huntington (1996) sometimes stressed population pressures. See also Homer-Dixon (2001).


                                                     14
between 15 and 24.

Using our ACD-based measures for onset of major and all conﬂicts, I ﬁnd that lagged population
growth rates are positively associated with conﬂict risk, but the estimated coefﬁcients are substan-
tively very small and not statistically signiﬁcant. When added to Model 1 in Table 1, lagged popu-
lation growth rate gets an estimated coefﬁcient of about .08 with a standard error of .11 (p = .45).12
Somewhat smaller standard errors and slightly higher estimated coefﬁcients can be found if one
drops various other variables from the model. Population growth rates are moderately well cor-
related with several other variables, such as 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” with these other variables, but not by much. Results are similar whether
we look at major or all conﬂict onsets.

Results for a measure of arable land per capita (from the World Development Indicators) are sim-
ilar, or even weaker. Arable land (in hectares) per capita is negatively related to civil war onset
when added to Models 1 or 3, but not close to statistically signiﬁcant. This measure of resource
scarcity, if that is what it really is, shows no consistent association with higher civil war risk.

Urdal (2006) found that “youth bulge” as measured by the share of 15-to-24 year olds in the total
population is not related to onset risk, which is what Fearon and Laitin (2003) and Collier and
Hoefﬂer (2004) also found. However, Urdal reported that “youth bulge” as measured by the share
of 15-to-24 year olds in the a ult population as related to civil war onset in his data. He argued
that “from a theoretical perspective” (p. 615) the latter indicator is to be preferred, but it is not
clear 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 substantively
small and not close to statistically signiﬁcant (p = .78). The estimated coefﬁcient for income
weakens slightly – low income and youthful populations are highly correlated – but remains close
to signiﬁcant at p = .11; otherwise nothing changes much. If we use all ACD onsets, “youth
bulge” as a share of adult population is marginally signiﬁcant at p = .097, taking away from
income whose coefﬁcient halves but keeps a p value of .078. It thus appears that “youth bulge” as
Urdal formulates it performs less well as a predictor when the dependent variable is based on this
ACD civil war list than in the formulation he used. 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. Population
  12
    I 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 inﬂuential
outliers. If these are left in, results are highly erratic depending on speciﬁcation.


                                                   15
Division data), which allowed an extension of the data to 2009. I also used the data for males
between the ages of 15 and 25, rather than both males and females (although this almost surely
would make no difference at all). I ﬁnd that neither young males as a share of adult or total
population is signiﬁcantly related to civil war onset in these data. Added to the speciﬁcations in
Model 1 or Model 3, or paired with income and other combinations of variables, it may take a
positive or negative coefﬁcient but the estimates are not close to being statistically or substantively
signiﬁcant. Once again, young males as a share of adult population (but not young males as a
share of total population) performs somewhat better 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 difﬁcult
to get stable or sharp estimates of the partial correlation with civil war onset controlling for other
variables. There is a tendency for income per capita to “trump” youth bulge in these data, a
tendency that is very strong when youth bulge is measured as youth over total population and
fairly strong 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 of these two
different measures should be particularly different. So I am inclined to think that the evidence is
not very good that population structure explains a big part of why poor countries are more civil
war prone, rather than something else about poor countries explaining why countries with young
populations happen to be more civil war prone.

“Opportunity cost” arguments about civil conﬂict hold that conﬂict is favored by anything that
makes joining a rebel band ﬁnancially more attractive for young men. Unemployment is sometimes
suggested as one such factor, and one might hope that “youth bulge” could provide a reasonable
cross-national proxy for youth unemployment rates. Direct cross-national data on unemployment
(from the WDI) is not usable for this purpose, as it is self-reported according to deﬁnitions that
vary greatly across countries, and the reporting is itself highly erratic across countries and years.

I considered whether youth bulge is related to the WDI unemployment rates for countries that
might be expected to have similar reporting standards and where data is available. Unfortunately,
there is no clear indication that higher unemployment tends to go along with many young males
as a share of the adult or total population. As a result, the above results simply cannot speak to
whether higher unemployment rates are correlated with higher risk of conﬂict.


52     ertical and hori ontal inco e ine uality

The bivariate relationship between income inequality and civil war onset in these data is actually
negative (more inequality, lower odds of conﬂict), although not statistically signiﬁcant. The nega-
tive sign persists when we add the covariates in Table 1, Model 1, or subsets of them, and remains


                                                  16
insigniﬁcant (p = .17) in the full model. For all ACD onsets, the estimates are essentially zero. Nor
is this a matter of inequality picking up regional “effects,” as results are the same when regional
dummies 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 claims
about the causes of civil conﬂict, not only is there no apparent positive correlation between income
inequality and conﬂict, but if anything across countries those with more equal income distributions
have been marginally more conﬂict prone.

Some have proposed that a more relevant form of income inequality is “horizontal,” meaning across
groups within a country as deﬁned by ethnicity or religion (Stewart 2002). Good measures of in-
equality across groups are hard to ﬁnd and construct, however. Early efforts to use the Minorities at
Risk data (Gurr 199x) to examine the relationship between economic disadvantage of “minorities
at risk” and propensity to rebel found inconsistent or no evidence (Gurr 199x, Moore 199x, Fearon
and Laitin 1999). More recently there have been some efforts to use the Demographic and Health
Surveys (http://www.measuredhs.com/) to construct measures of the relative economic standing of
different ethnic groups in subSaharan Africa. Results are inconsistent, though overall not very sup-
portive of the “horizontal inequalities” proposition. Østby (2008) ﬁnds at best very faint evidence
                                                                                                   a
that ethnic horizontal inequalities are related to conﬂict onset at the country level. Østby, Nord˚ s
and Rød (2009) ﬁnd positive but not signiﬁcant coefﬁcients on measures of the relative economic
and social deprivation of a region (relative to the rest of the country) and conﬂict onset in Africa.
Using similar measures of economic welfare but looking speciﬁcally at ethnic groups rather than
administrative regions, Condra (2009) ﬁnds that, if anything, it is the relatively better off ethnic
groups have been more prone to be involved in rebellions in sub-Saharan Africa.

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 more
likely 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 data
for 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 completely
driven by the four observations where the coded EGIP is a minority. Unclear on what the coding
rules were for EGIPs, Fearon, Kasara, and Laitin coded instead the ethnicity of the ruler for all
country years in all regions. They ﬁnd a positive but statistically insigniﬁcant relationship between
  13
   I have used the WIDER inequality measure, which is based on an updated version of Deininger and
Squire. The variable is constructed as the average of all observations within each country. Results are the
same if we interpolate.



                                                    17
rule by a member of an ethnic minority and civil war onset.14 In their analysis (which, like Ceder-
man and Girardin, used the Fearon and Laitin (2003) model and civil war list), ethnic minority 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 ethnic minority
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 systematic
effort they undertook to code EGIPs and MEGs. They used some process of expert surveys and
consultations to assess ﬁrst whether ethnicity was “politically relevant,” in a country year, meaning
that their coders perceived discrimination or politicians mobilizing based on 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, and thus 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 ﬁnd that the log
of the share 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 of ethnic civil wars.

Importing Wimmer et al.’s country-level codings into the data set considered here, I ﬁnd that the
log of the lagged share of “excluded groups” is positively related to major conﬂict onset odds
when added to Model 1.17 The coefﬁcient is .24 with a standard error of .08; substantively this
implies that moving from a country with no excluded groups to one where 23% of the population
of ethnopolitical groups are excluded (25th to 75th percentiles) approximately doubles the annual
risk of civil war outbreak. The unlogged version of the variable is much more weakly related.
With ﬁxed effects the estimated coefﬁcient is essentially zero, which suggests that “exclusion” is
picking up enduring characteristics of countries more than a switch to exclusion of a particular
group signiﬁcantly increases the average risk of civil war.

When we include all ACD conﬂicts, minor through major, the estimated coefﬁcient on the log of
  14
    Related 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.
  15
    There are also (as usual) concerns about endogeneity. Cross-sectional analysis could understate the
conﬂict-generating effects of ethnic minority rule, if it is more likely in precisely those places where it is
viewed as tolerable. On the other hand, ethnic minorities may in some cases work especially hard to attain
and hold onto power where they would be highly threatened if they lost power (e.g., Syria, Iraq). Fixed
effects models that try to control for country-speciﬁc characteristics like this yield unstable, though usually
positive, estimates, because there are so few cases to go on.
  16
     In all three cases politicians have mobilized along ethnic lines (in Papua New Guinea, this is essentially
all there is), but the broader concern is whether this is coding on the dependent variable or not.
  17
       I added .01 to avoid log of zero; I am not sure what Wimmer et al. do.


                                                       18
Wimmer et al.s’ excluded groups variable is one third as large and is not statistically signiﬁcant.

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 this
means). Using as a predictor the share of ethnopolitical groups that are “discriminated against”
by Wimmer et al.s’ codings yields similar results to “excluded,” except that the (logged) variable
is now also a signiﬁcant predictor of all ACD onsets. In sum, Wimmer et al.’s study suggests that
countries that raters judge to have bad ethnic relations and discrimination against relatively larger
groups are more prone to civil conﬂict.18

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 political
exclusion and discrimination are based on the subjective judgements of diverse coders, trying to
code somewhat impressionistic things. Countries where there has been no ethnic conﬂict and where
ethnic relations have been calm are for that reason judged to have a low value on “exclusion” –
thus, the dependent variable at least partly 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
country increases the probability that he or she judges that, earlier on, groups were discriminated
against or politically excluded. And codings of discrimination at time t may be based on earlier
experiences of 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 the
population that is “excluded” or discriminated against by government policy, we are now running
a olicy 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 results
of policy choices, as well. But they are produced by policy choices over longer periods of time,
and arguably much more indirectly, than a variable that tries to measure current government policy
with respect to an ethnic minority. If we have concerns about the endogeneity of income and ELF,
which we should, then we must have them far more strongly about a direct policy.19

It is very important to understand that the endogeneity of the policy choice might lead us to over-
or underestimate the average causal effect of introducing a more (or less) exclusionary policy in a
typical country. For example, if governments tend to calibrate levels of exclusion to what they can
get away with, then estimates from panel data may understate what would be the causal impact
  18
    Cederman, Wimmer and Min (2010) use their EPR data to present a similar analysis at the level of
ethnic groups in countries. They ﬁnd similarly that “excluded” groups are more likely to have been engaged
in rebellions than are groups judged by their expert raters to have some share of power in the central state.
At the group level of analysis, they report that this relationship still obtains when they include country ﬁxed
effects.
  19
       See Rodrik (2005) for a nice discussion of this issue in the context of studies of economic growth.


                                                        19
of arbitrarily switching to more exclusionary policies (Fearon and Laitin 2010). Alternatively, to
the 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 impact
of exclusion or discrimination.

Buhaug, Cederman and Rød (2008) consider determinants of ethnic conﬂict at the group (rather
than the country) level. They ﬁnd that Eurasian ethnic groups that are larger, live farther from the
capital, and in more mountainous terrain are more likely to be involved in ACD minor or major
level conﬂicts.20 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 least as plausible as measures of capability to sustain rebellion as of
motivation to rebel.21


5 3 Civil liberties hu an rights abuses

Are countries whose governments abuse human rights and restrict civil liberties more prone to civil
war onset? This certainly seems plausible and likely on its face. The question is motivated by an
intuition similar to that behind studies of horizontal inequality, though here the focus is on whether
repressive or restrictive government policies favor war even if they are not necessarily directed at
any particular ethnic or religious group.

Once again, however, we need to be careful about the interpretation of results, since abuse of human
rights and restriction of civil liberties are policy choices and thus almost surely are endogenous to
other causes of conﬂict. Even worse, there is a major danger that these indicators may simply pick
up onset of civil conﬂict before it happens to get coded by ACD or others. Is this government
abuse that is causing a conﬂict, or government abuse that is already part of a conﬂict we are trying
to explain? Still, it may be interesting to know how strong is the correlation between existing
measures and civil war onset.

Freedom House provides a 1-to-7 scale of government observance of civil liberties for a large
number of countries since 1972. Although procedures have varied over the years, for the most part
the 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, and
Personal Autonomy and Individual Rights.” Higher values on the scale indicate fe er civil liberties
for 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.
  20
       Condra (2009) reports similar ﬁndings for groups in Africa.
  21
    All 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.


                                                       20
For example, a country may also be judged to have worse civil liberties if it has “groups opposed
to the state [that] engage in political terror that undermines other freedoms.” Thus to some extent
the scale may incorporate a measure of civil conﬂict.

The civil liberties measure proves to be highly correlated with other measures of democracy – for
instance, r = −.85 with Polity – which we have already seen is not a signiﬁcant predictor of civil
war or conﬂict onset. When added to Model 1, the coefﬁcient on lagged civil liberties is positive
(worse civil liberties, higher conﬂict risk) and just statistically signiﬁcant if the Polity measures of
regime type are not included (p = .083). However, if lagged for two years, the estimate is close to
zero and not at all signiﬁcant. This suggests strongly that that the measure is picking up incipient
civil war, at least as much as bad civil liberties performance causes greater civil war risk.

In addition, estimates are smaller and insigniﬁcant (even with the one year lag) if we include the
variables for anocracy and democracy as measured by Polity; if we look at simpler speciﬁcations
provided income is included; and if the dependent variable is all conﬂict (minor to major). Ba-
sically, observance of civil liberties behaves similar to democracy indicators. Indeed, if we add
a squared term (and drop the Polity measures), we ﬁnd evidence of the inverted U. Other things
equal, civil war risk is highest for countries with a Freedom House civil liberties score of 5 out
of 7, which they describe as countries “with a combination of high or medium 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 concern
may be that including lagged civil liberties is like including an indicator for whether there was
conﬂict in the previous period. I also tried running Model 1 (and variants) only for cases in which
there was no civil war in the prior period. This leads to the coefﬁcient on civil liberties doubling
(or more), and marginally greater statistical signiﬁcance. Overall, it is difﬁcult to know whether
these results suggest that democracy has a direct effect of lowering conﬂict risk, or if government
abuse of civil liberties has a direct effect, or neither.

A potentially more focused measure of government abusiveness is the Political Terror Scale, an
annual 1-to-5 index based on coding of Amnesty International and State Department human rights
reports from 1976 to 2008 (Gibney, Cornett and Wood 2008). Mark Gibney writes that “Coders
are instructed not to turn a blind eye towards violence by non-state actors, but that their primary
goal is to measure levels of violence by the state.”22
  22
    There 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 levels
is: “(Level 5) Terror has expanded to the whole population. The leaders of these societies place no limits
on the means or thoroughness with which they pursue personal or ideological goals. (Level 4) Civil and
political rights violations have expanded to large numbers of the population. Murders, disappearances, and
torture are a common part of life. In spite of its generality, on this level terror affects those who interest
themselves in politics or ideas. (Level 3) There is extensive political imprisonment, or a recent history


                                                     21
The main difﬁculty that this measure poses is that it is hard to separate out government abusiveness
that is part of a civil war from government abusiveness that causes a civil war onset. Of the 774
country years coded as 4 or 5 on the PTS (the highest two levels of human rights violations), almost
half are ACD major conﬂict years, and almost 70% have some kind of ACD conﬂict in progress.
In other words, most of the worst human rights abuses, according to these data, are carried out
by governments during civil wars. The problem is then that if lagged human rights performance
predicts onset in a regression model, we don’t know how much this is because the onset of civil
war is coded with error, and how much it is because government abusiveness caused rebellion. A
second difﬁculty is that the lagged PTS 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, both
for major conﬂicts and for all conﬂicts. With a one-year lag the scale appears to be very strongly
related to subsequent civil war onset, both in terms of a substantively large estimated coefﬁcient
and statistical signiﬁcance. However, for major wars the estimate for PTS when lagged ﬁve years
is only one-fourth as large as for the one-year lag, which suggests that there may indeed be a
big problem with the one-year lag picking up conﬂicts that have already started (this may also
happen for the ﬁve-year lag, but hopefully less so). We also see this when the dependent variable
is all ACD conﬂicts, although here there is still a statistically signiﬁcant relationship between PTS
lagged ﬁve years and conﬂict onset.

Substantively, the estimated association for the one-year-lag with major conﬂicts is simply enor-
mous: the .88 estimate in Model 1 means that each ste up the Political Terror Scale associates
with a 2.4 times increase in the odds of civil war onset in the next year. This would imply that
country years with a 5 on PTS have about an 80 times greater odds of onset in the next year than
those with 1! If we use the one year lag estimate for all conﬂicts, each step on the scale asso-
ciates with an 43% increase in odds, and the difference between 5 and 1 is a factor of 6, which is
comparable to the effect of “new state.”

The extremely large estimate for the “effect” of lagged human rights abuses on civil war onset
suggests that something may be going wrong in the estimation. Closer investigation reveals that
this is the case: zero major conﬂict onsets are recorded for states with a 1 on the PTS scale, which
means that the relative risk in going from 1 to 2 is inﬁnitely large. The models in Table 3 implicitly
treat PTS as a linear scale, estimating an average “effect” of moving from 1 to 2, 2 to 3, and so
on. If we dummy out each of the ﬁve levels, we ﬁnd that we cannot estimate a logit model because
there were zero ACD civil war onsets for states at PTS level 1. (Level 1 states are described by

of such imprisonment. Execution or other political murders and brutality may be common. Unlimited
detention, with or without a trial, for political views is accepted. (Level 2) There is a limited amount of
imprisonment for nonviolent political activity. However, few persons are affected, torture and beatings
are exceptional. Political murder is rare. (Level 1) Countries under a secure rule of law, people are not
imprisoned for their views, and torture is rare or exceptional. Political murders are extremely rare.”


                                                    22
Gibney as countries “under a secure rule of law, [where] people are not imprisoned for their views,
and torture is rare or exceptional.”)

Table 4 shows the frequency of major and all conﬂict onsets for each PTS level, by both one- or
ﬁve-year lagged human rights abuses. We see that major conﬂicts almost never follow when prior
PTS levels were less than 4, but there is a major jump at level 4. This is also the case for all
conﬂicts, although when we include the minor conﬂicts there is also some evidence of increasing
risk as abuses go from 1 to 2 and perhaps 2 to 3.23

                Table 4: Annual frequency of onsets by levels of lagged human rights abuses

                                PTS      major      major           all         all
                                       1-yr lag   5-yr lag    1-yr lag    5-yr lag
                                1             0          0      .0059       .0044
                                2        .0046      .0094       .0249       .0335
                                3        .0092      .0045       .0408       .0309
                                4        .0202      .0181       .0927       .1069
                                5        .0387      .0127       .0718       .0506


According to Gibney, level 4 refers to countries in which “Civil and political rights violations have
expanded to large numbers of the population. Murders, disappearances, and torture are a common
part of life. In spite of its generality, on this level terror affects those who interest themselves in
politics or ideas.” As compared to levels 1-3, this seems likely to pick up mainly countries with
actual or incipient civil war.

Table 5 lists country years for which ACD codes the start of a civil war (a major conﬂict), and the
PTS scale was 4 or greater in the previous year. There are a number of cases for which the high
PTS score in the prior year is almost surely related to the fact that there was already a conﬂict in
progress in the country. For a few of the others, it is plausible to see the conﬂict arising in some
signiﬁcant way from oppressive prior rule – for example, Cambodia 1978, Uganda 1978, and
Yugoslavia 1998. But there are more cases where the lagged PTS coding is picking up incipient or
residual civil war activity that is for different reasons missed by ACD – for example, Mozambique
1977, El Salvador 1979, Sri Lanka 1984, Pakistan 1995 and 2007, Rwanda 1997, Russia 1999,
Iraq 2004, Israel 2006, and Somalia 2006.
  23
    These patterns hold up when we use a linear probability model (ordinary least squares with conﬂict
onset as the dependent variable), and control for the other covariates of onset. Using robust (and country
clustered) standard errors to try to deal with the heteroscedasticity that arises in the linear probability model,
only levels 4 (and in a couple of case 5) are signiﬁcantly different from level 1. In all of these models, I
group 1.5 with 1, 2.5 with 2, and so on, for observations where the State Department and Amnesty disagreed
so that there is an averaged value between integers.


                                                       23
Overall, the analysis suggests that very poor human rights performance is a very bad sign for a
government: 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. It
is plausible that government abuses would often encourage support for rebellion, but it is also
possible 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 this
literature consists of case studies that suggest that indiscriminate brutality by government forces
increased 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 a
bit mixed. Lyall (2009) ﬁnds, somewhat surprisingly, that random artillery shelling by Russian
troops in Chechnya suppressed rebel activity. Lyall and Wilson (2009), who examine the effect
of mechanized counterinsurgency on government success in a large sample of guerrilla wars and
in a case study of Iraq, would seem to suggest the opposite, as does Condra and Shapiro’s (2010)
incident study of Iraq.

                       Table 5:   ACD war onsets in countries that had PTSt−1 ≥ 4
                country             year wart−1 ? any conﬂictt−1 ? PTSt−1 PTSt−5
                Mozambique         1977        0                    0      4.0
                Uganda             1978        0                    0      5.0
                Cambodia           1978        0                    0      5.0
                El Salvador        1979        0                    0      4.5
                Sri Lanka          1984        0                    0      4.5    2.5
                Iran               1986        1                    1      5.0    4.5
                India              1989        1                    1      4.0    3.0
                Iraq               1991        1                    1      5.0    4.0
                Bosnia             1993        1                    1      5.0
                Pakistan           1995        0                    0      4.0    3.5
                DRC                1996        0                    0      4.0    3.5
                Rwanda             1997        0                    0      5.0    4.0
                Yugoslavia         1998        0                    0      4.0    4.0
                Angola             1998        0                    1      4.0    5.0
                Russia             1999        0                    0      4.5    4.5
                Iraq               2004        0                    0      5.0    5.0
                Somalia            2006        0                    0      4.0    4.0
                Israel             2006        0                    1      4.0    4.5
                Pakistan           2007        0                    1      4.0    4.0



                                                   24
54     olitical refor   and civil war onset

The results in Table 1 show a strong relationship between changes in governing institutions, as
measured by change in components of the Polity index, and an increased probability of civil war
onset in subsequent years. The coefﬁcient estimate is the same when we add country ﬁxed effects,
so it is not just a matter of more unstable countries having more civil wars. Rather, civil war
somewhat reliably follows institutional changes in the direction of greater democracy or greater
autocracy.

Does it matter whether the change is towards more democracy or autocracy? Previous work on
civil war onset ﬁnds no signiﬁcant differences, with a slight tendency for greater risk following
autocratizing change (e.g., Fearon and Laitin 2003). With the formulation of the onset variables
used here, we ﬁnd a slightly stronger tendency for conﬂict to follow changes in an autocratic
direction than in a democratic direction. This is shown in Table 6, where I have added to the core
speciﬁcation dummy variables marking whether there was a one-or-greater change 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). Autocratizing change increases the odds of war onset in the next year by
a factor of 2.54 on average, which is quite large. Democratizing change associates with increased
odds of about 60%, though the estimate is not statistically distinguishable from zero. The results
are similar though a bit smaller in magnitude for all conﬂicts.

The likelihood that democratic reform leads to civil conﬂict does not vary much across levels of
development. The risk is slightly lower as one moves up the income scale, but not signiﬁcantly
so. Nor does it matter much what the initial level of democracy is. Civil war onsets have been
somewhat more likely to occur following democratizing change in an autocracy or a democracy,
as compared to an “anocracy” (a partial democracy). For all conﬂicts, minor and major, the risk
of onset does not seem to vary much with initial regime type at all. One might have expected
that democratic reform would be more likely to end in violent conﬂict the more authoritarian the
reforming state. It is of course possible that 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). So these measures
of democratizing and autocratizing change should be understood to be “short term,” and not likely
to pick up more complicated political trajectories.


55     apid governance refor      and conﬂict onset

Some case studies prepared for the WDR 2011 have suggested that rapid a inistrative or olitical
refor s may actually heighten the risk of violent conﬂict in a country in the short run, even if the
broader effects could be quite beneﬁcial in the longer run. In this section I consider whether
conﬂict onset is correlated with large positive changes in several governance indicators that might

                                                25
                         Table 6: Change of political regime and conﬂict risk
                                              major wars       all conﬂicts
                          autoc. change              0.933 ∗∗          0.690∗∗
                                                    (0.348)           (0.265)
                          democ. change              0.481             0.324
                                                    (0.371)           (0.221)
                          log(gdpt−1 )             −0.454∗∗          −0.379∗∗∗
                                                    (0.165)           (0.094)
                          log(popt−1 )               0.205∗∗           0.245∗∗∗
                                                    (0.071)           (0.060)
                          log(% mountains)           0.329∗∗∗          0.149∗
                                                    (0.093)           (0.061)
                          new state                  1.771 ∗∗∗         1.429∗∗∗
                                                    (0.495)           (0.356)
                          oil producer               1.089 ∗∗∗         0.761∗∗∗
                                                    (0.310)           (0.219)
                          anocracyt−1                0.328             0.378∗
                                                    (0.277)           (0.187)
                          democracyt−1             −0.199              0.146
                                                    (0.388)           (0.228)
                          ELF                        0.475             1.080∗∗∗
                                                    (0.370)           (0.296)
                          prior war                                    0.046
                                                                      (0.242)
                          prior war                −0.533†
                                                    (0.318)
                          constant                 −4.181∗∗          −3.898∗∗∗
                                                    (1.370)           (0.908)
                          N                      7836              7780
                           Robust standard errors in parentheses
                           Regime changes are measured from year t − 2 to t − 1.
                           †
                               signiﬁcant at p < .10; ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001




potentially pick up attempts at major administrative reform.

The “reform” indicators used here are drawn from two main sources, both of which are presented
and discussed in more detail in the section below on governance and conﬂict. The ﬁrst set come
from the International Country Risk Guide (ICRG), a company that sells assessments of business
conditions and political risk to investors. They produce a number of indices that attempt to tap
various dimensions of political risk, and that appear to be derived primarily from expert surveys. I


                                                         26
try to identify major efforts at administrative reform by coding for large, fairly rapid changes in the
ICRG’s corruption index, “bureaucratic quality” index, and a variable called “military in politics”
that is said to measure “the degree of military participation in politics” (higher numbers indicate
less military involvement). The corruption and “military in politics” measures are scales from 1
to 6, while bureaucratic quality ranges from 1 to 4; higher values are better. We have annual data
from 1984 to 2006 for between 110 and 139 countries. Unfortunately, relatively conﬂict-prone
countries are more likely to be missing.

Table 7 lists the top 15 country years for two-year increases in the ICRG corruption, bureaucratic
quality, and military in politics measures. That is, columns 3, 6, and 9 in the table show the change
in the scale from year t − 2 to t, and these countries saw the largest such changes in a two-year
period in our sample. Inspection suggests that a large number of these are cases of substantial
reform during, or in the wake of, a successful episode of democratization. In some cases, it is
reform following the end of a civil war as well.

Inspection also suggests that very few of these country years saw, or were soon followed by, the
onset of violent conﬂict. Following major improvements in government corruption (the ﬁrst three
columns), only Liberia 2001 saw new armed conﬂict soon after, and this does not seem like a
compelling example of an anti-corruption drive stirring up conﬂict. On bureaucratic quality, there
is Angola 1996 (new war in 1998) and Algeria 1989-90; the usual narratives for these cases do not
mention destabilizing impacts of administrative reform, though clearly blocked democratization in
Algeria was a part of the sequence of events leading to the violence of the 1990s. The largest and
most rapid reductions in the political role of the military (columns 7-9) clearly tend to occur after
democratization and often civil war. The only contender here would be Yugoslavia 1997 (war in
1998).




                                                  27
                                  Table 7: Top 15 reform episodes by ICRG measures
            country     year   Δ corruption      country     year   Δ bur. qual.        country     year   Δ mil. in pol.
           Bahamas     1994            4.00       Algeria   1989           2.00           Niger    2001             4.00
           S. Korea    1992            3.00       Angola    1996           2.00         Algeria    2005             3.00
           Paraguay    1992            3.00    Indonesia    1993           2.00            Chile   1991             3.00
             Liberia   2001            2.50      Panama     1998           2.00      Guatemala     2005             3.00
          Indonesia    1992            2.42       Poland    1991           2.00           Israel   1992             3.00
           Bulgaria    1993            2.00     Paraguay    1993           1.75            Mali    1994             3.00
               Chile   2006            2.00        Ghana    1990           1.67           Korea    1990             2.92
              Egypt    1993            2.00     Uruguay     1997           1.67          Poland    1990             2.92
              Ghana    1990            2.00         Malta   1996           1.50        Lebanon     1992             2.92
         Guatemala     1997            2.00     Namibia     1992           1.46            Haiti   1996             2.75
            Guyana     1995            2.00        Gabon    1986           1.45            Peru    2002             2.67
           Lebanon     1994            2.00        Kenya    1986           1.36        Namibia     1993             2.50
               Mali    1994            2.00       Kuwait    1993           1.33    Sierra Leone    2003             2.50
          Nicaragua    1986            2.00   Philippines   1997           1.33      Yugoslavia    1997             2.50
         Philippines   1987            2.00        Egypt    1986           1.09          Kuwait    1993             2.42

         Numbers are change in governance index from t − 2 to t.



So there is no strong suggestion here that the most rapidly improving states on corruption, bu-
reaucratic quality, and role of the military (as coded by the ICRG) have often been followed by
increased civil conﬂict. Of course, we should really check the larger sample that has variation on
the independent variables. For this purpose I construct dummy variables marking country years
where the two-year increase or decrease in the relevant governance indicator is in the top and bot-
tom 10% of all changes in the variable. The dummy marking large increases in the variable thus
codes “reform” and the dummy marking a large decline marks “regression.” I then added lagged
versions of these variables two at a time to the baseline cross-national model. So, for example,
for corruption, I add a dummy marking country years such that the improvement in the corruption
index from t − 3 to t − 1 was in the top 10% of all changes in this measures, and also a dummy
marking country years where the corruption score fell so much from t − 3 to t − 1 that the change
was among the 10% largest drops. This means that the coefﬁcients mark the effect of a large
change against a baseline of zero or little change in the last several years.24

Table 8 shows the results, for models where the dependent variables are major conﬂicts and for
all conﬂicts. There are no reliably signiﬁcant results and no consistent patterns. There is a huge,
negative estimated coefﬁcient for positive change in the corruption index, which means that rapid
improvement in the corruption index associates with lo er odds of civil war onset. But not much
weight should be put on this because there are only 18 major conﬂict onsets in this ICRG subsam-
ple, and this variable comes just one case shy of perfectly predicting zero major onsets.25
  24
       I also checked results using unlagged versions of these variables; this made no consistent differences.
  25
       If we use a linear probability model instead of logit, there is little difference in estimated odds.


                                                               28
Apart from this instance, we do not see strong evidence that rapid improvement on these gov-
ernance indicators associates with increased risks of instability. The estimated coefﬁcients are
weakly positive for improvements in bureaucratic quality and military in politics, but not close to
signiﬁcant (except for major conﬂict onset and military in politics).

               Table 8: War/conﬂict onset and ICRG gov. reform and regression indicators
                                                        ACD wars           all ACD
                gov indicator reform/regression coef.         se    coef.         se
                corruption       big + change          -6.54 0.35 -0.33          0.36
                                 big − change          -0.74 1.06 -0.58          0.37
                bur. qual.       big + change          0.37 0.57 0.13            0.31
                                 big − change          -0.74 0.98 -0.98          0.47
                mil. in pol.     big + change          1.07 0.52 0.28            0.27
                                 big − change           0.61 0.64 0.32           0.26
                Indicators are dummies for top and bottom 10% on lagged two yr change.
                Coef’s from model with controls for prior war, income, population,
                mountains, oil, instability, anocracy, democracy, and ELF.



I next consider four component variables that go into the World Bank’s Country Policy Institutional
Assessment indicator: quality of public administration (qpa), quality of budget and ﬁnancial man-
agement (qbfm), policies for gender equality (gender), and equity in public resource use (epru).
These are scored on 1-to-6 scales, but we have the data only for 1998 through 2008, and only for
countries that received or have received IDA ﬁnancing. This is a short time series for variables that
change little over time within countries (around 80% of the variance is across countries for these
variables and these 11 years). We are also limited by the fact that the dependent variable, conﬂict,
is quite rare for any given country in an 11-year window.

Table 9 repeats the exercise of Table 8 for these variables, using a logit model where the dependent
variable is the onset of any level of ACD conﬂict. There are too few major conﬂict onsets to
estimate a logit model for this sample, and even for all onsets it is not advisable to estimate a
model with the full set of controls, due to the problem of overﬁtting. The model used here includes
as controls prior war in progress, lagged log of per capita income, lagged log of country population,
and ethnic fractionalization. In all the estimations, these take their usual signs (negative, negative,
positive, positive), and are either statistically signiﬁcant or close to it. This is despite the fact that
the number of onsets in these models varies between 24 and 30, out of between 780 and 1000
country years.

The same cannot be said for the dummy variables marking positive or negative change in each of
the four CPIA governance indicators. The thresholds used here are the top 15% for improvements
and the bottom 15% for “regressions” in governance, over a two year period and lagged one year.

                                                   29
We see that a large positive change in quality of public administration and quality of budget and
ﬁnancial management is followed by an elevated average onset risk, but the “effects” are not sta-
tistically signiﬁcant (though somewhat close for qbfm). Large improvement in gender policies is a
nearly perfect predictor of a sence of subsequent conﬂict onset, but little or nothing should made
be of this (in my opinion) due to the small sample and risk of overﬁtting. Improvement in equity
of public resource use associates with a slight decline in risk of conﬂict onset in the next year, but
again this is not statistically distinguishable from zero effect. So no very consistent pattern emerges
regarding the impact of a large, rapid positive change and conﬂict risk for these four indicators.26

                Table 9: All ACD conﬂict and CPIA gov. reform and regression indicators
                    gov indicator          reform/regression coef.     se  p value
                    qual. pub. admin.      big + change         0.7   0.87  0.42
                                           big − change        -0.25 0.64   0.69
                    qual. budget/ﬁn.       big + change        1.06 0.68    0.12
                                           big − change        0.58 1.14    0.61
                    gender                 big + change        -6.07 0.4      0
                                           big − change        -1.01 0.99   0.31
                    equity resource use big + change           -0.24 1.16   0.83
                                           big − change        -1.39 1.04   0.18
                    Coef’s from logit model for all conﬂicts, 1998-2008.
                    Controls: prior war, log(income), log(pop), ELF.


So at least for these seven indicators (three from ICRG and four from CPIA), we ﬁnd no indication
that rapid administrative or military reform is consistently followed by an elevated risk of the
outbreak of violent conﬂict. To the contrary, rapid reform often seems to accompany major and
successful episodes of democratization, which might for a variety of reasons mark countries that
are less likely to have a conﬂict begin. It is possible that case studies selected on certain conﬂicts
might give a misleading impression of general patterns, even if it is possible that in some cases
rapid reform does elevate conﬂict risk.


56      atural resources

As seen in Table 2, oil producers are on average at much greater risk of war and conﬂict onset – the
relative risk is about 3 for wars, and about 2 for all conﬂicts, using those estimates. The estimates
are stable even with ﬁxed effects for major conﬂicts, although statistical signiﬁcance declines and
  26
    I note also the results are somewhat unstable if we construct the variables the same way but don’t lag the
dummy variables (so looking at the impact of change from t − 2 to t instead of t − 3 to t − 1), or if we vary
the threshold window. This is not surprising given the sparsity of the onset data for an 11 year period.


                                                     30
there is less evidence for minor conﬂicts.

The measure used in Table 2 is from Fearon and Laitin (2003), and marks country years in which
at least one third of GDP was from fuel exports, based primarily on World Bank numbers. This
measure is open to various criticisms, and there are other plausible measures one could try. As to
criticism, a number of researchers have worried that oil exports as a percentage of GDP brings the
rest of the economy into the measure, so that it might be capturing an effect of prior conﬂict or the
anticipation of future conﬂict on GDP rather than a causal effect of dependence on oil production
on conﬂict. That is, conﬂict or its anticipation might cause greater resource dependence due to lack
of investment and greater mobility of the non-oil sector.

Humphreys (2005) developed measures of oil production per capita and oil reserves per capita,
which could lessen endogeneity concerns by removing GDP from the denominator. He again
found support for a correlation between both oil production and oil reserves per capita and civil
war risk, using panel data but no ﬁxed effects.27

Ross (2006) improved on Humphreys’ data and also used data on fuel rents per capita produced
originally by Hamilton and Clemens (1999). He again found a signiﬁcant relationship between oil
measures and conﬂict onset, in a large number of panel regressions that considered ﬁve different
civil conﬂict codings and also whether the conﬂicts were separatist or over the center, and ethnic or
ideological. In addition, he ﬁnds that fuel rents from off-shore oil deposits are unrelated to conﬂict
risk.

Overall, the evidence for a correlation between oil and conﬂict thus appears fairly good, although
this is not based on a lot of data and results with ﬁxed effects are fragile and often non-existent.
It is also not clear what the most important mechanisms are (if there is a causal relationship).
Ross (2006) argues that the cross-national patterns and case-study evidence strongly support the
idea that oil wealth sometimes encourage separatist and other regional rebellions to gain more
local control. He observes, however, that the correlation still obtains when we restrict attention to
center-seeking (non-separatist) civil wars, so the separatist mechanism is probably not the whole
story. As mentioned in the section on causes below, state weakness (relative to per capita income)
could be a factor, as could greater inequality (that is, higher prize value for capturing the central
government).

Researchers have also spent quite a bit of time on the question of whether precious gems and
minerals associate with higher conﬂict risk. Ross (2006) reviews and revisits this question as
well, using updated and improved data to estimate annual gem rents per capita, both for primary
  27
    Brunnschweiler and Bulte (2009) try to instrument for resource dependence, and ﬁnd when they do that
they can eliminate a signiﬁcant effect for resource dependence variables. In my view their set of instruments
could not possibly satisfy the exclusion restriction necessary to allow us to infer that they are estimating the
causal impact of resource dependence on conﬂict risk.


                                                      31
and secondary diamond sources. At least since Billon (2001) it has been conjectured that alluvial
diamonds (secondary) are more conﬂict-causing than diamonds from deep mines (primary), which
are more easily controlled by the state. However, contrary to some earlier studies (e.g., Lujala et
al. 2005), Ross ﬁnds fairly strong evidence of association between primary diamonds and conﬂict
risk; the evidence for secondary diamonds is much weaker. He notes, appropriately, that in both
cases these are generalizations based on rather few countries, considerably fewer than even the
case of oil.

The literature on natural resources and conﬂict was initiated and stimulated by early versions of
Collier and Hoefﬂer (2004), the main argument of which was that the signiﬁcance they found for
a measure of primary commodity exports as a share of GDP indicated that civil wars were to be
explained mainly by the ﬁnancing opportunities available to rebel groups. This measure, which
combines cash crops and oil but leaves out minerals, proved not to work very well as a predic-
tor outside of the particular Correlates-of-War based civil war list used by Collier and Hoefﬂer.
For example, when this measure (and/or its square) is added to the models in Table 2 above, the
estimated coefﬁcients are negative, usually with large standard errors.


5 7 Geography

As seen in Table 2, a measure of rough terrain is associated with higher probabilities of civil conﬂict
in the global sample. If one looks region by region, one ﬁnds that the measure takes a positive and
fairly substantial coefﬁcient in all regions (excluding the the West, where there are few cases of
conﬂict), and actually manages statistical signiﬁcance in a number of them. The same pattern is
evident for all ACD conﬂicts, though somewhat weaker.

The measure used here, as in most studies in the literature, is based on estimates of the percentage
of each country that is “mountainous” as judged by geographer John Gerrard, who was commis-
sioned by Paul Collier to construct the measure when Collier was at the World Bank.28 Since that
time, Geographic Information Systems tools have made it possible to construct measures of rough
terrain based on satellite imagery. Surprisingly, there are no published studies that use this ap-
proach to reassess the rough terrain factor at the country level (as far as I know), although Buhaug,
Gates and Lujala (2009) and ... have constructed GIS-based terrain measures at the level of “con-
ﬂict areas” within countries.

Matthew Kocher (2004) developed a country-level GIS-based rough terrain measure, by computing
the standard deviation in elevation of 1-by-1 kilometer cells and their neighbors for each country.
His measure is the percentage of cells in a country with neighbor standard deviation of at least 50
meters. This measure is correlated at .81 with the Gerard measure (in logs). When used in place
  28
    Fearon and Laitin (2003) produced estimates for additional countries by using the difference between
the highest and lowest elevation in a country, which turns out to be well correlated with Gerard’s numbers.

                                                    32
of the Gerard measure in the models in Table 2, the results are actually somewhat stronger overall:
the coefﬁcients and signiﬁcance levels are slightly larger for models 1 and 2 (major and FL civil
wars), and marginally smaller for all ACD wars.

As discussed more in the next section, there are multiple ways to interpret this association. It
could be that it is easier for the state to deter or eliminate nascent rebel groups in a relatively ﬂat
state, because it is harder to hide. Case studies of any conﬂicts point to how guerrilla groups
have survived by taking to the hills, or the jungle. Or it could be that, on average, central states
have developed less administrative control of the periphery in places where the physical obstacles
to extending government authority are relatively great. Or, it could be that because of the low
presence of a strong central state, countries with rough terrain tend to have ethnic or religious
groups with “cultures of honor” and/or a deep distrust and dislike of lowland authority. Or it could
be a combination of these factors, or others still.

Gallup, Sachs and Mellinger (1999) have argued that being landlocked causes a country to be
poorer and have a harder time developing economically, basically because of greater transporta-
tion costs. One might also wonder if landlocked countries are more likely to have civil conﬂicts.
Landlocked countries tend to have several other risk factors associated with conﬂict, as shown
in Table 10. On average they are poorer, more mountainous, and more ethnically fractionalized
(however, they are also smaller and less likely to be oil producers on average).

                     Table 10: Average values for landlocked versus other countries

                                                     Not landlocked landlocked
                        per cap income                       $8,691        $5,714
                        population                             33m           7.1m
                        % mountains                              16             30
                        oil producer                          0.147         0.048
                        political instability                 0.096         0.115
                        anocracy                              0.235         0.255
                        ELF                                   0.366         0.515
                        share of major war years              0.092         0.088
                        freq of major onsets                  0.011         0.009
                        share of all conﬂict years            0.195         0.132
                        freq of all onsets                    0.033         0.028
                        Entries are averages over all country years with data.


It is perhaps surprising, then, that since 1946 landlocked countries have actually been somewhat
less conﬂict prone, both in terms of numbers of major and minor conﬂict onsets, and in terms of
number of years with major or minor conﬂict. This bivariate observation holds up when we add
a dummy variable for “landlocked” the models in Table 2. The coefﬁcients remain negative. For

                                                   33
example, landlocked countries are estimated to have had 40% lower annual odds of major conﬂict
onset, with a p value of .12 in model 1, controlling for the other factors.

The absence of a positive relationship, and the possibility of a negative one, does not appear to
result from outliers or region-speciﬁc effects. Partially excepting Latin America (where landlocked
Paraguay and Bolivia had some violent conﬂicts in the late 40s and 50s, a long time ago), region by
region there is a tendency for the landlocked countries to have experienced less violent conﬂict. In
Eastern Europe and the FSU: Hungary, Czechoslovakia, Belarus, Armenia, all the “Stans” except
Tajikistan. In Africa, the landlocked countries of the Sahel have been generally more peaceful than
the coastal West Africa states, and the landlocked countries of southern Africa (Zambia, Malawi,
Botswana, Lesotho) have been more peaceful than most of their neighbors with coasts. In Asia,
Afghanistan and Nepal have seen a lot of conﬂict, but not so for Mongolia and Bhutan.

Whether there is any causal connection here is not clear. I have trouble even thinking of a plausible
story for why being landlocked would by itself cause a country to be less prone to internal conﬂict.
(On the other hand, it is not obvious why landlock would cause greater conﬂict risk either.)

One other geographic factor deserves mention: Evidence from multiple regions and time periods
suggests that within countries, groups farther from the center of power are more likely to see armed
conﬂict with the center than groups closer to the center. Tong (1988) found this to be strongly the
case for rural rebellions by Chinese peasants in the Ming Dynasty. Brustein and Levi (1987)
present very similar patterns for rebellion in early modern Europe. In the present day, using GIS
data Condra (2009) and Buhaug, Cederman and Rød (2008) ﬁnd that distance from the capitol
predicts higher rebellion risks for countries in subSaharan Africa and in Eurasia, respectively.


5        Gender ine uality and civil conﬂict

Caprioli (2005) and Melander (2005) ﬁnd several cross-national measures of gender inequality to
be related to civil war, in the context of incidence models where the dependent variable is whether
there was conﬂict in a country year.29 This subsection considers what happens when some similar
measures are added to the onset models in Table 2.

The measures examined are total fertility rate (births per adult woman), share of parliament who
are women, and the ratio of girls to boys completing primary school. The ﬁrst two are drawn
from the World Development Indicators. For the primary school ratio I use two different series:
the education data in the WDI, and the data produced by Robert Barro and Jong-Wha Lee (2000).
Coverage varies quite a bit. Fertility data is available for most countries at ﬁve-year intervals from
1960; I interpolate for the intervening years. The ﬁrst estimates for women in parliaments are for
    29
   It is not clear how Caprioli handles the issue of duration dependence. Melander uses cubic splines; I
have doubts about whether this method makes sense with incidence data.

                                                  34
1990, and then are provided each year from 1997 for most countries. The WDI primary education
data essentially starts in 1975 for a reasonable cross-section, then coming in ﬁve-year periods till
the 2000s. The Barro and Lee series, which has been very widely used in research on economic
growth, has data for only about 104 of the countries in our sample, starting in 1960 and for ﬁve
year periods to 2000-04. I interpolate estimates for the missing years in both cases, and extend the
2000 Barro-Lee estimates through 2009.

Table 11 shows correlations between these various indicators and income, both in raw form and
after netting out income (that is, looking at the correlations of the residuals after regressing on log
income). Not surprisingly, like so many aspects of “modernization,” gender equality is strongly
related to per capita income. As with other variables like population growth or “youth bulge,”
collinearity will make it difﬁcult to estimate the conditional expectation of civil war onset with
respect to gender equality measures. Share of women in parliament is the least well correlated
with income, perhaps reﬂecting international inﬂuences on new constitutions. After netting out per
capita income, the correlations are much weaker, and disturbingly so between the two measures of
sex ratio of primary school students, which ostensibly measure the same thing. Still, the reasonable
correlations between total fertility and the education measures after netting out income suggest that
these indicators may be tapping something like a common dimension on gender policies and the
standing of women.

                          Table 11: Correlations among gender inequality indicators

                                              correlations                  corr. net of income
                                log(gdp)      iwmpri wmpc         ifert   iwmpri wmpc ifert
                  iwmpri              53
                  wmpc                53            64                           44
                  ifert              -68           -66      -53                 -48   -21
                  iwmparl             27            21       20    -29            6     5   -12
                  iwmpri = interpolated WDI sex ratio in primary school
                  wmpc = interpolated Barro-Lee sex ratio, primary completion
                  ifert = interpolated WDI fertility rate
                  iwmpri = interpolated share women parliament



Tables 12 and 13 add these measures to the basic onset models of Table 2, for all conﬂicts and
for major conﬂicts as the dependent variables.30 The results are quite variable from indicator to
indicator and depending on which level of conﬂict we look at, although overall there is some
indication that higher levels of gender equality associate with a lower propensity for conﬂict.
  30
    It is important to keep in mind that because of variation in coverage, some of the variation in estimates
is due to the samples changing from model to model.



                                                            35
Higher fertility rates are positively related to conﬂict and war onset, but the estimates are not
statistically signiﬁcant and are substantively close to zero. The share of women in the national leg-
islature is negatively related to onset both for all and for major conﬂicts, but estimated coefﬁcients
are radically different in magnitude and only signiﬁcant (statistically and substantively) when the
dependent variable is all conﬂicts. The ratio of girls to boys in primary school is signiﬁcantly re-
lated to lower odds of all and major conﬂicts when the measure is from the WDI, but not when the
measure is from the Barro and Lee data.

Since per capita income, ethnic fractionalization, democracy, and each of the four measures of
gender inequality are fairly well correlated, it is pushing the data very hard to include them in the
same statistical model. Some of the instability of the results (and note also that the coefﬁcients
for income drop a lot in some cases when one of these gender measures are added) is due to
overﬁtting the one or two onsets that may or may not show up in the “off diagonal” cells (e.g.,
low income, high gender equality), depending on the speciﬁc measure employed. For example, the
WDI primary school ratio gains strength from the fact that ACD conﬂict are recorded for Iraq, Iran,
Lebanon, Oman, and Saudia Arabia, all of which are relatively high income (for conﬂict countries)
but don’t do well on the gender ratio in primary school; the Barro Lee measure is missing for some
of these cases or is somewhat different.

Overall, however, the results are at least somewhat supportive of the proposition that even compar-
ing countries at similar levels of economic development, those that have more women in parliament
or more gender equality in school attendance tend to be somewhat less likely to have violent civil
conﬂicts break out. If this is accepted, the next question is how to interpret the relationship. It
seems clear that we cannot infer that if a government implements greater gender equality as a re-
sult of international pressure or inducements, this would lower conﬂict risk. For all we know, this
could increase risk by increasing frictions between “traditional” and “modernist” factions.31 This
relationship, if it is meaningful, is particularly likely to have an interpretation such as “societies
that are okay with gender equality are the sort of societies that are less likely to have civil wars,”
rather than that the one causes the other in some straightforward way.32


6 Causes of civil war

Patterns of correlations like those presented above have been the basis for a new round of academic
and policy-maker arguments about the causes of civil conﬂict that started after the end of the Cold
  31
       Of course, it still might be worth doing, for the sake of improving the lot of women and girls in a society.
  32
    There is fairly little within country variation in these measures of gender equality, so that, not surpris-
ingly, ﬁxed effects models yield no signiﬁcant coefﬁcients.



                                                         36
                        Table 12: Gender inequality measures and all ACD conﬂicts
                                            Model 1 Model 2 Model 3 Model 4
                     fertility                  0.05
                                               (0.08)
                     share women parl.                   −5.00∗
                                                          (2.17)
                     WDI prim. sch. ratio                           −1.48∗
                                                                     (0.64)
                     BL prim. sch. ratio                                       −0.67
                                                                                (0.47)
                     ln(gdp)t−1               −0.26†     −0.47∗∗ −0.21†        −0.34∗
                                               (0.14)     (0.18)     (0.12)     (0.16)
                     ln(pop)t−1                 0.27∗∗∗    0.33∗∗     0.30∗∗∗    0.30∗∗∗
                                               (0.07)     (0.10)     (0.08)     (0.07)
                     ln(% mountains)            0.16∗      0.08       0.14†      0.07
                                               (0.06)     (0.10)     (0.07)     (0.07)
                     oil producer               0.71∗∗     1.21∗∗∗    0.64∗      1.02∗∗∗
                                               (0.24)     (0.35)     (0.29)     (0.29)
                     new state                  1.38∗∗∗               2.30∗∗∗    0.01
                                               (0.36)                (0.52)     (0.72)
                     pol instabt−1              0.36∗      0.11       0.41∗      0.28
                                               (0.17)     (0.30)     (0.18)     (0.19)
                     anocracyt−1                0.49∗    −0.27        0.29       0.36
                                               (0.20)     (0.38)     (0.26)     (0.26)
                     democt−1                   0.05     −0.04      −0.03        0.21
                                               (0.26)     (0.35)     (0.30)     (0.28)
                     ELF                        1.15∗∗∗    1.99∗∗∗    0.98∗∗     1.08∗∗
                                               (0.33)     (0.54)     (0.35)     (0.40)
                     prior war                  0.10     −0.12        0.04       0.06
                                               (0.24)     (0.34)     (0.26)     (0.29)
                     constant                 −5.33∗∗∗ −3.42∗       −4.21∗∗∗ −4.20∗∗
                                               (1.58)     (1.69)     (1.09)     (1.29)
                     N                      6429        2496       4854       4886
                     Standard errors clustered by country in parentheses
                     †
                         signiﬁcant at p < .10; ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001




War.33 In early versions of their inﬂuential paper “Greed and Grievance in Civil War,” Collier
  33
    There is of course a long history of theorizing about civil war and revolutions before this, mainly on the
basis of particular cases (e.g., Moore 1966, Skocpol 1979), but including an earlier round of quantitative


                                                              37
                   Table 13: Gender inequality measures and ACD major conﬂict onsets
                                            Model 1 Model 2 Model 3 Model 4
                     fertility                  0.10
                                               (0.11)
                     share women parl.                   −0.58
                                                          (1.74)
                     WDI prim. sch. ratio                           −2.33∗∗
                                                                     (0.88)
                     BL prim. sch. ratio                                       −0.84
                                                                                (1.01)
                     ln(gdp)t−1               −0.14      −0.62∗     −0.11      −0.41†
                                               (0.20)     (0.27)     (0.22)     (0.23)
                     ln(pop)t−1                 0.24∗∗     0.27       0.27∗      0.22∗
                                               (0.09)     (0.22)     (0.11)     (0.09)
                     ln(% mountains)            0.34∗∗∗    0.19       0.35∗∗     0.27∗
                                               (0.10)     (0.24)     (0.12)     (0.11)
                     oil producer               0.92∗∗     1.73∗∗∗    0.94∗      1.44∗∗∗
                                               (0.33)     (0.45)     (0.41)     (0.33)
                     new state                  2.30∗∗∗               2.47∗∗∗    0.36
                                               (0.48)                (0.69)     (1.09)
                     pol instabt−1              0.63∗    −0.20        0.48       0.27
                                               (0.32)     (0.79)     (0.39)     (0.38)
                     anocracyt−1                0.48       0.56       0.40       0.44
                                               (0.30)     (0.72)     (0.45)     (0.33)
                     democt−1                 −0.45      −0.78      −0.23      −0.09
                                               (0.43)     (0.84)     (0.47)     (0.43)
                     ELF                        0.92†    −0.04        0.43       0.78
                                               (0.48)     (0.74)     (0.55)     (0.51)
                     prior war                −0.64†     −0.64      −0.64      −0.68†
                                               (0.35)     (0.85)     (0.44)     (0.37)
                     constant                 −7.74 ∗∗∗ −3.28       −5.59 ∗∗∗ −4.16∗

                                               (2.26)     (3.70)     (1.59)     (1.65)
                     N                      6429        2496       4854       4886
                     Standard errors clustered by country in parentheses
                     †
                         signiﬁcant at p < .10; ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001




and Hoefﬂer (2004) interpreted the strong association they found between civil war onset and
measures of dependence on primary commodity exports and low education levels – versus the

cross-sectional analysis (e.g., Hibbs 1973).


                                                              38
weak 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 proﬁt
rather 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 civil
war would be located in the otivations of rebel groups.

Later versions of the paper converged (at least on this point) with the interpretation in Fearon and
Laitin (2003), who argued that “Surely ethnic antagonisms, nationalist sentiments, and grievances
often motivate rebels and their supporters. But such broad factors are too common to distinguish
the cases where civil war breaks out” (p. 76). The idea is that grievances that could potentially
motivate a rebellion are regrettably common (and reasonable) in much of the world, so that more
of the “action” in terms of explaining cross-national variation in civil war propensities is likely to
be found in variation in factors that affect the viability of, or o ortunity 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 of
violent civil conﬂicts in this period have been fought as guerrilla wars or militia-based conﬂicts,
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 intense
grievances to get going” (p. 76).

The prevalence of small-scale guerrilla or militia-based wars does not rule out the possibility that
variation in the level of broad social grievances across countries could be an important explanation
for civil war risk. In principle, it could be that social support from many sympathetic people is
necessary for even a small rebel band to operate successfully. In practice, however, if this is the
case 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 deep
social support, or they can coerce it.34

Fearon and Laitin (2003) give mainly “opportunity” interpretations for the patterns of correlation
described in sections 4 and 5 above. For example, when a colony receives independence, its
central government receives a negative shock to its capability to deter and ﬁght rebels, if and when
the colonial army leaves. Likewise, political instability may signal weakness at 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 governance problem harder
  34
   Cf. Weinstein (2007). Drawing on an exhaustive reading of “micro-level” accounts of particular civil
wars, Kalyvas (2006) provides almost innumerable examples of how the relative local strength of rebel
and government forces shapes local support for the rebels versus the government, more than the other way
around.


                                                  39
for the center (harder to develop reliable chains of principals and agents to monitor what is going on
at the local level; India is much harder to govern than, say, Mauritius). Rough terrain is argued to
associate with better opportunities for rebel groups to hide, and, often, less historical development
of central administrative structures throughout a region.

Along similar lines, Collier and Hoefﬂer (2004) found strong results for a variable measuring the
share of primary commodity exports in GDP. In the 2004 version they interpreted this as an indi-
cator of greater ﬁnancing opportunities for would-be rebel groups, principally through extortion
of producers. They also ﬁnd some evidence that having a large minority diaspora in the US as-
sociates with greater likelihood of war renewal, which they interpret as evidence that diaspora
funding provides opportunities for rebel groups to return to war.

As we saw above, lower per capita income is strongly related to a higher propensity to have civil
wars and conﬂicts across countries (although there is not much evidence that economic growth
or decline within countries over time strongly predicts change in conﬂict risk). Two main causal
interpretations of this pattern have been advanced in the literature. Collier and Hoefﬂer (2004)
stressed a la or ar et explanation: low income, they hypothesize, means lots of underemployed
youth who ﬁnd the opportunity costs of joining a rebel band to be small. Fearon and Laitin (2003),
by contrast, stressed a state ca a ilities explanation. In this view, per capita income is largely
a proxy for state administrative, military, and police competence, and thus the ability to deter or
defeat nascent insurgencies.35

The state capabilities explanation is arguably more consistent with the lack of a robust within-
country relationship between income and civil war: one wouldn’t expect a lock-step relationship
between state capabilities and income, whereas short-run income changes should affect labor mar-
ket conditions. The labor market explanation might be more consistent with Miguel, Satyanath
and Sergenti (2004), who found that exogenous variation in rainfall is related to civil war onset in
subSaharan Africa. They argue that a direct causal effect of income on civil war propensity can be
identiﬁed by using the variation in income related to variation in rainfall, on the assumption that
the only way rainfall variation affects civil war propensity is through income. As they note, it is
possible that heavy rainfall might provide a negative shock to state capabilities by making roads
less passable. But it is perhaps more likely that the effects would go through economic channels.
  35
    Both papers mention the alternative stressed by the other as a possibility. Fearon (2008) develops a
model of individual decisions to join a rebel movement, noting that the standard opportunity cost argument
neglects that while there may be less to lose by joining a rebel band in a poor country, there is also less to
gain. Unless one assumes a somewhat unusual form of utility function (increasing relative risk aversion),
the propensity to join a rebel group will be independent of per capita income in the country. An alternative
explanation, stressed in that paper, is that in more modern economies it may be more difﬁcult for rebel
groups to extract wealth using typical tactics, given that much of it is virtual wealth held in bank accounts,
human capital, and businesses that are more mobile than are farmers in poor countries.



                                                     40
As they are typically stated in the empirical literature, “opportunity” arguments are incomplete and
not coherent as explanations for civil war. Simply noting that some factor makes rebellion more
likely to succeed or be self-sustaining does not explain why it would occur. If this factor – such
as a weak central state, or lots of unemployed youth, or rough terrain, etc. – makes would-be or
actual rebel groups stronger relative to the central government, then why doesn’t this just change
the terms of implicit or explicit deal that the government offers to avoid a costly conﬂict?

Explanations for civil war that fully address this question (which also arise for “grievance” stories)
remain relatively rare in the literature; this is an area where further empirical progress probably
depends a lot on theoretical progress.36 The most developed answer in the academic literature
proposes that shocks to the relative capabilities of rebel groups and governments can create a
commitment problem that renders bargained solutions hard to reach and implement (Acemoglu
and Robinson 2001; Fearon 1994; Fearon 2004; Walter 1997). The idea is that shocks to the
relative capability of actual or potential rebel groups create “windows of opportunity” that cannot
be bargained away because the government would have an incentive to renege on the deal after the
window closes and the rebel group’s opportunity declines. This problem is most acute uring civil
wars, since disarming in a peace deal may leave one or both sides highly vulnerable to reneging
by the other (Walter 1997; Walter 2002). But it can also explain civil war onset. For example,
armed groups may have a window of opportunity to rebel successfully (either to take the center or
gain some regional autonomy) in the ﬁrst few years of independence; the center may not be able
to credibly commit to a long-term policy of favoring the region or satisfying the rebel’s demands,
if a peace deal would in the long run strengthen the central government. More generally, the idea
is that in poor countries with relatively weak central governments it will take less of a shock to the
relative strength of center versus opposition to get a would-be rebel group over the threshold that
makes ﬁghting a worthwhile gamble (Fearon and Laitin 2007).

A second possible approach to explaining the puzzle of why governments aren’t always successful
at deterring and buying off violent challenges would put the stress on incomplete information
about what it will take. Perhaps governments sometimes just guess wrong about what sort of
redistributive and other policies will be enough to prevent the mobilization of armed opposition.
There is then a further puzzle, in that if a government guesses wrong and armed conﬂict begins,
why doesn’t it then revise its beliefs and propose or implement a better deal, so that civil “conﬂict”
would be so short as to barely qualify? Arguments (and questions) of this sort are more developed
in the analysis of interstate conﬂict,37 and some of them may transfer straightforwardly to the
domestic setting (e.g., Cetinyan 2002). But this is not so clear, and it is also not clear how this
approach would explain the pattern of empirical ﬁndings. One possibility is that bargaining is
more likely to fail due to private information in poor countries because governments are less well
informed about popular or minority group preferences. Another is that weaker central states imply
  36
       Blattman and Miguel (2010) similarly note the underdevelopment of theory in this area.
  37
       For example, Fearon (1995, 2007), Powell (1999),and Wagner (2000).


                                                      41
that misjudgments of relative strength by rebel groups are more likely. Or perhaps the “prize” value
of capture of government (or regional autonomy) versus living in the regular economy is greater
on average in less developed countries, so that the bargainers are willing to make aggressive offers
that run a greater risk of costly conﬂict.

Theoretical arguments about “grievances” as causes of civil conﬂict are even more underdeveloped
than arguments about opportunities. As we saw above, a number of the variables that come to mind
ﬁrst as possible measures of “societal grievance” – such as lack of democracy, income inequality,
and various formulations of ethnic or religious demography – do not show a strong or consistent
relationship with a country’s propensity for civil war onset. But it is hard to know how to interpret
this. One possibility, of course, is that these are not good measures of “average level of grievance,”
and if we keep trying, we will ﬁnd something better that “works.” Research discussed above by
Østby, Cederman, and others has pushed in this direction, by attempting to come up with better
measures for “horizontal inequalities” and “ethnic exclusion” from political power.

Another possibility is that the enterprise of trying to correlate measures of societal grievance with
civil war risk is misguided, because “grievance” is a function of government policies, which are
themselves chosen partly in light of civil war risk. Thus we can’t learn about the importance
of “grievances” relative to “opportunities” by looking at cross-national data, or from time-series
within countries unless we have some plausibly exogenous variation in grievance (which is hard
to imagine). For example, as noted above, the exercise of correlating “ethnic exclusion” with
civil war risk might either underestimate the causal effect of policies of exclusion, if governments
choose what they can get away with, or overestimate it, if governments tend to exclude those
minorities that are more threatening based on “opportunity” variables such as relative strength. A
more developed theoretical analysis is probably necessary to allow us to see how to conceptualize
and understand the relationship between “grievance” and relative power and cost factors, rather
than posing them as distinct.


7 Governance indicators and civil conﬂict

As discussed above, several of the most striking cross-national patterns in civil war onset might
be explained by an interpretation that puts “state capabilities” at the center of the story. In this
view, low per capita income is strongly related to conﬂict onset because it is a proxy for the central
state’s capability to deter and suppress armed challengers, and possibly also to provide public
services. A set of alternative interpretations argues that there is a direct effect of low income on
civil war propensity through some labor market channel. For instance, it is argued that poverty
makes joining a rebel group relatively more attractive for young men.

This section uses three sets of governance indicators – the Worldwide Governance Indicators pro-
duced by Kaufmann, Kraay and Mastruzzi (2009), the International Country Risk Guide (ICRG)

                                                 42
indicators, and the World Bank’s aggregate Country Policy Institutional Assessment (CPIA) – to
try to better 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 about
140 countries. The WGI project began in the 1990s at the World Bank. Kaufmann and Kraay
assembled 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 into
sets that they argue correspond to six dimensions of governance: “government effectiveness,”
“voice,” “political instability,” “rule of law,” “corruption,” and “regulatory quality.”38 They then
use techniques akin to factor analysis to extract a common dimension in each area. This has
yielded 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 staff have coded Bank client countries on 16 or more dimensions
concerning the quality of policies and institutions. These codings are then aggregated to a summary
measure called the Country Policy and Institutional Assessment, which is used for various purposes
including decisions about aid allocation. The aggregate index varies from 1 to 6, with higher scores
indicating a better policy and governance environment from the Bank’s perspective. Unfortunately
the CPIA index is produced only for aid recipient countries, so we have nearly 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, including
inﬂuential papers by Acemoglu, Johnson and Robinson (2001), Knack and Keefer (1995), and
  38
    Kaufmann, Kraay, and Mastruzzi (p. 6) describe these six areas as follows: “(1) Voice and Accountability
(VA) capturing perceptions of the extent to which a country’s citizens are able to participate in selecting
their government, as well as freedom of expression, freedom of association, and a free media. (2) Political
Stability and Absence of Violence (PV) capturing perceptions of the likelihood that the government will
be destabilized or overthrown by unconstitutional or violent means, including politically-motivated violence
and 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 of policy
formulation and implementation, and the credibility of the government’s commitment to such policies. (4)
Regulatory Quality (RQ) capturing perceptions of the ability of the government to formulate and implement
sound policies and regulations that permit and promote private sector development. (5) Rule of Law (RL)
capturing perceptions of the extent to which agents have conﬁdence in and abide by the rules 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 perceptions of the extent to
which public power is exercised for private gain, including both petty and grand forms of corruption, as well
as “capture” of the state by elites and private interests.”


                                                      43
Mauro (1995). But the longer time series employed here appears not to have been used even in
that 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 not
completely clear what the expert ratings are measuring. This is partly due to lack of clarity about
what we are trying to measure. Just what are “good governance” and “good institutions”? Many
people have strong intuitions here, based on experiencing the relative efﬁciency, competence, and
corruption of public services and ofﬁcials in different countries. In more theoretical terms, the
tradition associated with North and Thomas (1973) identiﬁes “good institutions” with formal and
informal 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 the
efﬁciency 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 from
diverse observations of different things.

This fact makes expert surveys a natural approach for measuring the quality of “governance” and
“good institutions,” but it is also makes it hard to know exactly what the experts are doing. For
example, 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 going
these 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 include
many considerations that we would not associate conceptually with governance and institutions. In
sum, there are reasons to be concerned about both the validity and reliability of the expert-survey
based measures of governance, but it is also not obvious what a better approach would be.39

The second major problem one faces when trying to use governance indicators to assess the causes
of economic growth or civil war onset is endogeneity. If an indicator is well correlated with
contemporaneous growth or civil war onset, we cannot infer causality, because it could be that the
observation of growth is leading the experts to think that governance is good, or that the observation
of 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 single
observation of level or growth of income, or conﬂict performance, the only feasible solution is
to ﬁnd an instrument for governance – an exogenous variable that affects growth or conﬂict only
  39
   A signal advantage of Kaufmann and Kraay’s approach is that by drawing on a large number of different
expert-based measures, their measures may have greater reliability than any one source.


                                                  44
through its effect on governance. Such variables are very hard to ﬁnd and the exclusion restriction
is not testable.40

With data from time t on governance and from time t+ 1 on growth or conﬂict, we can ask whether
the former predicts the latter, controlling for other possible determinants of growth or conﬂict. An
important advantage of this design is that it cannot be that observation of the outcome (growth
or 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 enough
years of data on governance and growth or conﬂict, then we can ask whether expert assessments
of governance actually re ict subsequent conﬂict or growth experience.

If the answer is yes, this still does not settle the question of causality, since it could be that omitted
variables are causing both expert assessments of quality of governance at time t an conﬂict or
growth performance subsequently. In particular, as we will see below, all of the WGI and ICRG
indicators are highly correlated with per capita income. This is as it should be, if it is true that
income is a proxy for state capabilities and that good governance and good institutions cause
economic growth over the long term. But it raises the problem of how to separate out the causal
impact 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 about
the 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 other
countries at the sa e level of er ca ita inco e. The attempt to identify the causal impact of gov-
ernance quality on conﬂict then comes from seeing whether surprisingly good or bad governance
in one period of time predicts conﬂict onset subsequently. The strategy will be effective to the
extent that whatever determines surprisingly good/bad governance in one time period inﬂuences
subsequent conﬂict risk primarily through governance and institutions, rather than via some other
path.

With the ICRG and CPIA indicators, we have long enough time series and enough variation over
time 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.

The core strategy here is potentially subject to the concerns noted above concerning “policy regres-
sions,” where the researcher tries to infer something about the causal effect of a policy choice by
  40
    For studies of institutions and governance as causes of economic growth, Mauro (1995) used ethnic
fractionalization 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 is
corruption. Acemoglu, Johnson and Robinson (2001) famously used settler mortality in colonies hundreds
of years ago as an instrument for 1983 expropriation risk (ICRG). Knack and Keefer (1995) did not really
address the endogeneity issue.


                                                   45
measuring different policies across cases and putting them on the right-hand-side in a regression
model. “Surprisingly good governance” (SSG) is of course at least partly a policy choice by a
leadership or political regime. Then we will un eresti ate the positive impact of good governance
if on average leaderships tend to choose better governance when, for other unmeasured reasons,
they expect that the risk of civil conﬂict is high. In that case, SSG will be partly correlated with
unobserved factors that favor conﬂict, so that our estimates of the pacifying effects of SSG will be
biased downwards. But, on the other hand, what if leaderships are more able to implement SSG
in country years when, for reasons completely unrelated to governance, conﬂict is unlikely? To
this extent, we would tend to overesti ate the causal impact of good governance on conﬂict risk.
I ﬁnd it difﬁcult to think of plausible stories here – perhaps there are cultural trends that arise in-
dependent of governance but can enable better governance, and these directly determine civil war
propensity? But in principle the risk is there.

The efﬁcacy of the approach depends in the end on what explains variation in surprisingly good
governance. If, or to the extent that, leaderships and state bureaucracies “get their acts together”
for reasons that are largely independent of other, independent causes of civil strife, then the results
below suggest that good governance and institutions are indeed important factors in reducing a
country’s conﬂict risk. For instance, governance and institutions may improve when an old leader
dies and the new one is just more capable, or is politically situated so that he or she can implement
better policies and develop better institutions. Or governance may be fairly steady, but income
varies due to international shocks and other vagaries of economic growth, in which case “surpris-
ingly good governance” will appropriately estimate a causal effect. I ﬁnd these possibilities more
plausible than alternative arguments that would imply that this approach leads to overestimation
of the “governance effect,” but clearly more work on the determinants of surprisingly good gover-
nance is needed.


7 1 The      GI IC G and C I 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, and
regulatory quality. ICRG produces a large set of indicators, which have varied somewhat over the
years. 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 WGI
indicators 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 a
number of growth studies.

Because they are derived from a factor-analysis-like technique, the WGI indicators all have mean


                                                  46
zero and standard deviation of 1, with higher values indicating better quality governance on that
dimension.41 For ICRG, corruption and rule of law are on 1-to-6 scales. Investment proﬁle varies
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 been
grouped into four equally weighted clusters, described as “economic management,” “structural
policies,” “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 that
receive IDA loans, and that countries can “graduate” out of and enter into this category depending
on economic and government performance. As a result, the CPIA sample is already truncated by
having relatively poor countries, and, even worse, there is a built-in selection bias that will work
against identifying the impact of governance on conﬂict (or growth) outcomes. Namely, countries
that perform well are more likely to exit the CPIA sample, and countries that perform poorly may
enter it.

Figures 6 and 7 show scatterplots that illustrate how closely related are the different ICRG and WGI
governance indicators to each other and to (the log of) per capita income. The data are for 2005,
but the picture of course looks very similar for other years. The correlations are given in Table
14, which also shows that associations among the WGI, ICRG, and CPIA indicators are strong
as well.42 There is not much indication that correlations across the ICRG and WGI indicators are
higher within what should be the same dimension – for example, “rule of law” in the two different
data sets.

Table 15 shows the correlations between the residuals of the WGI, ICRG, and CPIA indicators
after regressing each of them on log per capita income. They remain substantial, which is encour-
aging in that it suggests that raters’ perceptions of quality of governance or institutions are not
completely determined by level of economic development. Instead, there appears to be some level
of agreement about surprisingly good or bad governance. However, there is not much indication
that agreement is markedly higher within categories (e.g., corruption, rule of law, etc.) than across
them. This suggests either that these various dimensions of governance quality tend in practice
to 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 performance
clearly.

One other descriptive statistic about these indicators is worth presenting before moving to the
  41
    One problem with this approach is that a country’s rating may change from one year to the next not
because anything changed in the country, but because other countries changed; these measures have more
validity as a ranking within a given year than as a time series measure.
  42
       This is partly mechanical, since ICRG indicators are one of the many inputs into the WGI indicators.


                                                       47
                        Table 14: Income and governance indicator correlations
                                                WGI                             ICRG
                   income ge voice pol. stab. corr. rol reg. qual. ip corr. rol            bq
     govt eff.       79
     voice           58     75
     pol. stab.      67     79     71
     corruption      74     94     72      77
     rule of law     77     95     79      83       94
     reg. qual.      75     94     79      75       88    91
     inv. prof.      72     82     73      73       79    83        88
     corruption      61     85     74      67       88    84        78       65
     rule of law     69     73     50      72       75    78        66       61  65
     bur. qual.      77     89     77      66       82    85        83       72  76  63
     cpia            51     78     58      47       62    67        82       73  51  39    62



                   Table 15: Correlation between WGI and ICRG, netting out income
                                                      WGI
                   ICRG        ge voice pol. stab. rol corruption reg. qual.
                   ip          38      36       33     40      35           47
                   corruption 60       51       41     61      66           52
                   rol         42      14       42     55      46           33
                   bq          67      54       27     58      53           57
                   cpia        61      41       24     51      47           64



analysis. Table 16 shows the percentage of variation for each indicator that is due to variation
across countries as opposed to over time within countries. Almost of all of the variation in the WGI
indicators is across countries, which makes sense given that the time period is just over a decade
and no one could think that “state capabilities” would change a great deal from year to year. There
is much more within-country variation for the longer ICRG and CPIA series, and especially for the
ICRG “investment proﬁle” indicator. This will allow us to consider a ﬁxed effects model with the
ICRG and CPIA data.


7 2 IC G governance         easures and civil war onset

Table 17 reports our baseline model (Model 1 of Table 2), for major conﬂicts, adding each of the
four ICRG governance indicators in turn. The indicators are lagged two years to try to lessen the
risk that their values are being caused by knowledge of a civil conﬂict already in progress.


                                                48
           Table 16: Percentage within vs. between country variation in governance indicators
                               variable          between % within %
                               log(income)               85           15
                               ACD war onset              6           94
                                            WGI: 1996-2008
                               ge                        96            4
                               voice                     97            3
                               pol. stab.                91            9
                               corruption                96            4
                               rol                       96            4
                               reg. qual.                94            6
                                           ICRG: 1984-2006
                               ip                        41           59
                               corruption                71           29
                               rol                       72           28
                               bq                        82           18
                                           CPIA: 1977-2008
                               cpia                      58           42



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 rule of law
strongly so). For investment proﬁle, moving from the 75th to the 25th percentile (8.5 to 5) is esti-
mated to associate with an increase in annual civil war odds of a factor of 5.9. Going from 12 to
1, the full range of the scale, increases the estimated odds by a factor of about 400. Moving from
the 75th to the 25th percentile on rule of law and corruption associates increasing the annual onset
odds of 4.6 and 1.9, respectively. Note that the estimates for investment proﬁle and rule of law are
remarkably large, in substantive terms.

Model 1 of Table 17 is Model 1 of Table 2 but restricted to the sample used when we add ICRG
indicators, which is limited to 1984-2006 and countries with ICRG data. It shows that in this sub-
sample the standard errors for several of the other covariates increase and some of the estimated co-
efﬁcients diminish. One reason is that the ICRG measures are systematically more missing for low
income and high conﬂict countries – presumably investors don’t need to be told that Afghanistan
has a poor investment proﬁle. It is plausible that the results would be generally stronger if we had
more complete data for the lowest income countries.

Even so, given the data we have we ﬁnd that the estimated coefﬁcients for per capita income
actually turn positive (though not signiﬁcantly difference from zero) when we add the governance
indicators in Models 2-4. This is consistent with the hypothesis that income “matters” because it



                                                  49
proxies for state capabilities, or governance.43

                     Table 17: ICRG governance and civil war onset (major conﬂicts)
                                   Model 1 Model 2 Model 3 Model 4 Model 5
                ipt−2                           −0.50∗∗∗
                                                (0.12)
                corruptt−2                                 −0.32†
                                                            (0.19)
                rolt−2                                                −0.59∗∗∗
                                                                       (0.17)
                bqt−2                                                            −0.23
                                                                                  (0.31)
                log(gdpt−1 )         −0.13       0.33        0.04       0.23        0.04
                                      (0.23)    (0.24)      (0.25)     (0.24)     (0.34)
                log(popt−1 )           0.23      0.30        0.21       0.27        0.25
                                      (0.20)    (0.19)      (0.20)     (0.21)     (0.20)
                log(% mountains)       0.11      0.04        0.14       0.13        0.10
                                      (0.17)    (0.17)      (0.18)     (0.18)     (0.17)
                oil producer           1.34∗∗    1.09∗       1.18∗      1.14∗       1.23∗
                                      (0.46)    (0.46)      (0.51)     (0.47)     (0.49)
                pol instabilityt−1   −0.02      −0.06        0.02     −0.11      −0.03
                                      (0.58)    (0.54)      (0.58)     (0.59)     (0.59)
                anocracyt−1            1.08∗     1.05∗       1.06∗      1.08∗       1.08∗
                                      (0.50)    (0.45)      (0.49)     (0.47)     (0.49)
                ELF                    0.92      1.55†       0.99       0.93        1.07
                                      (0.87)    (0.88)      (0.87)     (0.85)     (0.88)
                prior war            −0.06      −0.52      −0.13      −0.45      −0.13
                                      (0.70)    (0.76)      (0.69)     (0.69)     (0.68)
                constant             −7.55 ∗∗   −9.25 ∗∗∗ −7.92∗∗     −9.10 ∗∗   −8.73∗∗
                                      (2.83)    (2.62)      (2.86)     (3.07)     (3.12)
                N                   2777      2777       2776       2776       2776
                SE’s clustered by country.
                †
                    signiﬁcant at p < .10; ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001



Table 18 is the same as Table 17, except for all ACD conﬂicts (including minor, medium and small
ones). Here all four ICRG governance indicators are signiﬁcantly negatively related to conﬂict
onset, with the strongest effects again being for investment proﬁle and rule of law. Low income
is signiﬁcantly negatively related to conﬂict onset in this subsample (Model 1), and while the
  43
   This effect is stronger if we consider models that drop other covariates, for example leaving just prior
war, income, population, and ethnic fractionalization.


                                                             50
signs remain negative when governance indicators are added, the estimated coefﬁcients for income
shrink towards zero and in several cases are no longer statistically signiﬁcant. This again favors
the interpretation that income is proxying in a large part for state capabilities and administrative or
policy competence.

In principle, it could happen that when a country gets better ICRG (or WGI) ratings, foreign in-
vestment increases as a result, which then increases subsequent growth, which then has a direct
economic effect on the odds of civil conﬂict. Governance or institutions are still causally efﬁca-
cious for lower conﬂict risk in this argument, but the pathway is through the economy. A ﬁrst-order
check of the hypothesis can be done by adding the current economic growth rate to the models in
Tables 17 and 18; a better approach would be to check whether improvements in ICRG ratings are
strongly related to subsequent FDI. When we do this, the estimated coefﬁcients on the governance
variables change very little, although in some cases signiﬁcance drops slightly. This again argues
against the main effect of better governance indicators on conﬂict risk working primarily through
the economy.

Table 19 repeats the exercise for all conﬂicts, but using conditional ﬁxed effects logit and thus
controlling for all unmeasured (but temporally stable) country characteristics. Remarkably, given
the small number of countries in the sample now and the relative lack of temporal variation in
governance quality, all four of the ICRG indicators get negative coefﬁcients (though corruption is
essentially zero), and the estimates for investment proﬁle and rule of law are statistically signiﬁcant
(p = .002 and .091 respectively). As in Table 2, Model 5 (ﬁxed effects on the larger sample), per
capita income can takes “the wrong sign” and is never close to signiﬁcant.

Thus, within countries over time, civil war onset has been somewhat more likely when investment
proﬁle, corruption, and rule of law were judged worse in recently preceding years. This result
supports a causal interpretation of the relationship between governance quality and conﬂict onset
more than the previous models, because here identiﬁcation is based on within-country comparisons
(and because it is somewhat remarkable to ﬁnd anything given the lack of within-country variation
in governance indicators).44

One might still worry that a two-year lag is not enough to 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. I have constructed a panel with three waves, for the 80s, 90s, and 00s, asking if
average ICRG ratings in one decade forecast conﬂict in next decade, controlling for prior conﬂict
experience and lagged income levels. I ﬁnd that the results are quite similar: ICRG indicators
forecast conﬂict outbreak even in next decade.
  44
    Looking only at major conﬂicts, there are too few countries with a major onset in the ICRG subsample
(only 17) to get anything reliable out of ﬁxed effects. However, the coefﬁcient for investment proﬁle is
negative and very close to “signiﬁcant” at .10.



                                                  51
                   Table 18: ICRG governance and conﬂict onset (all ACD conﬂicts)
                                  Model 1 Model 2 Model 3 Model 4 Model 5
               ipt−2                          −0.27∗∗∗
                                               (0.07)
               corruptt−2                                −0.18†
                                                          (0.10)
               rolt−2                                                −0.20∗
                                                                      (0.09)
               bqt−2                                                           −0.29†
                                                                                (0.16)
               log(gdpt−1 )         −0.45∗∗ −0.17        −0.37∗      −0.33†    −0.23
                                     (0.17)    (0.15)     (0.18)      (0.20)    (0.19)
               log(popt−1 )           0.30∗∗    0.35∗∗∗    0.28 ∗∗     0.32∗∗    0.34∗∗
                                     (0.10)    (0.10)     (0.10)      (0.11)    (0.11)
               log(% mountains)       0.12      0.09       0.13        0.13      0.10
                                     (0.12)    (0.12)     (0.13)      (0.13)    (0.12)
               oil producer           0.91∗∗    0.72∗      0.83∗       0.81∗     0.75∗
                                     (0.33)    (0.32)     (0.35)      (0.34)    (0.34)
               pol instabilityt−1     0.12      0.10       0.12        0.07      0.11
                                     (0.25)    (0.26)     (0.26)      (0.26)    (0.26)
               anocracyt−1          −0.04       0.03     −0.06       −0.01     −0.03
                                     (0.31)    (0.30)     (0.31)      (0.30)    (0.31)
               ELF                    1.23∗     1.58∗∗∗    1.24 ∗      1.24∗     1.40∗∗
                                     (0.52)    (0.47)     (0.52)      (0.52)    (0.51)
               prior war              0.53      0.27       0.49        0.40      0.44
                                     (0.34)    (0.34)     (0.33)      (0.34)    (0.36)
               constant             −3.71∗    −5.02∗∗ −3.77∗         −4.19∗    −5.45∗∗
                                     (1.75)    (1.59)     (1.81)      (1.91)    (1.84)
               N                  2777       2777       2776       2776       2776
               SE’s clustered by country.
               †
                 signiﬁcant at p < .10; ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001




73      GI indicators and civil war onset

The WGI series is only for 1996 to 2008, with some of the early years missing. It also has very
little over time variation within countries, and the method of its construction raises some questions
about 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 in

                                                          52
                  Table 19: ICRG governance and all conﬂict onsets, country ﬁxed effects
                                      Model 1        Model 2       Model 3      Model 4
              ipt−2                      −0.228  ∗∗

                                          (0.075)
              corruptiont−2                             −0.038
                                                         (0.153)
              rule of lawt−2                                          −0.226†
                                                                       (0.134)
              bqt−2                                                                 −0.236
                                                                                     (0.215)
              log(gdpt−1 )                 0.287        −0.166          0.015       −0.024
                                          (0.541)        (0.505)       (0.512)       (0.517)
              log(popt−1 )                 0.517        −0.352          0.104       −0.429
                                          (0.948)        (0.901)       (0.923)       (0.894)
              oil                          0.528          0.441         0.383         0.239
                                          (0.964)        (0.976)       (0.965)       (0.982)
              pol. instab.                 0.184          0.163         0.155         0.161
                                          (0.324)        (0.325)       (0.326)       (0.325)
              anocracyt−1                −0.088         −0.131        −0.169        −0.126
                                          (0.398)        (0.396)       (0.399)       (0.398)
              democracyt−1               −0.587         −0.721        −0.746        −0.687
                                          (0.478)        (0.488)       (0.487)       (0.487)
              prior war                  −1.414  ∗∗     −1.271 ∗∗     −1.342 ∗∗     −1.295∗∗
                                          (0.359)        (0.355)       (0.362)       (0.357)
              N (N countries)        956(44)       955(44)       955(44)       955(44)
              Country ﬁxed effects      yes            yes           yes           yes
              †
                  signiﬁcant at p < .10; ∗ p < .05; ∗∗ p < .01




1996 or 1998 actually forecast conﬂict experience in the next decade, controlling for initial level
of income and prior conﬂict experience. Using the ACD civil war variable, only 10 countries had
onsets between 1997 and 2009. Using all ACD conﬂicts, 37 countries had a total of 63 onsets. We
control for income level in 1996, along with prior conﬂict experience and ethnic fractionalization.
Thus, we are asking if perceptions of “surprisingly good governance” relative to income level and
conﬂict history can still forecast civil peace.

Table 20 shows that the perceptions of of “government effectiveness,” “political stability,” and “rule
of law” in 1996 are indeed signiﬁcantly related to major conﬂict outbreak over the next 13 years.
This might not be too surprising for “political stability,” which is based on expert surveys intended
to capture “perceptions of the likelihood that the government will be destabilized or overthrown
by unconstitutional or violent means” (Kaufmann, Kraay and Mastruzzi 2009). But the results are


                                                                 53
also present for “government effectiveness” and “rule of law,” and the estimated coefﬁcients are
negative and substantively large for corruption and regulatory quality as well. It is also evident that
adding governance measures tends to turn the sign on the estimate for income slightly positive or
close to zero, again consistent with the hypothesis that income normally “stands in for” governance
or state capabilities. The pattern continues to hold if we add region ﬁxed effects (although the sig-
niﬁcance of government effectiveness weakens a bit), or measures of oil production or population
in 1996.

Table 21 repeats the exercise with the dependent variable of all ACD conﬂicts.45

I have run the same models using the WGI indicators from 1998, and the dependent variable
as onsets after 1998. The results are quite similar, though marginally weaker. This lowers the
likelihood that the results are a ﬂuke from one year of WGI data (which as we have seen are highly
stable over time anyway). I have also added average growth rate of GDP per capita after 1996,
again ﬁnding little change for the governance indicators.

There is not much evidence that different dimensions of governance as measured by WGI show
notably stronger or weaker relationships to subsequent conﬂict risk. “Political stability,” which
is 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. (This
is consistent with our earlier ﬁndings of little link between democracy and conﬂict risk in poor
countries.) Rule of law and “government effectiveness” – which KKZ describe as “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 of policy formulation and implementation, and
the credibility of the government’s commitment to such policies” – appear to be most predictive
after political stability. But overall, just as we saw that there are quite high correlations among these
different dimensions, no strong conclusions can be drawn about what dimension 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 in
countries’ quality of governance as measured by expert ratings in 1996 is essentially random with
respect to unmeasured other determinants of subsequent civil war risk. I ﬁnd it difﬁcult to think
of omitted variables distinct from “governance” or “institutions” that would plausible cause both
rater perceptions of governance quality and conﬂict performance over the subsequent ten years.
But the possibility still exists.
  45
    I log the number of prior onsets (plus one) because some countries have many prior low-level onsets and
the distribution is quite skewed. Also, note that because some countries have multiple onsets after 1996, a
negative binomial model could be used instead of logit with whether a country had at least one onset as the
dependent variable; results are quite similar.



                                                    54
The identiﬁcation strategy here will not do a good job of estimating the long run impact of good
governance or state capabilities, since some part of these may be incorporating into per capita
income in 1996. Depending on how we think about what causal effect we are trying to estimate –
relatively short run or long run impact of governance on civil war odds – this would tend to make
for underestimates rather than overestimates of the importance of governance and state capabilities.

The next step in this part of the investigation should be to examine the determinants of surprisingly
good/bad governance – what explains variation in expert perceptions once we have netted out
income level? In preliminary work, I ﬁnd that, interestingly, ethnic fractionalization is almost
completely unrelated to surprisingly good governance. This is surprising in light of the literature
in economics ﬁnding and arguing that ethnic diversity directly causes low public good provision
and corruption. Second, I ﬁnd that oil producers and autocracies are consistently judged to have
worse governance, even controlling for income.46


7 4 C I indicators and civil war onset

Table 22 is essentially our basic model for all ACD conﬂicts using the set of countries and years
with CPIA estimates (Model 1), adding one- and two-year lags of the CPIA index (Models 2 and
3), and then Model 2 with country ﬁxed effects. The estimated coefﬁcients for the lagged CPIA
index are similar in magnitude to those for the other governance indicators, despite the sample
truncated sample and the selection bias issues. However, they are barely statistically signiﬁcant
for all conﬂicts, not at all with ﬁxed effects, and not at all if we look only at major conﬂicts. The
estimate hardly changes when we use the two- instead of the one-year lag, which suggests that
there probably isn’t very much “coding of CPIA on civil war in progress” going on. There is less
indication here that the CPIA measure successfully competes with income.




  46
       As noted, the results in Tables 17 and 18 are not changed if we control also for oil or democracy.

                                                        55
                                      Table 20: WGI governance in 1996 and ACD civil war onset, 1997-2008
                                Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7
     gov’t effec. 1996                     −0.97†
                                            (0.58)
     voice 1996                                      −0.52
                                                      (0.44)
     pol. stab. 1996                                           −0.97∗
                                                                (0.44)
     rule of law 1996                                                    −1.40∗
                                                                          (0.60)
     corruption 1996                                                                −0.96
                                                                                     (0.66)




56
     reg. qual. 1996                                                                           −0.73
                                                                                                (0.45)
     onsets pre 1997               0.50∗     0.46∗     0.43†     0.24      0.38       0.40†      0.45∗
                                  (0.22)    (0.23)    (0.23)    (0.27)    (0.23)     (0.23)     (0.23)
     log(income) 1996            −0.42       0.11    −0.20       0.08      0.24     −0.02      −0.02
                                  (0.34)    (0.47)    (0.38)    (0.42)    (0.45)     (0.45)     (0.41)
     ELF                         −0.71     −0.63     −0.46     −0.59     −0.60      −0.62      −0.60
                                  (1.29)    (1.28)    (1.32)    (1.30)    (1.32)     (1.28)     (1.30)
     constant                      0.60    −4.16     −1.51     −3.95     −5.57      −3.11      −2.95
                                  (2.98)    (4.17)    (3.48)    (3.81)    (4.09)     (4.02)     (3.69)
     N                           156      156       156       156        156       156        156
     DV = ACD civil war onset after 1996. Logit, with standard errors in parentheses
     †
         signiﬁcant at p < .10; ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001
                                Table 21: WGI governance in 1996 and all ACD conﬂict onset, 1997-2008
                                 Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7
     govt effect. 1996                        −0.62
                                               (0.41)
     voice 1996                                         −0.47
                                                         (0.30)
     pol. stab. 1996                                              −0.93∗∗
                                                                   (0.32)
     rule of law 1996                                                       −0.67†
                                                                             (0.38)
     corruption 1996                                                                   −0.12
                                                                                        (0.41)




57
     reg. qual. 1996                                                                              −0.45
                                                                                                   (0.33)
     log(# onsets pre 1997 + 1)      1.26 ∗∗∗   1.19∗∗∗   1.15∗∗    0.72 †    1.12∗∗     1.22 ∗∗∗   1.17∗∗∗
                                    (0.35)     (0.35)    (0.35)    (0.39)    (0.35)     (0.37)     (0.35)
     log(income) 1996              −0.55∗     −0.24     −0.40     −0.23     −0.27      −0.51†     −0.34
                                    (0.23)     (0.31)    (0.25)    (0.27)    (0.28)     (0.29)     (0.28)
     ELF                             1.30       1.39      1.53†     1.58†     1.41       1.30       1.43
                                    (0.87)     (0.88)    (0.89)    (0.91)    (0.88)     (0.87)     (0.88)
     constant                        1.72     −1.13       0.18    −1.18     −0.89        1.29     −0.22
                                    (2.07)     (2.76)    (2.29)    (2.37)    (2.56)     (2.54)     (2.54)
     N                             156        156       156       156       156       156         156
     DV = at least one ACD conﬂict onsets after 1996. Logit, with standard errors in parentheses
     †
       signiﬁcant at p < .10; ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001
        Table 22: CPIA governance and all ACD conﬂict onset
                       Model 1 Model 2 Model 3           Model 4
prior war                −0.16     −0.26     −0.14          −1.89∗∗∗
                          (0.29)    (0.27)     (0.29)        (0.35)
log(gdpt−1 )             −0.48 ∗∗  −0.37  ∗  −0.33  ∗          0.01
                          (0.16)    (0.16)     (0.16)        (0.42)
log(popt−1 )               0.27∗     0.31∗∗     0.31∗∗         0.27)
                          (0.11)    (0.11)     (0.10)        (0.65)
log(% mountains)           0.17∗     0.18∗      0.18∗
                          (0.08)    (0.08)     (0.09)
oil producer               0.69†     0.58       0.65†          0.91
                          (0.37)    (0.37)     (0.37)        (0.77)
pol instabilityt−1         0.03    −0.01     −0.00             0.11
                          (0.25)    (0.25)     (0.27)        (0.29)
anocracyt−1              −0.04     −0.06     −0.08             0.11
                          (0.27)    (0.27)     (0.28)        (0.37)
democracyt−1             −0.01       0.04    −0.04             0.09
                          (0.34)    (0.35)     (0.35)        (0.43)
ELF                        1.22∗     1.23∗∗     1.32∗∗
                          (0.49)    (0.47)     (0.45)
cpiat−1                            −0.31  †                 −0.08
                                    (0.16)                   (0.17)
cpiat−2                                      −0.31  †

                                              (0.18)
constant                 −3.09∗    −3.26∗    −3.62∗
                          (1.55)    (1.54)     (1.45)
N                       3120      3120      3063       1652(61)
Country ﬁxed effects?     No        No        No           Yes
Standard errors in parentheses
†
    signiﬁcant at p < .10; ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001




                                            58
                           Figure 1. # of civil conflicts and wars, 1946−2008


                                   all conflicts (kia > 25/year)
                      50




                                   major conflicts (kia > 1000/year)
                                   war onsets
                                   war ends
                      40
# of wars/conflicts

                      30
                      20
                      10
                      0




                            1950    1960      1970      1980      1990   2000   2010

                                                      year




                                                        59
                                       % countries with conflict or war, 1946−2008
                               0.25




                                             all conflicts (kia > 25/year)
                                             major conflicts (kia > 1000/year)
                               0.20
% of countries with conflict

                               0.15
                               0.10
                               0.05
                               0.00




                                      1950     1960      1970      1980      1990   2000   2010

                                                                 year




                                                                   60
                             SSA                                                Asia
# conflicts




                                                  # conflicts
              15




                                                                 15
              5




                                                                 5
              0




                                                                 0
                   1950   1970      1990   2010                       1950   1970      1990   2010

                             year                                               year



                            NA/ME                                              LA/Ca
# conflicts




                                                  # conflicts
              15




                                                                 15
              5




                                                                 5
              0




                                                                 0




                   1950   1970      1990   2010                       1950   1970      1990   2010

                             year                                               year



                             EEur                                               West
# conflicts




                                                  # conflicts
              15




                                                                 15
              5




                                                                 5
              0




                                                                 0




                   1950   1970      1990   2010                       1950   1970      1990   2010

                             year                                               year




                                                                61
                                                 Figure 4. % of world population in conflict countries



                                                                              all conflicts
                                           0.5




                                                                              civil wars
share of world pop in conflict countries

                                           0.4
                                           0.3
                                           0.2
                                           0.1
                                           0.0




                                                    1950    1960    1970     1980       1990   2000   2010

                                                                           year




                                                                            62
             Figure 5. Distribution of conflicts by
                  income quartiles and years
1.0




                                                       top




                                                        3rd
0.8
0.6




                                                        2nd
0.4




                                                       bottom quartile
0.2
0.0




      1960       1970      1980        1990     2000

                          5 year periods




                                  63
                −2   0    WGI indicators0 and income, 2005
                           2         −2            −2  0 2




                                                                                 8 10
         lgdp




                                                                                 6
2




                     ge
0
−2




                                  voice




                                                                                 0
                                                                                 −2
                                          polstab
0
−2




                                                                                 1
                                                    corrupt




                                                                                 −1
2




                                                              rol
0
−2




                                                                    regqual      2
                                                                                 0
                                                                                 −2



     6   8 10                −2     0               −1   1          −2   0   2




                                               64
                  ICRG indicators and income, 2005
                  2  6  10             1  3 5




                                                                            10
         lgdp




                                                                            8
                                                                            6
10




                      ip
6
2




                                                                            6
                                                                            4
                            corrupt.1




                                                                            2
                                                                            0
5




                                                rol.1
3
1




                                                                            4
                                                                            3
                                                            bq.1
                                                                            2
                                                                            1
                                                                            0



     6   8   10            0   2   4        6           0   1   2   3   4




                                       65
References

Acemoglu, Daron and James A. Robinson. 2001. “A Theory of Political Transitions.” American
    Economic Review 91(4):938–63.
Acemoglu, Daron, Simon Johnson and James A. Robinson. 2001. “The Colonial Origins
    of Comparative Development: An Empirical Investigation.” American Economic Review
    91(5):1369–1401.
Balch-Lindsay, Dylan and Andrew J. Enterline. 2000. “Killing Time: The World Politics of Civil
     War Duration, 1820-1992.” International Studies Quarterly 4:615–642.
Barro, Robert J. and Jong-Wha Lee. 2000. “International Data on Educational Attainment: Updates
     and Implications.” Harvard University, CID Working Paper No. 42.
Besley, Timothy and Torsten Persson. 2009. “The Logic of Political Violence.” Ms., LSE and IIES,
     Stockholm University.
Billon, Phillipe Le. 2001. “The Political Ecology of War: Natural Resources and Armed Conﬂicts.”
     Political Geography 20:56184.
Blattman, Christopher and Edward Miguel. 2010. “Civil War.” Journal of Economic Literature
     48(1):3–57.
  u
Br¨ ckner, Markus and Antonio Ciccone. 2010a. “International Commodity Prices, Growth, and
     the Outbreak of Civil War in subSaharan Africa.” The Economic Journal .
  u
Br¨ ckner, Markus and Antonio Ciccone. 2010b. “Transitory Economic Shocks and Civil Conﬂict.”
     Unpublished paper, Universitat Pompeu Fabra.
Brunnschweiler, Christa N. and Erwin H. Bulte. 2009. “Natural Resources and violent conﬂict:
    Resource Abundance, Dependence, and the Onset of Civil Wars.” Oxford Economic Papers
    61:651–674.
Brustein, William and Margaret Levi. 1987. “The Geography of Rebellion: Rulers, Rebels, and
     Regions, 1500 to 1700.” Theory and Society 16:467–495.
Buhaug, Halvard, Lars-Erik Cederman and Jan Ketil Rød. 2008. “Disaggregating Ethno-
    Nationalist Civil Wars: A Dyadic Test of Exclusion Theory.” International Organization
    62(3):531–551.
                                    a
Buhaug, Halvard, Scott Gates and P¨ ivi Lujala. 2009. “Geography, Rebel Capability, and the
    Duration of Civil Conﬂict.” Journal of Conﬂict Resolution 53(4):544–569.
Caprioli, Mary. 2005. “Primed for Violence: The Role of Gender Inequality in Predicting Internal
     Conﬂict.” International Studies Quarterly 49:161–178.

                                              66
Cederman, Lars-Erik, Andreas Wimmer and Brian Min. 2010. “Why Do Ethnic Groups Rebel?
    New Data and Analysis.” orld Politics 62(1):87–119.
Cederman, Lars-Erik and Luc Girardin. 2007. “Beyond Fractionalization: Mapping Ethnicity onto
    Nationalist Insurgencies.” American Political Science Review 101(1):173–185.
Cetinyan, Rupen. 2002. “Ethnic Bargaining in the Shadow of Third-Party Intervention.” Interna
     tional Organization 56(3):645 – 677.
Collier, Paul and Anke Hoefﬂer. 2004. “Greed and Grievance in Civil War.” Oxford Economic
     Papers 56:563–595.
Collier, Paul, Anke Hoefﬂer and Mans S¨ derbom. 2004. “On the Duration of Civil War.” Journal
                                      o
     of Peace Research 41:253–273.
Condra, Luke and Jacob Shapiro. 2010. “Who Takes the Blame? The Strategic Effects of Collateral
    Damage.” Unpublished ms., Stanford and Princeton Universities.
Condra, Luke N. 2009. “Ethnic Rebellion against the State: Perils of the Periphery.” Annual
    Meetings of the APSA, Toronto, 2009.
Cunningham, David E. 2006. “Veto Players and Civil War Duration.” American Journal of Political
    Science 50(4):875–92.
Fearon, James D. 1994. Ethnic War as a Commitment Problem. Presented at the 1994 Annual
     Meetings of the American Political Science Association, 2-5 September, New York.
Fearon, James D. 1995. “Rationalist Explanations for War.” International Organization 49(3):379–
     414.
Fearon, James D. 2004. “Why Do Some Civil Wars Last So Much Longer Than Others?” Journal
     of Peace Research 41(3):275–301.
Fearon, James D. 2005. “Primary Commodity Exports and Civil War.” Journal of Conﬂict Resolu
     tion 49(4).
Fearon, James D. 2007. “Iraq’s Civil War.” oreign Affairs 86(2):2–16.
Fearon, James D. 2008. Economic Development, Insurgency, and Civil War. In Institutions and
     Economic Performance, ed. Elhanan Helpman. Cambridge, MA: Harvard University Press.
Fearon, James D. and David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” American
     Political Science Review 97(1):75–90.
Fearon, James D. and David D. Laitin. 2007. “Civil War Termination.” Unpublished paper, Stan-
     ford University.


                                              67
Fearon, James D. and David D. Laitin. 2010. “Sons of the Soil, Migrants, and Civil War.”   orld
      evelopment . Forthcoming.
Fearon, James D., David D. Laitin and Kimuli Kasara. 2007. “Ethnic Minority Rule and Civil War
     Onset.” American Political Science Review 101(1):187–93.
Gallup, John Luke, Jeffrey D. Sachs and Andrew D. Mellinger. 1999. “Geography and Economic
     Development.” International Regional Science Review 22(2):179–232.
Gibney, Mark, L. Cornett and R. Wood. 2008. “Political Terror Scale 1976-2008.” Date Retrieved,
    from http://www.politicalterrorscale.org/.
Hamilton, Kirk and Michael Clemens. 1999. “Genuine Savings Rates in Developing Countries.”
      orld an Economic Review 13:33356.
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.    ass Political iolence. New York: Wiley.
Homer-Dixon, Thomas F. 2001. Environment Scarcity and iolence. Princeton, NJ: Princeton
   University Press.
Humphreys, Macartan. 2005. “Natural Resources, Conﬂict, and Conﬂict Resolution: Uncovering
   the Mechanisms.” Journal of Conﬂict Resolution 49(4):508–537.
Huntington, Samuel. 1996. The Clash of Civilizations and the Rema ing of     orld Order. New
    York: Simon and Shuster.
Kalyvas, Stathis N. 2006. The Logic of iolence in Civil   ar. New York: Cambridge University
    Press.
Kaufmann, Daniel, Aart Kraay and Massimo Mastruzzi. 2009. “Governance Matters VIII: Aggre-
    gate and Individual Governance Indicators, 19962008.” World Bank, Development Research
    Group, Policy Research Working Paper 4978.
Knack, Stephen and Philip Keefer. 1995. “Institutions and Economic Performance: Cross-Country
    Tests Using Alternative Institutional Measures.” Economics and Politics 7(3):207–227.
Kocher, Matthew Adam. 2004. “Human Ecology and Civil War.” Ph.D. thesis, University of
    Chicago, Department of Political Science.
         a
Lujala, P¨ ivi, Nils Petter Gleditsch and Elisabeth Gilmore. 2005. “A Diamond Curse? Civil War
     and a Lootable Resource.” Journal of Conﬂict Resolution .



                                              68
Lyall, Jason. 2009. “Does Indiscriminate Violence Incite Insurgent Attacks? Evidence from
     Chechnya.” 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.
Melander, Erik. 2005. “Gender Equality and Intrastate Armed Conﬂict.” International Studies
    Quarterly 49:695–714.
Miguel, Edward, Shanker Satyanath and Ernest Sergenti. 2004. “Economic Growth and Civil
    Conﬂict: An Instrumental Variables Approach.” Journal of Political Economy 112(4):725–
    753.
             e
Montalvo, Jos´ and Marta Reynal Querol. 2005. “Ethnic Polarization, Potential Conﬂict, and Civil
    War.” American Economic Review 95(3):796–816.
Moore, Barrington. 1966. Social Origins of    ictatorship and   emocracy. Boston, MA: Beacon
    Press.
North, Douglass C. and Robert Paul Thomas. 1973. The Rise of the     estern   orld A ew Eco
     nomic istory. New York: Cambridge University Press.
Østby, Gudrun. 2008. “Polarization, Horizontal Inequalities and Violent Civil Conﬂict.” Journal
     of Peace Research 45(2):143–162.
                              a
Østby, Gudrun, Ragnild Nord˚ s and Jan Ketil Rød. 2009. “Regional Inequalities and Civil Conﬂict
     in Africa.” International Studies Quarterly 53:301–324.
Powell, Robert. 1999. In the Shadow of Power. Princeton University Press.
Rodrik, Dani. 2005. “Why We Learn Nothing from Regressing Economic Growth on Policies.”
    Unpublished, Kennedy School of Government.
Ross, Michael. 2006. “A Closer Look at Oil, Diamonds, and Civil War.” Annual Review of Political
     Science 9:265–300.
Sambanis, Nicolas. 2001. “Do Ethnic and Non-Ethnic Civil Wars Have the Same Causes? A
    Theoretical 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.
Tong, James. 1988. Rational Outlaws: rebels and bandits in the Ming Dynasty. In Rationality and
    Revolution, ed. Michael Taylor. New York: Cambridge University Press pp. 98–128.
Urdal, Henrik. 2005. “People vs Malthus: Population Pressure, Environmental Degradation, and
    Armed Conﬂict Revisited.” Journal of Peace Research 42(4):417–434.

                                              69
Urdal, Henrik. 2006. “A Clash of Generations? Youth Bulges and Political Violence.” International
    Studies Quarterly 50:607–629.
Vreeland, James R. 2009. “The Effect of Political Regime on Civil War: Unpacking Anocracy.”
     Journal of Conﬂict Resolution 52(3):401–425.
Wagner, R. Harrison. 2000.      “Bargaining and War.” American Journal of Political Science
    44(3):469–84.
Walter, Barbara. 2002. Committing to Peace. Princeton: Princeton University Press.
Walter, Barbara F. 1997. “The Critical Barrier to Civil War Settlement.” International Organization
     51(3):335–64.
Weinstein, Jeremy M. 2007. Inside Rebellion The Politics of Insurgent iolence. 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 Review
   74:316–337.




                                                70