UNIVERSITY OF AMSTERDAM
GRADUATE SCHOOL OF SOCIAL SCIENCES MASTER THESIS
Students as a cause for civil conflict?
A quantitative empirical research on the influence of students on
civil conflict.
Arthur de Wilde Student Number: 85942829 M.Sc. International Relations 2013-‐2014
Research Project: The Political Economy of Conflict
Supervisor: dhr. prof. dr. Brian Burgoon
Second reader: dhr. Prof. dr. Geoffrey Underhill
Abstract.
So far, academic research on the effect of higher education on civil conflict has been unclear and ambiguous. Primary and secondary education, are suspected to influence civil conflict in a negative way, and the effect tertiary education has on civil conflict remains ambiguous and vague. There has been research on the ambiguous influence tertiary education has on civil conflict, and Barakat and Urdal have sought to explain this influence trough large youth cohorts. I in this research will do an adaptation to their article, and will try and explain the influence secondary and tertiary education have on civil conflict through economic structures. I translate these economic structures in unemployment for this research. And this study suggests there is a positive relationship between the two. I argue that larger shares of higher education when combined with higher shares of unemployment correlate positively with civil conflict incidence. This could have offsetting implications for education policies since larger shares of students could induce a large risk of civil conflict when the unemployment is high. More positive policy implications could include a stronger unemployment policy which is by this paper suggest to reduce the change of civil conflict when the level of secondary and tertiary educated students is higher. Supporting this argument is a cross-‐sectional time series logit regression, suggesting that the influence of higher education on civil conflict is positive when controlled for unemployment.
Knowledge is Power, Power corrupts.
Study hard, Be evil?
Introduction.
In recent studies education has proved to be an important factor in influencing civil conflict. The higher the percentage of the population to have followed primary and secondary education, the lower the chances that civil conflict will break out, this is argued to happen through a mechanism that increases opportunity cost. The effects of primary and secondary education on civil conflict seem clear and undeniable; statistical research and case studies argue that primary and secondary education, both have a negative effect on civil conflict. In other words, we have good reason to believe that countries with higher average levels of primary and secondary education do indeed have a lower risk of experiencing internal armed conflict. These findings appear to coincide with qualitative case studies, which generally seem to suggest that low access to education explains participation in civil conflict (Barakat & Urdal, 2009; Thyne, 2006; Bussman, 2007). One example is Brett & Specht (2004) who have been conducting interviews with young soldiers, and have found strong micro-‐level support for the expectation that lack of schooling in addition to poverty, and low alternative income opportunities are important reasons for joining a rebel group.
The mechanisms through which education affects civil conflict are perceived to work in two ways. First, educational investment of all kinds, including for primary education, provides a strong signal to the people that the government is attempting to improve their lives, and this is apt to lower grievances even in desperate times. Second, secondary education can generate economic, political and social stability by giving people tools with which they can resolve disputes peacefully, making them less likely to incur the risks involved in joining a rebellion. So one could argue that primary and secondary education undoubtedly have a negative effect on civil conflict.
Yet the effect of tertiary education on civil conflict remains unclear. There are several studies arguing that the effects are positive, negative or simply insignificant. As Østby and Urdal (2010) argue in their review article, findings on the effect of expansion in higher education seem to have no clear effect on civil conflict. These studies all focus on the effect of expansion in tertiary education on civil conflict, including armed conflict, terrorism, and riots. They study the impact of expansions in higher education on the
levels of lethal and non-‐lethal urban social disturbances and came to several
conclusions. There is evidence that the interaction of youth bulges with expansion in higher education was associated with an increased risk of terrorism. Yet, other studies find that expansions in higher education seem to have no bearing on the risk of civil conflict, riots, or urban violence. Other scholars find individual-‐level studies reporting higher education can, in fact, be a factor in recruitment to terrorist organizations, suggesting in turn that tertiary education can increase rather than decrease the risk of violent conflicts (Østby and Urdal, 2010).
This article aims to deliver clearer explanations about the effect of tertiary education on civil conflict, It’s main theoretical and empirical innovation relative to prior research is to consider the added influence of an important mediating factor that plausibly conditions how education shapes conflict: is unemployment. This paper will try and clarify the influence that the percentage of people with tertiary education has on the outbreak of civil conflict. This influence is expected to be conditional, though – for the existing literature suggests, and my own expectation is, that the role of tertiary education to be an ambiguous one that can translate into either increased or reduced risks of conflict. I expect this conditionality to involve economic factors, particularly unemployment. Since I expect higher educated to become frustrated earlier by socio-‐ economic structural factors.
I aim to analyse the potential mediating role of such unemployment on the relationship between education and civil conflict. My ambition is to use the existing work by Barakt and Urdal (2009) and reshape their research on the mediating role of educational attainment and education reform on the relationship between youth bulges and conflict. By rearranging the mediating and explanatory variables I aim to come to a clearer effect of education on civil conflict. This will be done trough testing a global model for the time period from 1970 till 2010.
According to Barakat and Urdal (2009) higher levels of educational attainment are generally perceived to have a negative influence on civil conflict incidence, increasing the opportunity cost of rebel recruitment among young people, and hence reducing the likelihood that they are recruited to rebel organizations. According to Thyne (2006) educational investment sends out a signal, to society, that the government is attempting to improve the internal situation, this is expected to lower grievances. Next to that education is perceived to generate economic political and social stability by giving
people tools that enable them to resolve disputes peacefully. So these are arguments that opt for a negative influence education and educational investment on civil conflict. Yet as stressed by Barakat and Urdal (2009) there also might be positive effects linking secondary and especially tertiary education to civil conflict, they address claims that conflict may arise due to unmet expectations in the form of low progression ratios between different education levels. They argue that rapid expansions in secondary and tertiary education might produce an over-‐capacity of highly educated youth for which there are limited employment opportunities. These students will grow frustrated due to this shortage of jobs and are therefor expected to have a positive influence on civil conflict.
On this latter argument I will build my thesis, yet I will drop the mechanism Barakat and Urdal (2009) pose based on frustration due to an over-‐capacity of highly educated youth. I believe the role of tertiary educated, as having a positive influence on civil conflict to be one that is conditional upon the degree to which economic conditions in the economy generally, and with respect to the prospects of the tertiary-‐educated population in particular, are bad. To the extent that unemployment in the economy is high, one can hypothesize that the pacific effects of having a high percentage of the population with a tertiary education can turn from peaceful to violent-‐prone. Higher shares of tertiary-‐educated can be a cadre and organizing force for violent ferment and political discontent, to the extent that unemployment generally is high. This leads me to a research question stating the following: What are the implications for civil conflict when we encompass larger groups of higher educated in society? And does this influence differ with fluctuation in the level of unemployment?
As Gates (2012) argues, war is a development issue. War kills, but the
consequences extend far beyond these direct deaths. In addition to battlefield casualties, armed conflict often leads to forced migration, refugee flows, capital flight, and the destruction of societies’ infrastructure. Social, political, and economic institutions are indelibly harmed. The consequences of war, and especially civil war, for development are profound, as suggested by Gates (2012). War creates a development gap between those countries that have experienced armed conflict and those that have not. Through this suggestion by Gates I would like to put emphasis on the policy shaping value of this thesis, when we have a clear idea of the influence education has on civil conflict; policy makers can stimulate or retain educational policies in order to prevent civil conflict. The
academic value of this thesis lies in the answer it can grant to the on-‐going debate as depicted by Østby and Urdal (2010), that of the role higher educational attainment and enrolment play in civil conflict when accounted for structural socio-‐economic factors. And in particular be complementary to the work Barakat and Urdal did and enrich it with a new perception of tertiary and secondary education as an explanatory value for civil conflict.
I do so through the use of a combined dataset that consists out of the Unesco educational enrolment dataset measuring cohort-‐specific educational enrolment and participation rates and contains time-‐series data from 1950-‐2010 for 146 countries. For? the controlling, mediating and dependent variable I use the Quality of Government dataset made available by the university of Gothenborg Sweden. It is theorized that primarily young men are linked to civil conflict hence I will run models with the whole percentage of the country that is enrolled in tertiary education as well as the percentage of male enrolment within tertiary education (Collier, Hoeffler & Söderbom ,2004. Collier & Hoeffler, 2004). I will also run these two models encompassing the number of
secondary educational enrolment. Since one might expect the influence of higher
education and unemployment to have a delayed effect on civil conflict and for matters of endogeneity, I will run the models in such a matter that the effect of the independent variables is linked to civil conflict onset a year after the measured year for the
independent variables.
It turns out, after running the models that tertiary and secondary education enrolment combined, while mediated for unemployment has a positive influence on civil conflict incidence. This effect seems slightly stronger when only measuring male
educational enrolment. When only measuring tertiary enrolment numbers the effect does not hold up, the coefficient remains positive yet is no longer significant. So we know from the interactions that the direction of the interaction is as theorized above for tertiary and secondary enrolment numbers combined. Then through running two
models without the mediating variable, and including only the cases where
unemployment was above the median, I prove that educated population shares cause a significant raise of conflict risk where unemployment is high.
Theoretical framework.
The theoretical and methodological framework for this thesis is partially laid out by Barakat and Urdal (2009) in their paper “Breaking the Waves? Does Education
Mediate the Relationship Between Youth Bulges and Political Violence?” In this paper they try to address ways through which education may serve as a strategy to reduce the risk of political violence and civil conflict, particularly in the context of large cohorts of young males. Their results suggest that poor countries do have some leverage over reducing conflict potential through increasing educational opportunities for young people. And their study supports broad policy interventions in education by relaxing concerns about the consequences of rapid educational expansion. They suggest this happens through several mechanisms and place these along the lines of the grievances and opportunities debate (Collier & Hoeffler, 2004).
With regard to this thesis’ interest in tertiary educated students, Barakat and Urdal argue that frustration, political opposition and aggression among youthful citizens arise as a result of pressure within educational institutions and the labor market
(Moller, 1968; Choucri, 1974; Braungart, 1984; Huntington 1996; Goldstone, 1991; 2001; Cincotta et al., 2003). And for the mediating variable it should be noted that in some countries youth unemployment even is particularly high among educated youths (McNally et al., 2004: 162; Kabbani & Kothari, 2005). In the following I will discuss how education when related to unemployment might influence relevant determinants of conflict motives and opportunities.
When unemployment is low I expect the tertiary students to be satisfied and this combined with the fact that education increases the opportunity cost of rebellion to have a stabilizing effect on society. And by virtue of attracting and being a potential future basis of political economic development and harmony, which in turn ought to diminish the chances of violent conflict and other forms of discontent. Yet when
unemployment rises I expect this group of tertiary students as stated above to become frustrated and form a foundation for political violence and civil conflict. Since higher shares of tertiary-‐educated can be a cadre and organizing force for violent ferment and political discontent.
Education as a factor that increases opportunity cost.
Opportunity factors can be perceived of as the economic benefit that could have been accrued had there not been conflict, and so are directly linked to the protractors of the conflict itself and the structural conditions. High unemployment is theorized to be a factor that reduces recruitment costs through the abundant supply of rebel labour since there are no other “jobs” available. “Rebel recruits join to obtain a private good, weighing the potential gains against the expected costs represented by the risk of being killed or maimed. Relative gains are high either when outside options are poor, or when a rebel group can offer greater rewards through loot-‐seeking activities” (Gates, 2002: 116).
Education is generally perceived to heighten the opportunity cost; it is suggested and empirically proven that higher educated earn a higher salary, therefor their life and time is more precious than that of lower educated. Or as Collier and Hoeffler (2004) state: rebel recruitment is more costly and rebellion less likely the higher the level of education in a society, everything else remaining equal. So when economic conditions are good we can expect tertiary education to have a pacifying effect on civil conflict, and through rational behaviour and opportunity-‐based mechanisms the tertiary educated function as a force for good. In addition to this, tertiary educated, when economic conditions are good, also function as a force for good through creating a situation or cadre in which it is appealing for international corporations to settle down. This settling down of large TNC’s has proven to have a positive influence on the countries GDP per capita and therefor should have a negative influence on civil conflict. This because expansion in GDP per capita is known to have a negative influence on civil conflict incidence (Chatterji, 1988; Miguel et all., 2004). Henceforth it can be theorized that when one perceives the economic situation in a country as positive and unemployment is low, tertiary educated can be expected to be a force for good.
Education as a factor that creates frustration.
Yet as Barakat and Urdal point out education can also be expected to have a less positive or negative influence on civil conflict, when certain socio-‐economic factors are present. This moment of higher educated students turning violent instead of peaceful
prone is on the grievance side of the grievances and opportunity debate. The relative deprivation theory posits that grievances and frustration arise when the gap between people’s expectations and their actual situation widens (Gurr, 1970). Higher educated people can be perceived to have a lower relative deprivation threshold than lower educated people, for they are more aware of amenities throughout the world and therefor can perceived to be more critical when comparing themselves with these amenities. So when the socio-‐economic situation in a country declines and higher educated are affected, they can be expected to be among the pioneers broaching the subject in society. Furthermore political violence is argued to be a mean to redress these grievances (Sambanis, 2002), higher shares of tertiary and secondary educated people can be a cadre and organizing force for violent ferment and political discontent to the extent that unemployment generally is high.
Concerning the relative deprivation, increasing education leads to increasing expectations of employment and salary, so according to this when there is high
unemployment and this is affecting tertiary educated, grievances should arise. Choucri (1974: 73) argues that: “high unemployment among educated youth is one of the most
destabilizing and potentially violent political phenomena in any regime[.]” In addition to
this Lia (2005) argues that due to rapid expansion of higher education in the Middle East and a labour market that does not expand accordingly, radicalizing effects have been measured among the higher educated youths. This coincides with the empirical findings that Terrorists are often higher educated than is the mean in their country or region of origin (Lia 2005). So when there is a shortage of economic opportunities, satisfying jobs and the amenities of higher educated students are not met, higher educated youth segments may address frustrations that could motivate political violence. So contrary to the argument Barakat and Urdal (2009) make, I perceive the frustration among the higher educated not to be caused through rapid growth of the educational system and large youth bulges, but simply trough unemployment. I expect unemployment among tertiary educated to be the mediating factor when looking at the influence of tertiary education on civil conflict. The reason and cause for this unemployment as Barakat and Urdal put an emphasis on, is of no importance in explaining the effect. So because of higher educated people, their higher social and intellectual capital, they can be a more effective? cadre for political violence or civil conflict. When they feel that these are the
means they need to use in order to address the grievances they experience, earlier than lower educated in society due to their lower relative deprivation threshold.
Grievances, Opportunities and a recap.
We see these fields of tension as discussed above spread out against the background of the grievances and opportunities debate. As argued by Collier and Hoeffler (2004) and later by Gleditsch and Cederman (2011), civil wars are either caused by grievances, opportunities or a combination of both. We see that the case concerning the effect of education on civil conflict is one that encompasses the whole of this spectrum. As cited by Collier and Hoeffler (2004: 563) “Rebellion may be explained by atypically severe grievances, such as high inequality, a lack of political rights, or ethnic and religious divisions in society. Alternatively, it might be explained by atypical
opportunities for building a rebel organization.” When we apply this mechanism to
higher education, unemployment and the two theories stated above we see that the first is merely about opportunities. The opportunity cost of recruiting tertiary educated in an army or rebellion group is higher when unemployment is low and a perception of
relative deprivation is absent. There will be little or no reason for grievances or other frustrations through the mechanisms described above. Also higher educated can be perceived to create a positive effect on the economy and henceforth have a negative effect on civil conflict onset since these two have a statistically proven and undoubted negative relation. Yet when the economic situation changes and unemployment rises among tertiary educated, one can perceive a case of combined grievances and
opportunities. Grievances arise due to relative deprivation and opportunities arise due to the fact that political violence might result in more personal gain than remaining loyal to the regime, also the relatively high personal capital of higher educated when
compared to lower educated might contribute to a cadre better able to support civil conflict.
So to recap: The percentage of the population enrolled in higher education can be expected to have important implications for violent conflicts, but those implications are likely to be mediated by background economic conditions like unemployment. In general, it may be that higher tertiary-‐educated populations are a force for peace, by
virtue of attracting and being a basis of political economic development and harmony, which in turn ought to diminish the chances of violent conflict and other forms of
discontent. Second the opportunity cost of higher educated persons rises, which in effect will make it more difficult for rebels to recruit them. Yet these pacific implications are likely to be mediated or conditional upon the degree to which economic conditions in the generally, and with respect to the prospects of the tertiary-‐educated population in particular, are good. To the extent that unemployment in the economy is high, one can hypothesize that the pacific effects of having a high percentage of the population with a tertiary education can turn from peaceful to violent-‐prone. Higher shares of tertiary-‐ educated can be a cadre and organizing force for violent ferment and political discontent to the extent that unemployment generally is high. Furthermore one could hypothesize that the relative deprivation threshold of tertiary educated is lowered due to awareness of surrounding and international cases and examples. This might cause them to sound their discontent earlier then others. Hence, the observable hypothesis is the following: Tertiary-‐educated shares should have more positive or less negative implications for civil conflict incidence where unemployment is higher. When applying this to the statistical research and models in this paper, four testable hypotheses come forth:
H1: Tertiary and Secondary education enrolment shares should have more positive or less negative implications for civil conflict incidence where unemployment is higher.
H2: Tertiary education enrolment shares should have more positive or less negative implications for civil conflict incidence where unemployment is higher.
H3: Tertiary and secondary education enrolment shares among males have more positive or less negative implications for civil conflict where unemployment is higher.
H4: Tertiary education enrolment shares among males should have more positive or less negative implications for civil conflict incidence where unemployment is higher.
As can be drawn from the hypotheses, I chose only to focus on secondary and tertiary education and not on primary education. I did so for several reasons, primary
education could be perceived to have both a negative and positive influence on civil conflict, through the same mechanisms as described above. It can be expected that just like secondary and tertiary education, primary education provides offsetting
implications – lowering grievances and raising opportunity costs for violence, but on the other hand increasing cadres of militants. And since universal primary schooling is considered a given, ever since the Education for All and Millennium development goals processes, no policy will be changed. Also I build my argument as an addition and therefor on the work of Barakat and Urdal (2009), and they neither focus on primary education, mainly for the reason described above. In addition to that Barakat and Urdal (2009) show that in most countries primary education attainment rates are high and close to a 100 % between the age of 15 and 25. While the aim of universal primary schooling is considered a given since the Education for All (EFA) and Millennium Development Goals (MDG) processes
Empirical model and data.
This thesis analysis is based on a dataset combined from two sources and expanded through the calculation of additional derived variables. The dataset covers 146 countries for the 1950-‐2010 period. Data on the independent variable -‐ tertiary
educational enrolment’ is drawn from the Barro-‐Lee education database. Where Barakat
and Urdal use the IIASA dataset, I chose to use the Barro-‐Lee dataset over the IIASA for the fact that the IIASA only measures till 2000, but it is argued that since the end of the cold war we have seen a massive incline in civil conflict, therefor it seemed more relevant to use a dataset that also comprises the years 2000 till 2010 (Fearon & Laitin, 2003). The dependent variable data on civil conflict is drawn from the Quality of
governance dataset from which I used the Uppsala/PRIO dataset part, which is the same as Barakt and Urdal (2009) use. The dataset uses a low battle death threshold for the onset of minor armed conflict: a minimum of 25 battle deaths per year. Also used are thresholds of 1000 battle deaths yet I recoded these to match the same low threshold of 25 battle deaths, in order to be able to create a binary dependent variable.
Unemployment data is also drawn from the Quality of Government dataset and is comprised from the World Development Indicators dataset, some difficulties arise here,
since unemployment refers to the share of the labour force that is without work but available for and seeking employment and definitions of labour force and
unemployment differ by country. Ideally therewould have a database that grasps the unemployment in different educational attainment groups, yet this data is only available for highly developed countries, of which most in the last half-‐century did not experience civil conflict. Therefor, for this thesis, it would be no use to implement this data in the dataset, the WDI data is the most complete dataset available on unemployment
worldwide and encompasses a time period from 1980-‐2010 so it comprises? the most important years concerning civil conflict. These datasets combined –on education, civil conflict en unemployment-‐ show enough observations for the models to be significant and robust, yet I worked with different proxies for unemployment in order to back up my finding.
The dependent variable I measure is civil conflict, the Prio and UCDP databases have several measures for conflict in their databases and this data is present in the Quality of Government database. There are different measures of civil conflict so these had to be rewritten and combined in a new variable.
Civil Conflict measurement.
The PRIO database measures internal conflict in two ways, firstly they measure Internal Armed conflict, these conflicts occur between the government of a state and internal opposition groups without intervention from other states. Secondly the PRIO database measures internationalized internal armed conflict, these conflicts occur between the government of a state and internal opposition groups with intervention from other states. These two measures are relevant to this research since I in line with Barakat and Urdal (2009) only measure conflicts that arise inside the state and even though there could be international interference we should not drop the case from our research. These cases are valuable to the research since there is a small chance of actual civil conflict incidence; so ruling them out would unnecessary lower our observations of cases in which conflict did incidence.
Then PRIO categorizes these two measures of conflict in 4 different ways, they categorize: Firstly a situation where there is no war with a threshold of 25 battle related deaths per year. Secondly there is minor armed conflict that comprises at least 25
battle-‐related deaths per year for every year in the period. Thirdly there is intermediate armed conflict this exists out of more than 25 battle-‐related deaths per year and a total conflict history of more than 1000 battle-‐related deaths, but fewer than 1000 per year. Fourthly they measure war: which needs at least 1000 battle-‐related deaths per year. Since this research only measures conflict incidence the amount of battle related deaths is not relevant and only troubles the statistical research.
So in the new variable there is the data combined of internal conflict and
internationalized internal conflict, and I coded the measure on conflict incidence binary in which it either states zero in which case there is no conflict or it is coded as one which comprises all classifications of armed conflict.
Barro-‐Lee educational attainment data.
The educational attainment data originates from the Barro-‐Lee dataset, it contains information for 147 countries for the 1950-‐2010 period, “the benchmark
figures on school attainment (621 census/survey observations) used are collected from census/survey information, as compiled by UNESCO, Eurostat and other sources. The census/survey figures report the distribution of educational attainment in the population over age 15 by sex and by 5-‐year age group for most cases, in seven categories: no formal education, incomplete education, complete primary, lower secondary, upper secondary, incomplete and complete tertiary.” (Barro-‐Lee 2012) However for this research I use
only part of the Barro-‐Lee data, the data containing the figures on the percentage of enrolment within the country. I do not measure educational attainment as a whole since I expect students who are still studying to get frustrated through the mechanisms as described above. The data I? use is present in the Barro-‐Lee dataset and they derived it from the UNESCO education database. I use the data comprising the percentage of tertiary students per country and the percentage of tertiary combined secondary
students per country. This is done for both men and women combined as well as just for men singled out. This is done due to the perceived more aggressive nature of men when compared to women since men are believed to play a more aggressive and occupant role in civil conflict. The dataset covers a great number of developing countries, including many countries with recent conflict experiences.
Barakt and Urdal argue the IIASA data is prevalent over the Unesco data that partly comprises the Barro-‐Lee data set because they tend to miss earlier data, less consistent because of different definitions in countries and neither are they believed to be over time. They argue the IIASA dataset attempts to overcome these limitations “by
employing a novel methodology inspired by demographic methods to reconstruct historic attainment data from censuses performed around the year 2000.” (Barakat and Urdal
2009: p. 8) Besides their econometric argument for approaching earlier educational attainment data, the IIASA database only counts till the year 2000. But I have seen a huge incline in civil conflict since the end of the Cold War and the fact that the UNESCO data is argued to have grown more precise in the last two decades I? prefer the Barro-‐ Lee dataset above the IIASA dataset for the fact that it comprises data till 2010. Missing data in the Barro-‐Lee dataset is created by extrapolation and they use a time gap of five years for every measurement.
Unemployment.
Unemployment data was drawn from the quality of government dataset and is comprised from the World Development Indicators dataset, some difficulties arise though; unemployment refers to the share of the labour force that is without work but available for and seeking employment. Definitions of labour force and unemployment differ by country. Even more problematic proved the fact that the number of
observations is relatively low when doing cross section-‐time analyses with regard to education and civil conflict. In order to solve this and make the argument and research more reliable I decided to add one additional proxy for unemployment of which I expect it to influence the frustration of higher educated persons.
In order to proxy unemployment I used GDP growth and GDP per capita Growth, this because GDP growth is negatively related to unemployment. When GDP growth increases we see a decrease in unemployment and when GDP growth declines we see an increase in unemployment (Aghion & Howitt 1994). The data is available in the QoG dataset and is drawn from the World Development Index. It is set out as percentage of the GDP in order not to get a skewed view of countries with different size economies. GDP is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products.
So unemployment in the analysis is a percentage of society that is unemployed and in order to strengthen these figures GDP growth was used as a Proxy for
unemployment and together I expect the two variables to consist of enough observations to make the analysis reliable and significant.
Education mediated by unemployment.
In order to measure the mediating effect of unemployment on education a new independent variable had to be created. The mediating variable –unemployment-‐ is one that seeks to identify and explicate the mechanism that underlies the observed
relationship between the independent variable –education-‐ and the dependent variable –civil conflict-‐, instead of hypothesizing a direct causal relationship between the
independent and dependent variable. Thus the mediating variable serves to clarify the nature of the relationship between the independent and dependent variable (Judd &
Kenny, 1988).
By multiplying the WDI unemployment data with the Educational enrolment data I created this mediating variable; since there are four different groups of enrolled
students I measure, I created four different new variables. Nothing had to be recoded because both variables measure in percentages of the population and the country codes were similar. In theory, since I measure the difference gender makes on the effect of students on civil conflict, the unemployment data had to be transformed. Yet due to a lack of data on actual unemployment among different sexes I was not able to transform this variable in the correct way.
Due to the fact that in general most societies were civil conflict takes place the traditional male female structures are more rigid than in societies were civil conflict does not take place this does not have to be a problem for my research (Collier, 2008). Since females not looking for a job are not perceived of as unemployed, one could theorize that the amount of unemployed females is much smaller than the amount of unemployed males. In the case where I measure both sexes this is no problem, since we measure the whole of the population in both male and female cases. When I measure male enrolment though this data on unemployment will bias my statistical outcomes. I expect the mediating variable when only measured for male enrolment to be biased in such a way that it will show a weaker correlation. This because the analysis will
relatively measure more unemployed for the amount of enrolled male students than for the whole of the case, since this is what I suspect to be the frustrating variable the measured mediating effect will turn out lower. So since the Bias will only lower the measured correlation of my analysis I do not deem it problematic and the effect should be perceived to be stronger than the outcome of the analysis suggests.
Control Variables
For my control variables I decided to build on Barakat and Urdal (2009): I use the
Level of development, which is interchangeably proxied by infant mortality rate (IMR)
comprised in the QoG dataset that collected it from the World Population Prospects (UN, 1999). I use a log-‐transformed measurement for PPP adjusted GDP per capita collected from the Penn World Tables and comprised in the QoG dataset (Heston et al., 2002). And just as Barakat and Urdal I use the Polity IV data (Marshall & Jaggers, 2000) in order to measure regime type, this variable ranges from -‐10 (most autocratic) to 10 (most democratic). I use the measure of total population size that is drawn from the World Population Prospects and is log-‐transformed.
Combined dataset and model.
The above datasets where combined by year and Country code. In order to do so I extrapolated the Barro-‐Lee data, this seemed like the preferable way to do so since Barro and Lee also extrapolate their own missing data. By doing so I would remain within their mode of operation.
The combined dataset covers 147 countries for the years 1980-‐2010, the key variables used in the analysis are: Civil conflict, infant mortality rate, polity index, total population, GDP per capita, percentage of tertiary educated and unemployment.
Unemployment is used itself as a variable and is also proxied by GDP growth. I run four different models in which the variance is found in the student population I measure. I measure tertiary enrolment among the whole of the population and just among males, and tertiary and secondary population are measured among the whole population and just among the male shares of the population. These four models where all run
measuring civil conflict onset the year after the independent variable measurement. This is done because the independent variables could be considered to have a delayed effect on civil conflict and in order to solve problems of endogeneity (Blundell & Powell, 2003). One can expect civil conflict to influence all the independent variables, so in order to maintain a correct causal relation I measure the Independent variables a year prior to the dependent variable. In total 4 models are run; the models are random intercept maximum likelihood logit models with unstructured covariance in the estimation of the standard errors. This means that I have separate intercepts for each of the countries in the analysis, allowing me to take account of the country-‐specific process of politics being modeled econometrically. These analyses where run through the use of the statistical system STATA.
Analysis.
The core set of explanatory variables included in the model is drawn up
according to the explanatory variables Barakat and Urdal (2009) use for their models. I use the Infant mortality rate (IMR) as the indicator of ‘general development’ in the present analysis, this because as a measure it is more sensitive to social development and economic inequality and does not only focus on material living conditions. GDP per capita is used since it strongly correlates to civil conflict. Four different types of
educational enrolment groups are being used, both male and female are being measured and there are two tests where male share of the enrolled students is being measured. This due to the fact as described above that we expect males to be more aggressive and play a more occupant role in society. Furthermore we analyse the effect the addition of secondary education enrolment has on the independent variable. Measure 1 is the total percentage of males and females in secondary and tertiary education. Measure 2 is the ratio of both males and females enrolled in tertiary education, measure 3 is the ratio of males that are enrolled in secondary as well as tertiary education. And measure 4 is the percentage of males enrolled in tertiary education.
I involved secondary education in the models because it is found to provide the most suitable differentiator in assessing the role of education when taken
later years the proportion of the highest category -‐enrolment in tertiary education-‐ are small in most countries, while the share of secondary enrolment youth covers the full range of possible values. Next to that Barakat and Urdal (2009) argue, the measure of secondary enrolment is of policy importance, “the transition to secondary education is a
threshold for participation in the modern economic sector and as such is likely to mark a significant rise in opportunity costs for participation in violent conflict.” (P.12) So in line of
this though the effects of secondary schooling are of practical interest for policy formulation in developing countries. This in contrast to the aim of universal primary schooling since this is considered a given because of the Education For All and Millennium development Goals processes. Also one should realize that for many developing countries mass enrolment at the tertiary education level remains arduous. Because of all this there is genuine debate and a wide range of opinions, within
developing countries and international organizations, regarding the appropriate rate of expansion of secondary schooling (Alvarez et al., 2003).
Finally in order to counter problems of endogeneity, and to disambiguate the causational direction, I measure, the conflict occurrence a year later then educational enrolment data, unemployment data and the other control variables of which I know the relationship to be ambiguous. Due to the measurement a year later, enrolment numbers and unemployment cannot be directly influenced by the occurrence of conflict in the year of analysis. The possibility that an unknown third factor affects both education, unemployment and the occurrence of conflict, but with different time lags, can of course not be excluded.
Measured educational effects on civil conflict.
To start, the set of independent variables excluding the newly created variable education that was mediated for unemployment were tested. This was done in order to see if I could establish an underline, verifying whether a sort like effect as found in Barakat and Urdal (2009) could be replicated, while using different educational measures, and data.
The models are random intercept maximum likelihood logit models with
