We’re Not Gonna Take It: The Impact of Societal Support on Conflict Onset


Societal support for a regime is a critical element in the build-up to conflict. Even so, there exist no studies that attempt to directly measure societal support. In this analysis, I test the effect of support on civil conflict onset using World Values Survey data to measure support across a sample of 69 countries. I find that decreased support for a regime significantly increases the likelihood of civil conflict breaking out in subsequent years. These findings demonstrate that society, not just rebel and state actors, plays a role in the process of civil conflict.

Table of Contents: 

    Why Societal Support?

    From 2011 to 2013, Egypt experienced a degree of political turbulence that it had not seen in decades. Dissatisfaction with the government erupted into protest and the deposition of President Hosni Mubarak within a span of months. The protests succeeded in forcing change, but soon the tide of public outrage turned against the newly elected Mohamed Morsi. Within two years, two leaders lost the support of the Egyptian people, and conflict broke out between the competing factions. Though the situation in Egypt is far from settled, it demonstrates the powerful impact of societal support on conflict onset. Despite its key role in this process, societal support is a relatively understudied factor in civil conflict. In this analysis I develop a theory that attempts to answer the question: What is the role of societal support in civil conflict onset?

    Research into the causes of conflict has been expansive, examining economic, ethnic, and resource-based theories of conflict onset (Sambanis 2002). One theoretical approach to conflict onset has been through an escalating process of dissent and repression. Young (2012) demonstrates that repression is a significant factor in predicting civil war onset. He suggests that repression causes a decline in societal support, which leads to more dissident activity. Theories dealing with the strength of the state or economic factors are well-researched, but work dealing exclusively with societal support is non-existent. I attempt to fill this gap in the existing literature by examining the impacts of societal support on conflict onset.

    The State of Repression Research

    There is a wealth of literature examining repression. Here I break down the existing literature into several categories: causal factors of conflict onset, repression, decision-based models of repression and dissent, mobilization, and the existing literature on societal support.

    Conflict Onset

    Civil conflict has many causes. Collier and Hoeffler (2004) famously examined “greed and grievance” models of civil conflict onset, investigating a broad range of economic, geographic, and social influences. They performed regressions for “greed” and “grievance” models of civil war outbreak, finding that several factors are significant in determining the likelihood of intrastate conflict. Among the most significant of their findings are economic causes, measured in commodity exports divided by GDP and GDP growth; social causes, measured in population size and fractionalization; and government, measured using an index for democracy.

    Young (2012) uses repression to explain why some weak states experience civil war while others do not. While controlling for several of the major causal factors found in Collier and Hoeffler, Young uses a logit analysis to demonstrate that the use of state repression increases the likelihood of civil war onset in weak states. Young’s findings demonstrate the need for a better understanding of how repression contributes to civil conflict onset.


    Research into the factors that influence repression is vital for understanding the escalation that leads to civil war. Because repression plays an essential role in determining societal support, the factors that cause it must be taken into account. Poe and Keith (1999) identify 11 statistically significant factors for predicting repression: Civil war, international war, democracy, British cultural influence, population size, rate of population change, economic standing, percent economic growth, leftist government, military control, and previous history of human rights abuse. This analysis provides a basic understanding of the factors that lead to repression.

    Other studies have focused on the link between regime type and repression. Davenport (1999) examines the effect of regime change on repression. A regime change to democracy causes both a short-term and long-term reduction in repression, whereas a change to autocracy leads to a short-term increase in repression that subsides as time goes on. Poe and Tate (1994) find that loss of democracy leads to an increase in repression over time.

    In a 2010 analysis, Sabine Carey examined which factors provide a sufficient threat to warrant repressive tactics. Carey (2010) tests government responses to different types of dissent: demonstrations, strikes, riots, guerilla warfare, and revolution. The analysis finds that when controlling for regime type, guerilla warfare is the most likely form of dissent to provoke a repressive response. The theoretical reason behind the repressive response to guerilla warfare is that it is both highly organized and violent, and regimes view both of these characteristics as especially threatening. Carey also contributes to the role of democracy in repression by demonstrating that there is less active repression in democratic and authoritarian countries. Weak democracies are most likely to experience repression.


    Interactions between state and rebel actors can be conceptualized as a series of decisions. The process of conflict escalation, and the role of societal support in that process, can be more effectively understood by modeling repression and dissent in this way. Moore (2000) suggests that concession and repression are effectively substitutes for one another, as they are both tools used by the state to meet its ends. The state should, therefore, use whichever tactic proved more successful in past interactions. The state, according to this theory, is less likely to use whichever tactic (concession or repression) is met with dissent. He demonstrates this theory using a quantitative analysis of case studies in Peru and Sri Lanka.

    Carey (2006) uses a more expansive group of cases spanning several events in nine different countries to show that government and dissident tactics influence future interactions. Shellman (2006) likewise finds that previous interactions influence future decisions based on cases in Israel and Afghanistan. Shellman and Carey both provide evidence against one-way causality, instead lending support to a reciprocal relationship between dissent and repression. Decisions to escalate dissent or repression are not made in a vacuum; participants take prior experience with the other party into account when deciding on their actions.

    Gartner and Regan (1996) analyze the process of dissent and repression as a non-linear decision-making framework. They argue that the political context of the threat posed by dissenting groups is taken into account. States consider the potential international and domestic costs of repression when deciding how to respond to dissent. Additionally, governments take into account the seriousness of the threat posed by dissidents when deciding whether or not to repress. They argue that this series of decisions can often result in a cycle of repression and dissent that stops short of civil war.


    Research into the process by which citizens become rebels is important for understanding how a lack of societal support contributes to dissent and conflict escalation. One of the earliest and most influential works in this area is Why Men Rebel (Gurr 1970), which examines the factors that contribute to individual incentives to join demonstrations and rebel groups. Gurr introduces the concept of relative deprivation as an explanation for mobilization. He asserts that when citizens feel that their own economic and social position is low relative to their expectations, and when they are not able to change their position through the established political system, rebellion is their only option to achieve social and economic mobility. Gurr’s work suggests a potential mechanism through which societal support can contribute to conflict onset: a dissatisfied populace is more likely to rebel.

    Rasler (1996) examines the Iranian revolution of 1979 to determine how dissent and protest spread. She finds that repressive tactics initially reduced the number of protests, but caused them to grow in the long term. This suggests that the repressive tactics used by the Iranian government caused dissent to increase despite initially containing it, thereby contributing to the escalation of conflict.

    This “backlash” effect demonstrated in Rasler’s data provides evidence for a more active role of citizens in dissent. There is also evidence that states recognize the important role that citizen support plays in conflict. Valentino et al. (2004) demonstrate that mass killings often target civilian populations that support guerrilla groups. They propose that these targeted killings are aimed at reducing the amount of societal support available to rebel groups.

    Barbara Walter (2004) finds that civil war is more likely to recur if individuals have incentives to join rebel groups. In particular, she notes two factors that increase the likelihood of civil war recurrence: low quality of life, and absence of legitimate channels through which to enact political change. Unable to enjoy a sufficient quality of life or change their situation politically, individuals have no incentive to stop fighting. In order to prevent conflict recurrence, individuals, and not just institutions, must have reason to choose peace over continued violence.

    Societal Support

    The first attempt to measure societal support as a factor of repression is Young’s analysis on repression and civil war onset. His theory and tests claim that weak states are most likely to experience civil war when they are more repressive and lack societal support (Young 2012). Young conceptualizes societal support as being a component of state capacity because a low level of societal support creates an opportunity for dissident activity. In his analysis, Young uses Kugler and Domke’s measure of relative political capacity (RPC) as a proxy for societal support (Kugler and Domke 1986). RPC is measured by dividing the actual extraction of resources from the expected extraction of resources in a country, the idea being that more capable political systems are more capable of extracting resources from their citizens.


    My theory attempts to account for the impact of societal support in conflict onset while addressing weaknesses in the dominant conceptualization of support.


    I argue that Young’s measurement of societal support is actually a measure of state capacity to extract resources from its citizens. In fact, Kugler and Domke (1986, pg.45) created RPC as a way to measure political capacity of the government, not the support of its citizens. The ability of RPC to stand in for societal support is predicated on the assumption that tax payment is voluntary and that citizens will not pay taxes if they do not support the state. There are, however, some weaknesses in this assumption. First, a citizen who does not support the regime might be persuaded to pay taxes by some implicit or explicit threat of consequences from the government. Second, these consequences are not the same across all countries (See Figure 1.).

    Assuming that tax payment is voluntary and that citizens are rational actors, they should only pay taxes if the cost of paying taxes outweighs the costs of refusing to pay taxes c (t > c). Because c and t both vary across countries, using RPC to conceptualize societal support will bias results toward greater support for countries that have low tax rates or high consequences for not paying taxes. While this measure may accurately portray the ability of a state to collect taxes, it is not a substitute for approval ratings, as Young claims (Young 2012, pg.7).


    A lack of societal support for a regime increases the likelihood of civil conflict onset. For the purposes of this theory, I define societal support as the percentage of citizens who approve of the continued existence of the current regime. I assume for this argument that rebel groups are given; i.e.,  the government is never able to completely prevent rebel groups from forming or existing.

    A drop in support for the government increases the capabilities of dissenting groups relative to the government, allowing them to resist the government more effectively. A drop in support for the regime effectively lowers the capacity of the state. The reciprocal increase in support for rebel groups enhances rebel capacity, resulting in a situation where rebel groups and the state are more equally matched. As rebels are increasingly able to fight against the government, I expect conflict to escalate.

    Hypothesis: A lower level of societal support will result in greater likelihood of civil conflict onset.

    As support for rebel groups increases, so too does the ability of rebel groups to mobilize citizens. Because of the drop in support for government, rebel groups have a larger pool of citizens from whom to recruit. Citizens who do not approve of the government are more likely to join or support rebel groups, which effectively increases their social capital. The societal effects of support also include an acceptance of rebellion as it becomes more widespread.

    As support drops, more citizens feel that the costs of continuing to live under the current regime outweigh the costs of participating in rebel groups. Thus, a greater number of citizens are willing to join rebel groups as state support decreases. People feel that rebellion is worth the costs when government is not doing its job relative to what they expect.

    In summary, I expect to see the likelihood of civil conflict increase as government support drops. A lack of support increases individual and societal incentives to join or support rebel groups. This increases the capacity of the rebel groups relative to the state’s ability to repress them. This increase in rebel group capacity allows for increased conflict between rebel groups and government, leading to the onset of civil conflict.

    Research Design

    In order to test my hypothesis, I will be using a measure of conflict onset, societal support, and several controls. Conflict onset data is available from 1946 to present, but the temporal domain of the data is limited by the availability of public opinion data. Data on support is available from 1999 and onward at the earliest, depending on the country.

    Conflict Onset

    The dependent variable in my analysis will be civil conflict onset. I use the UCDP definition of civil conflict, defined as “conflict between a government and a non-governmental party, with no interference from other countries,” and resulting in 25 or more battle-related deaths (Themnèr and Wallensteen 2012). In order to measure the onset of civil conflict, I will use a variable from the UCDP Onset of Intrastate Armed Conflict database, 1946–2011. This variable measures conflict onset at the country-year level.

    Due to constraints on the availability of data on support, there are only 5 instances of conflict onset for country-years in which data is available, and two of these are resurgences of conflict after a single year of peace. Because this lack of samples could lead to unreliable statistical inferences, I also test to see whether conflict occurs in the 5-year and 3-year interval after the measure of support is taken. Instances in which conflict arises in a country up to three (or five) years after the support measure is taken will be coded 1, while instances in which no new conflict occurs will be coded 0. The full dataset for the 3-year test contains a sample of 123 observations, of which 11 result in conflict onset. The data set for the 5-year test contains 106 observations, of which 16 result in conflict.

    A variable for conflict onset from the UCDP database, onset2v412, will also be tested as a robustness check. This variable measures conflict onset in countries where at least two years have passed since the last conflict. This two year requirement eliminates the resurgence of existing conflict from interfering with the data. Using a 5-year standard for intervals between conflicts creates a dataset too small to be reliable.

    Societal Support

    The independent variable of interest in my analysis is the degree of societal support for the government. I define societal support as the percentage of citizens who support the continued existence of the current regime. In order to determine citizen support for the government, I will be using data from World Values Survey Integrated Dataset (European and World Values Surveys Four Wave Integrated Data File, 1981-2004, 2006).

    The dataset I will use for this analysis consists of responses to a survey question from several waves of worldwide surveys. The World Values Survey question of interest deals with confidence in government. (Full question text and response options are available in Appendix A.) The variable will be measured as a percentage of those who support the government.

    This approach allows for public approval to be directly measured without relying on proxies. I believe that using opinion data rather than tax collection data provides a more accurate picture of societal support, because it prevents other factors, such as state capacity, from affecting the data. Because surveys were conducted throughout the year in which support was measured, support of the previous year is used in order to better isolate causality. This prevents cases in which support falls because of conflict onset in the same year.


    There are a number of control variables that must be accounted for when assessing the likelihood of conflict onset in a country-year. I include only controls that are shown in the literature to have a consistent effect on conflict onset for the sake of parsimony and in order to avoid limiting the sample size of the data. Table 1 shows the distribution of all independent variables, and Figure 2 displays the geographic location of the data points.

    Repression. The level of repression carried out by the state is an important factor in whether a conflict escalates to civil war (Young 2012). I will control for the level of repression in a given country-year by including a measurement of repression from the Political Terror Scale. The Political Terror Scale rates the level of repression carried out in a country-year on a 1 to 5 ordinal scale, where 1 indicates countries “under a secure rule of law,” and 5 indicates that terror “has been extended to the whole population.” (Wood and Gibney 2010).

    The PTS provides two measures of state terror, one based on Amnesty International reports and the other based on State Department reports. I take the average of the two scores when possible, and use the available score for country-years in which both measures are not present.

    Economics. The economic standing of a country has been shown to have a significant effect on the likelihood of conflict onset (Collier and Hoeffler 2004). I will include a measure of GDP per capita as a control for the effects of economic differences in the likelihood of civil war onset. GDP per capita is measured using World Bank data (World Development Indicators 2013).

    Population. The size of the population in a country is a significant predictor of civil war occurrence (Collier and Hoeffler 2004). I will include a measure of population in the dataset in order to control for the population size of a country. I use population data gathered by the World Bank (World Development Indicators 2013). Population is logged for easier interpretation of results.

    Democracy. Previous studies have shown that democratic governance reduces the likelihood of civil conflict onset (Collier and Hoeffler 2004). To account for democratic governance, I control for democracy using Polity IV data on Democracy (Marshall and Gurr 2013). Democracy is measured on a scale from 0 to 10, with ten being the most democratic and 0 being the least democratic.


    Because civil conflict onset is a binary dependent variable, I will test my hypothesis using logistic regression. Logistic regression calculates the expected change in likelihood of civil conflict onset for each change in an independent variable. My hypothesis predicts a significant negative coefficient for support; that is, the higher the level of societal support, the lower the likelihood of conflict onset.

    I will perform three tests, each using a different measure for civil conflict onset. The first measure indicates whether conflict breaks out in the next 5 years; the second measure indicates whether conflict breaks out in the next 3 years; and the final measure indicates whether conflict breaks out in the next year, requiring that at least two years have passed since the last conflict in that country. I use these three measures so that the results can be reliably robust.


    First, I test to determine the probability of conflict onset in the five year period after a measure of support is taken. Table 2 displays the results of this test. Table 2 shows that the percentage of support for a regime is a significant factor in civil conflict onset. GDP per capita is a negative predictor of conflict onset, meaning that an increase in GDP per capita reduces the likelihood of onset. Population size is the only significant positive predictor of civil conflict, meaning that a larger population leads to more conflict. Neither repression nor democracy is significant in this model.

    Next I test for the probability of conflict onset in the 3-year period after the measure of support is taken. The results of this test are shown in Table 3. The 3-year model shows results similar to the 5-year model. Support and GDP per capita are negative predictors of conflict, while population is a positive predictor. Repression and democracy remain insignificant.

    Finally, I test for the probably of conflict onset in the year following measurement of support. The results of this test are shown in Table 4. In this model, percent support is the sole variable that remains significant. The diminished significance of other factors could be evidence that support is a better predictor of conflict in the short-term than in the long-term. I caution that this model may be less reliable than the other two, as only three of the 123 cases result in civil conflict within the next year.

    In each case, the results show support as a significant factor in civil war onset. The coefficient for support on conflict onset is negative in all models, meaning that an increase in support leads to a decrease in the likelihood of conflict onset. This confirms my expectations according to the hypothesis.

    Hypothesis: A lower level of societal support will result in greater likelihood of civil conflict onset.

    In order to understand the magnitude of societal support as a predictor for conflict onset, I estimate the probability of conflict onset as support changes. I apply a Monte Carlo simulation technique to determine the probability of conflict onset as societal support moves from 0 to 100 percent (King, Tomz, and Wittenberg 2000). Figure 3 shows how the likelihood of conflict onset within five years changes with support, with all other variables being held at their mean value.

    Figure 3 demonstrates that countries with very low percentages of societal support are much more likely to be the scene of conflict. A country with minimal support has a seven percent likelihood of experiencing conflict onset within a five year period. Increasing support to just 60 percent reduces the probability of civil conflict onset to less than two percent. Governments with support of over 80 percent almost never experience conflict.

    Next, I simulate the likelihood of conflict onset within the next year, with all other factors held at their mean value. This test is shown in Figure 4.

    Figure 3 and Figure 4 show similar trends. Note that the y-axis (probability of conflict onset) of Figure 4 is scaled down from that of Figure 3. This makes intuitive sense; the probability of conflict onset within a single year should be lower than the probability of conflict onset within a five year period. Figure 4 shows that countries with minimal amounts of support have just below a one percent likelihood of experiencing conflict within a year. Increasing support to 60 percent reduces the likelihood of civil conflict onset to below half a percentage point.


    The results of this analysis suggest a number of future paths for policy decisions and research. Most importantly, these results demonstrate on a theoretical level that societal support matters. Understanding how the populace behaves can be a valuable explanatory tool in understanding civil conflict onset. Academic focus in this area has traditionally been centered around policies, dollars, protests, and other units of analysis that can be easily and directly measured. I argue that public opinion is an important angle to conflict onset that has traditionally been neglected, largely because data is seldom readily available.

    The obvious solution to the neglect of public opinion in conflict literature is better data collection. In many cases, surveys do not consistently ask the same questions over multiple years. Survey data is also difficult and expensive to collect in countries where it is most relevant; in particular, this study is missing data from most countries in Africa. Increasing data collection would allow for more extensive, focused, and statistically sound analysis of public opinion in relation to conflict.

    A second solution to the neglect of public opinion in conflict literature is shifting to a model of dissent and conflict that focuses on the populace. There is no shortage of studies that focus on country-level traits that lead to conflict; there is also a large body of literature that concentrates on the decision-making processes of state and rebel actors. In order to more fully understand civil conflict, the role of the public must also be understood. What causes societal support for a regime to change? In what circumstances is support most likely to matter? How can support be restored after conflict? These are all questions for future research.

    The impact of societal support is not limited to academia; it also carries important policy implications. Examining the level of support for a regime, in combination with other factors, can be used to predict future conflict. It also stands to reason that increasing support can help prevent conflict. This is critical information for someone wishing to maintain or depose a regime.

    It is my hope that this finding encourages future research into societal support, in addition to general trends in public opinion and citizen behavior. State and rebel behavior are relatively well understood, and a better understanding of societal behavior is an important piece of the puzzle.

    Appendix A

    World Values Survey Question E069_11

    Confidence: The Government

    Question text: I am going to name a number of organizations. For each one, could you tell me how much confidence you have in them: is it a great deal of confidence, quite a lot of confidence, not very much confidence or none at all?

    The government

    • 5 Missing; Unknown
    • 4 Not asked in survey
    • 3 Not applicable
    • 2 No answer
    • 1 Don’t know
    • 1 A great deal
    • 2 Quite a lot
    • 3 Not very much
    • 4 None at all

    (European and World Values Surveys Four Wave Integrated Data File, 1981-2004 2006)


    • Carey, Sabine C. 2006. “The Dynamic Relationship between Protest and Repression.” Political Research Quarterly 59:1–11.
    • Carey, Sabine C. 2010. “The Use of Repression as a Response to Domestic Dissent.” Political Studies 58(1):167–186.
    • Collier, Paul and Anke Hoeffler. 2004. “Greed and Grievance in Civil War.” Oxford Economic Papers 56:563–595.
    • Davenport, Christian. 1999. “Human Rights and the Democratic Proposition.” Journal of Conflict Resolution 43(1):92–116.
    • European and World Values Surveys Four Wave Integrated Data File, 1981-2004. 2006. The European Values Study Foundation and World Values Survey Association. URL: www.worldvaluessurvey.org
    • Gartner, Sigmund Scott and Patrick M. Regan. 1996. “Threat and Repression: The Non- Linear Relationship between Government and Opposition Violence.” Journal of Peace Research 33:273–287.
    • Gurr, Ted Robert. 1970. Why Men Rebel. Princeton, NJ: Princeton University Press.
    • King, Gary, Michael Tomz, and Jason Wittenberg. 2000. “Making the Most of Statistical Analyses: Improving Interpretation and Presentation.” American Journal of Political Science 44(2):341 355.
    • Kugler, J. and W. Domke. 1986. “Comparing the Strength of Nations.” Comparative Political Studies 19:39–69.
    • Marshall, Monty G. and Ted Robert Gurr. 2013. “Polity IV Project: Political Regime Characteristics and Transitions, 1800–2012.” URL: http://www.systemicpeace.org/polity/polity4.htm
    • Moore, Will H. 2000. “The Repression of Dissent: A Substitution Model of Government Coercion.” Journal of Conflict Resolution 44(1):107–127.
    • Poe, Steven C. and C. Neal Tate. 1994. “Repression of Human Rights to Personal Integrity in the 1980s: A Global Analysis.” American Political Science Review 88:853–872.
    • Poe, Steven C., C. Neal Tate and Linda Camp Keith. 1999. “Repression of the Human Right to Personal Integrity Revisited: A Global Cross-National Study Covering the Years 1976-1993.” International Studies Quarterly 43:291–313.
    • Rasler, Karen. 1996. “Concessions, Repression, and Political Protest in the Iranian Revolution.” American Sociological Review 61:132–152.
    • Sambanis, N. 2002. “A Review of Recent Advances and Future Directions in the Quantitative Literature on Civil War.” Defence and Peace Economics 13(3):215–243.
    • Shellman, Stephen M. 2006. “Process Matters: Conflict and Cooperation in Sequential Government-Dissident Interactions.” Security Studies 15(4):563–599.
    • Themnèr, Lotta and Peter Wallensteen. 2012. “Armed Conflict 1946-2011.” Journal of Peace Research 49.
    • Valentino, Benjamin, Paul Huth, and Dylan Balch-Lindsay. 2004. “Draining the Sea: Mass Killing and Guerilla Warfare.” International Organization 58:375–407.
    • Walter, Barbara F. 2004. “Does Conflict Beget Conflict? Explaining Recurring Civil War.” Journal of Peace Research 41:371–388.
    • Wood, Reed M. and Mark P. Gibney. 2010. “The Political Terror Scale (PTS): A Reintroduction and Comparison to the CIRI Human Rights Database.” Human Rights Quarterly 32(2):367–400.
    • World Development Indicators. 2013. The World Bank, Washington, DC: World Bank Group. URL: http://data.worldbank.org/data-catalog/world-development-indicators
    • Young, J. K. 2012. “Repression, Dissent, and the Onset of Civil War.” Political Research Quarterly 66(3).

    Table 1: Distribution of Independent Variables

    Table 2: Conflict Within 5 Years

    Table 3: Conflict Within 3 Years

    Table 4: Conflict in the Next Year

    Figure 1: Cost of Paying Taxes Illustration

    Figure 2: Geographic Distribution of Data

    Figure 3: Probability of Conflict Onset Within Five Years

    Figure 4: Probability of Conflict Onset in the Next Year