The hunger games: Food prices,ethnic cleavages and nonviolent unrest in Africa

Abstract

Nonviolent movements are more successful when mobilizing large and diverse numbers of participants. However, while there has been considerable research on the outcomes of nonviolent campaigns, far less is known about the initial emergence of nonviolent action. A growing literature suggests ethnic divisions may undermine the ability of activists to engage in mass nonviolent mobilization across diverse social lines. Yet many large and diverse nonviolent movements have successfully emerged in various ethnically divided societies across the world. I argue that nonviolent mobilization is made possible in ethnically polarized contexts when broader cross-cutting grievances are present as they enable local activists to widen their appeal across social lines. I focus on food price spikes as an example of a cross-cutting issue that is likely to affect consumers from different ethnic groups. The unique and symbolic nature of food price spikes facilitates nonviolent mobilization across ethnic lines and provides clear short-term incentives for many people to participate in protests against the government. Using new spatially disaggregated data on government targeted nonviolent action, I analyse grid-cell years across 41 African countries (1990–2008). I find strong evidence that food price spikes increase the likelihood of nonviolent action in politically excluded and ethnically diverse locations.

Keywords

ethnicity food insecurity food prices nonviolent action spatial analysis

Introduction

In recent years, mass nonviolent resistance has become an increasingly prevalent form of anti-government dissent. Movements have been remarkably successful in achieving political change using unconventional nonviolent action (Chenoweth & Stephan, 2011), defined as a combination of persuasive tactics (i.e. protests and demonstrations) and non-compliant methods (i.e. strikes and sit-ins) (Sharp, 2005).¹ The growing civil resistance literature relates this success to a movement’s ability to mobilize large and diverse numbers of participants across social lines, which builds legitimacy and draws support away from the government (DeNardo, 1985; Schock, 2005; Sharp, 2005; Svensson & Lindgren, 2011; Chenoweth & Stephan, 2011).

Yet, while there has been considerable research on the outcomes of nonviolent campaigns, far less is known about the initial emergence of nonviolent action (Chenoweth & Ulfelder, 2017), in particular, how movements succeed in mobilizing large numbers of people across diverse support bases. There are many prominent examples where activists have engaged in mass nonviolent action after unifying otherwise disparate social groups, including the ‘Arab Spring’ and movements that emerged in countries with a history of ethnic conflict (e.g. Burundi and Bosnia). Yet, a number of recent studies have shown that ethnic cleavages undermine nonviolent mobilization, by reducing the ability of activists to attract mass numbers of people from different social groups (Svensson & Lindgren, 2011; Arriola, 2013; Vidovic & Gleditsch, 2015; Thurber, 2018).² How do nonviolent movements mobilize sufficient numbers in societies with ethnic divisions?

This article seeks to unravel this question and extend our understanding of emergence by highlighting a process through which nonviolent activists overcome political exclusion and local ethnic divisions to engage in mass and diverse mobilization. I argue that cross-cutting grievances, which transcend divisions within and between politically relevant ethnic groups, provide opportunities for activists to appeal to individuals across social boundaries, thereby greatly facilitating nonviolent mobilization in socially divided contexts. I focus on one example of a cross-cutting issue, spikes in domestic food prices, which are likely to impact consumers from all social groups, enabling a movement to mobilize individuals based on a common economic grievance.

To test these claims, I disaggregate the emergence of nonviolent action to the subnational level, using new geocoded events data. This approach advances existing quantitative research on emergence using country-level variables that do not reflect subnational realities (Chenoweth & Ulfeder, 2017; Butcher & Svensson, 2016). Ethnic cleavages are a social barrier faced by activists locally and vary considerably within countries along with other structural factors that influence the viability of nonviolent mobilization. I explore the emergence of nonviolent action across subnational grid-cells of 41 peacetime African countries (1990–2008). The findings provide strong evidence that the cross-cutting nature of food price spikes increases the feasibility of nonviolent mobilization in ethno-politically excluded and diverse areas that would otherwise be unlikely to observe nonviolent action.

The article begins by exploring existing research and ways ethnic cleavages may constrain nonviolent action, before theorizing food prices as a cross-cutting issue that facilitates mass and diverse mobilization. This is followed by the empirical analysis and concluding remarks.

Ethnic divisions and nonviolent mobilization

The strategic logic of mass nonviolent action is to generate enough leverage, through mobilizing greater numbers, in order to disrupt the state’s ability to rule or impose particular policies (DeNardo, 1985; Schock, 2005; Sharp, 2005; Chenoweth & Stephan, 2011; Svensson & Lindgren, 2011). Few governments confronting nonviolent action from 5% of the population have avoided political change (Lichbach, 1998).

Studies of civil resistance tend to explain the emergence of mass nonviolent action through forms of political agency, such as the skills and leadership of activists (Sharp, 2005; Schock, 2005). Yet, mass mobilization is also likely to depend on existing structural contexts that remain outside the control of political activists (Goldstone, 1994; Chenoweth & Ulfelder, 2017; Butcher & Svensson, 2016). Nepstad (2015), for example, points to three broad determinants of nonviolent action: widely held grievances against the government, intergroup coalitions, and space to organize. While the latter has synergies to opportunity factors and resources that provide ‘space’ for mobilization, the former largely correspond to two interconnected forms of mass mobilization unique to nonviolent action: the vertical mobilization of large numbers against the regime (i.e. building on widespread anti-government grievances) and horizontal mobilization that occurs across diverse social groups (coalition building).

Nonviolent movements seek to challenge the regime vertically by mobilizing large numbers of people in order to achieve political change. Leverage is achieved by drawing on widely held anti-government grievances and by dislocating the regime from its so-called ‘pillars of support’ (e.g. police, military, workers, civil servants, business, political parties), which if removed, limit the regime’s ability to rule (Sharp, 2005). Nonviolent activists target these pillars of government support by appealing to common ground and by encouraging loyalty switches within broad segments of the population, including pro-regime supporters and security forces that carry out state repression (Chenoweth & Stephan, 2011; Nepstad, 2015).

Nonviolent mobilization also occurs horizontally across social lines, as winning popular support is often dependent on the movement’s ability to appeal to various social groups. Mobilizing individuals from different social backgrounds, occupations and political ideologies, which have differing relationships with the state, provides a movement with greater leverage (Svensson & Lindgren, 2011; Butcher & Svensson, 2016). Yet, while nonviolent mobilization has fewer moral and physical barriers to participation (Chenoweth & Stephan, 2011), mobilizing thousands of people across social lines is extremely challenging, particularly in polarized societies with high social distance within and between ethnic groups (Svensson & Lindgren, 2011; Arriola, 2013; Bhavnani & Jha, 2014; Vidovic & Gleditsch, 2015; Thurber, 2018). Existent research remains poorly placed to explain the emergence of nonviolent action in divided societies, in which ethnic cleavages are rarely featured in existing explanations (see Thurber, 2018).

A large literature on grievances instead argues that collective participation can be achieved by solely appealing to ethnic grievances that derive from ethno-political inequalities (Stewart, 2008; Wimmer, 2013), with empirical evidence suggesting that ethnic exclusion motivates engagement in nonviolent action (Gurr, 1993; Olzak, 2006; Jazayeri, 2016), as well as armed rebellion (Stewart, 2008; Østby, 2008; Cederman, Gleditsch & Buhaug, 2013). By focusing on ethnic mobilization, the grievance approach has greatly extended our understanding of why excluded groups might be motivated to mobilize against the government (vertical mobilization), but has largely ignored the horizontal dimension of nonviolent mobilization and how movements mobilize across group lines. This relates, in part, to the wider research agenda that has emerged in isolation to civil resistance literature and has almost exclusively focused on civil war. This has led many grievance scholars to draw on civil war mechanisms that are poorly suited to explain mass nonviolent mobilization, since nonviolent and violent mobilization are distinct (Chenoweth & Lewis, 2013). For instance, while armed opposition only needs to recruit a few hundred fighters and may be aided by ethnic-based recruitment, nonviolent resistance requires tens of thousands of participants, meaning horizontal coalitions across group lines are often necessary to achieve the greater scale of mobilization.

More recent research suggests that nonviolent movements can bypass social barriers to horizontal mobilization when they have the support of a larger politically excluded ethnic group with a bigger recruitment pool (White et al., 2015; Dahl et al., 2016). Yet, this highlights two key theoretical limitations within existing grievance research.

First, even with the support of a large ethnic base, interethnic support is often essential because some ethnic groups are more tied to the state than others and are clustered in certain occupations, sectors and settlement types (DeNardo, 1985; Lichbach, 1998; Schock, 2005; Butcher & Svensson, 2016). This is illustrated by nonviolent struggle against apartheid in South Africa, where the non-racial United Democratic Front also appealed to Asians, Coloureds and Whites even though Blacks made up 85% of the population. Moreover, ethnic movements can alienate potential support from other ethnic groups and can be more easily isolated and repressed (Chenoweth & Stephan, 2011; Svensson & Lindgren 2011; Butcher & Svensson, 2016). Excluded groups are often less able to participate or engage in nonviolence campaigns due to ethnic divisions, meaning only 13% of nonviolent resistance campaigns are mobilized in pursuit of ethnic goals (Thurber, 2018). By focusing on ethnic mobilization, existing grievance literature misses the vast majority of cases where nonviolent mobilization involves multiple groups.

Second, while some movements are able to mobilize along ethnic lines, they are still likely to face internal divisions. The grievance literature often assumes ethnic preferences are monolithic, ignoring challenges to horizontal mobilization deriving from differing interests and divisions within ethnic groups (e.g. class and language). For example, although the Oromo represent the largest ethnic group in Ethiopia, intra-ethnic differences have limited the group’s ability to mobilize large numbers of co-ethnics and engage in nonviolent action (Arriola, 2013). To be successful nonviolent activists must overcome intra-ethnic divisions as well as cleavages between ethnic groups (Gorenburg, 2000; Goldstone, 2011). How a movement may unify subgroups as well as disparate ethnic groups remains poorly understood.

Other structural literatures on revolution and social movements provide a better indication of how movements may unify social groups to engage in horizontal mobilization, pointing to the importance of resources and political opportunities. Social networks and pre-existing organizations (e.g. trade unions) often provide ‘free space’ for civilians to organize (Ackermann & DuVall, 2000; Nepstad, 2015) and can facilitate meso-mobilization – temporary coalitions that underpin nonviolent movements (Gerhards & Rucht, 1992). Mass mobilization is also assisted by information and economic resources, most often found in urbanized and industrialized countries (Goldstone, 1991; Siegal, 2009; Butcher & Svensson, 2016), and political opportunities which signal that nonviolent action is feasible, for example, when the regime is unstable or increases its political openness (Tilly, 1978; McAdam, 1999; Meyer, 2004).

However, studies of revolution and social movements often disregard the importance of ethnic cleavages as a social barrier to nonviolent mobilization. While important, resources are likely to be distributed along social lines rather than across them and activists may be unable to act on political opportunities if ethnic and intra-ethnic cleavages divide the wider population that activists seek to mobilize.

Local structures and the emergence of nonviolent action

The preceding discussion provides the expectation that ethnic and intra-ethnic cleavages are an important social structure that undermines the feasibility of mobilizing large numbers of people (vertical mobilization) by restricting mobilization across social lines (horizontal mobilization). Grievance literature has shown that ethno-political exclusion produces particularly salient animosities between groups (Stewart, 2008; Wimmer, 2013) that have been found to have a particularly negative impact on the emergence of nonviolent action. Political exclusion diminishes the likelihood of broader coalitions, social diversity and loyalty switches (Svensson & Lindgren, 2011), and reduces the likelihood of having coethnics within key pillars of power, thereby decreasing the likelihood that excluded groups will participate in protest (Thurber, 2018).

Yet, nonviolent mobilization and activism are often carried out locally, with activists drawing on communal and informal networks that loosely bind would-be participants (Goldstone, 1994; Kuran, 1991). While we know ethno-political exclusion undermines nonviolent action at the national, group and movement levels (Svensson & Lindgren, 2011; Thurber, 2018), this should also be evident at the subnational level by hindering localized attempts to build coalitions.

First, subgroups within excluded ethnic groups often have differing goals against the government. Without superordinate goals that cut across social lines, it is difficult for subgroups to cooperate, find common ground and build coalitions, even when in close proximity (Tajfel, 1982; Goldstone, 1994). As seen in Oromia (Ethiopia), the inability to coordinate strategy, broaden alliances and engage in horizontal mobilization is more closely connected to ethnic in-fighting and a switch to violent strategies, rather than engagement in nonviolent action (Arriola, 2013; Cunningham, 2015). The preceding discussion provides the expectation that ethnic exclusion undermines local ethnic-based nonviolent mobilization.

H1a: Mass nonviolent action is less likely to occur in locations with politically excluded ethnic groups.

Second, broader ethnic identities often form the basis of membership in social networks in politically exclusive societies (Wimmer, 2013), increasing social distance between groups that live in close proximity. Political inequalities, in particular, increase the salience of divisions between ethnic groups, thereby reducing the viability of horizontal mobilization across ethnic lines and undermining the scope of mobilization (Thurber, 2018). In societies where high levels of political inequalities exist between groups, boundaries tend be salient and ethnic networks are more socially ‘closed’ (Gurr, 1993; Stewart, 2008; Wimmer, 2013). This limits intergroup interaction in everyday life and precludes the formation of intergroup ties – a distinct obstacle to horizontal mobilization across ethnic lines.

H1b: Mass nonviolent action is less likely to occur in diverse locations where politically excluded ethnic groups coexist with other groups.

With existing research providing little indication of how movements overcome social barriers, the next section highlights the importance of cross-cutting issues as a means to mobilize disparate groups and engage in both vertical and horizontal nonviolent mobilization.

Food price spikes and overcoming ethnic obstacles to nonviolent mobilization

Building on the preceding discussion, the main argument of this article is that the emergence of nonviolent action within ethno-exclusive environments is greatly facilitated by the existence of broader cross-cutting grievances that transcend divisions across and within ethnic groups. Sharp increases in domestic food prices, or ‘spikes’,³ are a good example of a cross-cutting issue; when large enough, they generate widespread economic grievances that impact consumers from all social groups. Regardless of social affiliations, access to affordable food is a basic need for all consumers (Barrett, 2013).

The cross-cutting nature of increases in food prices sets this apart from other types of issues. For example, while elections often spark unrest against the government (vertical mobilization), many major political parties in Africa gain the majority of their support from one or a few ethnic groups (Kitschelt & Wilkinson, 2007), reducing the likelihood that protests will attract a broad spectrum of participants across group boundaries (horizontal mobilization). Food price spikes also differ from other economic shocks: climatic disasters (see Buhaug, 2015), negative economic growth (see Blattman & Miguel, 2010), downturns in food production (Wischnath & Buhaug, 2014), or shock declines in food prices (Dube & Vargas, 2013). These types of shocks typically impact certain ethnic groups residing in specific, often rural, parts of a country, and have been related to armed conflict and communal violence as they reinforce ethnic-based grievances rather than cut across them.

In contrast, spikes in food prices provide an opportunity for movements to broaden their appeal and engage in horizontal mobilization across intra-ethnic and interethnic divides because they are what Tilly (1978) calls a ‘symbolic issue’. In Bosnia, protests in 2014 over high prices and low wages were framed around the powerful message – ‘We are hungry in three languages’ (Hopkins, 2014). In Egypt, the famous ‘bread helmet’ was a symbol that transcended ethno-religious lines, and was easily replicated by various segments of the population.

I argue that the unique and symbolic nature of food price spikes provides two simultaneous forms of motivation that facilitate the emergence of nonviolent action: common intergroup grievances, which ease horizontal mobilization in ethno-politically exclusive environments, and clear short-term incentives to join in anti-government protests, which fosters vertical mobilization.

First, the greater the magnitude of price spikes, the more likely a higher number of social groups, across and within ethnic groups, will be impacted. Higher prices make economic hardship more ‘visible’ and therefore display the hardship of others (Weinberg & Bakker, 2014), increasing the perception that large segments of the population share similar grievances and have a common interest in addressing them. Between 2007 and 2008, the Egyptian opposition was able to mobilize 400,000 people, using record high food price spikes to unify individuals from different classes across Arab and Coptic Christian communities (Bush, 2010). Higher spikes encourage the participation of affluent segments and groups that are closer to political power (Barrett, 2013), including professionals and intellectuals, who tend to purchase more of their food and often participate due to a higher sense of entitlement (Tilly, 1978). This, in turn, is likely to encourage the participation of poorer and often politically marginalized consumers, who are the most vulnerable to price spikes (Ivanic, Martin & Zaman, 2012), but are less likely to have the coordination required to engage in nonviolent action (Gamson, 1990; Tilly, 1978). In other words, food price spikes generate superordinate goals that unify normally divided groups.

Second, high food price spikes create short-term incentives to participate in nonviolent action because of their sudden economic impact. Households generally have very little warning of the impending economic hardships due to price rises, placing significant pressure on coping mechanisms. Attempts to adapt are often constrained by static wages (Wodon & Zaman, 2010) and poor social protection; African governments, on average, spend the least on social protection programs relative to their GDP (Ortiz et al., 2014). Faced with significant economic pressure and few avenues to redress economic hardships, individuals have extraordinary short-term incentives to participate in nonviolent action in order to force immediate concessions from the government.

Food price spikes provide movements with a unique opportunity to overcome ethnic obstacles to horizontal mobilization by conflating non-ethnic food-related grievances with anti-government and often ethnic-based sentiments, which facilitates vertical mobilization against the government which is solely to blame for widespread economic hardship (Barrett, 2013; Smith, 2014; Weinberg & Bakker, 2014). A movement can link the inability of the government to deal with food prices with its broader ineffectiveness and misuse of political power, and offer a viable alternative vision that advances its opposition (Chenoweth & Stephan, 2011), thereby drawing intergroup support away from the government. This is illustrated in Uganda, where the opposition remains severely weakened by the legacy of a 19-year ban on opposition parties. The main opposition leader, Kizza Besigye, has increasingly used food price spikes to appeal across group lines and unite the fragmented opposition in demonstrations against the government by conflating economic hardships with issues of government corruption and misuse of power (Kron, 2011). In 2008, food price spikes enabled movements in a number of ethno-politically exclusive countries to organize nonviolent action by politicizing various issues linked to increasing food prices, including high wages, employment, lower tax levels, social protection, subsidies and corruption (Bush, 2010; Smith, 2014). Based on the preceding discussion, I offer the final hypothesis:

H2: In ethnically excluded and diverse locations, spikes in domestic food prices increase the likelihood of mass nonviolent action.

Research design

To test these hypotheses, I undertake a subnational analysis of all African countries between 1990 and 2008. The unit of analysis is grid-cell-years, based on the PRIO-GRID data-structure which consists of 0.5 X 0.5 degree geographical grid-cells (approximately 55 X 55 kilometres) (Tollefsen, Strand & Buhaug, 2012). This

Table I.

Number of geocoded estimates across events and PRIO-GRID locations

	Total events	National estimates	Regional estimates	Unknown estimates	Other estimates	Total estimates
No. SCAD events	5,823	1,637	145	298	67*	2,080 (35.7%)
No. grids with SCAD events	1,494	802	89	180	56*	992 (56.2%)**

* The ‘other’ category (rural and dockland areas) overlaps with national, regional and unknown estimates.

**153 estimated locations overlap with geocoded SCAD events, meaning 839 locations are uniquely estimated.

approach advances existing studies on the emergence of nonviolent action that rely on country-level data, which assumes conditions are consistent across a given country and conceals a great deal of subnational variation. Not all parts of a country experience nonviolent action. Different locations have disparate types of ethnic groups, and varying levels of diversity, population and wealth that either facilitate or undermine local nonviolent action. This can only be captured by the use of subnational data, which enables the exploration of spatial variations in the occurrence of nonviolent action across distinct local ethno-political contexts. Moreover, compared to administrative units, grid-cells do not vary in size or change over time and therefore are consistently comparable across countries (Tollefsen, Strand & Buhaug, 2012).

Dependent variable

To capture mass nonviolent action against the government at the grid level, I use georeferenced events data from the Social Conflict Analysis Dataset (SCAD) (Salehyan et al., 2012). SCAD events are based on local news reports from the Associated Press and Agence France Presse newswires.⁴ SCAD provides the most comprehensive collection of georeferenced social conflict events in Africa, offering detailed information on dates, coordinates, numbers of participants and descriptions of the incidents and actors involved.⁵ Using this information, I include SCAD events coded as: (1) targeting the national government, (2) involving nonviolent methods of noncompliance or persuasion, that is peaceful protests, demonstrations and strikes (Sharp, 2005), (3) organized, not sporadic and (4) involving a minimum of 1,000 participants. The last criterion follows the widely used Nonviolent and Violence Campaigns and Outcomes (NAVCO 2.0) dataset (Chenoweth & Lewis, 2013), capturing mass events that were mobilized despite potential barriers to nonviolent mobilization. Since nonviolent action does not use physical violence to resist the government (Butcher & Svensson, 2016), this also excludes violent unrest (i.e. communal violence and armed conflict) and incidental violence (i.e. riots).⁶ While it is difficult to ascertain whether these nonviolent events are mobilized along ethnic or interethnic lines, the cross-cutting mechanism applies for ethnic movements, attempting to unify intra-ethnic groups (Hypothesis 1a), and movements mobilizing across ethnic lines (Hypothesis 1b).

The SCAD data have two limitations. First, in some events the number of participants is coded as unknown. Where participation numbers are unknown, I only include events which have clear evidence of involving at least 1,000 persons, using secondary information on the event, its geographical scope and the size of the organizations involved. For example, many strikes far surpass this threshold (i.e. the 1994 general strike in Burundi against military rule).

Second, many widespread events are often not accurately georeferenced. For example, events coded as nationwide are assigned to the geographical centre of the country, while regional events are georeferenced in the centre of the region. These events cannot be excluded as they represent one-third of all events (see Table I) and represent important examples of mass nonviolent action, including, for example, numerous anti-apartheid events in South Africa.⁷ Nationwide events also occur exclusively in 91 out of the 340 country-years and removing these events would wrongly code these years as a ‘false 0s’.

To avoid dropping these cases, I generate different specifications to approximate the location of the non-geocoded events to closely resemble the types of events that are geocoded, in which 93% occur in cities (with a population over 100,000).⁸ Approximations are also based on SCAD’s description of each event and its actors; in this way, I can ensure the geocoding reflects spatial realities. For example, tea-producer strikes in Kenya are coded as occurring in tea-producing areas.

For the main analysis, I use an estimated dependent variable (DV) that includes all geocoded events and estimates non-geocoded events as occurring in the capital and the country’s (or region’s) five largest cities (Top5_est). This is a conservative assumption since most nationwide campaigns occur in multiple urban centres (Chenoweth & Stephan, 2011), as illustrated by the January 2018 Iranian protests that occurred in at least 19 cities amid food price spikes. I generate other DVs that estimate these events as occurring more widely across a country as a robustness check. Using the most modest estimated DV (Top5_est) in the main analysis balances potential bias induced by excluding non-geocoded events with bias from including too many estimated locations (see Table I). Each event is aggregated to the relevant PRIO-GRID year. Grid-years with at least one nonviolent event are coded as 1 (otherwise 0).

Independent variables

To capture domestic food price spikes, I take the highest percentage change in domestic food indices from one month to the next and aggregate this to the year.⁹ Following Smith (2014), I use domestic food price data from indices collected by the International Labour Organization (ILO) (2014). Most peacetime African countries have monthly indices available from 1990 to 2008,¹⁰ based on the price of a basket of the most important foods in the capital or major city.¹¹ This is a continuous variable with price spikes ranging from minimal increases to very high changes that I expect are more likely to increase the risk of nonviolent action.¹² Using the highest percentage changes in food indices captures the short-term nature of spikes and provides a comparable indicator across countries.

While these national-level data do not capture potential subnational variations in prices, I remain confident the ILO data are appropriate for two reasons. First, current disaggregated data remain limited. For example, recent data from the World Food Programme (WFP) only comprehensively cover some African countries and are largely confined to recent years. Second, as Figure 1 shows, trends in the WFP data suggest that price spikes vary little across different urban areas, where the vast majority of nonviolent action occurs. Figure 1 displays two examples of Kenya and Malawi, comparing the monthly price changes of urban and rural markets (each line is a local market), for the year 2011 when prices achieved record highs and good data are available. This figure shows that price trends in urban markets tend to move in the same direction in Kenya (spike in July) and Malawi (dip around May and June), while much more price variation occurs across rural markets.

I focus on domestic rather than international food prices, as they are less distorted by state market interventions (Ivanic, Martin & Zaman, 2012; Hendrix & Brinkman, 2013; Smith, 2014) and more closely reflect the price that people pay, with 90% of all food consumed in sub-Saharan Africa being produced domestically (Barrett, 2013). I exclude country-years in which civil war is ongoing, because of its clear endogenous relationship with food prices (Gates et al., 2012).

To assess the impact that local ethnic exclusion and diversity have on nonviolent action, I create two variables at the grid level. First, grid-cells occupied by ethno-politically excluded groups, to capture ethnic exclusion and proxy potential intragroup ethnic differences. Second, locations with excluded groups that coexist with other ethnic groups, to measure diversity where ethnic groups co-exist. These areas should normally constrain nonviolent resistance when food prices spikes are small. I use group-level data from the Ethnic Power Relations (EPR) dataset to gain information on the level of group

Figure 1.

Food price variation across urban and rural areas of Kenya and Malawi (2011)

representation in government, coding groups as included or excluded from governmental power (EPR data, Version 3.0; Wimmer, Cederman & Min, 2009). The EPR dataset only includes politically relevant ethnic groups represented by at least one political actor in the national political arena, which ensures there are political actors in place to potentially engage in unrest.

These variables are translated to the grid level using georeferenced ethnic settlement patterns from the Geo-EPR dataset (Wucherpfennig et al., 2011). Grids containing no politically relevant groups are used as the reference category. These areas are either inhabited by no ethnic groups or by less politically relevant ethnic groups. These locations do not exhibit the social barriers associated with politically relevant groups (Posner, 2004) and therefore make a good comparison with more socially constrained local ethnic contexts. I also exclude grids with a population less than the requirement for mass civil resistance (<1,000). I include two variables to capture politically relevant ethnic groups: the locations of politically relevant excluded groups, considered oppositional, and areas with included groups, treated as pro-government – coded as 1, otherwise 0. Using Kenya in 2000 as an illustrative example, Figure 2 visualizes the data. In 2000, three Kenyan ethnic groups were excluded from the executive: the Kikuyu, Luo and Somali (highlighted). The settlement areas of these groups are proxies for opposition areas. The other included groups (shaded and not highlighted) represent pro-government areas.¹³ Politically excluded grids are then interacted with food prices to test the hypothesis that food prices increase the likelihood of nonviolent action in excluded areas.

To measure more challenging diverse areas, I generate a dummy variable for locations where excluded group

Figure 2.

Included and excluded ethnic group areas (Geo-EPR data) over the PRIO-Grid

settlement areas overlap with areas occupied by at least one other ethnic group. To test the hypothesis that high food price spikes increase the likelihood of nonviolent action in excluded and diverse areas, I generate an additional interaction term – Diverse excluded areas X Food price spikes.¹⁴

Control variables

At the grid level I control for poverty, using the grid-level equivalent of national GDP (Nordhaus, 2006), as this reduces adaptability to price spikes and reduces the cost of participation (Barrett, 2013). I include the grid population (logged) (CIESIN, 2005), since larger and more concentrated populations facilitate the coordination of nonviolent mobilization (Butcher & Svensson, 2016; Barrett, 2013). Also included is the size of the largest excluded group within a grid (% of population), as larger groups have a mobilization advantage (Dahl et al., 2016). As nonviolent action is largely an urban phenomenon, I also control for travel time (minutes) to the nearest urban centre and distance to the capital (kilometres) (Tollefsen, Strand & Buhaug, 2012). To account for spatial and temporal dependencies, I include a temporal lag for nonviolent action occurring in the previous year (t–1) and a spatial lag for nonviolence in neighbouring grids.

National-level controls are also included. These include the number of excluded groups in a country, as governments are less likely to compromise when facing multiple groups (Walter, 2009), a logged version of national population to proxy for the size of the country (World Bank, 2013), the Polity2 measure (Marshall & Jaggers, 2012) to control for regime types and the number of peace years to account for instability. Finally, I include a dummy variable for national election years which often spark unrest (Lindberg, 2009).¹⁵

Method

I run country-fixed-effects logistic regression models to restrict the analysis to the within-country variance. This allows the models to control for unobserved differences between countries and for certain country characteristics that influence food prices and nonviolent action, including environmental vulnerabilities, transport networks, food infrastructure and trade policies, such as tariffs, food assistant programs and subsidies (Smith, 2014). To aid post-estimation, all independent variables are standardized, so each model reports the effect of a one-standard-deviation increase in each variable.¹⁶

Results

In this section, I analyse my hypotheses; local ethno-political exclusion and diversity reduce the likelihood of nonviolent action (H1a, H1b) and higher food price spikes increase the feasibility of nonviolent action in these excluded and diverse areas (H2). All models displayed in Table II explore these propositions using the primary DV, which estimates SCAD’s non-geocoded events as occurring in a country’s five largest cities (Top5_est).

Model 1 (Table II) explores the baseline effect of local ethnic exclusion and diversity on the likelihood of nonviolent action. Model 1 reports no statistically significant effect of ethnic exclusion on the emergence of nonviolent action. Although it is extremely difficult to separate out ethnic and interethnic nonviolent events, this could relate to simultaneous effects, whereby ethnic exclusion generates ethnic protest in some areas, such as the Berber region of Algeria (Gurr, 1993; Jazayeri, 2016), but undermines more intergroup nonviolent activism in other areas, where the EPR variables may not pick up diversity, that is, cities within ethnic homelands (Thurber, 2018). In contrast, Models 1 and 4 show that nonviolent action is less likely to occur in ethnically excluded areas that are diverse (have at least one more EPR group). This provides strong evidence for Hypothesis 1b, confirming that diverse environments present an obstacle to nonviolent horizontal mobilization and reduce the feasibility of nonviolent action (p < 0.01 and p < 0.05).

I proceed to explore the mediating impact of food price spikes on ethnic barriers to nonviolent resistance. Model 2 explores the general impact of food prices on the incidence of nonviolent action. The coefficient is positive and highly significant, providing strong evidence that greater rises in food prices increase the likelihood of nonviolent action in diverse locations within states (p < 0.001). When increasing the percentage increase of food prices by one standard deviation (moving food prices from a 5.2% to a 10.5% increase), the risk of nonviolent action rises by 12.7%.

Model 3 then explores the likelihood of nonviolent action in locations occupied by excluded ethnic groups. The model shows a positive effect for this interaction, suggesting that higher food price spikes increase the likelihood of nonviolent action in areas home to these groups, even when controlling for other structural factors. In addition, ethnically excluded areas, a non-finding in baseline models, becomes significant and negative, suggesting that these areas are likely to constrain a movement’s efforts to engage in nonviolent action during times of small or no price spikes (p < 0.05). While we cannot be sure who is participating in nonviolent action in ethno-excluded areas, regardless of whether movements are attempting to mobilize subgroups within the same ethnicity, or across ethnic divides, food price spikes have a general impact on nonviolent action in excluded areas. In these areas, Model 3 reports an increase of 20.2% in the likelihood of nonviolent action with a one-standard-deviation increase in food prices (from 14.7% to 19.1%).

Lastly, Model 4 explores the initiation of protest within a more challenging and divided diverse local environment. The interaction effect of food prices and diverse locations is positive (p < 0.01), suggesting that while diversity undermines nonviolent action, food price spikes mediate this effect and increase the likelihood of nonviolent action in these areas. When food prices increase by one standard deviation (from 8.8% to 12.7%), the likelihood of nonviolent action in diverse locations increases by 15%. Overall, Table I supports my second hypothesis and suggests that food price spikes increase the feasibility of mass and diverse nonviolent mobilization, even in the most difficult multi-ethnic environments.

The control variables suggest there are other important drivers of nonviolent action. Grids that are more wealthy, have larger populations and are closest to urban centres are more likely to witness nonviolent resistance. This gives support to claims that urban-based resources and related networks facilitate nonviolent action (Butcher & Svensson, 2016; Ackermann & DuVall, 2000; Nepstad, 2015). Several control variables – distance to the capital, number of peace years, regime type and the number of excluded groups residing within a state – have no significant effects on nonviolence. Furthermore, while other studies argue that larger ethnic groups have a greater potential for mobilization (Dahl et al., 2016), I find little evidence of this at the subnational level. While large groups facilitate ethnic-based protest, this is unlikely to facilitate mobilization across ethnic lines to engage in intergroup action. Finally,

Table II.

Food prices, ethnic cleavages and mass nonviolent action, 1990–2008

	Model 1	Model 2	Model 3	Model 4
	w/FE	w/FE	w/FE	w/FE
Standardized increase in food prices		0.127***	–0.034	–0.013
		(0.037)	(0.065)	(0.058)
Excluded group area (opposition area)	0.0750	–0.023	–0.201*	–0.087
	(0.064)	(0.064)	(0.085)	(0.075)
Included group area (government area)	–0.032	–0.080	–0.091	–0.093
	(0.059)	(0.057)	(0.057)	(0.060)
Food price increases X Excluded areas			0.202**
			(0.065)
Food prices X Diverse excluded areas				0.150**
				(0.048)
Ethnically diverse excluded areas	–0.142**			–0.151**
	(0.053)			(0.054)
Grid wealth (GCP) (log)	0.379**	0.417***	0.464***	0.481***
	(0.117)	(0.118)	(0.118)	(0.119)
Grid population (log)	1.214***	1.201***	1.202***	1.223***
	(0.075)	(0.074)	(0.074)	(0.074)
Size of excluded group (%)	–0.031	–0.010	0.044	0.037
	(0.065)	(0.067)	(0.069)	(0.067)
Travel to urban centre (mins)	–4.376***	–4.460***	–4.461***	–4.403***
	(0.321)	(0.322)	(0.322)	(0.322)
Distance to capital (km)	0.075	0.081	0.081	0.070
	(0.055)	(0.054)	(0.054)	(0.054)
No. excluded groups	0.065	0.080	0.060	0.003
	(0.094)	(0.094)	(0.095)	(0.096)
National population (log)	–1.128*	–1.016*	–1.377**	–1.375**
	(0.441)	(0.440)	(0.456)	(0.455)
Regime type (Polity2)	0.020	0.004	0.020	–0.011
	(0.087)	(0.087)	(0.088)	(0.087)
National elections	0.122***	0.118***	0.107**	0.112**
	(0.035)	(0.035)	(0.036)	(0.035)
Number of peace years	–0.0190	–0.029	–0.036	–0.034
	(0.059)	(0.059)	(0.059)	(0.059)
Country fixed effects	Yes	Yes	Yes	Yes
Temporal and spatial lags	Yes	Yes	Yes	Yes
R2	0.524	0.524	0.525	0.526
Observations	86,203	86,203	86,203	86,203

^† p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001. Clustered standard errors by country. Reference category: grids with no ethnic groups.

nonviolent action is less likely to occur in larger countries (proxied by population size), which increases coordination costs.

Moving beyond the impact of one-standard-deviation increases in food prices, I explore in-sample simulated predictions using CLARIFY (Tomz, Wittenberg & King, 2003).¹⁷ Unfortunately, the post-estimations of fixed-effects models are limited and are not supported by CLARIFY.¹⁸ I therefore generate predictions based on logistic regression models clustered around country standard errors.¹⁹ The simulations show that the likelihood of nonviolent action is very low (at 0.0052%) as nonviolent action is very rare and that the effect exponentially increases with higher rises in food prices and when holding other variables at their means. When food prices rise by 15%, the probability of nonviolent action increases to 0.0071%, an increase of 38.1%. When moving to a 30% rise in food prices, the likelihood of nonviolent resistance is increased to 0.0114%, making a jump of 60.2%. The same occurs in ethnically excluded areas, where a 1%–15% rise in food prices increases the risk of nonviolent action by 51.1%. Finally, in more complex and ethnically diverse environments, the probability of nonviolent action increases by 31.4% when prices move from a 1% increase to a 15% increase and a further 37.8% when prices go up by 30%.

Following Ward, Greenhill & Bakke (2010), I then explore the predictive power of the above model in an out-of-sample analysis. Although more difficult to interpret substantively, out-of-sample prediction serves as a harder test for the predictive power of the model and whether key variables increase or reduce predictive power. Here I use K(10)-fold cross-validation, where nine random segments of the data are used to predict the final segment.²⁰ The output produces a size of area under the curve (AUC) ranging from 1.0 (perfectly predictive) to 0.5 (non-predictive). Higher rates have lower false positive and higher true positive outcomes. The above model performs very well at around 0.966, slightly less predictive than the in-sample (0.967). Moreover, inputting key explanatory variables into the model increases the model’s predictive power.²¹

Robustness checks

To check the robustness of my results I run additional analyses.²² The first issue I explore is whether food prices are indeed a unique cross-cutting issue and are not driven by the coding scheme. I suggested theoretically that elections should have a positive effect on protest (Salehyan & Linebarger, 2015), but not necessarily cross-cut ethnic divides, as electoral support in African countries is often drawn along ethnic lines. While I find elections do have a direct effect, I find no evidence that excluded and diverse areas are more likely to experience nonviolent action during election years. I explore this further by looking at riot outcomes. Rioting is directly impacted by food prices (see Barrett, 2013; Smith, 2014), but should not be dependent on a cross-cutting issue as the scope of mobilization is lower and sporadic. While price spikes have a direct effect, they do not mediate the likelihood of rioting in diverse areas. This suggests that the cross-cutting mechanism linking ethnic grievances to nonviolent action is unique to nonviolent mobilization. I then move on to other types of economic shocks that may cause nonviolent action or mediate the impact of food prices: deviations in average rainfall, economic growth and international food prices. The results do not impact the main findings.²³

Moving beyond the theoretical mechanism itself, the results could be driven by model choice. I rerun my results using other methods that address unit effects: random-effects and logistic regression with country corrected standards errors (Beck & Katz, 1995). For the latter, I employ rare-events logistical regression to better estimate rare outcomes, as nonviolence only occurs in 1.7% of all grid-years (King & Zeng, 2001). For both models the results are identical. However, when using rare-events logistic regression, the interaction of food prices with diversity remains positive, but falls just outside statistical significance.²⁴ I then look to see if the results can be reproduced at the country level as an alternative unit of analysis. Here I look at the number of excluded EPR groups in a country, which when interacted with high food price spikes increases the likelihood of nonviolent action, in line with results at the subnational level.²⁵

The next concern is the possible impact of time trends. Following Beck, Katz & Tucker (1998) I introduce time trend dummies and cubic splines (time since the last nonviolent event) using country fixed effects.²⁶ The results remain the same. Another method is to add further fixed effects, although there are suggestions this can produce unstable results with binary outcomes (Beck & Katz, 2001). Nevertheless, I introduce more restrictive models with year fixed effects and then grid-year fixed effects²⁷ where the results largely hold (p < 0.10). Next I rerun the results using included EPR groups as the reference category, excluding all grids without EPR groups. The results remain the same and support Thurber’s (2018) finding that excluded groups are less likely to protest than included groups.²⁸

I then turn to three alternative DVs, as the results may be driven by the primary DV – Top5_est. The first two extend my geocoded estimations to include the ten largest cities with 100,000 people (Top10_est) and I then include all other cities with a population over 300,000 (Full_est). The results actually become stronger, which suggests that the Top5_est DV may underestimate the location of many nationwide events. The results are less stable with the last DV, which removes the non-geocoded nation/regionwide events (No_est). However, as previously argued, removing the geo-estimates leads to the coding of ‘false 0s’ in one-quarter of country years witnessing nonviolent action and removes key widespread nonviolent action most likely related to price spikes.²⁹

Next, I explore urban-only nonviolent action, where I can be more confident that the food price data closely reflect what people are paying. I rerun the analyses with an urban-only sample (within five hours of an urban centre) and the full sample using an urban-only DV. I then also remove agricultural areas, which are less likely to be consumers and are less vulnerable to food prices than those in less agricultural areas (i.e. urban areas) (Barrett, 2013). While agricultural areas are associated with less nonviolent resistance (p < 0.05),³⁰ the results do not change.

In the final robustness checks I account for possible reverse causality (i.e. nonviolent action could increase prices) and other omitted variable bias.³¹ I rerun my analysis with lagged independent variables. Next I run additional models that account for other explanations of variations in the emergence of nonviolent resistance, including the CIRI Index to account for highly repressive states that may deter nonviolent action (Lichbach, 1998) and control for relatively richer and poor groups within a grid, as richer ethnic groups may have a higher mobilization potential. The findings remain the same.

Conclusion

Across various specifications, food price spikes, as a cross-cutting issue, have been shown to facilitate nonviolent mobilization, both vertically against the government and horizontally in forging coethnic and intergroup coalitions. These results provide strong evidence that the existence of cross-cutting grievances increases the emergence of nonviolent action in ethnically excluded and diverse locations, which are otherwise constrained by social divisions.

These findings extend our understanding of how and when nonviolent action is likely to emerge when activists are facing difficult social barriers to nonviolent mobilization. In doing so, this article contributes to two broader bodies of literature that have grown in isolation to one another: existing studies of civil resistance that have focused on outcomes and structural approaches that largely neglect the horizontal dimension of nonviolent mobilization when exploring its emergence. This study also builds on country and movement-level quantitative studies, by exploring nonviolent action at the subnational level, thereby accounting for the variation of nonviolent action within countries, using new geocoded events data.

Beyond academic contributions, this article also provides important policy implications. African regimes that deploy ethno-exclusive policies as a means to maintain political power are not immune from nonviolent action. Such regimes need to develop more inclusive institutions to reduce the likelihood of nonviolent action and other forms of unrest. Moreover, food price spikes are clearly important for political stability. While African governments have various options to implement stabilization mechanisms that alleviate the impact of rising food prices, safety net measures are either rare or ineffective. Only nine African countries implemented food-access programmes during record high prices in 2007 and 2008 (Berazneva & Lee, 2013). In many cases, governments are limited in how they can act, as was the case in Egypt where subsidies became simply unaffordable, accounting for 8% of the country’s GDP in 2011 (Hendrix & Brinkman, 2013). However, governments need to diversify their policies to implement creative long-term solutions for social protection.³²

Future research could look at other ‘unifying’ factors that cut across social boundaries and enable opposition groups to engage in nonviolent action, for example, currency devaluation, cross-cutting religious and language cleavages and other commodities such as fuel prices. The article is better placed to capture the grievances of consumers, but other research could consider decreases in food prices, particularly cash crops, which may trigger rural-based civil resistance which has its own unique mobilization challenges. Finally, research is also needed to understand how subnational variations in food prices impact unrest, but this is dependent on the availability of new data.

Footnotes

Replication data

The dataset and do-files for the empirical analysis, along with the Online appendix for this article, can be found at .

Acknowledgments

I thank the journal’s editor and reviewers for their comments and suggestions. I am also extremely grateful to Govinda Clayton, Lucy Barnes, Hugh Miall, Kristian S. Gleditsch, Halvard Buhaug, Roman-Gabriel Olar, Margherita Belgioioso, Robert Nagel and Daniel Kirkpatrick for valuable inputs.

Funding

I acknowledge the support of the Economic and Social Research Council (1353916).

ORCID iD

Luke Abbs

Notes

References

Ackerman

Peter

DuVall

Jack

(2000) A Force More Powerful: A Century of Nonviolent Conflict. New York: Palgrave.

Arriola

Leonardo

(2013) Protesting and policing in a multiethnic authoritarian state. Comparative Politics 45(2): 147–168.

Barrett

Christopher

, ed. (2013) Food Insecurity and Sociopolitical Stability. Oxford: Oxford University Press.

Beck

Nathaniel

Katz

Jonathan

(1995) What to do (and not to do) with time-series cross-sectional data. American Political Science Review 89(3): 634–647.

Beck

Nathaniel

Katz

Jonathan

(2001) Throwing out the baby with the bath water: A comment on Green, Kim, and Yoon. International Organization 55(2): 487–495.

Beck

Nathaniel

Katz

Jonathan

Tucker

Richard

(1998) Taking time seriously: Time-series–cross-section analysis with a binary dependent variable. American Journal of Political Science 42(4): 1260–1288.

Berazneva

Julia

Lee

David

(2013) Explaining the African food riots of 2007–2008: An empirical analysis. Food Policy 39(April): 28–39.

Bhavnani

Rikhil

Jha

Saumitra

(2014) Gandhi’s gift: Lessons for peaceful reform from India’s struggle for democracy. Economics of Peace and Security Journal 9(1): 76–88.

Blattman

Christopher

Miguel

Edward

(2010) Civil war. Journal of Economic Literature 48(1): 3–57.

10.

Buhaug

Halvard

(2015) Climate–conflict research: Some reflections on the way forward. WIREs Climate Change 6(3): 269–275.

11.

Bush

Ray

(2010) Food riots: Poverty, power and protest. Journal of Agrarian Change 10(1): 119–129.

12.

Butcher

Charles

Svensson

Isak

(2016) Manufacturing dissent: Modernization and the onset of major nonviolent resistance campaigns. Journal of Conflict Resolution 60(2): 311–339.

13.

Cederman

Lars-Erik

Gleditsch

Kristian Skrede

Buhaug

Halvard

(2013) Inequalities, Grievances and Civil War. Cambridge: Cambridge University Press.

14.

Center for International Earth Science Information Network (CIESIN) (2005) Gridded Population of the World Version 3 (GPWv3). Palisades, NY: SEDAC, Columbia University (http://sedac.ciesin.columbia.edu/gpw/).

15.

Chenoweth

Erica

Lewis

Orion

(2013) Unpacking nonviolent campaigns: Introducing the NAVCO 2.0 dataset. Journal of Peace Research 50(3): 415–423.

16.

Chenoweth

Erica

Stephan

Maria

(2011) Why Civil Resistance Works: The Strategic Logic of Nonviolent Conflict. New York: Columbia University Press.

17.

Chenoweth

Erica

Ulfelder

Jay

(2017) Can structural conditions explain the onset of nonviolent uprisings? Journal of Conflict Resolution 61(2): 298–324.

18.

Cunningham

Kathleen

(2015) Inside the Politics of Self-Determination. Oxford: Oxford University Press.

19.

Dahl

Marianne

Gates

Scott

Gleditsch

Kristian Skrede

González

Belen

(2016) Accounting for numbers: How group characteristics shape the choice of violent and non-violent tactics. Typescript, Peace Research Institute Oslo (PRIO).

20.

DeNardo

James

(1985) Power in Numbers: The Political Strategy of Protest and Rebellion. Princeton, NJ: Princeton University Press.

21.

Dube

Oeindrila

Vargas

Juan

(2013) Commodity price shocks and civil conflict: Evidence from Colombia. Review of Economic Studies 80(4): 1384–1421.

22.

Gamson

William

(1990) The Strategy of Social Protest, 2nd edition. Belmot, CA: Wadsworth.

23.

Gates

Scott

Hegre

Håvard

Nygard

Håvard Mokleiv

Strand

Håvard

(2012) Development consequences of armed conflict. World Development 40(9): 1713–1722.

24.

Gerhards

Jürgen

Rucht

Dieter

(1992) Mesomobilization: Organizing and framing in two protest campaigns in West Germany. American Journal of Sociology 98(3): 555–596.

25.

Goldstone

Jack

(1991) Revolution and Rebellion in the Early Modern World. Berkeley, CA: University of California Press.

26.

Goldstone

Jack

(1994) Is revolution individually rational? Groups and individuals in revolutionary collective action. Rationality and Society 6(4): 139–166.

27.

Goldstone

Jack

(2011) Cross-class coalitions and the making of the Arab revolts of 2011. Swiss Political Science Review 17(4): 457–462.

28.

Gorenburg

Dmitry

(2000) Not with one voice: An explanation of intragroup variation in nationalist sentiment. World Politics 53(1): 115–142.

29.

Gurr

Ted

(1993) Minorities at Risk: A Global View of Ethnopolitical Conflicts. Washington, DC: United States Institute of Peace.

30.

Hendrix

Cullen

Brinkman

Henk-Jan

(2013) Food insecurity and conflict dynamics: Causal linkages and complex feedbacks. Stability: International Journal of Security & Development 2(2): 1–18.

31.

Hopkins

Valerie

(2014) ‘We are hungry in three languages’: Citizens protest in Bosnia. Open Democracy, 13th February (https://www.opendemocracy.net/5050/valerie-hopkins/we-are-hungry-in-three-languages-citizens-protest-in-bosnia).

32.

Horowitz

Donald L

(1985) Ethnic Groups in Conflict. Berkeley, CA: University of California Press.

33.

International Labour Organization (2014) International Labour Organization Food Indices Database (https://ilostat.ilo.org/data/).

34.

Ivanic

Maros

Martin

Will

Zaman

Hassan

(2012) Estimating the short-run poverty impacts of the 2010–11 surge in food prices. World Development 40(11): 2302–2317.

35.

Jazayeri

Karen Bodnaruk

(2016) Identity-based political inequality and protest: The dynamic relationship between political power and protest in the Middle East and North Africa. Conflict Management and Peace Science 33(4): 400–422.

36.

King

Gary

Zeng

Langche

(2001) Explaining rare events in international relations. International Organization 55(3): 693–715.

37.

Kitschelt

Hebert

Wilkinson

Steven

(2007) Patrons, Clients and Policies: Patterns of Democratic Accountability and Political Competition. Cambridge: Cambridge University Press.

38.

Kron

Josh

(2011) Protests in Uganda over rising prices grow violent. New York Times 21 April (http://www.nytimes.com/2011/04/22/world/africa/22uganda.html?_r=0).

39.

Kuran

Timur

(1991) Now out of never: The element of surprise in the East European revolution of 1989. World Politics 44(1): 7–48.

40.

Lichbach

Mark

(1998) The Rebel’s Dilemma. Ann Arbor, MI: University of Michigan Press.

41.

Lindberg

Staffan

(2009) Elections and democracy in Africa 1989–2007 dataset (http://users.clas.ufl.edu/sil/downloads.html).

42.

Marshall

Monty

Jaggers

Keith

(2012) Polity IV Project, p4v2012d Version (http://www.systemicpeace.org/inscrdata.html).

43.

McAdam

Doug

(1999) Political Process and the Development of Black Insurgency, 1930–1970, 2nd edition. Chicago, IL: University of Chicago Press.

44.

Meyer

David

(2004) Protest and political opportunities. Annual Review of Sociology 30: 125–145.

45.

Nepstad

Sharon

(2015) Nonviolent Struggle: Theories, Strategies and Dynamics. Oxford: Oxford University Press.

46.

Nordhaus

William

(2006) Geography and macroeconomics: New data and new findings. Proceedings of the National Academy of Sciences of the USA 103(10): 3510–3517.

47.

Olzak

Susan

(2006) The Global Dynamics of Race and Ethnic Mobilization. Stanford, CA: Stanford University Press.

48.

Ortiz

Isabel

Hagemejer

Krysztof

Drouin

Anne

Bonnet

Florence

Behrendt

Christina

Guilbault

Emannuelle St-Pierre

Scheil-Adlung

Xenia

Woodall

John

(2014) World Social Protection Report 2014/15: Building Economic Recovery, Inclusive Development and Social Justice. Geneva: International Labour Office.

49.

Østby

Gudrun

(2008) Polarization, horizontal inequalities and violent civil conflict. Journal of Peace Research 45(2): 143–162.

50.

Posner

Daniel

(2004) The political salience of cultural difference: Why Chewas and Tumbukas are allies in Zambia and adversaries in Malawi. American Political Science Review 98(4): 529–545.

51.

Salehyan

Idean

Linebarger

Christopher

(2015) Elections and social conflict in Africa, 1990–2009. Studies in Comparative International Development 50(1): 23–49.

52.

Salehyan

Idean

Hendrix

Cullen

Hamner

Jesse

Case

Christina

Linebarger

Christopher

Stull

Emily

Williams

Jennifer

(2012) Social conflict in Africa: A new database. International Interactions 38(4): 503–511.

53.

Schock

Kurt

(2005) Unarmed Insurrection: People Power Movements in Nondemocracies. Minneapolis, MN: University of Minnesota Press.

54.

Sharp

Gene

(2005) Waging Nonviolent Struggle: 20th Century Practice and 21st Century Potential. Boston, MA: Porter Sargent.

55.

Siegal

David

(2009) Social networks and collective action. American Journal of Political Science 53(1): 122–138.

56.

Smith

Todd

(2014) Feeding unrest: Disentangling the causal relationship between food price shocks and sociopolitical conflict in urban Africa. Journal of Peace Research 51(6): 679–695.

57.

Stewart

Frances

, ed. (2008) Horizontal Inequalities and Conflict: Understanding Group Violence in Multiethnic Societies. Basingstoke: Palgrave Macmillan.

58.

Svensson

Isak

Lindgren

Mathilda

(2011) Community and consent: Unarmed insurrections in non-democracies. European Journal of International Relations 17(1): 97–120.

59.

Tajfel

Henri

(1982) Social psychology of intergroup relations. Annual Review of Psychology 33: 1–39.

60.

Thurber

Ches

(2018) Ethnic barriers to civil resistance. Journal of Global Security Studies 3(3): 255–270.

61.

Tilly

Charles

(1978) From Mobilization to Revolution. New York: McGraw-Hill.

62.

Tollefsen

Andreas Forø

Strand

Håvard

Buhaug

Halvard

(2012) PRIO-GRID: A unified spatial data structure. Journal of Peace Research 49(2): 363–374.

63.

Tomz

Michael

Wittenberg

Jason

King

Gary

(2003) CLARIFY: Software for interpreting and presenting statistical results. Journal of Statistical Software 8(2): 1–29.

64.

Uchida

Hirotsugu

Nelson

Andrew

(2010) Agglomeration index: Towards a new measure of urban concentration. WIDER Working Paper 2010/029. Helsinki: UNU-WIDER.

65.

Vidovic

Dragana

Gleditsch

Kristian Skrede

(2015) Ethnic diversity and nonviolent uprisings: Linking arms in solidarity. Paper presented at the 2015 Conflict Research Society Conference, Canterbury, UK.

66.

Walter

Barbara

(2009) Reputation and Civil War: Why Separatist Conflicts Are So Violent. Cambridge: Cambridge University Press.

67.

Ward

Michael

Greenhill

Brian

Bakke

Kristin

(2010) The perils of policy by p-value: Predicting civil conflicts. Journal of Peace Research 47(4): 363–375.

68.

Weinberg

Joe

Bakker

Ryan

(2014) Let them eat cake: Food prices, domestic policy and social unrest. Conflict Management and Peace Science 32(3): 309–326.

69.

White

Peter

Vidovic

Dragana

González

Belen

Gleditsch

Kristian Skrede

Cunningham

David

(2015) Nonviolence as a weapon of the resourceful: From claims to tactics in mobilization. Mobilization: An International Quarterly 20(4): 471–491.

70.

Wimmer

Andreas

(2013) Waves of War: Nationalism, State Formation, and Ethnic Exclusion in the Modern World. Cambridge: Cambridge University Press.

71.

Wimmer

Andreas

Cederman

Lars-Erik

Min

Brian

(2009) Ethnic politics and armed conflict: A configurational analysis of a new global dataset. American Sociological Review 74(2): 316–337.

72.

Wischnath

Gerdis

Buhaug

Halvard

(2014) Rice or riots: On food production and conflict severity across India. Political Geography 43(November): 6–15.

73.

Wodon

Quentin

Zaman

Hassan

(2010) Higher food prices in sub-Saharan Africa: Poverty impact and policy responses. World Bank Research Observer 25(1): 157–176.

74.

World Bank (2013) World Development Indicators (http://data.worldbank.org/data-catalog/world-development-indicators).

75.

Wucherpfennig

Julian

Weidmann

Nils

Girardin

Luc

Cederman

Lars-Erik

Wimmer

Andreas

(2011) Politically relevant ethnic groups across space and time: Introducing the GeoEPR dataset. Conflict Management and Peace Science 28(5): 423–437.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB