Nothing Compares? Investigating the Cost of Food as a Driver of Urban Unrest

Abstract Increasing food prices have implications for basic subsistence, have a strong price visibility and symbolic value, and are characterized by high volatility and inelasticity of demand. Research thus assumes that food price is an important trigger for unrest. Yet, whether food is an especially potent driver for people’s willingness to engage in collective action, or whether it concerns grievances about general inflation, is unknown. Does food have a greater effect on the willingness to participate in unrest? The paper investigates the relative importance of food in mobilization potential by using unique data from a survey experiment in Johannesburg, South Africa. The experiment collects information on how price increases in food, fuel, and electricity affects respondents’ willingness to engage in unrest. The results show a higher willingness to engage in collective action when presented with increasing living expenses, regardless of whether it is food, fuel or electricity, compared to stable prices. We also consider the level of risk exposure to price hikes, and find that those who report going hungry in the last year have a higher willingness to engage in unrest than those who do not. Thus, food access influences the willingness to partake in unrest during price hikes, also for commodities seemingly unrelated to food. This suggests that for those who are most affected by a price hike it is less important what type of commodity it is. The question is whether it introduces further strain on an already hard-stretched budget.

población a participar en disturbios sociales?Este artículo investiga este asunto utilizando datos únicos de un experimento de encuesta en Johannesburgo, Sudáfrica.Este experimento recoge información sobre cómo los aumentos de precios de los alimentos, del combustible y de la electricidad influyen sobre la voluntad de los encuestados a participar en disturbios sociales.Los resultados muestran una mayor disposición a participar en acciones colectivas cuando existe un aumento de los costes de vida, pero este efecto no es más fuerte para el precio de los alimentos en comparación con otros productos básicos.También tenemos en cuenta el nivel de exposición al riesgo de las subidas de precios, y descubrimos que quienes declaran haber pasado hambre en el último año tienen una mayor disposición a participar en disturbios sociales que los que no lo hacen.Por lo tanto, el acceso a los alimentos tiene potencial en materia de movilización, no solo debido a los precios, sino también en términos de cómo influye en la disposición a participar en disturbios sociales durante los aumentos de precios.Esto también ocurre para productos básicos aparentemente no relacionados con los alimentos.Esto sugiere que para aquellas personas que se ven más afectadas por un aumento de precios, el tipo de producto no resulta tan importante.La pregunta que nos hacemos es si esto crea más presión sobre una situación que ya es vulnerable y sobre un presupuesto muy ajustado.

Introduction
Extant academic literature suggests a positive association between higher food prices and unrest. 1 Studies of food and contentious action increased in the wake of the food price spikes in 2007-2008 and 2010-2011 and the corresponding unrest that followed (Bellemare 2015;Bush and Martiniello 2017;Sneyd, Legwegoh, and Fraser 2013).The interest in and importance of understanding the societal impacts of increasing living expenses has only further been strengthened after the COVID-19 pandemic and the Russian invasion of Ukraine.Concerns about high inflation has swept across the globe where countries grapple with a cost of living crisis, and increasing prices of commodities such as energy, food and fertilizer prices has corresponded with protests in a wide range of countries in the Americas, Asia, Africa, and the Middle East (Marsh, Rudolfsen and Aas Rustad 2022;Hossain and Hallock 2022).
The role of food prices is often pointed to as a particularly relevant phenomenon in the relative deprivation-conflict nexus (e.g.Gurr 1970;Hendrix and Haggard 2015).This is because food is assumed to have specific features that separates it from other commodities, related to the importance of food for basic subsistence, its price visibility, symbolism, and volatility and inelasticity of demand (Weinberg and Bakker 2015).Yet, while previous research points to food as one of the most fundamental aspect in the relative deprivation-conflict link, the question of whether food is a more important commodity compared to other living expenses in motivating participation in unrest remains so far unexplored.This article seeks to shed light on this question: Is the cost of food an especially compelling driver for people's willingness to engage in unrest compared to other commodities that make up a substantial share of the household budget?
To address this question, the study uses an individual-level approach by asking consumers about their willingness to engage in unrest following an increase in the price of food, fuel and electricity respectively.In the real world, the prices of these items often shift in tandem.As we seek to identify the importance of one commodity price hike over another, and since the outbreak of unrest may also influence the price of food and 1 We understand unrest as an overarching term for various forms of coordinated action involving multiple individuals, and focus here on such collective efforts occurring in urban areas.We use collective action and unrest interchangeably throughout the article.The action can be more or less legitimate, organized, legal and violent (Opp 2009;Tilly 1978).While we probe different types of collective action, we argue for similarities in the underlying process that might generate contentious action (McAdam, Tarrow and Tilly 2001).The different forms of actions included here are attending ward meetings to raise concern, demonstrations, strikes, and shutting down traffic.Armed conflict is not included in this definition.While the term unrest may have negative normative associations with it, we pass no judgment on collective action as a reaction to increasing costs.Unrest is a well-established term in the literature, and we use it here to encompass a variety of possible responses to increased household spending.
other items, we need a research design that can account for endogeneity and allow us to distinguish between cause and effect (Raleigh, Choi and Kniveton 2015).The study therefore makes use of a vignette experiment where we randomly assign respondents to treatment and control conditions consisting of fictional news stories of coming price hikes and subsequently map consumer reactions.We study how information about different commodity prices affects the stated willingness to engage in unrest.
We use data from a survey experiment conducted in Johannesburg, South Africa, and the case is particularly well suited to study the phenomenon of interest for several reasons.Johannesburg has experienced a relatively high number of both peaceful and violent events, especially since the mid 2000s (Hart 2013;Paret 2017;Tournadre 2018).Also, as in South Africa in general, Johannesburg exhibits widespread poverty, inequality and food insecurity (Chatterjee, Czajka and Gethin 2021;Rudolph et al. 2012;Statistics South Africa 2019).In addition, the phenomenon of interest is in a context of urban consumers, with corresponding population density and heterogeneity that may provide mobilization potential (Verpoorten et al. 2013).At the same time, the case exhibits significant variation with regards to both unrest activity and socioeconomic status.
The results show that respondents are more willing to engage in collective action when presented with increasing living expenses compared to stable prices, regardless of whether it is food, fuel or electricity.We also assess the level of vulnerability to price hikes for the individual, and find that those who report going hungry in the last year are more willing to engage in unrest over commodity price hikes than those who do not.Thus, food has mobilization potential not only due to its price, but also in terms of how food insecurity influences the willingness to partake in unrest during price hikes, also for commodities seemingly unrelated to food.This suggests that for those who are most affected by a price hike it is less important what type of commodity it is.The question is whether it introduces further strain on an already vulnerable situation and hardstretched budget.
We contribute to the literature on food and unrest in several ways.First, while the literature on food price and collective action is growing rapidly, previous studies have relied primarily on aggregated data.Filling this gap in the literature, the study utilizes individual-level data to assess the importance of increased cost of living for mobilization potential.Second, to our knowledge, this study is also the first to map reactions to different commodity price hikes, comparing food, fuel and electricity in an experimental design to assess the willingness to engage in unrest depending on variation in the source of increased living expenses.Finally, the paper contributes to the literature on relative deprivation and social conflict by accounting for not only reactions to increasing prices, but also the individual's vulnerability to it.While previous studies have largely used aggregate food prices as a proxy for relative deprivation for the individual, we measure risk exposure directly by accounting for food insecurity.Thus, the study seeks to grapple and understand the relative importance of food in collective action.While increasing living costs can have adverse impacts on urban consumers, we have little understanding of how the underlying vulnerability to such a shock impact mobilization potential.Some strands in the literature point to the importance of food because of its fundamental nature, others suggest that food functions as the last straw that ignites unrest due to general inflation.Our study suggests that it is both, where higher cost of living increases the willingness to engage in unrest, but this willingness is contingent on the vulnerability to price hikes.
In the following sections, we briefly present an overview of the existing literature.We then outline the theoretical framework and hypotheses, before introducing the research design and analysis.Subsequently, we discuss some of the threats to inference given the research design and external validity, before making some concluding remarks.

Increasing Cost of Living and Social Unrest
The body of research on the role of food in precipitating unrest is vast and have long historical roots (see e.g.Abbs 2020, Berazneva and Lee 2013, Bienen and Gersovitz 1986, Brinkman and Hendrix 2011, Bush 2010, De Winne and Peersman 2021;Hendrix and Brinkman 2013, Hertel 2014, Heslin 2021, Jones, Mattiacci and Braumoeller 2017, Koren and Bagozzi 2016, Popkin 1979, Rudé 1964, Scott 1976, Tilly 1971, 1983, Thompson 1971, Walton 1994, Weinberg and Bakker 2015), and there is an emerging consensus in the field that increasing food prices lead to unrest, especially in urban areas (Demarest 2014;Rudolfsen 2020).Recent quantitative studies, using both local (Raleigh, Choi and Kniveton 2015;Smith 2014) and global (Bellemare 2015;Hendrix and Haggard 2015;Sternberg 2012) food prices, have found that higher food prices increase the risk of unrest.Arezki and Brückner (2014), for example, point to the widening gap between rich and poor when prices increase.This inter-temporal change in relative deprivation subsequently increased the risk of conflict.In the same vein, Smith (2014) proposes that consumer grievances could increase unrest if the elite captures increased revenue, and the effect of rising prices on unrest is driven by consumer grievance over increased economic pressure.Economic grievances increase due to higher costs as a larger share of the household budget is spent on increasingly costly living expenses.
Relevant work in this context is also studies exploring the link between energy prices and protest.While the literature on electricity prices and unrest is, to our knowledge, scant2 , there has been an uptick in the number of studies focusing on fuel prices and various forms of social unrest in recent years (see e.g.Agbonifo 2023;Dube and Vargas 2013;Ishak and Farzanegan 2022;Jetten, Mols and Selvanathan 2020;Natalini, Bravo and Newman 2020).For example, Vadlamannati and de Soysa (2020) find that both high oil prices in oil-importing countries and low oil prices i oil-producing countries lead to protest, conditional on the level of foreign exchange reserves.In addition, investigating 41 countries between 2005and 2018, McCulloch et al. (2022) find that domestic fuel price shocks are a key driver of riots.They also find that countries with large fuel subsidies are more at risk of fuel riots when this policy becomes unsustainable to uphold due to fiscal concerns.In a meta study including a review of 350 studies, Blair, Christensen and Rudkin (2021) find that higher oil prices is positively associated with conflict.
Based on previous research, there is overall agreement in the literature that increasing living expenses (especially in the form of food and fuel prices) is positively associated with social unrest.However, we know much less about their relative importance in mobilization potential based on consumer preferences.In the following, we therefore outline two propositions on the motivation to partake in unrest based on increasing living expenses, and whether the increasing cost of food would be an especially potent driver in this regard.

Why Food Prices May be Especially Aggrieving
The link between higher living costs and social upheaval tends to be explained through relative deprivation, a central theory in conflict research (Gurr 1970).Relative deprivation is the experienced gap between a person's desired and actual situation, defined in terms of perceived entitlements.Relative deprivation increases with the loss of entitlements when living expenses rise.Feelings of deprivation could occur by a person's assessment of deteriorating welfare, or by interpersonal or intergroup comparisons.
Studies often point to food as something especially aggrieving, where feelings of relative deprivation should become particularly salient in the case of food price hikes.This is because food has well-known features that separate it from other commodities, such as the necessity of food purchase for basic subsistence, its price visibility, symbolism, and the volatility and inelasticity of demand.The first and most fundamental proposition for why food price would be an especially likely mechanism through which grievances lead to conflict is the necessity of purchasing food for basic subsistence: "The argument would seem particularly plausible for food, which is the most basic of all necessities and the one most likely to embody explicit or implicit political entitlements" (Hendrix and Haggard 2015, 145).According to Gurr, relative deprivation is most likely to occur over issues that people believe they are rightfully entitled, where it is likely that food would be one such entitlement.Thus, food prices would have a particular effect on the welfare of poor individuals who are net purchasers of food (see e.g.D'Souza and Jolliffe (2012)). 3Food purchase as an essential human necessity suggests that rising food prices lead to rising food insecurity and a reduction in the opportunity cost of violent unrest.This argument is based on the special nature of food as necessary for human survival compared to other commodities or costs of living.In the words of (Bush 2010, 119): "Food is a commodity and its use value, unlike those of most other commodities, provides for the maintenance and reproduction of life itself." Grievance-induced unrest would therefore seem particularly plausible for food (Pinstrup-Andersen and Shimokawa 2008).
Second, food price could be an especially compelling driver for unrest due to the visibility of increasing food prices for the consumer.While aggregate economic trends and smaller changes in the GDP may be difficult for individuals to track and make calculations about in order to identify potential grievances, food prices are much easier to observe as consumers make choices based on their own finances, rather than aggregate economic trends.Food is a good purchased regularly and is therefore according to Weinberg and Bakker (2015) the best available indicator of an individual's economic well-being.It provides a direct assessment of scarcity and is therefore both more easily and more frequently observed.
Third, food has a symbolic value as a threat to subsistence.For example, according to Simmons (2016), the threat to affordable access to water and tortilla in Mexico heightened solidarity within and between communities, where the threat to subsistence goods were not only material threats but also threats to community.The threat to subsistence helps bridge cleavages between social groups and facilitate widespread mobilization.Similarly, according to (Abbs 2020, 285), the cross-cutting nature of increasing cost of food sets it apart from other types of issues.Food price shocks provide an opportunity for movements to mobilize across intra-ethnic divides due to food prices being a unique and symbolic issue, where "[…] food price spikes generate superordinate goals that unify normally divided groups." Also, Hossain and Scott-Villiers (2017) write that food riots signal that the state is unable to meet the most basic condition of the social contract, which creates pressure on the state to address the most fundamental of failings.
Increasing food prices may be distinctive as a driver of unrest compared to other costs of living for two additional reasons.First, food prices are generally more volatile than other commodities.This volatility may create acute hardships during times of particularly sharp increases in ways that more predictably steady increases in other costs of living would not.For example, the cost of housing is generally much more stable than the cost of food.While the cost of housing is often the largest single expenditure, a family's rent or loan payment is often set for months or years in advance of when the expense will be incurred.While an increase in the cost of housing can have a significant impact on a household's budget, it is generally predictable in advance and, therefore, more easily adapted to.Food prices, however, can change daily, monthly or weekly, which can place unforeseen stress on household budgets (Tadesse et al. 2014).One commodity that may exhibit similar levels of volatility to food is the cost of fuel, particularly petrol or gasoline for transportation.The cost of fuel, however, is distinct from food in another important way: rising food prices tend to have a more substantial impact on the welfare of the poor than rising fuel prices (Chaudhry and Chaudhry 2008;Yu 2014). 4he inelasticity of demand for food means that increases in food prices do not affect all consumers equally.Engel's law states that as income increases, the proportion of household income that is spent on food decreases.Conversely, poor households tend to spend a higher proportion of household income on food purchases.Consequently, the economic hardship created by rising food prices may lead to a higher degree of perceived relative deprivation among poor households.Weinberg and Bakker (2015) note the disproportionate impact of food price increases on poor consumers, where the effect of price fluctuations on individual consumers is unambiguously unique due to the high proportion of expenditure on food and its nonsubstitutability, distinguishing food from any other commodity.If food prices are an especially compelling motivation for unrest because of their unique effect on consumers, we would expect to see a higher willingness to engage in collective action when faced with rising food prices.To test this, we formulate our first hypothesis: H1: Individuals will be more willing to engage in collective action when presented with the prospect of increasing food prices than when presented with the prospect of increases in the cost of fuel or electricity rates.

Why Food Prices May Not Be Especially Aggrieving
So far we've argued for why food prices may be an especially potent driver for urban unrest.However, there are also arguments for why food prices would not be especially compelling for unrest participation.Rather, it could be that higher inflation and increasing living cost in general after price hikes is linked to unrest participation.
Importantly, several contributions in the literature emphasize that food price-related unrest often is triggered by underlying grievances in society.This means that expressing grievances about the price of food can serve as a pretext for addressing other fundamental social issues, where food can function as a mobilization tool for political entrepreneurs such as opposition elites and other non-state stakeholders to address wider socioeconomic issues (Rudolfsen 2021).For example, in her work on Bangladesh and India, Heslin (2021) emphasizes that many food riots are not, in fact, directly caused by the issue of food.Rather, the outbreak of food riots is contingent on the local context, where the presence of existing actors who use decreased food access to address existing grievances shape the occurrence of riots.Thus, the issue of food can help mobilize collective action around a range of grievances.Also, Bush and Martiniello (2017) argue that food riots are part of larger political and economic crises.Looking at Uganda, Burkina Faso, Tunisia and Egypt, they claim that the unrest was inseparably linked to a political and economic struggle against state authorities.Also, (van Weezel 2016, 3) emphasizes that food-related protests are likely part of a larger struggle against issues such as inequality and political oppression: "As such, food price spikes provide a political trigger, often signaling other underlying factors of dissatisfaction, for the population to resort to violence."Thus, we might see a correlation between food price and unrest not because of food prices' unique effect on the willingness to engage in unrest, but simply because food prices fluctuate more often than other expenses and therefore more frequently presents itself as a circumstance where mobilization is possible.Food price shocks may provide a window of opportunity for mobilization towards causes that in principle can extend far beyond the availability of affordable food (McCarthy and Zald 1977).
More substantially, increasing food prices and unrest may be linked, not necessarily because food is the most fundamental element in the relative deprivation explanation, but because food price hikes represent a more general strain on the household budget, which subsequently increases grievances.While the relative deprivation argument seems very relevant to the issue of food, the explanation also applies to other expenses that cause a sudden increase in the cost of living.For example, the widespread unrest in Iran in November 2019 erupted after a surprise increase in the price of fuel, quickly escalating into the deadliest unrest since the 1979 Islamic revolution with more than 1500 dead (Fassihi and Gladstone 2019;Reuters 2019).Since 2000, there has been social upheaval over the cost of fuel and electricity in most parts of the world, such as India (Majumdar and Chakraborty 2020), Nigeria (Obiezu 2020), Zimbabwe (Marima 2019), Spain (Keeley 2008), Haiti (Associated Press 2018) and France (BBC News 2018).While these are highly different places with a variety of socioeconomic and political challenges, the protesters have the same dissatisfaction in trying to manage a hard-stretched budget.The resulting unrest could either be directed against the government, or the perceived source of the hardship, especially in authoritarian regimes where people have no other recourse to a change in leadership (Brancati 2014).This suggests that the mechanism linking increasing food prices to the willingness to engage in unrest is not necessarily due to the unique nature of food as a commodity.Rather, increasing living costs in general are a trigger for socioeconomic grievances and subsequent social upheaval.This leads us to our second hypothesis: H2: Individuals will be more willing to engage in collective action when presented with the prospect of increasing commodity prices, regardless of category.

Vulnerability to Commodity Price Hikes
To better understand the role of commodity price hikes for the willingness to participate in unrest, we also need to account for the level of vulnerability to a price hike, considering the level of susceptibility increasing commodity prices has on the individual.There are many different types of coping strategies that individuals may take to deal with the impact of a price hike, such as seeking substitutes, altering consumption patterns, rationing, and the like (FAO 2021).However, it is unlikely that all consumers would have the coping mechanisms needed to cushion a price shock, and the risk exposure to price hikes is not homogeneous.People prioritize what they spend money on and as expenditures increase, people cut the least valued items first, leaving food and water as the last items cut (Maslow 1943;Sen 1982).Even small increases in consumer prices can have severe consequences for people who already spend a considerable share of their money on food.
Therefore, consumers more sensitive to price changes might present a motivational force to engage in unrest due to limited coping mechanisms to put pressure on state authorities to intervene (Ivanic, Martin and Zaman 2012;Simmons 2016;Weinberg and Bakker 2015).
We proxy vulnerability to price hikes by comparing those who report experiencing hunger in the last year with those who report experiencing no hunger to account for that the impact of higher prices might vary between individuals.Indeed, based on previous literature it is assumed that food is relevant for mobilization potential because of its fundamental nature for basic subsistence.Thus, we also assess whether the willingness to engage in unrest is higher when this fundamental need is threatened, looking into whether risk exposure influences the willingness to engage in collective action due to increasing prices.We would expect those who are more heavily impacted by a price hike to be more willing to engage in collective action.
We calculated the mean percentage of household income spent on food, electricity and fuel by income quantile in South Africa in Table A2 in the Supplementary Appendix, showing that the share of household spending is substantially higher for food compared to other living expenses for poorer households.The question is then whether the source of the price spike matters, as long as it threatens a hard-stretched budget.For example, assessing the impact of the food, fuel and financial crisis in 2008-2011, Heltberg et al. (2013) ) find that the varying price hikes led to significant hardships, especially among the poor with limited coping response and resilience.We therefore assess the level of vulnerability to price shocks in predicting unrest participation.Lack of access to food may contribute to unrest participation as a more chronic grievance, rather than a specific food price shock that triggered participation.Thus, we formulate our third hypothesis: H3: Individuals reporting higher levels of risk exposure will be more willing to engage in collective action when presented with the prospect of increasing commodity prices.

South Africa
We investigate the motivational role of food prices compared to other household expenses using South Africa as an empirical case.Since the mid 2000s, there has been a substantial increase in the number of unrest events in South Africa, usually referred to as "service delivery protests," and concern mainly the perceived failure of the government to provide basic services (De Juan and Wegner 2019; Lodge and Mottiar 2016).Despite the African National Congress (ANC) government's investments in services, they have been unable to overthrow the structural inequalities from the apartheid era and establish inclusive citizenship for its population (Alexander 2010;Beresford 2016;Booysen 2007;Dawson and Sinwell 2012;Desai 2002;Royeppen 2016;World Bank 2018).
Prior to 1990, political protest played a significant role in the overthrow of the apartheid regime and the transition to democracy.Unrest, therefore, tend to be viewed as a legitimate tool to achieve political aims (Alexander et al. 2018;Ballard, Habib and Valodia 2006;Bedasso and Obikili 2016;Friedman 2012;von Holdt 2013;Welsh 2000).Most collective action is peaceful, and engaging demonstrations and strikes is entrenched in the South African Constitution (Lancaster 2016).However, violent collective action has increasingly been seen as a tactic successful in achieving political outcomes (Bohler-Muller et al. 2017).Community protests represent both a continuation of the social movements on the apartheid era and a new form of resistance: "The recent wave of community protests resembles the earlier wave of social movements in that it is rooted in the country's poverty-stricken townships and informal settlements" (Paret 2017, 9).While these service delivery protests tend to be fragmented and concern immediate grievances among community members, the claim making on behalf of the state is at its core about wider issues relating to socioeconomic rights and exclusion (Brown 2015;Duncan 2016).In Johannesburg, these unrest events mainly occur in working class areas referred to as townships (Bond 2012), that traditionally have been ANC strongholds where the party still enjoys relatively high levels of support.Thus, unrest in the area is usually not understood or framed as an attempt to remove the ANC, but rather part of a wider repertoire of action to get attention and redress from the ruling party.Indeed, these events tend to erupt after several attempts by the community to engage with the local government (Dawson 2014;Runciman 2017).
We believe South Africa is a particularly interesting context in which to study the roles of both food and unrest, and focus on apartheid-era black and colored townships in Soweto, south west of central Johannesburg, and the black township of Alexandra, north of central Johannesburg.The chosen case of Alexandra and Soweto is suitable for several reasons.As in South Africa in general, there is widespread poverty and unemployment in both townships (Seekings and Natrass 2015).According to Statistics South Africa, almost 20% of South African households had inadequate access to food in 2017 (Statistics South Africa 2019), and more than 60% of households that experienced hunger in 2017 were found in urban areas (Statistics South Africa 2019).Thus, the survey does not focus on the elite or better-paid segments in society.That being said, Soweto is a large and diverse township with different levels of wealth that should not be oversimplified.The residents in Soweto are better off than other South Africans on several indicators, although the unemployment rate is higher: "The township is neither homogeneously squalid nor generally rich" (Ceruti 2013, 55).The township of Alexandra, on the other hand, being the oldest township and one of the poorest areas in South Africa, is known for its over-crowded informal housing and widespread poverty (Bonner and Nieftagodien 2008).Meanwhile, as with Soweto, Alexandra is also known for its political engagement, using tactics such as demonstrations and boycotts to raise awareness about specific issues.Thus, Alexandra tends to be central in the emergence of social movements (Curry 2012).However, while the townships experience frequent unrest, there is large variation in the type of tactics and repertoires chosen to engage with authorities.The citizens make use of a variety of different strategies to engage in politics, either through institutionalized and legal formal channels, or outside the scope of conventional politics, and political engagement in the townships is certainly not reduced to spontaneous eruptions of violence (Booysen 2007; Oldfield and Stokke 2006).Thus, the case represents significant variation in both food access and unrest participation.
For the purpose of the study we have chosen what we believe is a most likely case, where the priors suggest that the case is likely to fit the theory (Levy 2008).In other words, if there is an additional effect of food prices on the willingness to participate in unrest, the characteristics should imply that we are more likely to uncover such an effect here, as Johannesburg is marked by characteristics including widespread unemployment, relatively poor households and high levels of inequality in an urban setting.Previous research suggests that these characteristics are often present when food price hikes lead to unrest.The approach would therefore also consider whether food price spikes leading to unrest would be especially likely when other socioeconomic grievances are present.This provides an opportunity to study the effect of food prices on individual level willingness to engage in unrest.
Figure 1 shows that food prices have steadily increased in South Africa in the last decade, and has had a higher increase compared to the global trend, seen in Figure A1 in the Supplementary Appendix.With this steadily rising trend, it is likely that poor consumers are facing an increasingly challenging situation to manage expenses.

Research Design
The analysis makes use of a survey experiment to investigate whether the price of food has a unique effect on individuals' willingness to participate in unrest.Survey experiments have many advantages, such as strong internal validity in that they clearly distinguish between cause and effect by randomly assigning respondents to treatment and control conditions (see e.g.Gibson 2008).This is especially important given that food price and conflict is characterized by an endogenous relationship (Raleigh, Choi and Kniveton 2015).The random assignment of the vignettes should ensure that the treatment and control groups are comparable (Angrist and Pischke 2015).Equally important, as we would like to isolate the effect of one commodity price increase over another, it is challenging to make use of a research design using observational data in the traditional sense.Vignettes, however, provide a useful way to elicit respondent preferences, which are crucial given our research question.Such vignettes facilitate the simulation of real-world scenarios while controlling the amount and type of information provided, making it possible to distinguish between different scenarios that might otherwise be overlapping (Kinder and Palfrey 1993).The experiment has been designed to maximize ecological validity, by modeling situations that is highly familiar to the participants.Hence, although the stories are fictional, they resemble what the participants experience in their natural settings, which is likely to elicit realistic responses.
We conducted the interviews in June and July in 2017, and Figure 2 shows the trend of unrest events in recent years in the province of Gauteng, where Johannesburg is located.According to the Armed Conflict Location & Event Data project (ACLED), there were 118 unrest events in Gauteng municipality in 2017 (ACLED 2021).We can see from the figure that unrest levels were relatively lower in 2017, before increasing again after 2018.Events that occurred around the time of the survey were violent demonstrations in Alexandra reported in early July, where residents blockaded parts of the highway with burning tires and rocks.There were also instances of unrest during the summer in Soweto.For example, in Lenasia South community members covered the road with rocks and tires in July.In May there were protests that turned violent in Eldorado park, and the Dobsonville municipal office was set on fire in late June during unrest over service delivery.Also, in Diepkloof protesters were blocking roads and burning tires in several instances in May.While unrest events are relatively frequent in the area, to our knowledge there were no major violent events at the time that would significantly influence the survey responses.
We surveyed a total of 608 respondents from Soweto and Alexandra between the ages of 18 and 65. 5 The respondents were first asked survey questions on previous unrest participation, political attitudes and community affiliations and lived poverty and food security, and the vignette experiment came at the end of the survey.We applied a multistage sampling design to ensure representation of the demographics in the target areas, and the Primary Sampling Units (PSUs) included Alexandra, and Bram Fischerville, central east, central west, Devland, Diepkloof, Eldorado Park, Meadowlands, Orlando and Protea Glen in Soweto.The secondary sampling units (SSUs) were randomly selected based on census enumeration areas, which were assigned to two teams of fieldworkers, consisting of one team leader and four enumerators per team.The two teams started at a predetermined intersection nearest to each center point based on the random sampling of enumeration areas.The interviewers alternated between male and female respondents and applied a three-house interval pattern to select households.The surveys were conducted using Android tables through Google Forms.
We presented each respondent with a randomly selected, hypothetical video news item concerning living expenses.The experimental conditions included three treatments, and one control.The treatments differ with respect to the type of living expense that is projected to increase in the coming month.In the first vignette ("control"), the respondent is provided with information that living expenses will be stable in the following month.The second treatment ("fuel prices") provides information about expected fuel price increases in the coming month, while the third ("food prices") and fourth ("electricity prices") treatments provide information about expected food and electricity price increases, respectively.The respondents were shown video clips of a person reading hypothetical news stories. 6To reduce bias in presenting different videos, the same person read out the four different vignettes, where the text only differed in the type of commodity that was presented. 7Table 1 shows a comparison of the different treatment groups by age, income, lived poverty index and union membership.
After the participants saw the vignettes, they were asked to answer which of the following actions they would be most likely to take: 1. Attend a local Ward Committee meeting to express concern 2. Engage in a labor strike demanding higher wages 3. Attend a demonstration march to Johannesburg City Hall 4. Shut down traffic on local roads to draw attention to your situation 5. Do nothing We present a range of different tactics to the respondent to better tease out the preferred response to a given price hike.As it is not necessarily the case that unrest related to food only manifests itself through violence, we wanted to include a range of possible outcomes, from conventional peaceful channels to more contentious behavior.Also, while a specific tactic such as going on strike may be relevant for respondents with a steady income, this is not a viable option for someone who is unemployed.More specifically, these actions were given as options because they are common actions that South African citizens take to express grievances.
First, local Ward Committee meetings are the most local level of government interaction.Municipal Ward Representatives hold regular meetings to hear from their constituents.Second, engaging in labor strike to demand higher wages is a possible route for employed respondents to take in order to tackle higher prices.Both attending a labor strike and a Ward meeting are not expressions of unrest, but they are nevertheless forms of collective action.Next, a demonstration march to Johannesburg City Hall is a peaceful and legal method of collective action that citizens might take if they feel that their grievances are not being heard through institutionalized political channels.Finally, blocking roads with rubbish or burning tires has become a common method for citizens in lower income areas to draw attention to issues that are not being addressed.
A dependent variable that measures a person's willingness to engage in unrest, and not their actual participation, is agnostic about the costs and constraints of such participation.Certainly, respondents who are willing to engage in contentious action may not act upon their stated response (see e.g.von Uexkull, d'Errico and Jackson 2020).However, previous studies suggest that there is a link between attitudes and behavior concerning an object, and that " […] factors that lead to participation in violent behaviors are also likely to affect the formation of attitudes towards violent behavior" (Rustad 2016, 109).For example, Linke, Schutte and Buhaug (2015) find a positive relationship between people's attitudes toward violence and an increase in the incidence of violence.While the survey experiment is not able to capture actual participation of unrest as a result of increased living expenses, we can test the first part of the causal chain that captures the mobilization potential for unrest through increased prices.We aim to capture grievances by asking about the respondents' willingness to participate in different types of collective action, and this perceived grievance is a necessary precondition for engaging in unrest.
Another aspect of the survey is that the respondents are faced with scenarios of higher prices, not lived experiences of the same.It is different to feel the consequences of higher prices directly, compared to being x 2 -statistic = 37.56; p = 0.000 presented with a hypothetical situation.If the mechanism from food price to unrest, for example, involves a psycho-social reaction to hunger through aggression it cannot be recreated in hypothetical scenarios.However, it is likely that the respondent has experience with price hikes and the corresponding consequences from the past and is therefore able to relate to the vignette and its possible impact.8Given the challenges with conducting surveys with sensitive issues such as direct participation, this study offers an important contribution regarding how grievances and willingness for unrest participation form mobilization potential.Also, we would arguably nonetheless be able to capture whether food is seen as something especially important by the respondent.Finally, while a survey vignette experiment facilitates comparison of different household expenditures, asking about individuals' preferences regarding participation in various forms of unrest necessitates being mindful of ethical aspects (Eck 2011).Although the vignettes cannot capture emotions as a mechanism directly, the vignettes may trigger an emotional response.To reduce the risk of emotional stress, the vignettes resemble news reports that occur in South African media outlets.Also, the enumerators all had previous training and experience with implementing surveys at the University of Johannesburg, and were trained and supervised before the data collection effort began to minimize potential risks and to ensure the respondents' informed consent and anonymity.The respondents were informed that they were not required to take part in the interview, that participation was completely voluntary, that they had the right to decline to answer any particular question(s), that no identifying information was gathered, and that they could end the survey at any time.Table A1 in the Supplementary Appendix displays summary statistics.

Results
The responses regarding most likely action by treatment group are presented in Table 2.A tabular analysis indicates that the treatment condition and response are not statistically independent (x 2 -statistic = 37.56; p = 0.000).Thus, a given treatment influences the probability of a given response.
Because we presented the respondents with a menu of possible reactions, we use a multinomial logit regression to compare the willingness to engage in each type of activity when "treated" with an increase in prices of fuel, food, or electricity respectively, compared to the baseline control, stable prices.Table 3 presents the results of the analysis.Model 1 includes only the treatment conditions as categorical explanatory variables.Because  assignment to treatment is randomized, control variables for respondent characteristics are unnecessary and would not provide any additional insight into willingness to engage in unrest.However, as it is important to consider the possibility that the results could be influenced by how the experiment was administered, Model 2 includes the enumerator as a control variable.The results are reported as exponentiated coefficients and should be interpreted as relative risk ratios (RRR).An RRR of over 1 indicates that respondents are more likely to be willing to take the stated action over the baseline treatment condition of stable prices and an RRR below 1 indicates a less likely response.Thus, for a unit change in the predictor variable, from stable prices to a price increase, the relative risk ratio of the willingness to take action over doing nothing is expected to change by a factor of the given parameter estimate while other variables in the model are held constant.This means that we contrast outcomes (different price hikes) with a common reference point (stable prices), and the effects (relative change in predicted probability) is always in relation to the reference category.For example, in M1 a respondent presented with fuel price increase is 80.8% more likely to be willing to attend a Ward meeting than a respondent presented with stable prices.
Overall, the results indicate that respondents presented with any of the treatment conditions, either fuel, food or electricity price increases, are more likely to be willing to attend a demonstration march and to shut down traffic than those presented with the control condition, stable prices.Also, we see that individuals consider attending a Ward Committee meeting to be the most appropriate response to an increase in electricity rates.Rosenthal (2010) points to the Soweto Electricity Crisis Committee as an organization addressing grievances related to electricity in a local context in Johannesburg.Resistance against the state-owned power utility Eskom is widespread in the townships due to blackouts caused by failures on electricity networks, but also deliberate power cuts from the company in areas because of non-payment of electricity bills (Deklerk 2021;Ngqakamba 2019).Notably, in South Africa, municipalities set the rates electricity distributors may charge for electricity, but municipalities have no control over fuel prices, which are set at the national level, or food prices, which are controlled by the government.This finding indicates that respondents are making strategic decisions about the action most likely to achieve a desired outcome.Additionally, individuals facing fuel price increases seem to be most likely to attend a demonstration march.
To ease interpretation, we present an aggregated panel where we collapse the categories in the dependent variable into Do something versus Do nothing and display the results graphically.Figure 3 shows the predicted margins from the multinomial logistic regression model by treatment, while Figure 4 shows predicted margins by response.These graphs were generated based on the results from model 3 in Table A3 in the Supplementary Appendix. 9We can see from Figure 3 that the effect of both fuel and electricity price rate increase is significantly different from doing nothing, while there is no such effect concerning food prices.Also, in Figure 4 the predicted margins by response shows that the model predicts a higher probability in the willingness to engage in some form of response when prices increase, but the effect is not more profound for the price of food.
The results indicate that individuals are more willing to engage in unrest activity when faced with increasing prices compared to stable prices, regardless of whether it is electricity, fuel or food. 10We do not find 9 Figure A3 in the Supplementary Appendix shows the predicted margins for all five outcomes by treatment, and figure A4 shows the predicted margins by response, generated based on the results from model 2 in Table 3. 10 As two positive and significant coefficients may still be statistically distinguishable from another even if their confidence intervals overlap, we also formally test whether the predictors for fuel, food and evidence that food price is an especially potent driver for respondents' willingness to engage in unrest.Thus, H2 is confirmed, H1 is not.11Finally, to test H3 we consider the level of risk exposure by comparing those who would be more heavily impacted by increasing prices to those who would not.Food access may be relevant for mobilization potential when a fundamental need is threatened, where experienced food insecurity influences the willingness to engage in collective action due to increasing commodity prices regardless of source.To account for risk exposure, we asked the respondents "In the last 12 months, since July of last year, were you ever hungry but didn't eat because you couldn't afford enough food?" Figure 5 includes respondents who report not going hungry (N = 389), and Figure 6 includes respondents who report going hungry (N = 177). 12We can see from Figure 5 that the effect of both fuel, food and electricity price is not significantly different from doing nothing for those who report experiencing no hunger in the last year.In contrast, looking at those that do report experiencing hunger in the last year in Figure 6, both the effect of fuel and electricity is significantly different from doing nothing.Thus, price shock vulnerability may function as an contributing factor in the willingness to participate in unrest.For example, the widespread unrest in South Africa during the summer of 2021 erupted because of the jailing of Jacob Zuma.At the same time, the instability also led to unrest and looting "[…] by desperate people with little or no connection to Zupta, to secure food and basic necessities" (Reddy 2021, 2).Thus, while ignited by political developments, food insecurity played an important role for why many were motivated to join the unrest.Also, while positive, the effect of food price is not significantly different from doing nothing.There may be several reasons for this.
First, the group of people setting the prices for fuel and electricity is smaller and therefore may be more susceptible to collective pressures and more easily targeted in the South African context.For example, the South African Federation Trade Union organized a demonstration in August 2022 over the rising cost of living.Eskom had then applied for an increase in electricity prices during rising inflation (Africanews 2022).The demonstration targeted rising cost of living in general, but Eskom was pointed electricity are statistically different from one another for the collapsed outcome category (Do something/Do nothing), as well as each outcome separately in model 6-10 in Table A4 in the Supplementary Appendix.
out as one of the sources of grievances in that campaign.Rising cost of living may influence levels of food insecurity, and food insecure individuals engage in mobilization over rising costs that are not restricted to food price.
Second, the results suggest that a hike in prices can have a significant impact on the household economy, regardless of whether it is the cost of food, fuel, or electricity (Heltberg et al. 2013).The implication is that rising food prices could lead to unrest, but so could increasing costs of living from any source.The importance of food in mobilization potential is seemingly not because of its price, but because the lack of secure access to food influence the willingness to partake in unrest in general.What the issue vulnerable individuals choose to mobilize over may vary, where the role of organizational ties in a given context is likely important (Rudolfsen 2023).For example, the Soweto Electricity Crisis Committee has been central in mobilizing against electricity price hikes in Johannesburg in the past.Civil society organizations in the area and the degree of presence of government agencies to address specific issues may influence patterns of mobilization potential on the individual level.Depending on the local context, price hikes can function as a mobilization tool for political entrepreneurs such as opposition elites and other non-state stakeholders to address wider socioeconomic issues (Heslin 2021).
The analysis has sought to assess the relative importance of food in collective action.We find that higher cost of living increases the willingness to engage in unrest.We also find that respondents experiencing hunger have a higher willingness to engage in unrest over price hikes regardless of source, affecting the willingness to engage in unrest over increased cost of living in general.

Discussion
Can these results be applied elsewhere, or is this finding limited to the South African context?Certainly, South Africa has an exceptional history.The legacy of apartheid is still present, and South Africa is one of the most unequal countries in the world, where poverty and hunger is widespread (Chatterjee, Czajka and Gethin 2021;World Bank 2018).Also, as protest played a central part in the ending of apartheid, it has come to be seen as a legitimate channel to achieve political goals.However, we believe that the findings presented here are of value in other contexts as well.For example, the global commodity price spikes in 2007-08 corresponded with unrest in around 50 countries, and the latest price spike in 2011 contributed to a wave of unrest in a variety of countries, such as Egypt, Mexico, Kenya and Bangladesh (Abbott and Borot de Battisti 2011;Brinkman and Hendrix 2011).Also, the Russian invasion of Ukraine in February 2022 exacerbated supply shortages in global energy and food markets.These in turn led to price increases and food and energy insecurity for millions of people, and protests over food and fuel occurred in a wide range of countries which include, among others, Brazil, Colombia, Chile, Venezuela, Peru, Sri Lanka, Indonesia, Sudan, Mauritius, Tunisia, Morocco, Turkey, and Pakistan (Marsh, Rudolfsen and Aas Rustad 2022;Abay et al. 2023).
Rising commodity prices have varying impacts in different countries, depending on trade regimes, price controls and domestic market structures.However, "[w]hile these pass-through effects vary considerably, it is clear that higher inflation has become a major socio-political issue and has led to social unrest across all continents" (Wodon and Zaman 2008, 4).Volatility in prices of commodities such as food and fuel are not unique to the South African context, nor is the use of unconventional political channels to raise issues and draw attention to grievances.The results therefore are likely to have relevance beyond the townships in Johannesburg, especially to other urban areas in middle-income countries that experience a high degree of socioeconomic inequality and widespread unemployment.However, to be able to assess the scope of the findings, it is necessary to investigate the relationship beyond the South African context.
With regards to the design, experiments are constructs that to a larger or lesser extent resemble reality, and there is a trade-off in terms of validity.While the random assignment of respondents to treatment and control facilitates causal inference and renders a high degree of internal validity, external validity may be a concern if the treatment does not in reality resemble the relevant phenomenon in question (Barabas and Jerit 2010).We argue that by constructing vignettes based on news reports, we present the respondents with something that is familiar, as media sources and politicians inform the South African public about price increases on different commodities frequently.The experiment has been designed to maximize ecological validity, by modeling situations that is highly familiar to the participants.Hence, although the stories are fictional, they resemble what the participants experience in their natural settings, which is likely to elicit realistic responses.
Another potential threat to inference is that the experimental condition exaggerates the stimuli of the treatment by introducing an atypical manipulation that rarely occurs in a natural setting (Kinder and Palfrey 1993).Although the concern that the clean environment in which the treatment is given could introduce a positive bias in the results, an arguably bigger concern for our study is the threat of contamination of the experimental setting by prior exposure.Meaning that, in the likely event that the respondents have been exposed to news items about price hikes in the past, this may produce an overly conservative estimate.However, the experiment strives to simulate events that are likely to occur in the real word: "[…] either there is a likelihood of contamination from real-world experience or the survey experiment explores a nonexistent or politically irrelevant phenomenon" (Gaines, Kuklinski and Quirk 2007, 12).Therefore, the estimates may have a negative bias, which comes at the cost of studying phenomena that have real-world relevance.
Also, we aimed to include commodities that are relevant for the consumer (e.g.expenses they are likely to spend money on), as we sought to capture the relative importance of food compared to other expenses that make up a large share of the household budget.However, in the current research design the treatments may be too thematically similar in how consumers weigh the impact of a price shock for each commodity to be able to pick up differences in effects.The somewhat differing levels of specificity in the vignettes may also influence our findings.Whereas fuel and electricity are specific commodities, staple prices might be perceived as more abstract by not referring to a specific commodity.However, while it can be challenging to assess how respondents perceive the different vignettes, we found it more coherent to refer to staple food in general than to choose a specific food commodity that might not be relevant for a given respondent.In addition, the presented commodity scenarios are unlikely to occur only in isolation, where commodities such as food and fuel can impact each other and rise simultaneously, and increasing prices of both has led to protests in the past (McCulloch et al. 2022).For example, higher fuel prices can negatively affect economic status by increasing cooking, heating and transportation costs that in turn increases prices of other basic commodities, including food (Boyd-Swan and Herbst 2012).
To assess whether these factors influence the results would require knowledge on the extent of previous news exposure by the respondents and the effect of a potential exaggeration of the treatment in a survey context, both of which are impossible to assess.As the pool of respondents remains and we only change the type of commodity, we are nonetheless able to consider and compare the relative impact of the different commodities on the willingness to engage in unrest, regardless if there is general upward or downward bias in the sample.While experiments can be more challenging to generalize from, we avoid spurious correlation as the random sampling allows us to estimate the average causal effect of the treatment in the population.However, it is still important to be explicit and keep these potential biases in mind when conducting a survey experiment.An avenue for future research would be to compare basic necessities with other commodities not defined as such, as well as measuring directly how basic commodities may influence each other and out a strain on the affordability on other basic commodities.

Conclusion
This article has sought to investigate the role of food as a driver of unrest relative to other living expenses using a survey experiment conducted in the areas of Soweto and Alexandra in Johannesburg, South Africa.The results suggest that individuals are more willing to engage in unrest activity when presented with the prospect of increasing commodity prices, regardless of whether it concerned food, fuel, or electricity, than when presented with the prospect of stable prices.Also, the findings presented here suggests that those with a higher risk exposure to price hikes have a higher willingness to engage in unrest over commodity price hikes than those who do not.This indicates that food prices are not more likely to cause unrest compared to other commodities, while vulnerability to price hikes, however, predicts higher willingness to engage in unrest regardless of shock.Seemingly, for those who are most affected by a price hike it is less important which commodity it is, what matters is whether it introduces further strain on an already hard-stretched budget.
Governments, international organizations and policy agencies are increasingly paying attention to the consequences of higher inflation and increasing food insecurity in an urbanized setting, where a growing urban population spend a substantial share of their household budget on food.Also, it is evident that individual-level vulnerability to price hikes increases the willingness to engage in unrest.Therefore, the societal impacts that may result from the mounting challenges of increasing urbanization and impact of rising living costs must be taken seriously and directly addressed in policy making, also in terms of its impact on social instability.This implies proactive and targeted policies that aim to stabilize and reduce prices of basic commodities.While governments are faced with difficult trade-offs in tackling higher prices and inflation pressures, this study emphasizes the need for tailored safety nets and subsidies that explicitly target the poor.

Figure 3 .
Figure 3. Predicted margins by treatment: do nothing vs. do something.

Figure 4 .
Figure 4. Predicted margins by response: do nothing vs. do something.

Figure 5 .
Figure 5. respondent reporting no hunger: do nothing vs. do something.

Figure 6 .
Figure 6.respondent reporting hunger: do nothing vs. do something.

Table 1 .
treatment group comparison.

Table 2 .
response by treatment group.