Responding to Catastrophe

Motives for Repression

We begin with a brief overview of the motives for regime repression. As we articulate in greater detail below, natural disasters create conditions that previous research suggests should increase the regime’s incentives to repress the population, particularly in those areas most affected by the disaster. While repression is a common feature of all states, its severity varies significantly across states, often as a function of largely static factors such as regime type and development level (Davenport and Armstrong 2004; Bueno de Mesquita et al. 2005; Poe and Tate 1994). Repression is usually relatively stable within a state from one time point to the next, suggesting that it assumes a state of punctuated equilibrium ³ in which some (largely unanticipated) shock leads to changes in regime behavior. Consequently, the central question is less whether repression occurs, but rather what factors drive changes in repression within a state from one time point to another.

From a decision theoretic perspective, an incumbent’s decision to alter the level of repression at a given time point is a function of the perceived ratio of its strength to that of potential challengers (Poe and Tate 1994; Poe 2004). In brief, as the perceived strength of threats increases, the incumbent is expected to increasingly resort to repression in order to (re)assert its authority. This and other recent research highlight the impact of both endogenous events and exogenous shocks on state repressive behavior. Most notably, overt challenges to the economic, social, or political status quo often provoke increased repression. Incumbents respond to behavioral challenges such as protests and strikes by increasing their reliance on coercive strategies such as arbitrary detentions, physical abuse, disappearances, and killings of dissident activists (e.g., Davenport 1995; Moore 2000; Rasler 1996). This relationship deepens as challenges to the state increase—a process that scholars often refer to as the “law of coercive responsiveness” (see Davenport 2007a). Repression reaches it acme during periods of violence dissent, rebellions, and insurrections. Indeed, violent dissent and internal political conflict can weaken the constraints that factors such as democratic institutions impose on the incumbent’s use of repression (Davenport 2007b). Taken together, the literature strongly suggests that repression increases when regime authority is challenged or when incumbents are unable to exert control over their territory and their citizens.

Other recent studies extend this core idea to examine the manner in which other types of “shocks” provoke spikes in repression. Unanticipated changes in the social or political status quo threaten regime stability by adversely impacting the incumbent’s ability to maintain the support of coalition members or by increasing the relative power (e.g., threat) of opposition groups. For instance, the imposition of economic sanctions contributes to increased repression in the target state by eroding regime control, emboldening political opposition, and constraining the ability of the incumbent to provide the resources necessary to prevent defections from the winning coalition (Wood 2008; see also Peksen 2009). In a related manner, rapid demographic changes such as the rapid expansion of the youth population (particularly males) within a society can increase regime threat perception and provoke repression or large-scale political unrest (e.g., Nordas and Davenport 2013; Urdal 2006).

Disasters and State Repression

Severe natural disasters have the potential to significantly alter the political status quo within a state. In some cases, these events produce positive changes in the political and social landscapes, such as the 2004 Indian Ocean Tsunami contributing to the termination of the secessionist conflict in Indonesia (Beardsley and McQuinn 2009; Enia 2008; Le Billon and Waizenegger 2007). ⁴ Yet, on balance, both case-specific and cross-national evidence suggest that natural disasters often create conditions of instability, dissent, and possibly violence. As we noted previously, the 2004 Tsunami also reignited violent conflict in Sri Lanka, and natural disasters have been implicated in the onset or escalation of civil violence in Nicaragua (1972), Guatemala (1976), and Egypt (1992) among others. While several recent analyses link rapid-onset disasters to civil conflict, we argue that these disasters often contribute to rising repression for two reasons. First, severe disasters provide a focal point for popular grievances and mobilization against the regime. Second, state policing capacity and sanctioning ability decline following rapid-onset disasters, which itself facilitates challenges to the regime and creates opportunities for violence. These factors, in turn, contribute to the escalation of regime repression.

Existing literature on disaster and conflict point to the manner in which disasters create conditions of scarcity and thus lead to competition (often violent competition) over increasingly scarce resources (Brancati 2007; Nel and Righarts 2008; see also Homer-Dixon 2001). ⁵ Natural disasters also serve as a focal event around which a government’s performance can translate to an evaluation of the legitimacy of the regime (Olson and Gawronksi 2010; Pelling and Dill 2010; Poggione et al. 2012). Disasters shift public focus from how well a government provides for higher-order needs to how well it provides basic needs like food, shelter, and human security. Developed states are not immune to disaster-related violence and instability, as the case of Hurricane Katrina in the United States illustrates. Even in decentralized political systems, the focus for political attention can shift from local governments to the national government (Schneider 2008). Disasters also have the potential to create a lens through which political actors examine preexisting cleavages within the state and serve to highlight the unequal distribution of state resources among groups, particularly where the disaster produces large-scale human suffering and state response is uneven (Pelling and Dill 2010). By creating conditions of scarcity and highlighting government deficiencies, disasters produce popular discontent with the regime and exacerbate existing grievances and factionalism within society (Le Billon and Waizenegger 2007).

While grievances may be pervasive within a society, disasters can facilitate the opening of political windows of opportunity for their expression—both those generated by the disaster itself and those that precede the disaster. Previous research on contentious politics and social movements suggests that structural and political changes within a polity can promote popular mobilization. Rapid mobilization often occurs in the context of temporary reductions in the state’s repressive capacity (e.g., Brockett 1991; McAdam 1996, 29-30). Natural disasters stretch state capacity and create space for resistance to state rule (Gawronski and Olson 2013), leading to both violent (Berrebi and Otswald 2011; Nel and Righarts 2008) and nonviolent challenges (Bueno de Mesquita and Smith 2010; Nardulli, Peyton, and Bajjalieh 2015; Olson and Drury 1997; Drury and Olson 1998) to the state. ⁶ Following large-scale disasters, such challenges to state authority may also undermine regime stability and threaten incumbent survival (Quiroz Flores and Smith 2013).

To the extent that disasters create opportunities for both violent and nonviolent challenges to the state, they are also likely to provoke coercive responses by the state. Moreover, as state control diminishes, the incentive for employing coercive violence against real and perceived threats is likely to increase. It is important to note that the capacity to effectively police the population and suppress dissent is not necessarily the same as the regime’s ability to deploy violence against perceived threats. Indeed, as Hannah Arendt (1970, 56) observes, violence often emerges where power is threatened. As legitimate authority and state control collapses, repressive violence increasingly becomes the default strategy adopted by the state. Severe disasters hinder the regime’s ability to communicate and coordinate policing and security efforts (Le Billon and Waizenegger 2007). As such, the regime is increasingly unable to effectively monitor potential threats and to preemptively intervene to prevent them from growing into stronger challenges. As the severity of a disaster increases (both in costs and geographic range), the policing and monitoring capabilities of the state can quickly erode. Previous research suggests that this is central to the onset of organized dissent and anti-regime violence.

As the ability to monitor opposition declines, incumbents increasingly rely on more repressive responses in an attempt to maintain authority and deter emerging threats. Such threats might include violent opposition, nonviolent dissent, or the emergence of other actors that supplant (or attempt to supplant) state authority. For instance, the 1999 Marmara earthquake in Turkey killed upward of 20,000 people and eroded the regime’s ability to exert effective control over the area. Moreover, the government’s disaster response was initially slow, which provided an opportunity for civil society organizations to fill the relief void, thereby usurping authority normally reserved by the state. In this context, the delivery of aid and provision of services to citizens impacted by the disaster became highly politicized, as it highlighted state weakness and elevated the status of civil society organizations. The ascendance of these actors in turn created a political crisis for the state. Public criticism of the state’s response led to increasing state hostility toward civil society groups, and the government began to target specific groups for repression and in some cases seized their assets (Pelling and Dill 2010).

The October 1992 Cairo earthquake also illustrates this dynamic, although in a more dramatic manner. The earthquake led to the deaths of over 500 people, injured 10,000, and rendered more than 40,000 persons homeless. As was the case in Turkey, the Egyptian government was initially slow to react to the disaster while civil society organizations—including the Muslim Brotherhood and others Islamist groups—responded immediately. According to one observer, “the government was totally paralyzed. [President Hosni] Mubarak was traveling abroad, and for two days the regime did absolutely nothing, nothing at all. Within hours, though, the Islamists were on the streets—with tents, with blankets, with food, with alternative housing” (emphasis in original; Berman 2003, 260-61). Arguably, the response was an effort by the Islamists to establish their own legitimacy while delegitimizing the regime—a strategy that appeared successful as popular support for these groups increased in the wake of the disaster (Berman 2003; Kepel 2002, 277, 293-94). The Egyptian government responded by cracking down on nongovernmental groups providing aid. The regime invoked military decrees allowing it to freeze any funding from outside the country directed to nongovernmental organizations (NGOs) active in the relief effort (Singerman 2002). Throughout the remaining months of 1992 and through the following year, the Mubarak regime intensified repression of civil society groups, particularly Islamist organizations. In December, the police initiated an intensive search-and-arrest campaign in Imbama, one of the areas hardest hit by the earthquake and stronghold of Islamist groups. The weeks-long campaign resulted in the detention of more than 700 people by the end of the year. Within the first two months of 1993, security forces detained an additional 300 supporters of Islamist groups. ⁷ Consistent with the argument presented above, the disaster increased grievances within the affected area and created opportunities for rival actors to supplant state authority. Facing these threats, the regime responded by escalating repression.

As these examples suggest, repression is likely to increase in the wake of disasters. Specifically, we argue repression becomes more likely when leaders face the threatening combination of rising popular dissent and reduced policing and sanctioning capacity in the wake of natural disasters. By providing both a focal point for opposition and reducing government authority and policing capacity in affected areas, natural disasters lead to an overall weakening of state control. Declining control in turn increases incentives for incumbents to repress real or perceived threats. This produces our central testable hypothesis.

Hypothesis 1: State-sponsored repression increases in the wake of natural disasters.

Global Response and Domestic Institutions

The international community often responds to large-scale disasters by contributing substantial financial aid and other assistance to disaster-affected areas. Previous studies find that disaster aid may exacerbate frictions and political competition among rival social groups within the state as well as between the state and society (Beardsley and McQuinn 2009; Cohen and Werker 2008). Other recent studies similarly find that inflows of humanitarian aid to unstable political environments may promote various forms of organized political violence (Hoffman 2004; Nunn and Qian 2014). In line with these recent studies and the general argument we presented above, we believe that post-disaster humanitarian aid may also influence the likelihood that states respond to disaster-related instability with coercive repression. On one hand, aid inflows may augment the government’s ability to respond to disasters, alleviating grievances and thus reducing incentives for repression. Alternatively, leaders may use aid for personal enrichment or divert it to key constituencies rather than directing it to the most needy communities. In this case, aid inflows may increase tensions and potentially contribute to increased repression. ⁸ We contend that institutional features of the recipient state likely condition the efficiency and effectiveness of aid, and by extension the relationship between aid and repression.

Prior research has consistently noted a strong relationship between the presence of democratic political institutions and constraints on state repression (Davenport and Armstrong 2004; Bueno de Mesquita et al. 2005). While repression exists across states, all else equal, democratic regimes tend to be less repressive than their more autocratic counterparts. That said, previous studies suggest that democratic regimes are willing to (and quite capable of) employ coercive repression when they face high levels of instability and violent internal dissent (Davenport 2007b). As we discussed above, large-scale disasters create conditions that may facilitate dissent and instability. As the events in Turkey after the Marmara Earthquake and the United States following Hurricane Katrina demonstrate, democracies are not immune to rising repression in the wake of severe disasters that overwhelm state resources and policing capacity.

The constraints on repression found in democracies are unlikely to disappear following disasters. However, like other conditions that create significant instability, disasters may weaken these constraints, leading democracies to increase repression above normal level. In democratic systems, particularly severe natural disasters suggest a failure of leadership, which in turn leads to popular dissent and pressures for leader ouster (Achen and Bartels 2004; Gasper and Reeves 2011). Indeed, recent survey research suggests that as the destructiveness of disasters increases, perceptions of regime legitimacy decline, dissent increases, and support for leader removal through nondemocratic channels such as coups increases (Carlin, Love, and Zachmeister 2014). Recent research demonstrates that severe disasters are more likely to produce large-scale dissent and instability in more democratic systems and by extension increase the likelihood of leader removal (Quiroz Flores and Smith 2013). This effect is likely compounded where the regime is unable to rapidly and effectively respond to the damage and displacement created by the disaster. Severe disasters may therefore produce particularly acute challenges for democratic states.

Disasters appear no more frequent in autocracies compared to democracies but impose a more substantial toll (Quiroz Flores and Smith 2013; Stromberg 2007). Despite this, citizens in autocratic states are comparatively less likely to engage in dissent activities (Quiroz Flores and Smith 2013). This occurs in part because preexisting regime repression and expectations of future repression are likely to condition public responses. In autocratic states, public dissent—both violent and nonviolent—are likely to be met with disproportionate repression. Moreover, autocratic regimes are often more easily able to preempt popular mobilization by employing tactics generally unavailable to democratic regimes. For example, following Cyclone Nargis in 2008, the Burmese junta prevented citizens from gathering aid sites and closed makeshift camps for survivors, effectively limiting the likelihood that such locations could become sites for anti-regime activity (Quiroz Flores and Smith 2013, 7; Larkin 2010). Consequently, during the period immediately following disasters, democracies are more likely to escalate repression in response to dissent while autocracies are more likely to escalate repression in order to deter dissent.

How a government responds to these challenges is partially determined both by the severity of the threat and the resources that are available to the regime. Foreign aid represents a key resource potentially available to incumbents after natural disasters strike. While typically intended as a humanitarian tool and often driven by the severity of the disaster, factors such as donors’ strategic interests, existing relationships among states, and the intensity of media coverage influence the levels of humanitarian aid committed to a state in the wake of a disaster (Drury, Olson, and Van Belle 2005; Kevlihan, DeRouen, and Biglaiser 2014; Stromberg 2007). Regardless of the specific motives of donors, inflows of foreign aid have the potential to shape domestic political interactions with recipient states. ⁹ Notably, aid—like other forms of nontax revenue—influences relationships between incumbent leaders and their supporters, affects leader tenure, and exacerbates or alleviates disparities among groups within the state. Incumbent leaders often use nontax revues generated by natural resources or foreign aid to pursue personalistic goals and retain their positions of power rather than for the provision of public goods (Wright and Winters 2010). Recent studies suggest that this strategy often succeeds, and aid is an effective tool through which incumbent leaders maintain power (Bueno de Mesquita and Smith 2009; Kono and Montinolla 2009). Disaster aid is no different in this respect. Like other aid resources, disaster assistance creates incentives for corruption and manipulation and creates conditions of moral hazard (Cohen and Werker 2008).

Extant political institutions heavily influence the impact of aid. Democratic leaders are not necessarily more benevolent by nature. Whereas the survival of autocratic leaders hinges on the continued flow of resources to a small group of supporters, the survival of democrats is contingent on satisfying the demands of a large winning coalition (Bueno de Mesquita et al. 2003). Consequently, autocratic incumbents are comparatively more likely to treat aid as a personal resource to stockpile and to use it as a tool to pay off supporters while democratic leaders are more likely to use aid to invest in public goods (Bueno de Mesquita and Smith 2009; Kono and Montinolla 2009). ¹⁰ This is particularly true during periods of unrest and political instability. As potential threats to the regime increase, autocratic leaders are increasingly likely to forgo deploying foreign aid to produce public goods and to instead use aid for private consumption (Wright 2008). The availability of these resources therefore allows leaders in more autocratic systems to suppress dissent and better ensure their survival (Bueno de Mesquita and Smith 2010). In the wake of natural disasters, autocrats can therefore successfully use disaster aid to increase their political control by suppressing dissent through both repression and private payoffs. By contrast, leaders of states that are more democratic are more likely to utilize aid to alleviate grievances created by the disaster and to compensate for temporary deteriorations in policing capacity produced by the disaster.

This discussion suggests that the different approaches autocratic and democratic leaders take toward the management of aid resources in the wake of disasters may also shape the repressive policies that leaders employ to maintain or reassert control over the population. As Wright (2008) notes, during periods of instability, autocrats increasingly devote aid resources to private payoff of political challengers and the repression of dissident threats. Similarly, Conrad and DeMeritt’s (2013) recent study highlights the manner in which the presence of democratic institutions moderates the relationship between natural resource rents and state repression. In the same vein, we argue that after disasters, aid inflows exert differential influences on the levels of repression employed by democracies and nondemocracies. The successful and efficient deployment of aid following disasters should help quell popular dissatisfaction, thereby reducing dissident challenges the state and minimizing both opportunities and incentives for the state to engage in coercive repression. By contrast, where incumbents allocate aid resources to regime supporters while denying assistance to large numbers of aggrieved persons, or where the government’s response is viewed as ineffectual, grievances are likely to increase, both increasing dissent and increasing the odds of either preemptive or responsive repression. Indeed, where aid inflows are large, and public knowledge of international assistance is high, corruption and aid misallocation may exacerbate public grievances. Similarly, if specific groups within a state receive disproportionate benefits from this aid, social tensions are likely to increase. In both cases, regime repression is likely to increase in response.

This argument suggests that the distinct institutional features associated with democracies and autocracies condition the manner in which the incumbent deploys aid and consequently the likelihood that aid influences regime repression following disasters. Thus, disaster aid is likely to condition the relationship between disaster severity and repression, but the conditioning effect differs across regime types. More specifically, it suggests that democratic states are less likely to escalate repression in the wake of disasters when they receive large inflows of foreign disaster relief. By contrast, this aid should have little positive impact (and potentially a negative one) on repression in post-disaster autocracies.

Hypothesis 2: Aid mitigates the impact that disasters have on repression in democratic states.

Data

We conduct our analysis on a sample of approximately 166 states for which we have yearly data on both state-sponsored repression and the occurrence and severity of natural disasters. After accounting for missing observations and data limitation for some control variables, this produces a sample with 4,738 country-year observations spanning the years 1977 to 2009. Our hypotheses predict that state-sponsored repression increases in response to the severity of natural disasters. Our dependent variable, Repression, is taken from the Political Terror Scale (PTS; Wood and Gibney 2010), which explicitly captures state-sponsored abuses of physical integrity rights such as extrajudicial killings, disappearances, torture and beatings, and political imprisonment and arbitrary detentions within a state’s borders. PTS scores are based on information obtained from annual human rights reports published by Amnesty International and the US Department of State (USDS). The resulting scale is a five-point ordinal measure of physical integrity violations within a country in a given year. Countries receiving a score of 1 are under the secure rule of law and rarely engage in observable acts of abuse against citizens (e.g., Canada or Finland most years) while countries scoring a 5 engage in widespread, systematic abuses of these (e.g., Syria in recent years). We present results from models using PTS scores based on information from the annual reports of the USDS but include results using the Amnesty International-based scores in the online appendix. ¹¹

Data on natural disasters are taken from the Center for Research on the Epidemiology of Disasters (CRED) Emergency Events Database (EM-DAT) International Disaster Database (CRED 2012). For inclusion in this data set, a disaster must meet one of the following criteria: (1) ten or more persons reported killed, one hundred or more persons reported injured; (2) the declaration of a state of emergency; or (3) a call for international assistance. Not all types of disasters are relevant to our analysis. Our argument focuses on the manner in which severe, unanticipated, rapidly occurring disasters impact state repression. It does not address the influence of disasters related to climatic changes such as droughts, which are typically slower to emerge and provide greater time for governments to respond and the population to adapt. We therefore limit our analysis to rapid-onset disasters, including earthquakes, floods, tropical storms, hurricanes, cyclones, and volcanic eruptions. ¹² Even with these inclusion criteria, disasters within the data set vary tremendously in terms of their intensity and the severity of the damage they inflict. Because our theoretical argument focuses on the manner in which disasters produce public dissatisfaction as well as declining state policing capacity, we require a variable that captures variation in the severity of disasters rather than simply its occurrence. To account for disaster severity, we rely on EM-DAT’s estimate of Affected persons, ¹³ which it defines as “people requiring immediate assistance during a period of emergency, i.e. requiring basic survival needs such as food, water, shelter, sanitation, and immediate medical assistance.” ¹⁴ We log transform these values in order to reduce the influence of states with extremely large populations and to account for the possibility that disaster severity exerts a nonlinear influence on repression.

Previous studies identify a set of robust control variables for use in the analysis (e.g., Poe and Tate 1994; Davenport 1995; Davenport and Armstrong 2004). Given the strength of the expected relationship between challenges to the regime and repressive responses, we control for both the number of antigovernment protests and the presence of armed civil conflict within the state producing at least twenty-five battle-related deaths during the year. Data on antigovernment protests come from Banks (2005) while data on civil conflicts are taken from the Uppsala Conflict Data Program/Peace Research Institute Oslo Armed Conflict Data set v.4-2014 (Themner and Wallenesteen 2014). Each is expected to increase the likelihood of state repression. We also include measures of a state’s per capita gross domestic product (GDPpc) and Population size (K. S. Gleditsch 2002). Previous studies also demonstrate that more democratic states are more likely to respect the physical integrity rights of the citizens compared to autocratic states. We proxy the level of Institutional Democracy with the twenty-one-point Polity2 measure from the Polity IV data set (Marshall, Jaggers, and Gurr 2011). For ease of analysis, we rescale the indicator by adding 10 to the −10 to 10 Polity2 measure in order to create a variable ranging from 0 to 20.

We also control for the level of Humanitarian Aid flowing into a state. We scale the value of this aid to the size of the recipient state economy (% GDP). Unfortunately, available sources do not explicitly differentiate disaster aid from other types of humanitarian assistance. Our variable therefore includes aid flows to cases in which no disasters occurred. This includes aid such as emergency relief, material assistance, emergency health services, food aid, relief coordination, and reconstruction relief. ¹⁵ To aid interpretation, we rescaled the aid variable to convert percentages into whole numbers by multiplying it by 100 for the main analyses. ¹⁶ These data come from the recent Aid Data (2.0) data set (Tierney et al. 2011). We rely on the interaction of this measure and the Institutional Democracy scale to test the contingent relationship described in Hypothesis 2.

Because newer regimes are perhaps more likely to face dissent and instability and to respond with coercive repression, we also include a control for regime Durability, which we take from the Polity IV data set mentioned above. Finally, we control for the Cold War period in order to account for possible bias in the original source material used to construct the PTS measures (Poe, Carey, and Vazquez 2001) as well as possible variation in disaster reporting across the two time periods. We code years prior to 1991 as Cold War years.

Model and Results

We rely on ordered probit models to evaluate the hypotheses posited above. The ordered probit model is a generalization of the common binary probit model that allows us to model the likelihood of observing one of multiple possible discrete outcomes on an ordinal scale. We choose this statistical method because the PTS is a five-category scale in which each category represents the observed level of state repression in a calendar year. In all models, standard errors are clustered on the country to account for within unit correlation. Finally, because we have panel data, we must consider the effects of temporal dependence within these panels. We therefore employ a series of binary lags of our dependent variable, where each dummy represents a category of the dependent variable (excluding the highest category as the reference). ¹⁷

We report our results in Table 1. According to the results presented in model 1, the logged value of Affected is positively and significantly related to increasing state repression. While some states rarely experience large-scale disasters, others experience them quite frequently, and it is possible that states prone to natural disasters simply possess characteristics such as low-state capacity or chronic underdevelopment that increase their reliance on repression. In model 2, we therefore restrict the sample to only cases that experienced a rapid onset natural disaster during the year. Restricting the sample in this ways allow us to isolate the influence of disaster severity on regime repression among states experiencing disasters and also helps us to distinguish disaster-related aid from other types of humanitarian assistance. As with the previous models, disaster severity is correlated with an increased likelihood of repression among states that experienced disasters. Interestingly, humanitarian assistance as a percentage of GDP appears to have little independent influence on state repression in either the full or restricted sample. ¹⁸ These results provide evidence of the general relationship we posited in Hypothesis 1.

OPEN IN VIEWER

Table 1.

Ordered Probit Analysis.

	Model 1	Model 2	Model 3	Model 4	Model 5
	Full sample	Disaster-years	Disaster-years	Demo cracies	Non democracies
Affecteda	0.009** (0.005)	0.012** (0.006)	0.012** (0.006)	0.018** (0.008)	−0.002 (0.006)
Disaster aidb	0.009 (0.024)	0.002 (0.041)	0.149** (0.071)	0.008 (0.021)	0.041 (0.032)
Institutional democracy	−0.039*** (0.005)	−0.033*** (0.007)	−0.032*** (0.007)
Democracy × disaster aid			−0.010** (0.005)
Affecteda × disaster aid				−0.006*** (0.002)	0.008† (0.005)
Populationa	0.091*** (0.021)	0.077*** (0.027)	0.078*** (0.027)	0.078*** (0.041)	0.123*** (0.021)
GDP per capitaa	−0.167*** (0.024)	−0.141*** (0.036)	−0.140*** (0.036)	−0.210*** (0.049)	−0.130*** (0.031)
Antigovernment demonstrations	0.051*** (0.013)	0.045*** (0.014)	0.046*** (0.014)	0.050* (0.026)	0.043*** (0.012)
Regime durabilitya	−0.056** (0.023)	−0.077*** (0.030)	−0.076** (0.030)	−0.071* (0.039)	0.003 (0.027)
Civil conflict	0.904*** (0.082)	0.772*** (0.100)	0.765*** (0.100)	1.008*** (0.167)	0.865*** (0.083)
Cold War	−0.392*** (0.050)	−0.401*** (0.078)	−0.397*** (0.078)	−0.568*** (0.096)	−0.233*** (0.063)
Wald χ2	2,094.39***	1,110.27***	1,113.10***	850.16***	1,175.34***
Observations	4,738	2,295	2,295	2,078	2,660

Note: Coefficient estimates with standard errors (clustered on country) in parentheses.

^a Natural log.

^b % gross domestic product (GDP).

*p ≤ 0.10. **p ≤ 0.05. ***p ≤ 0.01. ^† p = .107 (all p values are two-tailed).

In order to facilitate substantive interpretation of these results, we predicted probabilities for the relationship identified in model 1 and present them in Figure 1. In the figure, the X-axis reflects the severity of the disaster as measured by the logged number of Affected persons while the Y-axis represents the estimated probability that the state received a score of 3 on the PTS index (a moderate level of repression), given that it had received a score of 2 (a relatively low level of repression) during the previous year. Put otherwise, the figure shows the predicted probability that state repression increases by one category from the previous year as a function of the severity of disasters. For instance, all else equal, the probability of a one-unit increase in repression from the previous year to the current is approximately 25 percent when Affected equals zero. However, following a disaster that adversely affects approximately 20,000 to 50,000 persons (between about 10 and 11 on the logged scale), the probability of a one-unit increase in repression increases to about 28 or 29 percent. This reflects roughly a 12 to 16 percent change in the risk of repression escalation.

OPEN IN VIEWER

Figure 1.

Effect of disaster severity on repression escalation.

In addition to the general argument, we argue that aid can influence repression after disasters by allowing the government to respond to public grievances and compensate for the diminished capacity created by the disaster. However, this relationship is likely to vary across regime type. Specifically, we posit that aid moderates the impact of disasters on repression in more democratic states. This implies a multiplicative relationship between the relevant variables. As such, we examine the interactive effects identified in models 3 through 5. Model 3 presents results for the effect of Institutional Democracy on the relationship between Humanitarian Aid and Repression within the sample of states that experienced a disaster during the year. ¹⁹ Models 4 and 5 limit the sample to the subsets of democracies and nondemocracies, respectively, and interact the continuous Humanitarian Aid and Affected variables in order to better isolate the impact of foreign assistance provided in the wake of disasters on repression in each type of state. ²⁰ Because neither the coefficient values for the interaction terms nor their constituent terms independently provide useful information regarding the statistical or substantive impact of the relationship (Brambor, Clark, and Golder 2006; Braumoeller 2004), we present the marginal effects of these relationships in Figures 2 and 3. ²¹ The panels present the predicted change in likelihood of repression escalation (Y-axis) for a unit increase in the specified predictor over the range of the constituent term (X-axis).

OPEN IN VIEWER

Figure 2.

Conditional effect of democratic change by disaster aid.

OPEN IN VIEWER

Figure 3.

Conditional effect of aid by disaster severity: democracies versus nondemocracies. Note: Left panel calculated from model 4, Table 1. Right panel calculated from model 5, Table 1. To aid interpretation, the aid variable is converted to percentage points. Each panel displays the conditional effect of a one percentage point increase in humanitarian aid on the likelihood of increasing repression by one category over the range of the number of persons affected by natural disasters in a given year.

Figure 2 illustrates the conditioning impact of Institutional Democracy on the relationship between Humanitarian Aid and regime Repression. Specifically, the graph shows the impact of a one-unit increase in Institutional Democracy at varying levels of aid. As the panel illustrates, as a state becomes more democratic, aid exerts an increasingly negative impact on the likelihood of regime repression. For example, an average one-unit increase in the democracy measure reduces the likelihood of repression escalation by approximately 1.5 percent when the disaster-affected state receives aid equivalent to 1 percent of GDP. Consequently, a ten-unit increase in democracy—roughly the difference between a highly autocratic state and a nominally democratic state—would reduce the likelihood of repression escalation by 15 percent. Moreover, this suppressive effect grows as the volume of aid increases. This provides support for our second hypothesis by showing that more democratic states are less likely to use repression when they receive significant inflows of aid following natural disasters. This is good news from a policy perspective, as it highlights specific conditions under which foreign disaster aid is likely to benefit citizens in recipient states.

While the results from model 3 and these predictions provide support for Hypothesis 2, they only demonstrate that disaster aid is generally more effective at reducing aid in more democratic states. However, they provide little insight into the effect of aid at various levels of disaster severity across regime types. As we argue, post-disaster humanitarian aid should help moderate the adverse impact of more severe natural disasters, particularly in democratic states.

In order to assess these relationships, Figure 3 presents the results for separate analyses of the moderating impact of aid on the relationship between disaster severity and repression in democratic and nondemocratic states, respectively. More specifically, it illustrates the predicted effect of a one-unit increase in aid as a percentage of GDP on repression over the range of the disaster severity. The left panel shows this relationship for the subset of democracies while the right panel shows the relationship for the subset of nondemocracies. As the results in the left-hand panel illustrate, increases in humanitarian aid contribute to a declining likelihood in repression by democratic states as the severity of a disaster increases. For example, in cases where no persons are adversely impacted by a disaster, a one-unit increase in aid (1 percent of GDP) is predicted to contribute to a slight increase in the likelihood of repression—although the result is not statistically significant. However, for a disaster that adversely affects roughly 20,000 persons (∼10 on the logged scale), the same increase in aid is anticipated to reduce the risk that the target state increases repression by about 1 percent. The suppressive effect may initially appear quite modest, but this suggests that following a disaster of this magnitude, each percentage point increase in aid reduces the risk of repression escalation by a point. The effect continues to strengthen as the severity of the disaster increases.

By contrast, the right-hand panel demonstrates that the reverse is true for aid directed toward disaster-affected autocracies. As natural disasters become more severe, increases in aid contribute to higher risks of state-sponsored repression. Where no persons are affected by a disaster, a unit increase in aid raises the risk of repression approximately 2 percent. For a disaster that adversely impacts 20,000 persons, the same one-unit increase in aid is predicted to increase the risk of repression by 4 percent. Again, the effect increases as the severity of the disaster increases. The results of both figures thus show that not only does aid impact democracies and nondemocracies differently, but within each regime type, aid influences the relationship between disaster and repression in opposite ways.

The results for the control variables are generally consistent with previous studies. Across the models, higher levels of Institutional Democracy exert an independent negative influence on repression. Similarly, the durability of the regime is related to repression such that more durable (and presumably more stable regimes) are less likely to escalate repression. Level of economic development, as proxied by per capita wealth, exerts a similar suppressive influence on repression. By contrast, larger populations generally increase the likelihood of repression. Antigovernment demonstrations and civil conflict are also both correlated with an increased likelihood of repression. Importantly, this result holds for the subset of democracies analyzed in model 4, supporting the claim made above that democracies are not immune to escalating repression in the context of instability and rising threats. Finally, the Cold War variable is negative and statistically significant across model specifications. In our view, it is unlikely that repression was actually systematically lower during the Cold War period. Rather, this effect is likely driven by changes in data availability and reporting over the period of study, underscoring the utility of including this variable as a control. ²²

Conclusion

In addition to generating significant economic and human costs, rapid-onset natural disasters may contribute to political instability as well as both violent and nonviolent social and political conflict within affected states. While much of the recent literature within this emerging research agenda focuses on potential connections between disasters and civil conflict or war, we believe prior research often has overlooked the way disasters shape more fundamental political processes within states. Among these are patterns of dissent and repression. While few states experience large-scale civil conflicts, dissent and coercive state responses are observed in all states, even if their intensity differs greatly across cases and over time. Addressing this oversight is therefore important, given the role that state coercion plays in influencing the manner in which dissent unfolds and the likelihood that it culminates in large-scale civil conflict. We contend that natural disasters provoke a general response from states to increase repression, and we find support for this argument. However, we also explore ways in which domestic political institutions and the influx of post-disaster humanitarian aid—a common response from the international community—may influence this more general relationship. This analysis therefore represents an important contribution to a rich and rapidly evolving research agenda exploring the manner in which environmental and climatic events affect political and social interactions (e.g., Brancati 2007; N. P. Gleditsch 2012; Reuveny 2007; Salehyan and Hendrix 2014).

We first explored the relationship between natural disasters and contentious politics. Specifically, we argued that disasters often lead to observable upticks in state repression. We believe this relationship emerges as a result of two interrelated processes: an overall loss of bureaucratic and institutional control by the state and rising state–society tensions. The combination of these factors creates potential windows of opportunity for both violent and nonviolent challenges to state authority. Where such threats emerge, we expect incumbent regimes to increase violence in attempt to deter challenges and reestablish control. The results of our quantitative analyses provided support for the hypothesis that recent disasters contribute to escalating regime repression.

We also investigated the link between the international community’s response to natural disasters and the domestic impacts of such intervention. Specifically, we examined the potential conditional relationship between post-disaster humanitarian aid and domestic political institutions on state-sponsored repression. We contend that because democracies are more reliant on effective public goods provision to maintain political support, they will be more effective at allocating disaster aid than autocracies. Because of this, aid can be an effective tool to alleviate grievances within the population after disasters and can help minimize threats to the incumbent leadership. As a result, aid helps mitigate the impact that disasters have on increases in repression in democratic states. By contrast, the misallocation of aid in more autocratic states may exacerbate existing tensions, leading to increased conflict and repression. Exploring this multiplicative relationship in several ways, we found empirical support for this argument.

This analysis demonstrates that disasters can create conditions conducive to increasing state repression. As such, it suggests that disasters may provide a signal for the locations of increasing state repression and where the international attention should direct its attention and resources. Importantly, however, our argument and findings underscore the importance of carefully monitoring the manner in which aid flows are likely to interact with domestic institutions in the recipient state. The findings reveal that humanitarian aid has an uneven influence on repression in democratic and nondemocratic states. Consequently, donor states and international NGOs should more fully consider the potentially adverse consequences of directing large volumes of aid to the governments of disaster-ravaged autocratic states. This should not justify refusing humanitarian assistance to these states in the wake of severe natural disasters. Rather, it should serve as a call for donors to carefully monitor how and where aid dollars are directed and spent in order to help ensure that disaster aid reaches those groups who need it most rather than those individuals most loyal to the incumbent.

While our argument and findings suggest a relationship between disasters and repression, we recognize numerous additional avenues for exploring the impact of natural disasters on social and political conflict. More fully examining the conditioning influence of domestic political institutions represents one potentially fruitful avenue for future research. We suspect that the relationship may not be entirely intuitive. As noted above, democratic institutions may influence state responses to disasters, but it is not entirely clear that they would completely constrain state repression. For instance, following Hurricane Katrina in 2005, human rights groups reported numerous abuses in and around New Orleans, including imprisonment for minor infractions like curfew violations, often without formal charge or processing (Amnesty International 2010; Metzger 2006–2007). Security forces were also accused of firing on unarmed civilians—the shooting of six persons (and deaths of two) by New Orleans Police Department officers on Danziger Bridge represents the best documented of such incidents (Amnesty International 2010, 21). While this response is consistent with part of our argument—repression occurring as a function of lost monitoring and policing capacity—the disaster did not produce the kinds of organized competition and threat that we hypothesize produce state violence.

While developed democracies may experience an immediate loss of bureaucratic control, and possibly increased repression, the political effects may be different from other types of states. Indeed, recent studies find that disasters may destabilize both democratic and nondemocratic states, though the mechanisms differ, and the former are more sensitive to particularly severe disasters (Quiroz-Flores and Smith 2013). While our results indicate there are differences between democracies and autocracies in terms of how they respond repressively in light of disaster aid, a deeper investigation into the role of domestic political institutions and repressive choices after disasters is required. Consequently, further investigation of the manner in which institutions condition state responses to disasters could improve our understanding of the human and social costs of natural disasters.