Show Summary Details

Page of

PRINTED FROM the OXFORD RESEARCH ENCYCLOPEDIA, NATURAL HAZARD SCIENCE ( (c) Oxford University Press USA, 2018. All Rights Reserved. Personal use only; commercial use is strictly prohibited (for details see Privacy Policy and Legal Notice).

date: 15 December 2018

Natural Hazards and Voting Behavior

Summary and Keywords

Natural hazards have repercussions that reverberate to the political level. Their adverse socio-economic impacts could undermine political support from key fractions in society. Governments, aware of this, have incentives to ease the adverse social impacts of natural hazards. However, the channels of impact from natural hazards to voter and government behavior are complex, indirect, and nonlinear. More than their immediate impact, therefore, major natural hazards contain important symbolic and mythological power that can sway public opinion and influence disaster policies for years to come.

Keywords: blind retrospection, mediated retrospection, narrative retrospection, disaster relief, democratic competitiveness, blame games, Hurricane Katrina, Elbe flooding, Indian Ocean tsunami


This article examines the effect of natural hazards on voting behavior in pluralistic regimes. Natural hazards often have repercussions that reverberate to the political level, as voters can punish governments that do not do their utmost to ease the adverse impacts of natural hazards, which is why government responsiveness in the wake of disasters is stronger in more competitive and pluralistic political settings. Recently, there has been an increased interest in analyzing voter and government behavior in the wake of natural hazards. The advantage is that natural hazards—as opposed to fiscal and economic indicators—can be treated to a greater extend as exogenous shocks beyond direct government control. This makes it easier for voters to evaluate government responses to these hazards compared with government responses to crises plagued by multiple inference challenges, such as financial crises, budget deficits, and unemployment. While there is ample empirical evidence to suggest that natural hazards affect voting behavior, evidence also strongly suggests that voters cannot be reduced to rational actors who, based on perfect information, automatically punish or reward governments for their disaster management efforts. Voting behavior depends crucially on the symbols, images, and narratives that emerge in the wake of the natural hazard. It matters that the First Lady of the United States went onboard Air Force One to inspect the destruction of the 2017 Hurricane Harvey in Texas wearing black pegged trousers, black shades, an olive-green bomber jacket, and 5-inch Manolo Blahnik stilettos. And it matters even more that she had actually dressed down when reemerging from Air Force One in Texas, wearing white sneakers, a T-shirt and, a black baseball cap. The first attire signals vanity, elitism, and indifference to the suffering on the ground while the second outfit suggests a First Lady committed to meeting disaster victims even in the most rugged and inaccessible terrains.

The first part of this article discusses the theories of voting retrospection that most prominently link natural hazards to voting behavior. It distinguishes between traditional theories of voter retrospection, which emphasize the impact and the management of the natural hazard, and a newer strand, which places more emphasis on the political symbols and narratives that emerge from the natural hazard. Empirical evidence is presented linking natural hazards to voting behavior. The second part of the article moves attention to the implications of voter retrospection for government behavior in the wake of natural hazards. Variations of government behavior according to political systems are discussed, and how voter retrospection could end up distorting effective disaster management.

Natural Hazards and Voter Retrospection

A theoretical section introduces different state-of-the-art theories linking natural hazards to voting behavior, and an empirical section presents the existing evidence from different parts of the world.

Theories of Voter Retrospection

The link between natural hazards and voting behavior is principally shaped by the theory of retrospective voting. Essentially, the theory of retrospective voting assumes that voters place emphasis on past occurrences when deciding their vote. In the context of natural hazards, retrospection will most often involve an assessment of the nexus between natural hazard impacts and government responses to those impacts. There are three different ways that voters can respond to the occurrence of natural hazards. The first is blind (or simple) retrospection. Here, the electorate judges the incumbent government based on each voter’s own situation regardless of whether the government actually had an influence on that situation. The second is mediated retrospection. This refers to the fact that voters judge incumbent governments based on their policies. Voters behaving blindly would thus punish governments for merely preseding over natural hazards, while mediated retrospective voters would evaluate the quality of the governments’ disaster management efforts. These two types of retrospective voting behavior have been bundled together under the heading traditional theories: they share the view that voters are bounded rational and that they respond to tangible impact factors (fatalities, economic damages, and so on) as well as policy factors (government preparedness, response, and reconstruction efforts). The third is narrative retrospection. Here voting behavior is primarily determined by the dominant political narratives that surround particular issues. Voters respond less to the actual impacts of natural hazards and more to the way the hazards end up being framed in the political processes where different interpretations of events fight for dominance. This perspective is more post-modern: voters react not to tangible factors such as disaster damage or relief expenditure but to the prevalent discourses of the natural hazards. It is important to emphasize that these three different types of retrospective voting behavior are not mutually exclusive. On the contrary, it appears that voting behavior often contains dynamics from all three approaches. Voters react to the severity of the natural hazard, the implemented disaster policies, and the prevalent narrative emerging from the hazard. The relative weight of each of these three different dynamics in forming voting behavior appears to be an empirical question. Likewise, it should also be stressed that voting behavior is determined by much more than retrospective voting, which will be discussed in greater detail in the section “Traditional Theories of Voter Retrospection” on traditional theories of voter retrospection.

Traditional Theories of Voter Retrospection

Traditionally, retrospection has been divided into two separate dynamics: blind (or simple) retrospection, referring to the fact that the electorate evaluates the incumbent government based on outcome regardless of whether the government’s policies have actually influenced the outcome; and mediated retrospection, where voters judge the incumbent government based on its policies. Retrospective voting can be traced back to Anthony Downs’ 1957 monograph, An Economic Theory of Democracy, where retrospection was seen as a cost-effective way to evaluate candidates. The argument was that, while prospective voting (what the incumbents intend to do) would be more relevant for voters, the lack of credible information makes retrospective voting the only feasible alternative. Valdimer Key (1966) augmented retrospective voting in his monograph The Responsible Electorate from 1966. Rather than seeing it as a cost-effective solution to the information problem, retrospective voting was understood as an accountability mechanism. Key (1966, p. 75) argued that voter retrospection was beneficial for society because it meant, “that a political party cannot avoid accountability for its past performance.” Yet, the problem with Key’s understanding of retrospection is that voters appear to make judgments about their personal situation rather than the actual government policies implemented. Hence, the term blind retrospection. In situations of natural hazards, therefore, voters will behave “blindly” by retrospectively punishing governments for the mere occurrence of natural hazards. There are several reasons for this—that voters are ignorant is not necessarily one of them. The reason is rather that political activism is not cost-free. It requires effort to immerse yourself in politics to the extent that you have a comprehensive overview of government policies as well as the insights to evaluate these policies. Voters have many other contrasting claims on their time. They tend to prioritize survival, food, work, friends, family, or hobbies over probing complex political questions. Voters are therefore bounded rational and will not possess the required information to evaluate government policies in the wake of natural hazards. Information is more readily available and can be more easily processed if voters just ask the simple question: is my situation now better than before the last election? Some political slogans explicitly tap into this kind of mindset. Reagan famously urged voters at the 1980 US presidential debate to ask themselves whether they were better off today than four years ago (Reagan, 1980). As a shrewd politician, he was looking to activate the blind side of the retrospective voter that would punish incumbent president Carter for the adverse consequences of the 1979 energy crisis.

Morris Fiorina (1981) refined Key’s retrospective voting theory in his monograph Retrospective Voting in American National Elections, by adding an extra layer of voting complexity. While voters would still evaluate candidates in an election based on past personal experiences, voters would also base evaluations on mediated judgments on the actions taken by the government on particular issues. This extra layer of voter retrospection was termed mediated retrospective evaluations referring to the fact that voters would make judgments on the policies of the government based on information provided by the media, friends, or opinion leaders. Kiewiet and Rivers (1984) added to the theory of retrospective voting by empirically documenting that economic indicators in time periods closer to the election were better predictors of voting behavior compared to the past (Kiewiet & Rivers, 1984, p. 372). This indicates that, even when voters take into account government actions (and not just outcomes), they appear to be myopic, discounting actions and events that are not close to the election. More recent studies of voter movements have collaborated this claim: recent events have taken precedence when voters decide their vote (Achen & Bartels, 2004; Bechtel & Hainmueller, 2011; Cole et al., 2012; Lenz, 2011).

Both traditional theories of retrospection acknowledge that prospective voting, partisanship, and ideology influence voting behavior. Although issue voting is on the increase in many mature democracies, there is still inertia in voter movements (Ansolabehere, Rodden, & Snyder, 2008; Söderlund, 2008). Many voters identify with a specific party, and it takes a lot to trigger major shifts in voting patterns. Cook’s Partisan Voter Index points to even greater partisan polarization in US politics in recent years as the number of swing districts have declined 20% in just four years (Wasserman & Flinn, 2017). Through an experimental design in which respondents ranked seven public officials in order of how much they should be blamed for the property damage and loss of life in New Orleans after the 2005 Hurricane Katrina, Malhotra and Kuo (2008) found a clear partisan bias in the attribution of. These types of emotional attachments to particular parties or politicians will prevent voters from objectively and accurately attributing responsibility for events and actions to specific incumbents. The point here is that, despite voting inertia and partisan polarization, retrospection also matters. Even small voter movements based on this retrospection could ultimately determine an election. Fluctuations in precipitation across key states could very well have cost Democrat Al Gore the US presidential election to Republican president-elect George Bush in 2000 (Achen & Bartels, 2004; Gomez, Hansford, & Krause, 2007). Had the Elbe not flooded a month prior to the German elections in 2002, chances are that Edmund Stoiber would have been the new Chancellor rather than Gerhard Schröder, who secured a last-minute win (Bechtel & Hainmueller, 2011). Was it not for 2012 Hurricane Sandy, Velez and Martin (2013) shows that Obama would likely have lost the important swing-state of Virginia at the 2012 US presidential election. Thus, despite the fact that different dynamics impact voting behavior (prospection, partisanship, and ideology), retrospection remains a key factor in explaining voting behavior after a natural hazard.

Narrative Theories of Voter Retrospection

Several scholars have argued that any electoral effect of a disaster depends crucially on the extent to which an extreme event, in this case a natural hazard, is turned into a political narrative around which different political actors position themselves (Althaus & Kim, 2006; Boin, Hart, Stern, & Sundelius, 2005; Boin, McConnell, & Hart, 2008; Drake, 2016; Johnston & Goggin, 2015; Kamradt-Scott & McInnes, 2012; Kuttschreuter, Gutteling, & de Hond, 2011; Rubin, 2015). Although scholars highlight different factors responsible for constituting the narrative (the press, opposition parties, friends, culture, religion or the social media), these perspectives can be compiled under the heading narrative retrospection (Rubin, 2015, 2017). With an emphasis on governance, this article focuses on the political dimension of narratives. Politically, major natural hazards (and other crises) give rise to conflicting interpretations based on a plurality of political values and interests. This type of retrospection is less concerned with the natural hazard itself and more on the way it is used (or abused) politically. The causes and consequences of major natural hazards are not crystal clear for voters, not just because of bounded rationality, but because they are communicated in a political arena where different myths, narratives, and symbols fight for dominance. Dominant political actors have an interest in shaping the political narratives to fit their agenda. Boin, Hart, and McConnell (2009, p. 83) argued that what determines policy outcomes are not events on the ground but their public perception and interpretation. Pelling and Dill (2006, p. 3) argue along the same lines when they conclude that disasters have important symbolic power and that sociocultural contexts play a key role in shaping political outcomes. Myths and symbols should not be understood as imaginary or unreal folklore, but instead as sociopolitical narratives that provide meaning and comfort in the wake of a natural hazard (Hart, 1994). Political narratives can be understood as narratives that emerge from formal political forums, produced by politicians and public officials and reproduced in the public debate (Shenhav, 2006, p. 247). For a narrative to be considered political (and not just trivial), it must be contested by other key political actors and prioritized in political arenas in competition with other issues. As Shanahan, McBeth, and Hathaway (2011, p. 539) argued, “political narratives are persuasive stories for some political end (e.g., to win an election).”

In some religious circles today, natural disasters are still believed to be God’s punishment for humankind’s immoral behavior. Some indigenous populations also believe that natural hazards are messages and signs (not just punishments) from their ancestors or from Mother Nature. However, these understandings are less widespread today. They were put under serious pressure during the Age of Enlightenment, when the science behind natural hazards became generally accepted. The 1755 Lisbon earthquake is commonly seen as the turning point from perceiving disasters as a divine force to applying a more naturalistic/scientific perspective to disasters (Dynes, 2000). The prevailing narrative was still, however, one of an exogenous force overwhelming the unsuspecting society. Today, natural disasters are perceived to be much more endogenous to society. Ulrik Beck’s (1992) Risk Society was particularly influential in spreading the view of hazards and risks as endogenous to modern society. This view became mainstream in current disaster research (Dahlberg, Rubin, & Vendelø, 2015; Shroder, 2014). Deforestation might cause greater floods, climate change might produce more severe hurricanes, conflicts and water mismanagement could exacerbate minor droughts to states of famine, and unsustainable mining might cause lethal mud flows and avalanches. The fact that natural hazards can be the result of human action is part of the present political narrative, just as it is commonly accepted that the extent to which natural hazards turn into disasters depends crucially on socio-political factors (Blaikie, Cannon, Davis, & Wisner, 2014). This does not mean that citizens expect to be fully protected against the adverse impacts of natural hazards, but they do expect their government to act. Governments should act with regard to concrete relief policies, but they should also “put on an act”—be a uniter in times of chaos, grieve with the surviving relatives, convey the nation’s compassion to the victims of the disaster, and instill hope and optimism for the future. These symbolic functions are integral parts of a modern political narrative of natural hazards.

The link between natural hazards and voting behavior, therefore, needs to capture not just the physical impacts of the hazard but also their disaster narratives. Symbol politics (what politicians say and how they behave) play just as important a role as real-politik (what the politicians actually achieve). The natural hazard may awaken feelings and emotions that mature and evolve in the aftermath of the hazard, leading to political dynamics that are not necessarily imminent but could materialize down the road. Hence, natural hazards might have greater long-term political consequences than immediate short-term political consequences. The strength of symbols and narratives is that they often transcend different political issues, and they are quite resilient over time. Mediated retrospection would evaluate policies on a case by case basis while narrative retrospection would draw on more rigid symbols and narratives. Retrospective voters might not be able to assess in great detail the government’s disaster policies, but they would intuitively rely on sentiments, pictures, and symbols from the natural hazard.

In their review of voter retrospection theory, Healy and Malhotra (2013, p. 290) argued “that voting behaviour often introduces a considerable amount of noise into democratic accountability, potentially distorting policy-maker incentives in some cases.” It is certainly true that voting behavior might lead to suboptimal disaster management decisions where governments are electorally rewarded for responding to natural disasters instead of preventing disasters (see how government responds to rather than prevents disasters in the section “Natural Hazards, Voting Behavior and the Behavior of Governments”). However, narrative retrospective theory turns Healy and Malhotra’s argument upside down by focusing on how democratic accountability mechanisms make political actors purposely introduce noise into the political process. Narrative retrospection is not concerned with whether voting behavior distorts the incentives of policy-makers but rather on how the incentives of policy-makers distort the basis for “value-free” retrospection. One could speculate that this analytical perspective on retrospection has become even more salient in a political climate of “fake news,” where the echo chambers of politically charged social media newsfeeds and discussions explicitly aim to convey narratives rather than facts. As objective facts of natural hazard impacts and policies become more difficult to ascertain, political narratives naturally play a more dominant role in retrospective voting behavior. Research has yet to look at the link between natural hazards and voting behavior in the context of this changed landscape of political communication.

Summing Up

Retrospection can be divided into three complementary dynamics. Table 1 provides an overview of the three different types of retrospection together with their basic assumptions and implications. Blind retrospection is based on a high degree of bounded rationality because voters judge the incumbent government solely on the outcome. Thus, voters would electorally punish governments for any maladies that they experience from a natural hazard. Research based on this type of retrospection usually gauges voting behavior by linking the impact of the natural hazard measured in economic damages or fatalities with disaggregated electoral data. Voters possess more advanced bounded rationality in mediated retrospection. Here voters judge incumbent governments based on their policies. It is still bounded, however, because the theory assumes that the myopic voters will value recent events higher than events in the past. Research drawing on this theoretical perspective explicitly includes an account of disaster policies as well as measures of voter satisfaction with these policies. Both blind and mediated retrospection rely primarily on short-term electoral impacts and hypothesize that voting behavior will change immediately following the natural hazards. Narrative retrospection predicts that voters are primarily intuitive at the ballot box and will not critically assess the pros and cons of different government policies, but that the voting decision will be strongly rooted in sentiment. Thus, voters will respond to the prevailing discourses that materialize after a natural hazard: who emerged as the heroes, and who emerged as the villains? Methodologically, it is more difficult to quantify the independent variable under narrative retrospection because it relies on less tangible discourses and public perceptions. However, mixing qualitative and quantitative analyses of newspaper articles, social media feeds, graphical illustrations, transcripts from parliamentary debates, and national values surveys could help piece together a picture of the political narratives that surrounded the natural hazard. The depended variable, the impact on voting behavior, is also slightly more difficult to capture. Whereas blind and mediated retrospection assume a short-term effect right after the natural hazard, narrative retrospection hypothesizes that the electoral impacts could materialize long after, as voting behavior based on symbols and sentiments is more enduring. Retrospection could remain dormant until activated by similar situations or political behavior long after the disaster. Ultimately, the effects of natural hazards on voting behavior are an empirical question, one that will be discussed in the next section.

Table 1. Retrospective Voting Behavior and Natural Hazards

Voting Logic

Voting Behavior

Analytical Approach

Temporal Scope

Blind retrospection

Bounded rational

Punishes the government for preceding over a natural hazard

Investigates causal link between hazard impacts and voting behavior

Short-term electoral impacts

Mediated retrospection

Advanced bounded rational

Punishes/rewards the government for the quality of disaster management

Investigates causal link between government policies and voting behavior

Short-term electoral impacts that diminish over time

Narrative retrospection


Punishes/rewards political actors for their perceived actions

Investigates the causal link between political symbols and narratives and voting behavior

Short- and/or long-term impacts

The Empirical Evidence of Natural Hazards and Voter Retrospection

Traditional Voter Retrospection

Studies have documented that voters electorally punish governments blindly for complete exogenous events such as shark attacks (Achen & Bartels, 2004); the international price of oil (Wolfers, 2002); the defeat of local sports teams (Busby, Druckman, & Fredendall, 2017; Healy, Malhotra, & Mo, 2010), and bad luck at the national lottery (Bagues & Esteve-Volart, 2016). With regard to blind retrospection and natural hazards, Achen and Bartels (2004) analyzed the effect of torrential rain and drought on the US presidential elections in the period 1826–2000. They find that in constituencies that experienced an extreme natural hazard prior to the election, the incumbent government’s voter support decreased with about 1.5 percentage points. Such relatively modest drop could have substantial political consequences. Recall the very close presidential election in 2000, where only 537 votes separated the incumbent Democratic candidate Al Gore from Republican George Bush in the all-important swing-state of Florida. Achen and Bartels (2004) suggested that the volatile weather in 2000 with droughts in many Southwestern states and downpours in several Eastern states is likely to have cost Al Gore the election. Healy and Malhotra (2010) also found evidence of blind retrospection in their study of the effect of tornado damage on US presidential elections 1950–2004. They found that the effects of tornado damage in US counties decreased the incumbent presidential party’s vote share in the worst affected counties, with between 1 and 2 percentage points (depending on the cut-off point for damage), compared to the national average. Based on a historical study of The Great Mississippi Flood in 1927, Heersink, Peterson, and Jenkins (2017) found strong support of blind retrospection. Although disaster aid was dispersed at an unprecedented level, and despite presidential candidate Herbert Hoover’s numerous visits to the flood stricken areas, the study found that the flood ended up decreasing voter support for Hoover by more 10 percentage points in the affected counties. Arceneaux and Stein’s (2006) study of the 2001 Tropical Storm Allison renders further support to the theory of blind retrospection. The storm occurred just a few months before the election and caused a 500-year flood that wreaked havoc in Houston. Houstonians who lived in neighborhoods hard hit by the flood were more likely to blame the government for inadequate flood preparation, and at no level of government (city, county, and federal) did the flood lead to increased government support. Moving outside the United States, Cole, Healy, and Werker (2012) documented that Indian voters displayed the same dynamics of blind retrospection, as voters across 28 Indian states appeared to electorally punish the incumbent government coalitions for extremes in rainfall. Eriksson (2016) analyzed the 2005 storm, Gudrun, that hit the southern parts of Sweden with great force: 8 percent of the total population was left without electricity for up to 45 days, and the damages amounted to 1 percent of the country’s gross domestic product. Eriksson concluded that the storm caused a decline of almost 4 percentage points in government support in the affected areas at the subsequent Swedish election more than 1.5 years later, and that the decline (albeit smaller) persisted in both the 2010 and 2014 elections. A study of voting behavior, by Bodet, Thomas, and Tessier (2016), in the wake of a 2013 flood in the Canadian city of Calgary also found a substantial vote gap of 6 percentage point between the 15 flooded subdivisions and the 127 divisions that were not flooded. However, the authors highlight that the socio-political conditions in the flooded regions, vis-à-vis the non-flooded regions, were not equivalent, and the voting difference became insignificant once the regions were statistically matched, leading them to advocate for caution and skepticism when testing for blind retrospection (Bodet et al., 2016, p. 93). Although natural hazards could be considered exogenous shocks, simple comparisons of voting behavior in affected versus unaffected regions are likely to be flawed due to the inherent socio-political heterogeneity across regions.

While empirical evidence generally supports the proposition that voters electorally punish the incumbent government for merely presiding over natural hazards, more nuanced studies that include temporal dimensions have found that two conflicting phases underlie voting behavior. In the first phase, the incumbent government starts out at a disadvantage because it might instantly lose voter support after a natural hazard (blind retrospection). There is also a second phase, however, where the government can regain some of the lost votes by handling the disaster effectively (mediated retrospection). The previously mentioned study by Healy and Malhotra (2010), for example, found that tornado damage can actually increase voter share for the president’s party in the given county if the incumbent president issues a disaster declaration and releases federal funds. Although this result is not statistically significant (in part due to the infrequency of disaster declarations), voters appear not to punish government for random tornadoes when those events elicit responses. The same authors reached a similar conclusion with regard to natural disasters in general: the US counties in which natural hazards are followed by relief aid generate a higher share of votes for the incumbent party (Healy & Malhotra, 2009). Eriksson (2016) also did not ascribe the decline in voter support for the Swedish government to blind retrospection, arguing instead that it was more convincing that the affected public evaluated the government response (where no funds were provided to the affected areas) as a failure. There is ambiguous empirical evidence as to how many votes governments can gain by effectively responding to a natural hazard. The Indian study of Cole et al. (2012) indicated that, even though voters do reward politicians who provide emergency aid to disaster-struck states, the governments could only regain a fraction of what they initially lost from just presiding over the natural hazard. A similar study of the 2010–2011 Pakistani floods by Fair, Kuhn, Malhotra, and Shapiro (2017) found only weak evidence of voters rewarding politicians for their relatively effective handling of the floods. Thus, blind retrospection appears to be the key determinant of voting behavior in these cases. Other studies indicate that the government could gain more votes than it lost in the first phase. In Bechtel and Hainmueller’s (2011) analysis of German voting behavior after the Elbe flooding in 2002, the incumbent government was rewarded quite substantially for its effective disaster response, with a vote share increase of seven percentage points in the affected areas. The assumption that mediated retrospection dominates voting behavior is also supported by Velez and Martin’s (2013) study of Hurricane Sandy. Taking into account the sociopolitical heterogeneity across affected and unaffected counties by statistical matching, Velez and Martin (2013) estimated that President Obama received about a four-percentage point increase in vote share from Hurricane Sandy in the affected areas. One speculative explanation for the difference between the findings of these different studies could be that widespread poverty and illiteracy rates in Pakistan and India make voters more myopic and blindly retrospective compared to wealthier democracies such as Germany and the United States (see section: “The Role of the Media”). However, Remmer’s (2014) analysis of voting behavior in less mature democracies across 21 Caribbean islands found evidence that Caribbean voters appeared to avoid blind retrospection, instead rewarding and punishing politicians for their competence in managing the disaster. This suggests that whether governments gain or lose electoral support is an open and context-dependent question. The next strand of retrospective voting that will be discussed in the next section, on narrative retrospection, argues that this ambiguous result might be caused by the fact that voters react primarily to something other than disaster policy and natural hazard impact.

Narrative Retrospection

The fact that voting behavior appears so unpredictable might point to the relevance of including political narratives to explain voting in the aftermath of natural hazards. The study of Lazarev, Sobolev, Soboleva, and Sokolov (2014), of voting behavior in the wake of the 2010 forest fires in Russia, renders support to the presumption that another type of retrospection might be driving much of voting behavior. The wildfires destroyed more than 1,200 houses and killed 50 people. The population was generally highly critical of the government policies, and more than 50% blamed the government for the disaster. However, Lazarev et al. (2014) found that support for authorities remained the same or increased in the burned areas. Both blind and mediated retrospection would have predicted a decline in support. To explain this discrepancy, Lazarev et al. (2014) included the symbolic effect of government performance by proxying it with President Putin’s visits to the affected villages. This narrative retrospection had strong explanatory power for Putin’s popularity. Compared to his popularity in unburned villages, Putin’s popularity increased by about 15 percentage points in the villages that he visited despite them not having received any additional relief aid. Johnston and Goggin’s (2015) study of US voting behavior in the wake of the BP oil spill suggests that voters’ confidence in the Obama administration was unaffected by the oil spill itself (blind retrospection). Only after the media coverage started to focus on blame (spurred on by several Republican politicians) could Johnston and Goggin identify a drop in the public’s confidence of President Obama, leading them to rhetorically ask (2015, p. 468): “if voters blindly attribute blame for a disaster, then what accounts for the delayed reaction in response to the Deepwater Horizon oil spill?” More information of the disaster and the policies (mediated retrospection) was not sufficient to impact voting behavior, but media framing was the relevant independent variable that actually translated the oil spill into voting behavior.

Blind and mediated retrospection appear to have limited explanatory power for voting behavior in wake of the 2005 Hurricane Katrina. Despite substantial media attention and criticism of the government’s disaster management, it is very hard to discern any (positive or negative) movements in President Bush’s approval ratings. Bush’s downward spiral in approval ratings (since peaking in 2001) merely continued along the same trajectory after the natural hazard (Gallup, 2006). Even when looking at the opinion polls from the state of Louisiana, which was particularly hard hit by the hurricane (the entire state was declared a disaster zone when New Orleans was flooded) it is hard to identify a clear change in people’s opinion of Bush. Among the ethnic group that suffered the most from the hurricane, African Americans in New Orleans, Bush’s approval ratings did not appear to be affected. His approval ratings were low at around 20%, but rather constant before, during, and in the months after the hurricane (Survey USA, 2005). At the state level, all seven members of Louisiana’s House of Representatives were re-elected, as was the mayor of New Orleans. One of the greatest natural hazards in modern times in America thus did not have any measurable short-term political consequences, which somewhat undermines traditional theories of retrospection. President Bush admitted in his memoir that his handling of hurricane Katrina was probably his biggest mistake as president (Bush, 2010). He emphasized how a photo taken of him peeking out the window of Air Force One, hovering high above the destruction in New Orleans, came to embody the public’s perception of his disaster management efforts. He regretted that he did not touch down to shake hands with the Louisiana governor and ensure the people of New Orleans that help was on the way. While his failure to do this was not immediately apparent in the approval ratings, it did contribute to certain narratives and myths about the President. Concretely, the photograph might have helped cement the narrative engrained in the public from another photograph taken a few years earlier—the 2003 photograph of Bush standing in front of a huge “mission accomplished” banner on board the aircraft carrier USS Abraham Lincoln, proclaiming an end to major combat operations in Iraq. It was a narrative of a president out of touch with the reality on the ground. Like any skillful politician, Bush was well aware of the power of such narratives, and it is notable that his greatest regret looking back was not that he failed to provide adequate relief to the victims of Hurricane Katrina but that he failed on a symbolic and iconic level. The gravity of the mistake is confirmed in the following 2009 statement from Bush’s chief strategist in the 2004 presidential campaign: “Katrina to me was the tipping point. The president broke his bond with the public. Once that bond was broken, he no longer had the capacity to talk to the American public. State of the Union addresses? It didn’t matter. Legislative initiatives? It didn’t matter. P.R.? It didn’t matter. Travel? It didn’t matter” (Murphy & Purdum, 2009).

Narrative retrospection also was prevalent in the 2002 German election. Before the flooding of the Elbe River, the Conservative party leader Edmund Stoiber was a sure bet for an election victory: the August voting polls gave his coalition 51% of votes for the German election less than a month away, while the incumbent Social Democrats, led by Gerhard Schröder, could muster only 44%. Then came the worst flooding in contemporary German history. Several cities along the river were completely under water, 30,000 people had to be evacuated, and economic damages were substantial at around 15 billion EURO (Bechtel & Hainmueller, 2011). One would expect—based on theories of blind retrospection—the flooding to put Schröder’s government under increased pressure. The flooding was widespread (affected a large share of the electorate), and it occurred in the immediate run up to the election (it would still be fresh in the mind of voters). However, a couple of weeks after the flooding, political opinion changed drastically in favor of the incumbent government, and on election day, 53% now preferred the current government, while only 43% preferred the opposition (Roberts, 2003). The government took control of the political agenda. Schröder immediately visited the affected areas, and in the following days, the media could report how he flew from one affected city to another in a helicopter. Several media carried stories and photographs of Schröder on the front line of the government’s relief efforts. Stoiber, on the other hand, waited a whole week before visiting the areas hit. By then it was too late, and the dominant political narrative of a powerful and caring Schröder had been cemented in the minds of voters.

The last example of the benefits of working with political narratives can be found in the implications of the 2004 Indian Ocean tsunami for the Danish and Swedish governments. The 2004 tsunami was the deadliest natural hazard for Denmark and Sweden in more than a hundred years. With many vacationing Scandinavians in Thailand, Sweden lost 543 citizens and Denmark lost 46 to the tsunami (EM-DAT, 2017). The media was highly critical of the governments’ disaster response. Several editorials and opinion articles pointed to the absence of large-scale evacuation plans for citizens abroad; the lack of ministerial cooperation with the police; inadequate information dissemination; flawed institutional setups for the emergency force; and a widespread passivity in all levels of the administration in the first days after the disaster (Rubin, 2017). The negative media coverage had a clear impact on the general public’s perception, with several surveys indicating that the majority of voters in both countries were critical of the management of the disaster. However, the electorate did not punish the Danish and Swedish governments for the adverse impacts of the tsunami or the inadequate evacuation policies. The election polls showed no discernable impact from the disaster (Rubin, 2017). Narrative retrospection could explain this by the fact that the dominant political narrative did not encompass the tsunami in any forceful manner. In Denmark, there was no pressure to make the government’s disaster management part of a political narrative, and the critique was restricted to the systemic level of administrative (technocratic) flaws. In Sweden, the brunt of this critique occurred more than a year after the tsunami. The narrative surrounding the Swedish Foreign Minister was particularly unfavorable, as she appeared to make all the wrong symbolic choices. Several media outlets reported how the Foreign Minister went to the theater 20 hours after the tsunami, despite initial reports of more than 100.000 fatalities; that she did not show up for work in the Ministry until 31 hours after the tsunami; that she admitted having absolutely no idea where Phuket was; and that she explained the belated response with the fact that she did follow news on her day off. While there is strong evidence that the unfavorable tsunami narrative stuck to the Foreign Minister and was a contributing factor in her resignation in 2006, Eriksson (2016) identified only a small decline of support for the government as a whole in the municipalities affected by the tsunami (0.7%).

Summing Up

There is much empirical evidence to suggest that natural hazards influence voting behavior and that natural hazards can indeed determine an election. There is also enough evidence to suggest that the impact of natural hazards on voting behavior will contain elements from all three types of retrospection. A relevant question is whether it is possible to separate out the effects. Table 2 summarizes the methodological challenges involved in distinguishing the three different retrospective dynamics from each other. If a natural hazard leads to an increase in government support, then blind retrospection most certainly cannot be the determining dynamic. An increase would imply that mediated and/or narrative retrospection are likely to be the dominant dynamics. The methodological challenge is that effective disaster policies and favorable natural hazard narratives are highly correlated, which makes it difficult to separate out the effects. It is very rare that politicians are all talk but no action, and even rarer that they do not flaunt the successful policies that they have implemented. Schröder’s visits to the flood-affected areas, for example, were immediately followed up with concrete disaster policies: 45,000 soldiers were dispatched to the affected areas, and an emergency aid fund was established within the first two days (Bechtel & Hainmueller, 2011). One way to separate out the two effects would be to analyze differences in voting behavior in affected areas that were visited by heads of state compared to the other affected areas, while keeping the disaster policies constant (Lazarev et al., 2014). Another way would be to exploit the temporal dimension, where the impact of narrative retrospection does not need to be instantaneous. In the initial phase, voters are likely to respond only to the immediate impact of the natural hazard; in the next phase, they might be responding to the disaster policies; and in the last phase, they might respond to the prevailing political narrative of the natural hazard. Such a temporal approach would need to draw on panel data from multiple elections (Cole et al., 2012; Healy & Malhotra, 2010) or on surveys (Arceneaux & Stein, 2006; Lazarev et al., 2014) or on opinion polls (Johnston & Goggin, 2015; Rubin, 2017).

If a natural hazard leads to a decrease in support, then all three dynamics might be at play, and it could be difficult to ascribe the drop to any particular type of retrospective behavior. Many studies of mediated retrospection merely establish that government disaster policies have been criticized, but they fail to establish the causal link between an unfavorable public perception of government policies and the decline in government support, thus not eliminating the possibility that the decline might be caused by the natural hazard itself (Eriksson, 2016; Velez & Martin, 2013). Often overlooked are cases where the major natural hazards do not impact voting behavior at all. This defies both blind and mediated retrospection, which hypothesizes that the hazard will have some kind of impact. However, narrative retrospection assumes that natural hazards must be part of a political narrative to retain retrospective power. When major hazards coupled with highly criticized government response have no discernable impact on voting behavior, it strongly suggests that narrative analyses are necessary to account for this finding (Rubin, 2017).

Whether a natural hazard might prove to be an advantage or disadvantage for governments is difficult to answer simply because most studies do not analyze the overall electoral gains for the government. Focus is often on voter gaps between the affected and unaffected areas (see Bodet et al., 2016; Eriksson, 2016; Lazarev et al., 2014). Such a gap, however, says little about the overall electoral impact from the natural hazard. In light of a major natural hazard, it is unlikely that the voting behavior control group (the unaffected areas) will remain constant. Barring truly catastrophic natural hazards, most hazards strike locally, and the people affected often comprise only a small percentage of the overall electorate. The extent to which the majority of voters will sympathize with generous relief given to the victims depends on many factors, such as whether they risk being impacted by a similar hazard in the future; the extent to which they share key socio-economic characteristics with the victims; whether the victims could be blamed for engaging in moral hazards; or the way politicians frame the disaster policies. It is worth noting, for example, that the incumbent government’s very active relief policy during the 2002 Elbe flood actually cost the government 3 percentage points of support from the unaffected districts (Bechtel & Hainmueller, 2011). Although the net voting result ended positively for the government, this will most certainly not always be the case.

It should also be emphasized that most studies focusing on voting behavior implicitly do so in the context of rapid-onset, high-impact natural hazards, such as earthquakes, flash floods, hurricanes, and tsunamis. Slow-onset disasters, such as droughts or floods, as well as creeping and transboundary disasters such as desertification, climate change, or antimicrobial resistance are very likely to lead to different political processes and outcomes (Birkmann, 2007; Staupe Delgado et al., 2018). While the three different voting behavior dynamics have merit when applied to such hazards, the lack of a clearly identifiable and demarcated hazardous event makes it harder to isolate the hazard from other concurrent socio-political developments. The methodological benefit of rapid-onset natural hazards, and the reason most empirical studies analyze these rapid-onset disasters, is that they produce a distinct shock to the political system, which allows for more robust analyses of cause and effect.

All three retrospective dynamics are problematic as forces for effective disaster management in the wake of natural hazards. Blind retrospection could undermine government incentives for providing humanitarian relief because of the negligible political rewards to be had. Governments’ incentives under mediated retrospection would depend crucially on the timing of the election and on the level of competition. If narrative retrospective voters mainly register symbolic acts, then governments might enact empty rituals in the wake of natural hazards rather than the nuts and bolts of effective disaster management. The next section “Natural Hazards, Voting Behavior, and the Behavior of Governments” explicitly addresses the link between voting behavior and government responses in the wake of natural hazards.

Table 2. Linking Empirical Evidence With Theories of Voting Behavior in the Wake of Natural Hazards




Voter support increasing

Blind retrospection: Unlikely (falsified).

Mediated retrospection:

Likely, but must be linked to positive public perception of disaster policies.

Narrative retrospection: Likely, but must be linked to the nature of the dominant narrative of the hazard.

Healy & Malhotra, 2009

Healy & Malhotra, 2010 (secondary effects)

Bechtel & Hainmueller, 2011

Velez & Martin, 2013

Lazarev, Sobolev, Soboleva, & Sokolov, 2014

Remmer, 2014

Johnston & Goggin, 2015 (secondary effects)

Voter support constant

Blind retrospection: Unlikely. The implications of the hazard should decrease voter support.

Mediated retrospection: Unlikely. If the public has formed an opinion of the disaster policies, we would expect to see an impact.

Narrative retrospection: Likely, but must demonstrate that no political narratives formed around the natural hazard.

Rubin, 2017

Voter support decreasing

Blind retrospection:

Likely, but must linked to the impacts of the hazard only.

Mediated retrospection: Likely, but must be linked to a negative public perception of disaster policies.

Narrative retrospection: Likely, but must be linked to the nature of the dominant narrative of the hazard.

Achen & Bartels, 2004

Arceneaux & Stein, 2006

Healy & Malhotra, 2010

Cole, Healy, & Werker, 2012

Bodet, Thomas, & Tessier, 2016

Eriksson, 2016

Heersink et al., 2017

Natural Hazards, Voting Behavior, and the Behavior of Governments

This section links retrospective voting behavior in the wake of natural hazards to government behavior. The first four sections focus on voting behavior in pluralistic regimes while the last two sections stretch the initial mandate of this article to briefly include the impact of disasters on authoritarian regimes and the prospects of regime change. The first four sections address: (i) why governments might prefer reactive policies to proactive ones; (ii) the effect of democratic competitiveness on government disaster policies; (iii) the importance of the free media; and (iv) the link between voter retrospection and general political survival strategies. The last two sections investigate the impact of natural hazards on government behavior in authoritarian regimes, and discuss the opportunities for natural hazards to overthrow such regimes.

Responding Rather Than Preventing

Healy and Malhotra’s 2009 study of US relief expenditures across counties documented that incumbent governments are more prone to provide disaster relief in politically supportive constituencies. These findings have been replicated by several other studies. Atkinson, Hicken, and Ravanilla (2014), in a study of storm relief in the Philippines, found that party match between the congressperson and the mayor in a given municipality increased the per capita reconstruction funds allocated to that municipality, and Aldrich (2016), in a study of the 2011 Japanese tsunami, found that the most powerful predictor of recovery for a given area was the number of powerful politicians representing the area in the national government. However, Healy and Malhotra (2009) took their analysis one step further. The authors showed that governments are electorally rewarded only for responding to natural disasters and not for preventing disasters, which leads to perverse political incentives in disaster management. Despite obvious financial cost-benefit advantages in favor of preventive disaster measures (a dollar spent on disaster preparedness would reduce future damages within the same election cycle by more than 7 dollars), there are few political advantages in implementing preventive measures (Healy & Malhotra, 2009). In other words, politicians are not rewarded politically for a well-maintained dam; the lack of a disaster is a silent occurrence most likely to slip under the radar screen of the voters. Should the levee break, however, politicians would suddenly have the voters’ full attention, and great political rewards and penalties might come into play. Retrospective voting behavior could thus lead to suboptimal disaster policies.

Democratic Competitiveness

Reeves (2011) analyzed US presidential elections in the period 1981–2004 to examine whether electoral competitiveness influences natural disaster declarations. Electoral competitiveness was proxied by the margin of victory in the three preceding elections. It turns out that highly competitive states receive twice as many disaster declarations from the president after controlling for the actual impact of the natural hazards. According to Reeves’ analysis, a presidential disaster declaration can bolster the incumbent president’s vote share with 1 percentage point (Reeves, 2011, p. 1149). Further studies into disaster declarations in the United States reveal slightly more complex political dynamics. For the president to declare a disaster and dispatch federal disaster relief, the governors of the disaster-affected states must first make a formal request. Gasper and Reeves (2011) have analyzed the governors’ willingness to do just this. In the period 1972–2006, governors from battleground states appear to have asked above and beyond objective measures of disaster need. The exceptions were term-limited governors who did not face reelection, which appears to have undermined their incentives for such behavior. Sainz-Santamaria and Anderson (2013) have analyzed the electoral politics of disaster preparedness across U.S. counties from 1985 to 2008. Their results indicate that higher electoral returns “incentivize higher disaster preparedness spending until the county is safe enough that electoral patterns are irrelevant for the investment” (Sainz-Santamaria & Anderson, 2013, p. 244). Thus, disaster spending peaks in the most competitive counties. The current tendency for lower electoral competition and fewer swing districts in the U.S. elections (as laid out in the section “Traditional Theories of Voter Retrospection”) might, therefore, have very real repercussions for disaster relief.

Blame Games

In their meta-study of retrospective voting, Healy and Malhotra (2013) discussed whether voter retrospection appeared to incentivize politicians to maximize social welfare. In the context of natural hazards, the discussion would be whether voter retrospection impels governments to implement effective disaster management. It is already apparent that responding to natural hazards takes priority over preventing them. More generally, the fact that voters are critical of the government’s handling of natural hazards does not necessarily compel the government to implement effective disaster policies. A humanitarian disaster is not necessarily a political disaster. And humanitarian assistance is not necessarily good politics (Flores & Smith, 2013. p. 843). The government can use resources to place blame for the disaster on other relevant political actors, such as the opposition, authorities, former politicians, or companies. Hood (2002) described several defensive risk management strategies of blame. The government can use its political capital to try to (a) reverse blame (“the opposition did not allow us to pass the necessary legislation”), (b) displace blame (“the disaster is caused by climate change inflicted on us by the rich countries”), (c) shift blame (“it is primarily a matter for the local governor”), or (d) share blame (“we are all in this together”). In other words, governments might be more concerned with contriving an effective blaming narrative rather than implementing effective policies. While these blame strategies certainly are compatible with effective disaster management, not only do they not require such effective policies but they also make them less likely as they direct political attention away from concrete disaster policies.

Beside these blame games, there are other pluralistic dynamics that could undermine effective disaster management. These dynamics could include everything from direct lobbying conducted by outside groups to “log rolling” and “vote trading” from factions within the political system. Disaster relief could be further prevented by “pork barrel politics” or the corresponding “not in my backyard”, where locally elected politicians place the interests of their home constituents ahead of humanitarian needs (Rubin, 2010; Sainz-Santamaria & Anderson, 2013). A working paper by Collier and Hoeffler (2006) also points to the fact that incumbent governments might engage in illicit electoral tactics (interfering with voter registration, campaigning, and vote counting) to win elections. The response to political pressure from disasters might therefore be translated into covert and illicit government behavior, and not a humanitarian response. Furthermore, it is not always straightforward to identify accountable decision makers in a pluralistic system. Such a system can very well consist of a complex web of democratic institutions, many of which might have mutual/overlapping powers of authority. A multilayered democratic system can dilute accountability, and foster political blame games (Birkland & Waterman, 2008). When the population of Louisiana was asked who was mainly responsible for the lack of emergency aid in New Orleans following hurricane Katrina, 35% said the federal government (the president and the agency for disaster management), 33% pointed to the state government (governor), and 18% said the local government (mayor) (Gomez & Wilson, 2008). It was clearly hard or the voters to ascertain which democratic institution had the ultimate responsibility. This is also supported by Arceneaux and Stein’s (2006) study of the 2001 Tropical Storm Allison, which caused massive floods in Houston. Respondents from Houston who were less knowledgeable about how the local government works were less likely to attribute responsibility correctly. While county-level administrations are responsible for flood preparation, many respondents incorrectly assigned responsibility to other levels of government, such as city or federal levels.

The Role of the Media

As mentioned in the introduction, government responsiveness in times of distress appears stronger in more pluralistic settings (Boin et al., 2005; Flores & Smith, 2013; Kahn, 2005). Nobel Laureate Amartya Sen famously argued that democracies have effectively eradicated the threat of famine (Sen, 1999, 2009). Although the conclusiveness of the statement has been contested (see Devereux, 2007; Rubin, 2009), there is reason to believe that democracies contain safeguards that would effectively prevent major famines akin to the 1990s North Korean famine, estimated to have claimed the lives of a million people (Haggard & Nolan, 2007). An independent and active opposition, a free media, and an informed electorate are integral parts of such safeguards. A free media carries out two essential functions. First, it increases the government’s capability to respond effectively to natural hazards by providing information on distributional impacts of the hazard as well as the performance of the bureaucracy in addressing grievances. Second, a free media exposes mistakes and flawed policies, thus providing incentives for governments to respond effectively to disasters. Much research supports the notion governments respond with greater force to citizen demands when information flows and media penetration are relatively high. Based on a cross-country analysis of 97 countries, Djankov, McLiesh, Nenova, and Shleifer (2003) found that more press freedom is associated with better health and education outcomes. Reinikka and Svensson (2005) concluded that higher newspaper penetration in Uganda appears to spur greater parental engagement in the running of schools , which in turn led to increased school funding after controlling for other factors. Islam (2006) explored the link between information flows and governance quality through an extensive cross-country study that pointed to a strong positive and likely causal relationship between increased transparency measures (for example the adoption of freedom of information acts) and better governance (bureaucratic efficiency). Similarly, Bauhr, Grimes, and Harring’s (2010) cross-country study concluded that transparency deters abuses and mistakes in both democracies and non-democracies. Snyder and Strömberg’s (2010) study of newspaper coverage across U.S. congressional districts also underscored the importance of the free media for electoral accountability: local congressmen worked less for their constituencies in areas with less press coverage and federal spending was lower. Leigh (2009) showed that voters in more educated countries are less likely to be blindly retrospective and punish the politicians for fluctuations in the global economy beyond their control. Media penetration appeared to amplify this effect. Few studies have looked specifically at these dynamics in the context of natural hazards or crisis responses, but those that have support the relationship between media reporting and effective government disaster responses. Strömberg (2004) found that, during the great depression, more relief funds (from the Federal Emergency Relief Administration) were allocated to United States counties with high rates of radio ownership compared to areas with low rates after controlling for the county unemployment and wealth. Besley and Burgess (2002) found that newspaper circulation combined with higher literacy rates across Indian states increased disaster relief after controlling for other factors. Interestingly, this finding could not be replicated in the Indian study of Cole et al. (2012), where voters’ responses to rainfall shocks did not vary with their level of education. Still, the evidence overwhelmingly suggests that literacy rates, media penetration, and a pluralistic political system in synergy have positive effects on government responsiveness in the wake of disasters.

The Behavior of Authoritarian Regimes

Voting behavior in the wake of natural hazards is often not analyzed in authoritarian or semi-authoritarian settings. While almost all regimes hold elections, the existence of a strong opposition, an informed electorate, and a free media are vital elements in retrospective voting. A free media can cover the extent of the natural hazard, and can critically assess the government’s handling of the hazard. The existence of vibrant opposition parties can present the voters with a credible ruling alternative; they might even be able to overturn the government directly though parliamentary procedures (a vote of non-confidence, for example); and they will generally be highly critical of the governments’ handling of the natural hazard and eager to expose any policy mistakes or belated responses. With these institutions underdeveloped in authoritarian regimes, government behavior in the wake of natural hazards primarily rests on political dynamics other than voter retrospection. Many authoritarian regimes build their legitimacy to a large extent on the perception of a strong government capable of taking care of its own population, and they would be most adamant to quickly neutralize the chaos ensued by a natural hazard. Flores and Smith’s (2013) quantitative study (going back to the 1800s) suggested that the political leadership in autocracies might be more vulnerable to disaster occurrences but more resilient to disaster fatalities compared to democracies. The danger for leaders in autocracies, the argument goes, is the potential for disasters to concentrate displaced people and enhance the ability of the disenfranchised to organize (Flores & Smith, 2013, p. 843).

There are some advantages with authoritarian regimes with regards to disaster management that might compensate for a limited retrospective political pressure. Authoritarian regimes can potentially provide a much quicker and more extensive mobilization of resources, as the government does not need to abide by the same political rules as democracies. In democracies, disaster management policies often need to be negotiated in parliament, which can lead to delays and dilutions of disaster relief (Rubin, 2010). The democratic government might have to engage in compromises, negotiations, electioneering, lobbyism, and political games. One of the most famous cases of effective disaster management in authoritarian regimes is Cuba. Cuba has been repeatedly praised by the United Nations for the way it manages natural hazards (UN, 2004, 2011). Cuba has succeeded in organizing a mass-mobilization in case of natural hazards, with all Cubans being obliged to attend disaster drills every year (Bermejo, 2006). The results are noteworthy. In 1998, 600 people died in the Caribbean as a result of hurricane George, of which only 4 were from Cuba. When hurricane Jeanne hit Haiti in 2004, more than 3,000 people died, but when it later hit Cuba with even greater force no one succumbed to the disaster. In certain cases, Cuba has even appeared to provide better protection than the United States, When Hurricane Charley hit Florida in 2004, 30 people died while only four died in Cuba from the same category of wind gusts (Bermejo, 2006).

Regime Change

The shock from natural hazards could be hypothesized to have a destabilizing effect on the authoritarian regimes themselves. Fair et al.’s (2017) study of the 2010–2011 Pakistani floods concluded that hazards can highlight the importance of investing in political knowledge and becoming more politically engaged. The study showed that voters exposed to the floods knew more about politics 5–17 months after the floods. This, however, appears not to have undermined existing political institutions in Pakistan but rather bolstered them. Indeed, empirical evidence indicates that it is rare that natural hazards destabilize regimes. Cavallo, Galiani, Noy, and Pantano (2010) claimed that only two political revolutions in modern times can be identified that were partly triggered by natural hazards. The Iranian revolution in 1979, which replaced an authoritarian monarchy with an authoritarian theocracy, took place in the aftermath of the 1978 earthquake that claimed the lives of 25,000 people. Even though there are few direct links from the earthquake to the revolution, the state’s belated and inadequate response to the earthquake might have provided a much needed window of opportunity for the Islamic revolutionary guard to improve their level of organization and popularity. The 1972 Nicaraguan earthquake, which killed 5,000 people and made 250,000 homeless, may also have paved the way for the 1979 revolution, which dissolved the dictatorship with an authoritarian socialist government. Once again, the connection is not direct, but the lack of emergency aid might have undermined the government’s legitimacy, paving the way for the ousting of the regime seven years later (Cavallo et al., 2010). However, in none of the cases covered did natural hazards lead to fundamental shifts in the political systems; dictatorships did not transform into blossoming democracies. The empirical evidence thus points to the fact that indeed political systems are quite resilient to natural hazards. It is clearly easier for natural hazards to impact political processes and alliances than to lead to major macro shifts in the political system. Such fundamental political changes are only possible in situations where the country’s broader socio-economic factors support such a change. That appears to have been the case with regards to the 1970 Bhola cyclone that, together with the inadequate disaster response of the Pakistani military, facilitated the regime transition to what is now Bangladesh (Hossain, 2018). The cyclone claimed the lives of more than a quarter million people, and it became the main voting issue in the elections held just three weeks later. It strongly contributed to electing the leader of the national opposition, whom the Pakistani military had no intention of allowing to take office. In other words, the cyclone kick-started a chain of events that led to a war of independence and ultimately to the establishment of Bangladesh (Hossain, 2018). Thus, natural hazards can act as a catalyst that either accelerates (or derails) existing processes, but they are not strong enough to directly create the processes (Omelicheva, 2011). The analysis of Carlin, Love, and Zechmeister (2014) of the impact of the 2010 earthquake on trust in the local Chilean government suggested that natural disasters could erode legitimacy in such democracies (at least in the vulnerable transition phase), but that fundamental political changes necessitate that broader socioeconomic factors are conducive to such a change. The empirical evidence thus points to the fact that political systems are indeed quite resilient to natural hazards.


This article has discussed state-of-the art research when it comes to the nexus between natural hazards and voting behavior. Three different strands of voting behavior in the wake of natural hazards were laid out: (a) blind retrospection, where natural hazards will always be a burden for those in power because voters will blindly punish the incumbent government for adverse consequences; (b) mediated retrospection, where natural hazards can both increase or decrease voter support depending on the disaster policies implemented; and (c) narrative retrospection, where the prevailing discourse surrounding the natural hazards is the most significant driver for voting behavior. There is convincing empirical evidence for all three types of voting behavior, and it is likely that retrospective voting behavior includes all three dynamics. There is some indicative evidence, however, that a natural hazard needs to be part of a political narrative to impact voting behavior. It is rare that a political leader exhibits symbolic power only in the wake of natural hazards, or that he or she implements extensive disaster policies without any regard for the resulting political narrative. How to best address the link between natural hazards and voting behavior is ultimately an analytical choice.

The article also presented evidence that voting behavior affects government policies in times of distress. Retrospective voting behavior favors responsive policies over preventive policies, and increased pluralistic competition provides incentives for governments to provide more timely and extensive disaster relief. Media penetration and literacy also have positive effects on the government’s disaster management efforts. Studies have also shown that governments have other viable response-strategies, such as deflection of blame, to retrospective voting behavior in the wake of natural hazards. Rather than expecting a humanitarian response in the wake of natural hazards, therefore, one should expect a political response. Empirical evidence also suggests that political systems in general are very resilient to major natural hazards. Voting behavior and government behavior in the wake of natural hazards are difficult to predict ex ante, but there are some recurrent patterns and dynamics that make it easier to understand and expect certain types of voting behaviors in the wake of natural hazards. While laying no claims of presenting an exhaustive account of all possible contributions that could be related to natural hazards and voting behavior, this article has identified and discussed some of the core patterns and themes that should make it easier to gauge the impact of natural hazards on voting behavior going forward.


Achen, C., & Bartels, L. (2004). Blind retrospection: Electoral responses to drought, flu, and shark attacks. Presented at the Annual Meeting of the American Political Science Association, Boston.Find this resource:

Aldrich, D. (2016). It’s who you know: Factors driving recovery from Japan’s 11 March 2011 disaster. Public Administration, 94(2), 399–413.Find this resource:

Althaus, L., & Kim, M. (2006). Priming effects in complex information environments: Reassessing the impact of news discourse on presidential approval. Journal of Politics, 68(4), 960–976.Find this resource:

Ansolabehere, S., Rodden, J., & Snyder, J. (2008). The strength of issues: Using multiple measures to gauge preference stability, ideological constraint, and issue voting. American Political Science Review, 102(2), 215–232.Find this resource:

Arceneaux, K., & Stein, R. (2006). Who is held responsible when disaster strikes? The attribution of responsibility for a natural disaster in an urban election. Journal of Urban Affairs, 28(1), 43–53.Find this resource:

Atkinson, J., Hicken, A., & Ravanilla, N. (2014). Pork and typhoons: The influence of political connections on disaster response.

Bagues, M., & Esteve-Volart, B. (2016). Politicians’ luck of the draw: Evidence from the Spanish Christmas lottery. Journal of Political Economy, 124(5), 1269–1294.Find this resource:

Bauhr, M., Grimes, M., & Harring, N. (2010). Seeing the state: The implications of transparency for societal accountability. QoG Working Paper Series, 15.Find this resource:

Bechtel, M., & Hainmueller, J. (2011). How lasting is voter gratitude? An analysis of the short- and long-term electoral returns to beneficial policy. American Journal of Political Science, 55(4), 851–867.Find this resource:

Beck, U. (1992). Risk society: Towards a new modernity. New Delhi: SAGE.Find this resource:

Bermejo, P. (2006). Preparation and response in the case of natural disasters: Cuban programs and experience. Journal of Public Health Policy, 27(1), 13–21.Find this resource:

Besley, T., & Burgess, R. (2002). The political economy of government responsiveness: Theory and evidence from India. The Quarterly Journal of Economics, 117(4), 1415–1451.Find this resource:

Birkland, T., & Waterman, S. (2008). Is federalism the reason for policy failure in Hurricane Katrina. The Journal of Federalism, 38(4), 692–714.Find this resource:

Birkmann, J. (2007). Risk and vulnerability indicators at different scales: Applicability, usefulness, and policy implications. Environmental Hazards, 7(1), 20–31.Find this resource:

Blaikie, P., Cannon, T., Davis, I., & Wisner, B. (2014). At risk: Natural hazards, people’s vulnerability and disasters. Abingdon, UK: Routledge.Find this resource:

Bodet, A., Thomas, M., & Tessier, C. (2016). Come hell or high water: An investigation of the effects of a natural disaster on a local election. Electoral Studies, 43, 85–94.Find this resource:

Boin, A., Hart, P., & McConnell, A. (2009). Crisis exploitation: Political and policy impacts of framing contests. Journal of European Public Policy, 16(1), 81–106.Find this resource:

Boin, A., Hart, P., Stern, E., & Sundelius, B. (2005). The politics of crisis management: Public leadership under pressure. Cambridge, UK: Cambridge University Press.Find this resource:

Boin, A., McConnell, A., & Hart, P. (2008). Governing after crisis: The politics of investigation, accountability, and learning. Cambridge, UK: Cambridge University Press.Find this resource:

Busby, E. C., Druckman, J. N., & Fredendall, A. (2017). The political relevance of irrelevant events. The Journal of Politics, 79(1), 346–350.Find this resource:

Bush, G. (2010). Decision points. New York, NY: Crown.Find this resource:

Carlin, R., Love, G., & Zechmeister, E. (2014). Natural disaster and democratic legitimacy: The public opinion consequences of Chile’s 2010 earthquake and tsunami. Political Research Quarterly, 67(1), 3–15.Find this resource:

Cavallo E., Galiani, S., Noy, I., & Pantano, J. (2010). Catastrophic natural disasters and economic growth. Inter-American Development Bank. Department of Research and Chief Economist, Working Paper No. IDB-WP-183.Find this resource:

Cole, S., Healy, A., & Werker, E. (2012). Do voters demand responsive governments? Evidence from Indian disaster relief. Journal of Development Economics, 97, 167–181.Find this resource:

Collier, P., & Hoeffler, A. (2006). Democracy’s Achilles heel or, how to win an election without really trying. CSAE WPS/2009-08.

Dahlberg, R., Rubin, O., & Vendelø, M. T. (Eds.). (2015). Disaster research: Multidisciplinary and international perspectives. Abingdon, UK: Routledge.Find this resource:

Devereux, S. (Ed.). (2007). The new famines: Why famines persist in an era of globalization. Abingdon, UK: Routledge.Find this resource:

Djankov, S., McLiesh, C., Nenova, T., & Shleifer, A. (2003). Who owns the media? The Journal of Law and Economics, 46(2), 341–382.Find this resource:

Downs, A. (1957). An economic theory of democracy. New York, NY: Harper & Row.Find this resource:

Drake, P. (2016). Indonesia and the politics of disaster: Power and representation in Indonesia’s mud volcano. Abingdon, UK: Routledge.Find this resource:

Dynes, R. (2000). The dialogue between Voltaire and Rousseau on the Lisbon earthquake: The emergence of a social science view. International Journal of Mass Emergencies and Disasters, 18(1), 97–115.Find this resource:

EM-DAT. (2017). The International Disaster Database. Centre for Research on the Epidemiology of Disasters.

Eriksson, L. M. (2016). Winds of change: Voter blame and storm Gudrun in the 2006 Swedish parliamentary election. Electoral Studies, 41, 129–142.Find this resource:

Fair, C. C., Kuhn, P. M., Malhotra, N., & Shapiro, J. (2017). Natural disasters and political engagement: Evidence from the 2010–11 Pakistani floods. Quarterly Journal of Political Science, 12(1), 99–141.Find this resource:

Fiorina, M. (1981). Retrospective voting in American national elections. New Haven, CT: Yale University Press.Find this resource:

Flores, A., & Smith, A. (2013). Leader Survival and Natural Disasters. British Journal of Political Science, 43(4), 821–843.Find this resource:

Gallup (2006). Job-approval survey: “Do you approve or disapprove of the way George W. Bush is handling his job as president?.

Gasper, J., & Reeves, A. (2011). Make it rain: Retrospection and the attentive electorate in the context of natural disasters. American Journal of Political Science, 55(2), 340–355.Find this resource:

Gomez, B. T., Hansford, T. G., & Krause, G. A. (2007). The Republicans should pray for rain: Weather, turnout, and voting in US presidential elections. Journal of Politics, 69(3), 649–663.Find this resource:

Gomez, B., & Wilson, J. (2008). Political sophistication and attributions of blame in the wake of Hurricane Katrina. The Journal of Federalism, 38(4), 633–650.Find this resource:

Haggard, S., & Noland, N. (2007). Famine in North Korea. New York, NY: Columbia University Press.Find this resource:

Hart, P. (1994). Symbols, rituals, and power: The lost dimensions of crisis management. Journal of Contingencies and Crisis Management, 1(1), 36–50.Find this resource:

Healy, A., & Malhotra, N. (2009). Myopic voters and natural disaster policy. American Political Science Review, 103, 387–406.Find this resource:

Healy, A., & Malhotra, N. (2010). Random events, economic losses, and retrospective voting: Implications for democratic competence. Quarterly Journal of Political Science, 5(2), 193–208.Find this resource:

Healy, A., & Malhotra, N. (2013). Retrospective voting reconsidered. Annual Review of Political Science, 16, 285–306.Find this resource:

Healy, A., Malhotra, N., & Mo, C. (2010). Irrelevant events affect voters’ evaluations of government performance. Proceedings of the National Academy of Sciences, 107(29), 12804–12809.Find this resource:

Heersink, B., Peterson, B. D., & Jenkins, J. A. (2017). Disasters and elections: Estimating the net effect of damage and relief in historical perspective. Political Analysis, 25(2), 260–268.Find this resource:

Hood, C. (2002). The risk game and the blame game. Government and Opposition, 31(1), 15–37.Find this resource:

Hossain, N. (2018). The 1970 Bhola cyclone, nationalist politics, and the subsistence crisis contract in Bangladesh. Disasters, 42(1), 187–203.Find this resource:

Islam, R. (2006). Does more transparency go along with better governance? Economics & Politics, 18(2), 121–167.Find this resource:

Johnston, T. M., & Goggin, S. N. (2015). Presidential confidence in crisis: Blame, media, and the BP oil spill. Presidential Studies Quarterly, 45(3), 467–489.Find this resource:

Kahn, M. E. (2005). The death toll from natural disasters: The role of income, geography, and institutions. Review of Economics and Statistics, 87(2), 271–284.Find this resource:

Kamradt-Scott, A., & McInnes, C. (2012). The securitisation of pandemic influenza: Framing, security, and public policy. Global Public Health, 7(2), 95–110.Find this resource:

Key, V. (1966). The responsible electorate: Rationality in presidential voting, 1936–1960. Cambridge, UK: Belknap Press.Find this resource:

Kiewiet, R., & Rivers, D. (1984). A retrospective on retrospective voting. Political Behavior, 6(4), 369–393.Find this resource:

Kuttschreuter, M., Gutteling, J. M., & de Hond, M. (2011). Framing and tone-of-voice of disaster media coverage: The aftermath of the Enschede fireworks disaster in the Netherlands. Health, Risk & Society, 13(3), 201–220.Find this resource:

Lazarev, E., Sobolev, A., Soboleva, I., & Sokolov, B. (2014). Trial by fire: A natural disaster’s impact on support for the authorities in rural Russia. World Politics, 66(4), 641–668.Find this resource:

Leigh, A. (2009). Does the world economy swing national elections? Oxford Bulletin of Economics and Statistics, 71(2), 163–181.Find this resource:

Lenz, G. (2011). Understanding and curing myopic voting. APSA 2011 Annual Meeting Paper.Find this resource:

Malhotra, N., & Kuo, A. (2008). Attributing blame: The public’s response to Hurricane Katrina. The Journal of Politics, 70(1), 120–135.Find this resource:

Murphy, C., & Purdum, T. (2009). Farewell to all that: An oral history of the Bush White House.

Omelicheva, M. (2011). Natural disasters: Triggers of political instability? International Interactions, 37, 441–465.Find this resource:

Pelling, M., & Dill, K. (2006). Natural” disasters as catalysts of political action. Chatham House, ISP/NSC Briefing Paper 06/01.Find this resource:

Reagan, R. (1980). October 28, 1980: Debate with Ronald Reagan. University of Virginia, Miller Center.

Reeves, A. (2011). Political disaster: Unilateral powers, electoral incentives, and presidential disaster declarations. Journal of Politics, 73(4), 1142–1151.Find this resource:

Reinikka, R., & Svensson, J. (2005). Fighting corruption to improve schooling: Evidence from a newspaper campaign in Uganda. Journal of the European Economic Association, 3(2–3), 259–267.Find this resource:

Remmer, K. (2014). Exogenous shocks and democratic accountability: Evidence from the Caribbean. Comparative Political Studies, 47(8), 1158–1185.Find this resource:

Roberts, G. (2003). “Taken at the flood”? The German general election 2002. Government and Opposition, 38(1), 53–72.Find this resource:

Rubin, O. (2009). The merits of democracy in famine protection: Fact or fallacy? European Journal of Development Research, 21(5), 699–717.Find this resource:

Rubin, O. (2010). Democracy and famine. Abingdon, UK: Routledge.Find this resource:

Rubin, O. (2015). Natural disasters and politics. In R. Dahlberg, O. Rubin, & M. T. Vendelø (Eds.), Disaster research: Multidisciplinary and international perspectives (pp. 82–96). Abingdon, UK: Routledge.Find this resource:

Rubin, O. (2017). Government disaster response and narrative retrospection: A comparative study of the electoral impacts in Denmark and Sweden of the 2004 tsunami. Disaster Management and Human Health Risk V: Reducing Risk, Improving Outcomes, 173, WIT Press.Find this resource:

Sainz‐Santamaria, J., & Anderson, S. E. (2013). The electoral politics of disaster preparedness. Risk, Hazards, & Crisis in Public Policy, 4(4), 234–249.Find this resource:

Sen, A. (1999). Development as freedom. New York, NY: Knopf.Find this resource:

Sen, A. (2009). The idea of justice. London: Allen Lane.Find this resource:

Shanahan, E., McBeth, M., & Hathaway, P. (2011). Narrative policy framework: The influence of media policy narratives on public opinion. Politics & Policy, 39(3), 373–400.Find this resource:

Shenhav, S. (2006). Political narratives and political reality. International Political Science Review, 27(3), 245–262.Find this resource:

Shroder, J. F. (2014). Hazards, risks, and disasters in society. Amsterdam: Elsevier.Find this resource:

Snyder, J., & Strömberg, D. (2010). Press coverage and political accountability. Journal of political Economy, 118(2), 355–408.Find this resource:

Staupe Delgado, R., Kruke, B. I., Ross, R. J., & Glantz, M. H. (2018). Preparedness for slow onset environmental disasters: Drawing lessons from three decades of El Niño impacts. Sustainable Development.Find this resource:

Strömberg, D. (2004). Radio’s impact on public spending. The Quarterly Journal of Economics, 119(1), 189–221.Find this resource:

Söderlund, P. (2008). Retrospective voting and electoral volatility: A Nordic perspective. Scandinavian Political Studies, 31(2), 217–240.Find this resource:

Survey USA. Clifton, NJ.

UN. (2004). Cuba: A model in hurricane risk management. IHA/943. Meetings Coverage and Press Releases.

UN. (2011). On the occasion of the informal thematic debate of the General Assembly on disaster risk reduction.

Velez, Y., & Martin, D. (2013). Sandy the rainmaker: The electoral impact of a super storm. Political Science & Politics, 46(2), 313–323.Find this resource:

Wasserman, D., & Flinn, A. (2017). Introducing the 2017 Cook Political Report partisan voter index.

Wolfers, J. (2002). Are voters rational? Evidence from gubernatorial elections. Stanford GSB Working Paper No. 1730.Find this resource: