Barking Up the Wrong Tree: How Political Alignment Shapes Electoral Backlash from Natural Disasters

Abstract

While scholarship on “retrospective voting” has found that incumbent politicians can be punished for a range of events outside their control, the literature has paid scant attention to the role of political alignment between the different levels of government in disaster responses and its implications for voting decisions. We argue that retrospective voters punish only opposition incumbents (candidates in office but not aligned with the government leader), who have limited access to government resources for relief, for natural disasters. We use monthly data on precipitation and evaporation to capture droughts and floods in India’s four thousand State Assembly electoral constituencies over the years 1977–2007. Consistent with our hypothesis, we find that Members of State Assembly from the party of the Prime or Chief Minister do not face an electoral backlash under bad weather conditions during the monsoon season, whereas opposition politicians face major losses.

Keywords

retrospective voting natural disasters drought India alignment

Introduction

Scholarship on “retrospective voting” has found voters punish governments for exogenous events, such as natural disasters, terrorist attacks, or crime waves (Fiorina, 1981; Healy & Malhotra, 2013; Key, 1966). In studies from the United States and India, for example, there is a robust relationship between floods or droughts and an electoral backlash against the incumbent (Cole et al., 2012; Malhotra & Kuo, 2008). However, the relationship appears to be complicated by the government’s ability to provide voters relief (Bechtel & Hainmueller, 2011; Chen, 2013; Gasper & Reeves, 2011; Healy & Malhotra, 2009, 2010).

We argue that certain incumbent politicians are systematically better positioned to provide relief to their constituents. This variation shapes incumbents’ likelihood of being punished for natural disasters. In particular, because local politicians are often dependent on the assistance of policymakers at the state or national level, local officials who are aligned with politicians at higher levels of government are advantaged in efforts to obtain relief funding. As a result, incumbents who are aligned with the party in power at the state or national level will be less likely to be electorally punished in the aftermath of a disaster.

Existing studies tend to focus either on the effects of disasters on electoral outcomes (Cole et al., 2012; Malhotra & Kuo, 2008) or on the political determinants of relief (Bastos & Miller, 2013; Reeves, 2011). Other studies find that relief mitigates electoral backlash (Bechtel & Hainmueller, 2011; Chen, 2013; Cole et al., 2012; Gasper & Reeves, 2011; Healy & Malhotra, 2009, 2010), and investigate voter preferences over relief allocation (Bechtel & Mannino, 2017). However, existing studies leave open the question of why some politicians are able to provide more relief than others. Our theory provides a unified explanation, at least within this institutional context, that links these disparate phenomena—natural disasters, the ability to provide relief, and electoral outcomes—by addressing a key condition under which relief is more likely: when incumbents have allies in the party in power in the state or national government.

To test this theory, we use monthly values of the climatic water balance for India’s 4,313 state electoral constituencies during the 1977–2007 period. The advantage of our approach is that we can measure weather patterns in political constituencies—the relevant electoral unit—instead of the much larger administrative districts. We can also support our identifying assumptions by showing that constituency characteristics, based on a comprehensive enumeration of all Indian villages from the 2001 Census of India, are not correlated with abnormal weather.

Estimating both parametric and non-parametric regressions, we show voters penalize incumbent politicians (i.e., Members of State Assembly, MLAs) for natural disasters. However, the logic of these penalties is fully consistent with retrospective voting: the punishment is limited to opposition incumbents—those who are not aligned with the nation’s prime minister (PM) or the state chief minister (CM). Moreover, the effects are large—non-aligned incumbents during severe weather anomalies lose 2% to 5% of their vote share. Finally, results are not driven by differences in incumbent quality; the results are robust to including incumbent candidate fixed effects.

We also shed light on the mechanisms of electoral backlash. First, we demonstrate that drought and heavy rainfall do not affect voter turnout. Instead, voters move their support from non-aligned incumbents to other candidates. Second, using household data, we show that drought and heavy rainfall are associated with a reduction in people’s confidence in politicians. Finally, the same household data demonstrates that people who have access to relief are less likely to punish politicians for natural disasters.

The findings provide new insight into natural disasters and voting behaviors in a federal structure by showing the role of political alignment across different levels of governments in disaster response. While several studies have theorized how a federal system affects voting decisions (e.g., Arceneaux, 2006; Carsey & Wright, 1998) and suggested the conditions under which voters attribute blame to local governments in the wake of natural disasters, such as salience and direct experience of the events (Arceneaux & Stein, 2006), existing studies have generally overlooked the implications of political alignment. Our study, however, shows that structural advantages from political alignment, leading to greater relief assistance, is an important explanatory variable for voting decisions after extreme weather events.

The implications for democratic accountability are troubling. If people punish non-aligned incumbents, they are not rewarding disaster relief in a rational, forward-looking sense. Instead, they are naively punishing non-aligned incumbents for forces beyond their control. In Cohen and Werker’s (2008) model of responding to natural disasters, governments have incentives to withhold relief from areas controlled by the political opposition, and we provide evidence for these incentives. Not only does the government have an incentive to focus on areas that it controls, but it also has an incentive to penalize areas that have elected members of opposition parties to the State Assembly.

Natural Disasters and Electoral Politics

Natural disasters are not caused by government, but the damage they cause depends on the government’s preparedness and relief efforts. To account for this crucial distinction, we theorize about the assignment of blame and the resulting electoral backlash for different types of incumbents.

Related Literature

Studies on the political effects of natural disasters suggest that voters react by reducing support for incumbent politicians (Cole et al., 2012; Malhotra & Kuo, 2008). Such behavior is consistent with “retrospective voting” (Stokes, 2016), whereby voters assess changes in welfare during the incumbent’s tenure and vote accordingly. Natural disasters have been found to undermine the legitimacy of the state and erode public confidence in its democratic institutions (Carlin et al., 2014; Córdova & Seligson, 2010). A recent body of work, however, adds that disaster relief can mitigate these negative effects. Cole et al. (2012), for example, note that the negative effect of droughts and heavy rainfall on Indian incumbents is mitigated by the provision of calamity relief. Other studies find similar results in the United States and Germany, showing that voters reward disaster relief (Bechtel & Hainmueller, 2011; Chen, 2013; Gasper & Reeves, 2011). In a pair of related studies, Healy and Malhotra (2009, 2010) show that people reward the government for disaster relief but not for (more cost-effective) disaster preparedness.

These results are consistent with the standard notion of retrospective voting, but the implications for democratic accountability are less troubling. Although incumbents are punished for natural disasters, this attribution failure does not mean that voters are unable to hold the incumbents accountable. Incumbents can still act to control the damage. Ideally, voters would also reward disaster preparedness (Healy & Malhotra, 2009), but the rewards for disaster relief still encourage governments to act on natural disasters.

However, other studies suggest the provision of relief can be politicized. Reeves (2011) finds that US presidential disaster declarations are more common in “swing states,” while Bastos and Miller (2013) find that Brazilian mayors who are members of the President’s party are more likely to receive drought declarations. These findings raise questions about democratic accountability, as they suggest local-level politicians not connected to the party in power or not in politically pivotal areas may suffer at the voting booth, not because of their own failures, but because of an unwillingness on the part of national- or state-level politicians to provide relief.

Droughts, Rainfall, and Elections in India

The Indian electoral system is based on single-member electoral constituencies in state-level legislative assemblies and the national-level parliament. In 2019, there were 543 parliamentary constituencies and 4,116 legislative assembly constituencies. The electoral constituency boundaries are drawn by delimitation commissions based on population shares. Some seats are reserved for candidates from Scheduled Castes or Scheduled Tribes, which are historically underrepresented demographics protected under the 1951 Constitution. Elections are typically held every 5 years, and after the elections a ruling coalition is formed.

Uppal (2009) documents a severe incumbency disadvantage in India, wherein incumbents tend to fare worse simply by virtue of their currently being in power. A contributing factor to this disadvantage could be that incumbents are being blamed for factors outside their control, such as the effects of extreme weather conditions, rather than for lack of effort. Studies of retrospective voting are, therefore, particularly relevant to Indian politics because frequents droughts and floods caused by heavy seasonal rainfall make incumbents vulnerable to electoral backlashes.

It bears remembering that India’s population remains predominantly rural and dependent on agriculture. While India’s rapid urbanization has decreased agriculture’s portion of GDP to about 16.1%, it remains the primary livelihood for 70% of the population (Food and Agriculture Organization of the United Nations, 2016). This agricultural production is heavily dependent on rainfall during the monsoon season. India also suffers from insufficient water storage—with capacity for storing only 30 days of rainfall, compared to 900 days in developed countries.¹

Droughts and extreme rainfall cause significant damage in India. While droughts since India’s independence in 1947 have not resulted in mass famines, they remain devastating. In April 2016, India experienced a drought that affected 330 million people, and left millions without a steady water supply.² Nor are droughts uncommon. In Rajasthan, for example, one analysis suggested that there were 48 droughts between 1901 and 2003, and that some part of Rajasthan is affected by drought almost every year (Rathore, 2004). India is also susceptible to massive flooding. In 2015, a flood in South India displaced more than a million people and caused an estimated USD 3 billion in damages.³ In the state of Bihar, with a population over a hundred million, three in four people live “under the recurrent threat of flood devastation.”⁴

Efforts of political parties at the state and central level play a major role in the declaration of disaster and subsequent relief (Rathore, 2004, p. 16). Normally the state is charged with declaring a disaster and requesting the central government’s assistance. Such a declaration is supposed to be based on the directives in the Famine Code, but these guidelines are often violated for political and administrative reasons (Rathore, 2004). Once the decision is made to provide relief, the State Relief Manual sets the guidelines for relief efforts. Relief efforts include monetary relief through loans and grants, supply of seeds and fertilizers, food and water supplies, animal feed, and cash grants for vulnerable populations (children, elderly, etc.). The affected population can also be engaged in work efforts, such as deepening ponds, checking dams, and repair of roads (Prabhakar & Shaw, 2008).

Incumbency, Alignment, and Electoral Outcomes

The key distinction underpinning our theory is the difference between incumbency in the assembly constituency and political alignment with the state or central government. The combination of punishing incumbents and rewarding relief means that natural disasters hurt the electoral prospects of incumbents not aligned with state or central governments, but does not hurt incumbents with these affiliations. If this expectation holds, voter behavior is consistent with simple retrospection. We offer the intuition here, but Section A1 offers a political economy model of disaster relief and its electoral consequences. In the model, we consider both naive and sophisticated variants of retrospective voting to ensure that our hypotheses are based on first principles.

While the expectation that incumbents are punished follows the basic logic of retrospective voting, the distinction between aligned and non-aligned candidates is unique. We define alignment on the basis of partisanship, following a large body of literature in India (Arulampalam et al., 2009; Uppal, 2009) and elsewhere (Fouirnaies & Mutlu-Eren, 2015). An incumbent MLA is said to be aligned with the chief minister (prime minister) if he or she is from the same party. Within the context of India, it is co-partisanship that structures the relationship between MLAs and higher levels of government when it comes to relief efforts (Golden & Min, 2013). Also, because co-partisanship determines control of the state legislature, most state and central officials will care primarily about an MLA’s partisanship.⁵ In the Indian system, political alignment occurs across multiple levels. Both the central and state governments can allocate funds and implement various disaster relief schemes, and we expect that political alignment on either level helps citizens hold the incumbent accountable for outcomes.

Because local politicians have no autonomous control over disaster relief, they need assistance from the state or national government in order to help their constituencies. Historically, relief expenditures have come largely from the central government (Rathore, 2004, pp. 15–16). Since 1991, Indian drought relief is handled by state-level Calamity Relief Funds (CRFs) and, if those funds are insufficient, by the central government’s National Calamity Contingency Fund (NCCF). Seventy-five percentage of the states’ CRFs come from the national government.

Whether an incumbent is a member of the same party as the chief minister or prime minister is important because being connected to the higher-level party can facilitate incumbents’ efforts to obtain relief funding. This argument reflects both the institutional design of the Indian disaster relief system and the role of MLAs in the Indian political system more generally. Because of India’s extremely strict party discipline rules, the average MLA has little ability to influence policy and can be instead considered a broker between the government and the population living in his or her constituency (Chopra, 1996; Jensenius, 2013). In the case of disaster relief, the MLA’s role is to help the local population use the various relief mechanisms offered by the central and state governments. The MLA achieves this goal by requesting resources, monitoring the work of the local bureaucracy, and channeling citizen demands to the bureaucracy and political officials at higher levels.⁶

The role of political alignment can be readily in seen in fall 2015 droughts in Madhya Pradesh, a state governed by the Bharatiya Janata Party (BJP) at the time. The party leader of the Indian National Congress in the national parliament made public accusations that “the BJP government in the state has failed to make serious efforts providing relief to farmers . . . Referring to his Parliamentary constituency, Guna, the senior leader claimed that due to farm woes, farmers in Nai Sarai, Pachavali and Bamori villages have committed suicide.”⁷ Although the accusation was made by a high-level national politician, the basic logic was one of blaming a state government for not providing relief to farmers in dire need in one’s own electoral constituency.

Although voters may punish aligned incumbents for a natural disaster, they have better access to resources for disaster relief that can neutralize the negative effect of natural disasters on their political survival under naive retrospective voting.

Hypothesis 1. Under naive retrospective voting, natural disasters have no effect on the electoral performance of incumbent politicians aligned with state or federal government.

On the other hand, non-aligned incumbents suffer under bad weather and their lack of political connections prevents them from securing disaster relief. For example, Cohen and Werker (2008) find that governments have incentives to withhold disaster relief from areas controlled by the political opposition. Thus, contra accounts such as Cole et al. (2012), we expect non-aligned incumbents to bear the brunt of the negative electoral effects of natural disasters.

Hypothesis 2. Under naive retrospective voting, natural disasters have a negative effect effect on the electoral performance of unaligned incumbent politicians.

The non-aligned incumbents are, of course, not helpless in the face of natural disasters. As noted above, the Indian media often criticizes the government for neglecting certain areas in disaster relief. Non-aligned incumbents can complain in public that the government is neglecting his/her constituents’ plight. The effectiveness of this strategy is limited, however, if citizens primarily use their own current situation as the primary yardstick for assessing incumbent politicians. Theories of sophisticated voting that apply in the United States (e.g., Arceneaux & Stein, 2006) are unlikely to apply to Indian politics between 1977 and 2007, considering the low level of educational attainment; in 1991, the Census revealed a literacy rate of 52%. As Banerjee et al. (2014, p. 404) note based on a series of randomized controlled trials informing Indian voters of candidate characteristics, inability access information is a major problem for campaigns that attempt to shape voting by informing voters. Indian voters do react to relevant information, but low levels of literacy and media consumption make the use of such information very difficult (Chandra, 2007; Verma, 2012).

It is thus unlikely that most voters could distinguish between aligned and non-aligned incumbents’ ability to act. We hypothesize that they vote based on their current situation, as in models of retrospective voting, without sophisticated attribution of blame or efforts to distinguish between different levels of government (in India, state versus central). Indeed, if voters were informed and voted prospectively, we might expect to see non-aligned incumbents be replaced with aligned incumbents in anticipation of future natural disasters. We hypothesize, instead, that the vast majority of Indian voters use far simpler heuristics consistent with retrospective voting. If they face a natural disaster and suffer greatly from it because relief is not made available by government, they vote against their non-aligned incumbent. This form of electoral behavior is likely when access to reliable information is constrained, as one would expect to be true for India over the four decades of study we study.

The Role of Turnout

One important alternative hypothesis in the Indian context is a mechanical turnout effect: if bad weather conditions prevent people from voting, an electoral backlash against non-aligned MLAsmay simply reflect changes in the composition of voters.⁸ To assess this explanation, we also estimate the effects of natural disasters on turnout.

Research Design

To estimate the effects of droughts and heavy rainfall on Indian elections, we use data from 4,313 MLA constituencies over the years 1977–2007. The unit of analysis is a constituency-election, with Indian state elections (N = 195) typically held every 5 years. This is the appropriate unit of analysis, as the outcomes are measured at this level. Even if an MLA must rely on district or state administration for relief, voters can only influence partisan control of the state legislative assembly through their own MLA, given India’s first-past-the-post electoral system. Moreover, research on MLAs (Chopra, 1996; Jensenius, 2013) shows that they have a lot of authority, both formal (e.g., discretionary spending accounts) and informal (e.g., political connections), that they use to capture relief resources in a drought or flood.

Using this dataset, we investigate the effects of precipitation conditions in the constituency during the previous monsoon period on electoral outcomes (vote shares, turnout). As detailed in Table A1, this results in 29,566 constituency elections held throughout the year. While elections are spread throughout the months of the year, most are held in May (54), February (49), November (26), or March (20).

The general idea behind our identification strategy is that we only exploit variation in quasi-random weather conditions within constituencies while also controlling for election-specific shocks. In other words, we estimate the average effects of variation in weather conditions within a constituency but also assume that all constituencies within a state face a new common shock in every election. To account for heterogeneity by constituency i and election t, we include constituency and election fixed effects.⁹

We allow a non-linear relationship between precipitation and electoral outcomes in a quadratic specification, following existing studies that found a quadratic specification performs relatively well in accounting for nonlinear effects of weather (e.g., Cole et al., 2012; Lobell et al., 2011).¹⁰ We also conduct non-parametric locally weighted scatter-plot smoothing (LOWESS) analysis (Cleveland, 1979). The advantage of the LOWESS approach is that it fits the data without strong assumptions about functional form. It is also particularly well-suited to data in which much of the interest is at the extremes of the independent variable.

For the LOWESS analysis, we plot 95% confidence intervals around the estimates for statistical inference. These confidence regions are calculated using a symmetric nonparametric bootstrap with 1,000 simulations, evaluated on a fixed number of equally-spaced x values (Efron & Tibshirani, 1993).¹¹ Also, we test for influential observations by performing a diagnosis of the main models through partial residual plots, which represent the residual after the other variables have been taken into account (Myers, 1990). The partial residual plots show weak evidence of the presence of strong outliers.¹²

Additionally, in Sections 5 and 6 we provide micro-level evidence on the causal mechanisms using household-level survey data on the relationship between extreme weather and individual households’ confidence in politicians. We present evidence showing that this relationship is contingent upon whether a households receives government benefits, with those households that do exhibiting higher confidence in politicians and state and local governments.

Dependent Variables

Our primary dependent variables are the incumbent party’s vote share (0–100), the PM-aligned candidate’s vote share (0–100), and the CM-aligned candidate’s vote share (0–100).¹³ Our hypotheses lead us to expect that there will be a heterogeneous effect of SPEI on incumbent vote shares, with non-aligned incumbents suffering at the polls more than aligned incumbents.

We employ a split-sample approach in order to test for whether incumbents who are PM- or CM-aligned are electorally punished differently from those who are not. To identify PM-aligned and CM-aligned candidates, we compare the partisanship of a candidate with the partisanship of the PM and the CM at the time of the election.¹⁴ We demonstrate the results are robust when coalition members are coded as aligned (Figures A21–A22).¹⁵ In addition, we estimate our models with voter turnout (0–100) as the dependent variable Jensenius (2013).

Explanatory Variables

Measuring precipitation can be difficult on a large scale. While drought is defined on the basis of impacts, these are frustrating to quantify and reports of drought are likely to be plagued by subjectivity. Instead, we utilize a climate drought index from the global Standardized Precipitation- Evapotranspiration Index (SPEI) dataset. This data is derived from the monthly precipitation and potential evapotranspiration measurements on a 0.5 degree global gridded scale (Begueria et al., 2014; Pozzi et al., 2013). The SPEI is calculated on a monthly scale representing the climatic water balance—precipitation minus potential evapotranspiration. The values are standardized over time and across space, allowing us to capture the deviation of the climatic water balance from the long-term mean, with a mean of zero and a standard deviation of one.

SPEI has been used in a variety of situations and has shown better performance than some other drought indices (Begueria et al., 2014). While all of these are generally correlated, SPEI has been found to have a better correlation with ground conditions than the Standardized Precipitation Index (SPI) because it takes into account the evapotranspiration (Vicente-Serrano et al., 2012). It also generally out-performs the Palmer Drought Severity Index (PDSI), or the self-calibrating PDSI (SC-PDSI) for capturing both short- and long-term drought situations, while both versions of PDSI tend to be relatively insensitive to both short-term droughts and drought recovery (Vicente-Serrano et al., 2012).

The main advantage of SPEI is that it encompasses global data, on a monthly level, for the entire period of our study. Figure 1 shows examples of the SPEI data for India in 1977 and 2007. Negative values show a greater water deficit, while positive values suggest a higher than normal water balance. The resulting data produces 1,090 grid values in India per month recording the SPEI.¹⁶

Figure 1.

Examples of SPEI data from 1977 and 2007.

For larger constituencies, we calculated the mean monthly SPEI from 1957 to 2007. Most constituencies, however, were relatively small relative to the grid size for the SPEI data. For these constituenties, we recorded the SPEI value at the constituency centroid (see Burlig & Preonas, 2016).

Our explanatory variable is the average monthly SPEI value during the previous monsoon period. Rainfall in the monsoon season accounts for more than 75% of annual rainfall in India and is the most critical for agricultural products (Jayachandran, 2006). Recent studies on weather shock in India focus on the monsoon season to gain more precise estimates by excluding the weather events of relatively little economic importance (Bhavnani & Lacina, 2017; Blakeslee & Fishman, 2018; Cole et al., 2012).

India experiences two distinct monsoons—the South-West monsoon (July-September) and the North-East monsoon (October—December). Though rainfall received during the South-West monsoon accounts for a large portion of the annual rainfall in most regions, some states in south peninsular India, such as Tamil Nadu, Andhra Pradesh, Karnataka, Kerala and Orissa, receive considerable amount of rainfall during the North-East monsoon season (Rajeevan et al., 2012). To account for rainfall during these two monsoon periods, we calculate the average values of SPEI from July to December in the constituencies in Tamil Nadu, Andhra Pradesh, Karnataka, Kerala and Orissa and from July to October in those in other states.¹⁷ For elections between October in year t (January in year t + 1 for the five states affected by the North-East Monsoon) and July in year t + 1, we use the value of the last monsoon in year t. For elections during the monsoon period, we use the average value of the ongoing monsoon.

Using drought and election data at the constituency level allows us to conduct an analysis that has more power and is more fine-grained than previous district-level analysis (e.g., Cole et al., 2012). Many districts within India are quite large, and feature significant variation in weather conditions within them that.

In order to demonstrate the validity of our measurement, we compare the SPEI values to official Indian rainfall data at the weather zone-level for the monsoon months of 1987, 2000, and 2002, as well as rainfall data at the district-level for all of 1977–2002. The results shown in the appendix (Tables A6–A8) suggest an extremely strong and positive relationship between SPEI and rainfall. We also test whether SPEI is correlated with crop production on the district level and Table A9 demonstrates a strong, curvilinear relationship.

Identifying Assumptions

In Section A3, we provide balance statistics to test the assumption that the variation of SPEI measure is quasi-random. We, unfortunately, do not have time-varying covariates for electoral constituencies, but we can still test our identifying assumption by investigating whether the SPEI—and its extreme values in particular—predict village characteristics in the 2001 Census of India. If we are correct to argue that the SPEI can be considered quasi-random, we should see no correlation. The evidence supports the assumption. Tables A3 and A4 show that observations with higher and lower than average levels of SPEI for their constituency—both during the previous month and during the previous monsoon—do not vary significantly from observations with average SPEI levels on the following dimensions: area (hectares); household numbers; population; male and female population; distance to town (km); number of children; literate population; population of scheduled castes and tribes.

There are four primary threats to inference. The first is climate change over time, which has already affected drought and rainfall conditions in India (Turner & Annamalai, 2012). If changes in climate differ across states and are correlated with state characteristics, then we may draw incorrect inferences from the association between weather conditions and electoral outcomes. Thus, we account for unobserved heterogeneity across elections so that our estimations only exploit variation in outcomes across constituencies within a state in a given election-year.

The second threat is within-year variation. Given India’s large agricultural population, precipitation is much more important during the monsoon season, when rainfall is used to water crops, than during other times of the year. Fortunately, the SPEI provides us with monthly, as opposed to annual, estimates, and we can thus focus on the monsoon period in particular.

The third is across-constituency variation. Weather conditions differ significantly across constituencies, and conditions that count as normal in the desert of Rajasthan might be disastrous in areas with typically abundant rainfall. As noted earlier, we control for time-invariant features of MLA constituencies. Additionally, we show that the results are also robust to the use of incumbent party fixed effects rather than constituency fixed effects (see Tables A16–A18), and to controlling for the number of candidates running in each constituency-election, which may be systematically related to weather conditions (see Table A19).

Finally, the functional form of the relationship between precipitation and electoral outcomes is potentially complex. To deal with this issue, we not only report results from quadratic specifications but also estimate non-parametric local regressions to check whether the assumption of a quadratic polynomial function is reasonable.

Findings

We report our estimation results in graphical form. In the figures, we present LOWESS and quadratic prediction results utilizing demeaned independent and dependent variables with regard to constituency and election to remove heterogeneity by constituency and election. We present our results in two sets. First, we show that while voters punish non-aligned politicians for extreme weather, aligned incumbents do not suffer the same penalty. Second, we show that extreme weather does not result in a major reduction in turnout, suggesting that electoral backlash must be largely ascribed to a change in the proportion of voters voting against the incumbent.

Effects on Incumbent Vote Shares

We begin with a graphical illustration of our main results in Figure 2. On the left side, panels (a), (c), and (e) show the LOWESS estimations and a histogram of the SPEI index. On the right side, in panels (b), (d), and (f), we show the quadratic relationship from the parametric regressions. To facilitate comparisons, the x-axes and y-axes are identical across the six panels.¹⁸

Figure 2.

Effects of precipitationon electoral results: (a) vote share (incumbent): LOWESS plot, (b) vote share (incumbent): quadratic prediction plot, (c) vote share (PM): LOWESS plot, (d) vote share (PM): quadratic prediction plot, (e) vote share (CM): LOWESS plot, and (f) vote share (CM): quadratic prediction plot.

Beginning with the incumbent vote, we see an inverted U-shape in both estimations. In the LOWESS estimations, there is no systematic relationship between the incumbent vote share and SPEI values in previous monsoon period remain between ‒0.5 and 0.5. More extreme values, however, cause rapid decreases of the incumbent’s vote share, with drought conditions (more than below the expected value) resulting in a loss of around 2 percentage points and heavy rainfall (more than 0.5 above the expected value) resulting in a loss of up to around 4 percentage points. Considering the average vote share of first-place candidates in our dataset is 47%, while that of second-place candidates is almost 33%, these are significant effects.

The quadratic estimation reveals a similar pattern with a statistical significant second-order polynomial, but the relationship is less clear. In particular, drought conditions are not associated with reduced vote shares and heavy rainfall is only associated with about a 1 percentage point loss of votes. The dampening of the effect reflects the large number of data points in the middle, where no relationship exists between the SPEI index and voting.

The remaining four panels show that there is no relationship between the SPEI index and vote shares of PM-aligned or CM-aligned candidates. The LOWESS estimates are essentially flat lines in both cases, and the quadratic estimations confirm the lack of both linear relationships and curvature.

These relationships are inconsistent with theories such as Cole et al. (2012), which emphasize voters’ rational decisions to punish candidates aligned with the state or central governments. However, they are entirely consistent with standard accounts of retrospective voting: voters punish their own representative, even if this representative is essentially powerless to provide relief.

Effects on Aligned and Non-Aligned Incumbents

Our theory predicts that incumbents aligned with the government are not punished by retrospective voters because these incumbents can secure relief funding.

In Figure 3, we distinguish between incumbents aligned and not aligned with the PM’s party. Panels (a) and (c) show the LOWESS and quadratic estimations under alignment; panels (b) and show the same estimates without alignment. The results show that incumbents not aligned with the PM’s party receive lower vote shares in the aftermath of extreme weather. However, there is no such relationship with the electoral fortunes of aligned incumbents. This pattern is consistent with the retrospective voting theory.

Figure 3.

Effects of precipitation on PM-aligned and non-PM-aligned incumbents: (a) vote share (PM incumbent): LOWESS plot, (b) vote share (non-PM incumbent): LOWESS plot, (c) vote share (PM incumbent): quadratic plot, and (d) vote share (non-PM incumbent): quadratic plot.

Figure 4 replicates the analysis focusing on incumbents aligned or unaligned with the state CM. If anything, the effect of droughts and heavy rainfall on unaligned incumbents is larger than in the main or PM-aligned analysis. At the same time, there is no robust relationship between droughts, heavy rainfall, and the vote share of CM-aligned incumbents. Again, voters’ punishment is focused on those incumbents who can influence neither the weather nor the relief effort.¹⁹

Figure 4.

Effects of precipitation on CM-aligned and non-CM-aligned incumbents: (a) vote share (CM incumbent): LOWESS plot, (b) vote share (non-CM incumbent): LOWESS plot, (c) vote share (CM incumbent): quadratic plot, and (d) vote share (non-CM incumbent): quadratic plot.

In Table A11, we also present quadratic regression outputs in table format. The results suggest null effects for aligned incumbents with the PM’s party or with the state CM’s party. In contrast, for non-PM-aligned and non-CM-aligned incumbents, the coefficients for the quadratic terms of the average monthly SPEI value during the previous monsoon period are negative and statistically significant at the conventional level, which provides evidence of concave relationships between precipitation and their vote shares. The results from a joint test of the linear and quadratic co-efficients in each group also show that the effect of droughts and heavy rainfall is statistically significant only for the non-aligned incumbents at the conventional level. Hence, we perform a statistical test of the presence of a non-linear relationship in finite samples, suggested by Lind and Mehlum (2010). The test results provide evidence for inverse U-shaped relationship in the non-aligned samples (Tables A13 and A14).²⁰ We also examine whether the differences in the effects between the aligned and non-aligned incumbents are statistically distinguishable from 0 by conducting a Wald chi-square test. The test statistics suggest that the effects are statistically different between the groups: the χ² statistics for the difference between PM-aligned and non-PM-aligned are 4.58 (p = .03) and 2.72 (p = .09) for the difference between CM-aligned and non-CM-aligned incumbents.²¹

We also assess the marginal effects of the previous monsoon SPEI value on electoral outcomes (Section A7). The results suggest the negative electoral effects are clear when the previous monsoon SPEI is more than one standard deviation (0.5) below or above the mean value (–0.1) of the previous monsoon SPEI: vote shares of non-aligned incumbents are expected to decrease by around 1.2–1.5 percentage points for drought conditions (up to 2.4 percentage points) and by around 1.3–1.8 percentage points (up to 3.8 percentage points) for heavy rainfall conditions.²² The estimated effect sizes are substantial, considering that around 10% (21%) of MLA elections falls within the 2% (4%) margin of victory for the period of 1977–2007.

These findings provide strong support for our theoretical expectations. Incumbents are only punished for flood and drought conditions when they are not aligned with the party in control of either the chief minister or prime minister. We expect that these relationships are the result of the greater ease with which CM- and PM-aligned incumbents can provide relief assistance. The results in Table A21 further suggest that when the incumbent is not aligned, aligned candidates’ vote share tends to increase in periods of extreme weather.²³

Some studies on economic voting in India suggest that incumbency and party system institutionalization play important roles (Chhibber et al., 2014; Jensenius & Suryanarayan, 2020; Ravishankar, 2009). In particular, Jensenius and Suryanarayan (2020) show that economic voting is weak in the places where the incumbent party did not re-nominate the incumbent candidate. This suggests that our findings of weak electoral punishment for the aligned incumbent parties can be driven by weak party-candidate linkages which enable the incumbent parties to recruit new candidate to attenuate blame under extreme weather conditions. To investigate this possibility, we check if our findings hold for constituencies where the incumbent candidate was rerunning for the same party aligned with PM or CM. Overall, the LOWESS and quadratic estimation plots (Figure A23 and A24) show weak evidence that these aligned rerunning candidates are electorally punished in the wake of extreme weather conditions, suggesting that weak party-candidate linkages do not drive our main findings of weak electoral punishment for the aligned incumbents.²⁴ Moreover, we estimate our main analyses by controlling for the strength of party organization measure from Chhibber et al. (2014). The results (Section A16) show that our main findings remain substantively the same when accounting the strength of party-incumbent linkages and party organization.²⁵

We further consider the more detailed specifications on party alignment across the federal structure of India. Specifically, we account for four different kinds of incumbents: full opposition incumbents (not aligned with either the Prime or Chief Minister), two types of semi-government incumbents (aligned with either the Prime or the Chief Minister, but not both), and PM-CM incumbents (aligned with both the Prime and the Chief Minister). The results, presented in Figures A19–A20, suggest substantively similar findings: voters punish only non-aligned incumbents for flood and drought conditions. Finally, we also run the analysis by defining alignment in terms of ruling party coalitions at the national- and state-level, rather than alignment with just the prime or chief minister’s party. The results in Figures A21–A22 show that our main finding holds.

Effects on Turnout

Figure 5 tests for the possible mechanism of electoral effects through turnout. Interestingly, both the LOWESS and the quadratic estimation suggest an asymmetric effect. Drought conditions seem to reduce turnout, consistent with the notion that the distress caused by the drought reduces people’s ability to participate in the democratic process. However, heavy rainfall has no such effect. Given that we saw large negative effects of both droughts and heavy rainfall on non-aligned incumbent vote shares, it seems that changes in turnout thus cannot explain changes in the vote shares of incumbent MLAs. At least some of the effect can be attributed to people voting against non-aligned incumbents.

Figure 5.

Effects of precipitation on turnout: (a) turnout: LOWESS plot and (b) turnout: quadratic prediction plot.

While we can rule out a purely mechanical effect as an explanation, we must remember that turnout can still play a role in a model of retrospective voting. It is possible, for example, that the potential supporters of the incumbent MLA are less likely to vote because of natural disasters, while some people choose to vote exactly because they want to punish the incumbent MLA (see Chen, 2013). This mechanisms is entirely consistent with retrospective voting, as people choose whether and how to vote based on their past experiences.

The data on MLA elections that we have used so far allow us to show the effect of extreme weather on the electoral fortunes of aligned vs. unaligned incumbents. However, these data cannot tell us how individual households react to extreme weather or to the presence of government relief benefits. In the next two sections, therefore, we rely upon household-level data on confidence in politicians and state and local governments, as well as households’ access to government benefits, in order to present evidence on both. Section 5 describes the effect of extreme weather on household confidence in political figures, while Section 6 shows that this effect is moderated by the presence of government relief benefits.

Mechanisms: Drought, Rainfall, and Household Public Opinion

In this section, we use household survey data to examine whether drought and heavy rainfall reduce people’s confidence in politicians, state governments, and local governments. Our expectation is that there will be a negative effect of bad weather on confidence in politicians as a whole, but not in local and state governments as they have the ability to offer relief.

Data and Methods

The India Human Development Survey (IHDS) is an India-wide social and economy household survey. In 2004–2005 and 2011–2012, 41,554 households were interviewed about a wide range of issues, including public confidence in various institutions.²⁶ While the survey only reveals the district (of which there are 375) of the household for privacy reasons, we can see if district-level weather conditions during the pre-survey monsoon period influence the household’s responses. We provide summary statistics for the variables in Table A24 in the appendix. Because most house- holds were interviewed twice, we have both between-household and within-household variation.

Respondents were asked about their confidence on a 1–3 scale (with three indicting a great deal of confidence and 1 indicating hardly any confidence) in a number of political institutions. The specific institutions we considered are the following:

Politicians—to fulfill promises;

State government—to look after the people;

Village panchayats / nagarpalika (village councils / municipal corporations) to implement public projects.

Our theory implies that people’s confidence in politicians in general decreases, whereas confidence in state government or local authorities remains unchanged (or possibly even improves). Similar to our main analysis, we estimate models including a quadratic specification as well as constituency and election fixed effects. Standard errors are clustered by district. We also estimate LOWESS regressions.

Results

On the left of Figure 6, we show the LOWESS estimates; on the right, we show the quadratic specifications. The histogram of SPEI conditions is superimposed on the estimates. Since we include household fixed effects, these results should be interpreted as the average within-household effect of changes in weather over time.²⁷

Figure 6.

Mechanisms: effects of precipitation onconfidence in politicians, state governments, and panchayats: (a) confidence (politicians): LOWESS plot, (b) confidence (politicians): quadratic plot, (c) confidence (state govt.): LOWESS plot, (d) confidence (state govt.): quadratic plot, (e) confidence (panchayats): LOWESS plot, and (f) confidence (panchayats): quadratic plot.

The findings are consistent with our expectations. Confidence in politicians is sensitive to drought and heavy rainfall, as moving a standard deviation above or below the mean is associated with a clear reduction in confidence in politicians. However, when focusing on a specific level of government—local and state government, such negative effects become significantly weaker or turn to be opposite. Confidence in state governments seems to decrease with the SPEI index, but the relationship is not curvilinear. If anything, it seems that moving away from drought conditions reduces confidence, consistent with the notion that relief provided by the government is a critical variable. The results on local governments are even more striking, as there is a clear curvilinear relationship: extreme weather increases confidence in local governments.²⁸

Mechanisms: The Effect of Relief

We conclude the empirical analysis with a direct test of the theorized relief provision mechanism, again using the IHDS data, to show that households who received government aid displayed higher confidence in politicians after extreme weather. Specifically, we identify respondents who are eligible for India’s National Rural Employment Guarantee Scheme. While ideally we would use data that capture specific relief allocations, such data are only available at the state-year level (Cole et al., 2012). However, we argue that the rural employment data are a useful proxy for reasons discussed below. What evidence does exist at the state level, however, suggests that partisan alignment is an extremely important driver of government transfers. Arulampalam et al. (2009) find that “state which is both aligned and swing in the last state election is estimated to receive 16% higher transfers than a state which is unaligned and non-swing.” Nevertheless, because we do not have relief data at the MLA constituency level, we cannot test the causal mechanism by showing that constituencies with aligned MLAs receive more relief than those with non-aligned MLAs. This is a limitation on our findings, and one that could be fertile ground for future research.

Data and Methods

We use the National Rural Employment Guarantee Act (NREGA) as a proxy for relief because India has historically used rural employment as a crucial means of redistribution and providing assistance for the poor (Dreze & Sen, 1991, esp. 123). It has also been explicitly used to provide relief in times of agricultural stress in rural areas, and is credited with doing more for as an insurance and stabilization system in times of drought than as a long-term remedy for rural poverty (Vatsa, 2006). During the early 1970s Maharastra drought, for example, the central government relied on guaranteed employment schemes to compensate for lost crop income (Dreze & Sen, 1991, pp. 126–130). More recently, floods in the state of Kerala led the central government in 2018 to expand the number of guaranteed days’ work under NREGA for those affected.²⁹

However, although NREGA is supposed to be universal, in practice there has been variation in people’s ability to actually obtain job cards through the system. Das (2015) shows that individuals are more likely to benefit from the rural employment scheme when they are both politically active and affiliated with the local ruling party. Because the Indian relief effort relies on the public distribution system, we can directly test of the role of relief in protecting politicians against backlash. Specifically, we split the sample by whether the respondent’s household has access to an NREGA job card.The goal of this analysis is to test whether relief insulates public officials from backlash in the wake of droughts or floods. The question we use to split the sample asks how many NREGA cards are held by members of the respondent’s household, and is only found in the IHDS-II survey (2011–2012). As a result, we do not use household fixed effects. Responses range from 0 (75% of households) to 5, and we split the sample using a dummy variable indicating whether the house- hold has any NREGA cards. The assumption is that households with at least one NREGA card are more likely to depend on and benefit from disaster relief provided by the government, and thus more likely to make a distinction in elections between non-aligned politicians and the state government. We would thus expect NREGA card holders to exhibit stronger responses to bad weather conditions on average than other households. We estimate ordinary least squares models that include a quadratic term and cluster standard errors by district, and LOWESS plots of the data points in order to detect non-linear relationships between our independent and dependent variables.

Results

Results from the quadratic specification are shown in Figure 7. The panels on the left show the relationship between weather conditions and people’s confidence in politicians, conditional on access to NREGA job cards. On the right, we show the same results for households without NREGA job cards. The results show that there is indeed still a quadratic relationship between precipitation and confident in public officials. In particular, extreme drought and extreme rain result in higher confidence in state governments and panchayats and lower confidence in politicians in general, compared to moderate precipitation. But as the figures show, the quadratic relationship between the SPEI index and confidence suggests that households with NREGA job cards are more confident in politicians in general, state government, and panchayats at given levels of SPEI. Admittedly, the differences are quite small and are much stronger for droughts (i.e., low SPEI) than for excessive rain. Nevertheless, this result is consistent with the notion that relief is a key mechanism protecting politicians from disaster-induced backlash.

Figure 7.

Sub-sample depending on NREGA card: effects on confidence in politicians / state governments / panchayats: (a) confidence (politicians): NREGA job card =1, (b) confidence (politicians): NREGA job card =0, (c) confidence (state govt.): NREGA job card=1, (d) confidence (state govt.): NREGA job card=0, (e) confidence (panchayats): NREGA job card =1, and (f) confidence (panchayats): NREGA job card =0.

Conclusion

The analysis above paints a clear picture of how retrospective voting conditions the political effects of natural disasters in India. Our results show that disasters negatively affect the electoral fortunes of incumbents, but only those who are unable to provide relief. This inability is itself a function of politics, as incumbents who are members of the same party that controls the chief minister or prime minister’s office can reap the rewards of their alignment.

The findings have implications for a growing literature on the politics of climate change. Climate change is expected to increase the frequency, duration, and intensity of weather-based natural disasters (Field et al., 2012). The change in the amount of water vapor due to global warming will lead directly to more frequent and extreme drought or flood conditions (Hansen et al., 2012). Such an increase in extreme weather events implies that local electoral punishment for natural disasters will become more common, but our results suggest that the political alignment of local officials with higher levels of government, which affects their ability to get disaster aid, will play an important role in determining the degree to which local leaders will face electoral punishment.

The results also speak to the theoretical literature on electoral accountability (Ashworth et al., 2017). Our results are difficult to reconcile with models that emphasize voters’ use of optimal re-election rules based on a performance benchmark, as in such models voters should not punish relatively powerless opposition legislators for their inability to offer relief.

Our findings also cannot be attributed to forward-looking voters who react to natural disasters by electing aligned candidates: in our specifications with constituency and state-election fixed effect, recent natural disasters are simply not informative of future natural disasters. Finally, our models also go against the hypothesis that voters use natural disasters as an opportunity to learn about candidate quality, given that there are very little voters can learn about the qualities of relatively powerless opposition legislators even in severe natural disasters.

Our core finding is troubling for democratic politics in India and other developing countries. It suggests that non-aligned incumbents are being punished for events beyond their control and supports the hypothesis that governments have incentives to target relief funds to areas controlled by co-partisan legislators. If disaster relief helps aligned incumbents survive, then governments have incentives to forgo disaster preparedness and instead selectively provide relief to constituencies controlled by co-partisan incumbents. Such a political economy encourages distortions in the allocation of relief. Given that social capital and trust play an important role in disaster recovery (Aldrich, 2012), such distortions in relief provision could, over time, contribute to a vicious cycle of reduced resilience, as citizens lose their trust in and willingness to cooperate with authorities. Institutional constraints on opportunistic allocation of relief would yield a triple dividend: less biased elections with competitive opposition, better targeting of relief, and less distrust between government and society in the long run.

One question that arises naturally from our study is whether the role of political alignments in electoral politics is peculiar to natural disaster scenarios. We believe that the theory applies more generally in a federal structure where financial relations among levels of government play an important role in resource distribution. A voluminous research in distributive politics and fiscal federalism shows that subnational jurisdictions aligned with the higher level of government better secure fiscal transfers in many federal states (Arulampalam et al., 2009; Bhavnani & Lacina, 2017; Brollo & Nannicini, 2012). These fiscal transfers can be used in attempts to neutralize a wide range of negative shocks, helping aligned officials hold their offices, though the efficacy of these resources in other types of shocks would require empirical evaluation.

Still, further research is certainly warranted to establish the scope conditions of our theory. India does have a set of institutional and social features that might complicate generalizations, including an extreme incumbency disadvantage (Uppal, 2009). India’s limited state capacity and heavy dependence on agriculture for livelihoods could also amplify differences in access to relief relative to countries with better resources. What is more, India’s federal structure enables partisan alignment across two levels. In countries with different electoral systems and socio-economic circumstances, our assumptions about retrospective voting, importance of alignment in access to relief, and the general importance of weather conditions might not hold. Future empirical studies in different countries could offer useful insights into the scope conditions of our argument.

Supplemental Material

Droughts_APPENDIX_REVISED_FINAL – Supplemental material for Barking Up the Wrong Tree: How Political Alignment Shapes Electoral Backlash from Natural Disasters

Supplemental material, Droughts_APPENDIX_REVISED_FINAL for Barking Up the Wrong Tree: How Political Alignment Shapes Electoral Backlash from Natural Disasters by Brian Blankenship, Ryan Kennedy, Johannes Urpelainen and Joonseok Yang in Comparative Political Studies

Footnotes

Acknowledgements

We thank Francesca Jensenius, Meir Alkon, Chao-yo Cheng, and Michaël Aklin for their generous assistance with the data. We thank seminar audiences at Rochester University and the Delhi School of Economics for insightful comments. We are also grateful to Allison Carnegie, John Marshall, and Giancarlo Visconti for their comments. Replication materials and code can be found at Blankenship et al. (2020). Joonseok Yang thanks the Korean Ministry of Education and the National Research Foundation of Korea (NRF-2020S1A3A2A02092791) for support for this research.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Joonseok Yang

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

Brian Blankenship is an assistant professor of Political Science at the University of Miami in Coral Gables, FL, a Stanton Nuclear Security fellow at the Council on Foreign Relations, and associate director for the Institute for Sustainable Energy Policy (ISEP). He specializes in international relations, and my research and teaching interests lie at the intersection of international security and international cooperation.

Ryan Kennedy is an associate professor of Political Science, associate director for data analytics for the Institute for Sustainable Energy Policy (ISEP), and a research associate with the University of Houston’s Hobby Center for Public Policy. His research has been published in Science, the American Political Science Review, the Journal of Politics, and International Studies Quarterly, among others.

Johannes Urpelainen is the Prince Sultan bin Abdulaziz professor of Energy, Resources and Environment at the Johns Hopkins School of Advanced International Studies and the Founding Director of the Initiative for Sustainable Energy Policy. His primary research focuses on environmental politics, energy policy, and global governance.

Joonseok Yang is an assistant professor in the Department of Political Science & Diplomacy at Sungkyunkwan University, Seoul, South Korea. His research explores the political economy of business-government relations.

References

Aldrich

D. P.

(2012). Building resilience: Social capital in post-disaster erecovery. The University of Chicago Press.

Arceneaux

(2006). The federal face of voting: Are elected officials held accountable for the functions relevant to their office? Political Psychology, 27(5), 731–754.

Arceneaux

Stein

R. M.

(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.

Arulampalam

Dasgupta

Dhillon

Dutta

(2009). Electoral goals and center-state transfers: A theoretical model and empirical evidence from India. Journal of Development Economics, 88(1), 103–119.

Ashworth

de Mesquita

E. B.

Friedenberg

(2017). Accountability and information in elections. American Economic Journal: Microeconomics, 9(2), 95–138.

Banerjee

Green

D. P.

McManus

Pande

(2014). Are poor voters indifferent to whether elected leaders are criminal or corrupt? A vignette experiment in rural India. Political Communication, 31(3), 391–407.

Bastos

Miller

(2013). Politics under the weather: Droughts, parties and electoral outcomes (IDB working paper 455). Inter-American Development Bank. https://www.econstor.eu/bitstream/10419/89180/1/IDB-WP-455.pdf

Bechtel

M. M.

Hainmueller

(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.

Bechtel

M. M.

Mannino

(2017). Retrospection, Fairness, and Economic Shocks: How Do Voters Judge Policy Responses to Natural Disasters? (Working paper). https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2943046

10.

Begueria

Vicente-Serrano

S. M.

Reig

Latorre

(2014). Standardized Precipitation Evapotranspiration Index (SPEI) revisited: Parameter fitting, evapotranspi- ration models, tools, datasets and drought monitoring. International Journal of Climatology, 34(10), 3001–3023.

11.

Bhavnani

R. R.

Lacina

(2017). Fiscal federalism at work? Central responses to internal migration in India. World Development, 93, 236–248.

12.

Blakeslee

D. S.

Fishman

(2018). Weather shocks, agriculture, and crime evidence from India. Journal of Human Resources, 53(3), 750–782.

13.

Blankenship

Kennedy

Urpelainen

Yang

(2020). Replication data for barking up the wrong tree: How political alignment shapes electoral backlash from natural disasters (Comparative Political Studies Dataverse). https://doi.org/10.7910/DVN/NDHS9Q

14.

Brollo

Nannicini

(2012). Tying your enemy’s hands in close races: The politics of federal transfers in Brazil. American Political Science Review, 106(4), 742–761.

15.

Burlig

Preonas

(2016). Out of the darkness and into the light? Development effects of rural electrification in India (Working paper 268). Energy Institute at Haas. https://ei.haas.berkeley.edu/research/papers/WP268Appendix.pdf

16.

Carlin

R. E.

Love

G. J.

Zechmeister

E. J.

(2014). Natural disaster and demo- cratic legitimacy the public opinion consequences of Chile’s 2010 earthquake and tsunami. Political Research Quarterly, 67(1), 3–15.

17.

Carsey

T. M.

Wright

G. C.

(1998). State and national factors in gubernatorial and senatorial elections. American Journal of Political Science, 42, 994–1002.

18.

Chandra

(2007). Why ethnic parties succeed: Patronage and ethnic head counts in India. Cambridge University Press.

19.

Chen

(2013). Voter partisanship and the effect of distributive spending on political participation. American Journal of Political Science, 57(1), 200–217.

20.

Chhibber

Jensenius

F. R.

Suryanarayan

(2014). Party organization and party proliferation in India. Party Politics, 20(4), 489–505.

21.

Chopra

V. K.

(1996). Marginal players in marginal assemblies: The Indian MLA. Orient Longman.

22.

Cleveland

W. S.

(1979). Robust locally weighted regression and smoothing scatterplots. Journal of the American Statistical Association, 74(368), 829–836.

23.

Clogg

C. C.

Petkova

Haritou

(1995). Statistical methods for comparing regression coefficients between models. American Journal of Sociology, 100(5), 1261–1293.

24.

Cohen

Werker

E. D.

(2008). The political economoy of “natural” disasters. Journal of Conflict Resolution, 52(6), 795–819.

25.

Cole

Healy

Werker

(2012). Do voters demand responsive governments? Evidence from Indian disaster relief. Journal of Development Economics, 97(2), 167–181.

26.

Conley

T. G.

(1999). GMM estimation with cross sectional dependence. Journal of Econometrics, 92(1), 1–45.

27.

Córdova

Seligson

M. A.

(2010). Economic shocks and democratic vulnerabilities in latin America. Latin American Politics and Society, 52(2), 1–35.

28.

Das

(2015). Does political activism and affiliation affect allocation of benefits in the rural employment guarantee program: Evidence from West Bengal, India. World Development, 67, 202–217.

29.

Dreze

Sen

A. K.

(1991). Hunger and public action. Clarendon.

30.

Efron

Tibshirani

R. J.

(1993). An introduction to the bootstrap. Chapman and Hall.

31.

Fair

C. C.

Kuhn

P. M.

Malhotra

Shapiro

J. N.

(2015). Natural disasters and political engagement: Evidence from the 2010-11 Pakistani floods (Working paper). Georgetown University. http://dro.dur.ac.uk/21820/1/21820.pdf

32.

Field

C. B.

Barros

Stocker

T. F.

Dahe

(2012). Managing the risks of extreme events and disasters to advance climate change adaptation: Special report of the intergovernmental panel on climate change. Cambridge University Press.

33.

Fiorina

M. P.

(1981). Retrospective voting in American national elections. Yale University Press.

34.

Food and Agriculture Organization of the United Nations. (2016). India at a glance. Author. http://www.fao.org/india/fao-in-india/india-at-a-glance/en/

35.

Fouirnaies

Mutlu-Eren

(2015). English bacon: Copartisan bias in inter- governmental grant allocation in England. Journal of Politics, 77(3), 805–817.

36.

Gasper

J. T.

Reeves

(2011). Make it rain? Retrospection and the attentive electorate in the context of natural disasters. American Journal of Political Science, 55(2), 340–355.

37.

Golden

Min

(2013). Distributive politics around the world. Annual Review of Political Science, 16, 73–99.

38.

Golub

G. H.

Heath

Wahba

(1979). Generalized cross-validation as a method for choosing a good ridge parameter. Technometrics, 21(2), 215–223.

39.

Gulzar

Pasquale

B. J.

(2017). Politicians, bureaucrats, and development: Evidence from India. American Political Science Review, 111(1), 162–183.

40.

Hansen

Sato

Ruedy

(2012). Perception of climate change. Proceedings of the National Academy of Sciences, 109(37), E2415–E2423.

41.

Healy

Malhotra

(2009). Myopic voters and natural disaster policy. American Political Science Review, 103(3), 387–406.

42.

Healy

Malhotra

(2010). Random events, economic losses, and retrospective voting: Implications for democratic competence. Quarterly Journal of Political Science, 5, 193–208.

43.

Healy

Malhotra

(2013). Retrospective voting reconsidered. Annual Review of Political Science, 16, 285–306.

44.

Jayachandran

(2006). Selling labor low: Wage responses to productivity shocks in developing countries. Journal of Political Economy, 114, 538–575.

45.

Jensenius

F. R.

(2013). Power, performance and bias: Evaluating the electoral quotas for scheduled castes in India (Doctoral dissertation, University of California, Berkeley). https://cloudfront.escholarship.org/dist/prd/content/qt7qt35859/qt7qt35859.pdf

46.

Jensenius

F. R.

Suryanarayan

(2020). Party system institutionalization and economic voting: Evidence from India (Working paper). https://pavisuridotcom.files.wordpress.com/2019/06/econ_voting_final-1.pdf

47.

Key

V. O.

(1966). The responsible electorate: Rationality in presidential voting, 1936–1960. Harvard University Press.

48.

Lind

J. T.

Mehlum

(2010). With or without U? The appropriate test for a U-shaped relationship. Oxford Bulletin of Economics and Statistics, 72(1), 109–118.

49.

Lobell

D. B.

Schlenker

Costa-Roberts

(2011). Climate trends and global crop production since 1980. Science, 333(6042), 616–620.

50.

Malhotra

Kuo

A. G.

(2008). Attributing blame: The public’s response to hurricane katrina. Journal of Politics, 70(1), 120–135.

51.

Myers

R. H.

(1990). Classical and modern regression with applications (2nd ed.). Duxbury Press.

52.

Paternoster

Brame

Mazerolle

Piquero

(1998). Using the correct statistical test for the equality of regression coefficients. Criminology, 36(4), 859–866.

53.

Pozzi

Sheffield

Stefanski

Cripe

Pulwarty

(2013). Toward global drought early warning capability: Expanding international cooperation for the development of a framework for monitoring and forecasting. Bulletin of the American Meteorological Society, 94(6), 776–785.

54.

Prabhakar

S. V. R. K.

Shaw

(2008). Climate change adaptation implications for drought risk mitigation: A perspective for India. Climatic Change, 88(2), 113–130.

55.

Rajeevan

Unnikrishnan

C. K.

Bhate

Niranjan Kumar

Sreekala

P. P.

(2012). North- east monsoon over India: Variability and prediction. Meteorological Applications, 19(2), 226–236.

56.

Rathore

M. S.

(2004). State level analysis of drought policies and impacts in Rajasthan, India (Working paper 84). International Water Management Institute.

57.

Ravishankar

(2009). The cost of ruling: Anti-incumbency in elections. Economic and Political Weekly, 44, 92–98.

58.

Reeves

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

59.

Stokes

L. C.

(2016). Electoral backlash against climate policy: A natural experiment on retrospective voting and local resistance to public policy. American Journal of Political Science, 60, 958–974. https://doi.org/10.1111/ajps.12220

60.

Turner

A. G.

Annamalai

(2012). Climate change and the South Asian summer monsoon. Nature Climate Change, 2(8), 587–595.

61.

Uppal

(2009). The disadvantaged incumbents: Estimating incumbency effects in Indian state legislatures. Public Choice, 138(1–2), 9–27.

62.

Vatsa

K. S.

(2006). Employment guarantee scheme in Maharashtra: Its impact on drought, poverty, and vulnerability. The Indian Journal of Labour Economics, 49(3), 429–452.

63.

Verma

(2012). What determines electoral outcomes in India? Caste, class, or voters’ satisfaction with government performance? Asian Survey, 52(2), 270–297.

64.

Vicente-Serrano

S. M.

Beguería

Lorenzo-Lacruz

Camarero

J. J.

López-Moreno

J. I.

Azorin-Molina

Revuelto

Morán-Tejeda

Sanchez-Lorenzo

(2012). Performance of drought indices for ecological, agricultural, and hydrological applications. Earth Interactions, 16(10), 1–27.

Supplementary Material

Please find the following supplemental material available below.

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

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

0.00 MB

1.57 MB