Natural Disasters and Electoral Consequences: Evidence on Voter Preferences and Participation from U.S. Presidential Elections

Technical Summary: Natural Disasters and Electoral Consequences

Problem Statement
The literature on the electoral consequences of natural disasters remains divided regarding whether such events politically harm or help incumbents. Theoretical frameworks offer competing predictions: "blind retrospection" suggests voters indiscriminately punish incumbents for adverse outcomes regardless of responsibility, while "attentive retrospection" posits that voters reward incumbents for effective crisis management and relief efforts. Furthermore, existing empirical studies often focus exclusively on vote share or voter turnout, failing to distinguish whether disasters influence elections by altering voter preferences (persuasion) or by changing the composition of the electorate through mobilization or suppression (participation). Additionally, the specific role of federal intervention—specifically Presidential Disaster Declarations (PDDs)—and the moderating effect of partisan alignment between the President and state governors remain underexplored and yield mixed evidence.

Methodology
This study utilizes a county-level panel dataset covering 3,016 U.S. counties across seven presidential election cycles from 1996 to 2020. The unit of analysis is the county-election year.

• Data Sources:

  • Election Results: Congressional Quarterly (CQ) Press, used to calculate the change in the incumbent party's vote share relative to the previous election.
  • Disaster Exposure: The Spatial Hazard Events and Losses Database for the United States (SHELDUS), measuring disaster damage as the log of total inflation-adjusted property and crop losses per 10,000 residents.
  • Federal Relief: FEMA records of Presidential Disaster Declarations (PDDs) for the period 1994–2020.
  • Controls: Socioeconomic and demographic variables from the Bureau of Economic Analysis, Bureau of Labor Statistics, and U.S. Census Bureau (e.g., income, unemployment, race, education).

• Econometric Framework:
  The authors employ a Two-Way Fixed Effects (TWFE) model:
  $$y_{it} = \alpha_i + \delta_{st} + \beta_1 x_{it} + \beta_2 x_{i,t-1} + H'_{it}\Gamma + \varepsilon_{it}$$
  Where $y_{it}$ represents either the change in incumbent vote share or voter turnout in county $i$ at election $t$. The model includes county fixed effects ($\alpha_i$) to absorb time-invariant local characteristics (e.g., baseline political preferences, long-run risk) and state-by-year fixed effects ($\delta_{st}$) to control for common shocks within a state during a specific election year. Identification relies on within-county variation in disaster severity over time, assuming that conditional on fixed effects, disaster timing and intensity are plausibly exogenous.

• Robustness and Diagnostics:

  • Placebo Randomization: Disaster intensity is randomly reassigned across counties within state-year cells to generate a null distribution; the actual estimates lie well outside this distribution.
  • Event Studies: Tests for pre-treatment dynamics show no evidence that future disaster exposure predicts current electoral outcomes.
  • Categorical Analysis: To uncover non-linearities, the authors estimate models using categorical groups based on disaster severity (small/large), PDD status, and partisan alignment.
  • Sensitivity Checks: Analyses include alternative disaster intensity measures, extensive-margin specifications (including zero-damage counties), controls for domestic migration flows, and leave-one-election-out exercises.

Key Contributions

1. Joint Analysis of Preferences and Participation: Unlike prior studies that isolate vote share or turnout, this paper simultaneously examines both outcomes. This allows for the disentangling of preference shifts (retrospective evaluations) from participation changes (mobilization/suppression).
2. Broad Empirical Scope: The study moves beyond isolated case studies (e.g., Hurricane Katrina) to analyze all types of natural disasters across all U.S. counties over multiple decades, enhancing external validity and capturing systematic variation in political responses.
3. Disentangling Mechanisms: The research explicitly tests the roles of PDDs and partisan alignment, addressing gaps in the literature regarding how federal aid and intergovernmental political coordination shape electoral responses.

Results

• Impact on Incumbent Vote Share: More severe natural disasters significantly reduce the incumbent party's vote share, consistent with retrospective voting theory. This negative effect is driven primarily by disasters occurring in the election year rather than cumulative exposure from previous years.
• Role of Federal Relief (PDDs): Counties receiving PDDs exhibit a partial mitigation of electoral losses compared to similar counties without declarations. The interaction between disaster severity and PDDs is positive and significant, supporting the attentive retrospection hypothesis that visible federal assistance can offset some electoral penalties.
• Impact on Voter Turnout: Disasters exert weak and inconsistent effects on voter turnout. Neither disaster severity nor PDD issuance systematically alters turnout rates. This suggests that the political consequences of disasters operate primarily through shifts in voter preference rather than changes in aggregate participation.
• Partisan Alignment: Partisan alignment between the President and the state governor plays a more consistent and dominant role than PDD issuance alone. Alignment amplifies the political benefits of federal aid and is associated with higher turnout in aligned counties, though it does not necessarily eliminate the vote-share penalty for severe disasters.
• Categorical Findings: The categorical analysis reveals that while large disasters consistently reduce incumbent support regardless of PDD or alignment, partisan alignment is a stronger determinant of turnout responses than disaster severity or aid alone.

Significance and Claims
The paper claims to provide a unified framework demonstrating that the electoral costs of natural disasters arise primarily from changes in voter preferences toward the opposition rather than from the demobilization of incumbent supporters. It challenges simple retrospective voting accounts by highlighting that political context—specifically the coordination between federal and state governments—is a critical mediator of democratic accountability.

The authors assert that while federal disaster assistance (PDDs) offers a partial buffer against electoral losses, partisan alignment is a more robust determinant of electoral outcomes. These findings suggest that as climate-related disasters become more frequent, the electoral incentives for governments depend heavily on the political alignment of leadership and the visibility of relief efforts, rather than on the mere occurrence of the disaster itself. The study concludes that understanding these mechanisms is essential for evaluating the resilience of democratic institutions in the face of increasing environmental shocks.