Civic Participation and Lake-Effect Weather Exposure in Cook County, 1998–2024

1. Introduction

Weather is among the most widely studied exogenous shocks to voter turnout, with a literature dating to Knack (1994) and Gomez et al. (2007). Most of this work has been conducted at the county-day level and has been limited in its ability to identify the marginal effect of weather exposure during the hours in which voters actually make the decision to travel to the polls. We extend this literature using precinct-level data and hourly weather measurements from a single contiguous urban county over twenty-seven elections.

Cook County offers an unusual research setting: it is densely instrumented, demographically diverse, and subject to a localized meteorological phenomenon — lake-effect snowfall — that is sharply discontinuous across short distances and largely orthogonal to underlying political preferences. This combination permits unusually precise identification of the weather–turnout relationship.

2. Data

Precinct-level turnout records were obtained from the Cook County Clerk's office (suburban precincts) and the Chicago Board of Election Commissioners (city precincts). Records were geocoded to polling-place address and matched to the nearest of three NWS observation stations: Romeoville, Midway, and O'Hare. Hourly snowfall accumulation was constructed from station-reported one-hour precipitation and surface-temperature records.

Our analytic sample comprises 38,712 precinct-election observations across 1,558 precincts and 27 elections. After excluding precincts that were redistricted or consolidated mid-period and elections with documented administrative anomalies, the working sample is 36,104 observations.

3. Identification

Our primary specifications include precinct, year, and month fixed effects, controlling for time-invariant precinct characteristics and broad seasonal patterns in turnout. To address residual concerns about endogenous polling-place selection in response to anticipated weather, we instrument for polling-hour snowfall with Lake Michigan surface-water temperature measured five days prior to the election, leveraging the well-documented physical relationship between lake-surface temperature and lake-effect precipitation intensity.

4. Results

Each additional inch of accumulated polling-hour snowfall reduced precinct turnout by 1.42 percentage points on average (Table 2). Heterogeneity analyses suggest the effect was approximately 2.4 times larger in precincts in the bottom quartile of historical turnout than in the top quartile, consistent with prior findings that weather effects are concentrated among lower-propensity voters.

We find no statistically detectable effect of pre-election snowfall (snowfall in the 24-hour window ending six hours before polls opened), suggesting that voters' decisions are responsive primarily to current rather than anticipated conditions. Effects were larger in midterm and municipal elections than in presidential years.

5. Discussion

Our estimates are consistent with the broader literature on weather and turnout but provide more precise identification of the relevant exposure window. The magnitude of the effect — a 1.42 percentage-point reduction per inch — implies that a typical heavy lake-effect event could reduce a precinct's turnout by 4–6 percentage points, with substantial heterogeneity by precinct composition.

Several limitations bear noting. Our research design cannot distinguish between voters who would have voted later in the day (intra-day delay) and voters who would have voted in any case but stayed home (true loss). Cook County's experience may not generalize to jurisdictions with different voting infrastructure or different climatic regimes.