Life expectancy and voting patterns in the 2020 U.S. presidential election

Abstract

Introduction

In the 2016 U.S. Presidential election, voters in communities with recent stagnation or decline in life expectancy were more likely to vote for the Republican candidate than in prior Presidential elections. We aimed to assess the association between change in life expectancy and voting patterns in the 2020 Presidential election.

Methods

With data on county-level life expectancy from the Institute for Health Metrics and Evaluation and voting data from a GitHub repository of results scraped from news outlets, we used weighted multivariable linear regression to estimate the association between the change in life expectancy from 1980 to 2014 and the proportion of votes for the Republican candidate and change in the proportion of votes cast for the Republican candidate in the 2020 Presidential election.

Results

Among 3110 U.S counties and Washington, D.C., change in life expectancy at the county level was negatively associated with Republican share of the vote in the 2020 Presidential election (parameter estimate −7.2, 95% confidence interval, −7.8 to −6.6). With the inclusion of state, sociodemographic, and economic variables in the model, the association was attenuated (parameter estimate −0.8; 95% CI, −1.5 to −0.2). County-level change in life expectancy was positively associated with change in Republican vote share 0.29 percentage points (95% CI, 0.23 to 0.36). The association was attenuated when state, sociodemographic, and economic variables were added (parameter estimate 0.24; 95% CI, 0.15 to 0.33).

Conclusion

Counties with a less positive trajectory in life expectancy were more likely to vote for the Republican candidate in the 2020 U.S. Presidential election, but the Republican candidate's share improved in some counties that experienced marked gains in life expectancy. Associations were moderated by demographic, social and economic factors.

Highlights

• •Counties with stagnating life expectancy were more likely to vote Republican in the 2016 U.S. Presidential election.
• •We found a similar trend in the 2020 election: as life expectancy increases, the Republican share of the vote decreases.
• •The Republican candidate's share improved in some counties with marked life expectancy gains.
• •All associations were moderated by demographic, social, and economic factors.

1. Introduction

Although life expectancy in the U.S. increased by almost 10 years over a nearly 60-year period from 1959 to 2016, the rate of increase slowed over time and life expectancy has declined after 2014 (Woolf & Schoomaker, 2019). Large disparities in life expectancy among counties in the U.S. have increased as well. In 2014, a 20-year gap existed between counties with the lowest and highest life expectancy. A gap of more than 10 years in life expectancy existed between counties in the 1st and 99th percentile, an increase of 2.4 years from 1980 (Dwyer-Lindgren et el, 2107).

Recent reports have described the association between changes in life expectancy and voting behavior, noting a shift toward the Republican presidential candidate in counties where life expectancy has declined or increased the least (Guo, 2016; Bor, 2017; Bilal et al., 2018). One analysis found a 9.1 percentage point increase in the Republican vote share from 2008 to 2016 in counties in which life expectancy stagnated or declined from 1980 to 2014 (Bor, 2017). Voters in those counties were more likely to vote for Republican candidates in 2016, irrespective of prior voting patterns.

Following the 2020 Presidential election, we examined the election results to determine whether the pattern observed in 2016 persisted. In this brief report, we examine the association between change in county-level life expectancy from 1980 to 2014 and the proportion of votes cast for the Republican party in the 2020 U.S. Presidential election. We also assess the association between the change in county-level life expectancy and the change in the proportion of votes cast for the Republican candidate from 2016 to 2020.

2. Methods

We obtained county-level data on life expectancy from the Institute for Health Metrics and Evaluation (Institute for Health Metrics and Evaluation, 2017) and calculated the absolute difference in life expectancy from 1980 to 2014. We obtained county-level voting results from the 2020 Presidential election from GitHub (McGovern, 2020), which were scraped from results published by Fox News, Politico, and the New York Times, and calculated the Republican candidate's share of votes as the proportion of total votes cast for the Democratic and Republican candidates in each county and Washington, D.C. Across 3109 counties with any votes not attributed to the Democratic or Republican parties, the median percentage of votes for the third-party candidate was 2%, so we excluded votes for third-party and other candidates from the denominator. We also excluded one county due to missing data and counties in Alaska and Kalawao County, Hawaii due to a discrepancy in county classifications between the data sources, so 3110 counties and Washington, D.C. were included in the final analysis.

We used multivariable linear regression models, one without and one with state fixed effects, weighted by the number of two-party votes to assess the association between the absolute change in life expectancy from 1980 to 2014 and the proportion of votes for the Republican candidate in the 2020 Presidential election. We used the same approach to examine the association between the absolute change in life expectancy from 1980 to 2014 and the change in the proportion of votes cast for the Republican party from 2016 to 2020. As before, we ran one multivariable model without and one with state fixed effects. The multivariable models included adjustments for urban-rural classification using the 2013 NCHS Urban-Rural Classification Scheme for Counties (National Center for Health Statistics, 2013), the following sociodemographic and economic characteristics from the 2015–2019 5-year American Community Survey: percent college educated, Gini coefficient, unemployment rate, median house value, poverty rate, percent Black, and percent Hispanic (U.S. Census Bureau, 2020), and state fixed effects. Median home value was transformed to the logarithmic scale and percent Black and Hispanic were transformed to their square root to address skewed data. We used heteroskedasticity-consistent standard errors to calculate 95% confidence intervals. We generated maps of the national distribution of voting patterns and life expectancy and scatter plots displaying the relationship between change in life expectancy and voting patterns. This study was determined exempt by the Duke Health Institutional Review Board.

3. Results

From 1980 to 2014, the change in county-level life expectancy ranged from −2.2 years to 10.8 years with a median change of 3.9 years (inner quartile range [IQR] 3.0 to 4.8) (Fig. 1). The county-level proportion of the two-party vote for the Republican candidate in the 2020 Presidential election ranged from six percent to 97% with a median of 70%. (Fig. 2). After weighting for the total number of two-party votes, the weighted median percent of the two-party vote for the Republican candidate was reduced to 46% (IQR 36–60%). The county-level change in the percent of votes cast for the Republican candidate from 2016 to 2020 ranged from −11.3 to 28.1 with a median of −0.7 (IQR -2.1, 0.8) (Supplemental Fig. 1). The median change in the percent of votes cast for the Republican candidate after weighting for the total number of two-party votes in 2020 increased to −1.8 (IQR -3.0, −0.07).

Fig. 1.

Change in life expectancy in U.S. counties from 1980 to 2014

Caption: Change in life expectancy in years from 1980 to 2014. Smaller increases are shown in red, larger increases are in blue. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 2.

Republican Candidate's Share of the Vote in the 2020 U.S. Presidential election.

Caption: Red counties indicate Republican's share of the vote is greater than 80%, and blue counties indicate Republican's share of the vote is less that 53%. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Change in life expectancy at the county level was negatively associated with Republican share of the county-level vote in the 2020 Presidential election. For each year of increase in life expectancy, the Republican vote share at the county level decreased by 7.2 percentage points (95% CI, −7.8 to −6.6) (Fig. 3a). The association was attenuated (parameter estimate −1.3; 95% CI, −2.1 to −0.6) when sociodemographic and economic variables were included in the model, and further moderated when state fixed effects were included (parameter estimate −0.8; 95% CI, −1.5 to −0.2) (Fig. 3b, Supplemental Table 1).

Fig. 3a.

County-Level Life Expectancy and Republican Share of the County-level Vote in the 2020 Presidential Election

Relationship Between Change in County-Level Life Expectancy and Republican Share of the County-level Vote in the 2020 Presidential Election

Caption: In the 2020 election, as life expectancy increases, the Republican share of the vote decreases.

Fig. 3b.

Residuals of Change in Life Expectancy vs. Control Covariates and Residuals of Proportion of Republican Votes vs. Control Covariates

Caption: When state, sociodemographic, and economic variables are included in the model, the association between change in life expectancy and the proportion of Republican votes is attenuated.

Change in life expectancy at the county level was positively associated with the change in the proportion of votes cast for the Republican candidate from 2016 to 2020. For each year of increase in life expectancy, the change in the Republican vote share at the county level increased by 0.29 percentage points (95% CI, 0.23 to 0.36) (Supplemental Fig. 2a). The association was strengthened (parameter estimate 0.4; 95% CI, 0.2, 0.6) when sociodemographic and economic variables were included in the model, but attenuated when state fixed effects were added (parameter estimate 0.24; 95% CI, 0.15 to 0.33) (Supplemental Fig. 2b, Supplemental Table 2).

4. Discussion

Compared to prior elections, the 2016 U.S. Presidential election witnessed a shift of votes toward the Republican candidate in communities that have experienced recent stagnation or decline in life expectancy (Bor, 2017; Bilal et al., 2018). A recent analysis examined the association between community health and the shift from 2016 to 2018 in Republican share of the vote in U.S. House of Representatives elections (Wasfy et al., 2020). Although there was a general shift toward the Democratic candidate in the 2020 election, we found that voting in less healthy communities continued to favor the Republican candidate. When we adjusted for demographic, social and economic factors, the difference was attenuated, suggesting that these factors moderated the observed association between change in life expectancy and voting status. The association remained significant with the addition of state fixed effects, suggesting that within a given state, counties with declines in life expectancy were more likely to vote for the Republican candidate in 2020.

In contrast to Bor's analysis of the 2016 U.S. Presidential election (Bor, 2017), we found that the change from 2016 to 2020 in the proportion of votes cast for the Republican candidate increased with the change in life expectancy. These findings likely reflect that the Republican candidate's share improved in some counties (e.g., Hudson County in New Jersey, Rockland County in New York, Los Angeles and Santa Clara Counties in California, and Cook County in Illinois [Kolko, 2020]) that also experienced marked gains in life expectancy. Although these counties voted in favor of the Democratic candidate, the Republican candidate's share increased relative to 2016.

Compared with other high-income countries, the U.S. has the lowest life expectancy with a trajectory that is diverging from other high-income countries (Ho & Hendi, 2020). In addition, substantial disparities in life expectancy as a function of geographic location are well documented within the U.S., and growing evidence highlights the relationship between conservative state policies and reductions in life expectancy (Montez, 2020). Fig. 1, Fig. 2 offer graphic visual representation of the congruence of life expectancy and voting patterns. While this epiphenomenon is interesting, our analysis suggests that community voting patterns are likely to represent manifestations of differences in demographics, economic well-being, and educational status. Regardless of which political party is in power, effective policies must be implemented to improve health outcomes in counties and regions where life expectancy has stagnated.

This study is limited by its ecological nature and the inability to associate individual voting behaviors with health outcomes. Additionally, the study is limited by available data sources and would benefit from more detailed social and economic data. Updated estimates of life expectancy will be critical to guide policies to improve the well-being of negatively affected counties, but these data are not yet available.

5. Conclusion

Counties with a less positive trajectory in life expectancy were more likely to vote for the Republican candidate in the 2020 U.S. Presidential election, but the Republican candidate's share improved in some counties that experienced marked gains in life expectancy. The associations were moderated by demographic, social and economic factors that should drive health policy priorities over the coming years.

Ethical statement

The manuscript in part or in full has not been submitted or published anywhere. The study did not involve human participants.

Data availability statement

Data were obtained from public sources.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Author Statement

Lesley Curtis, Molly Hoffman, Robert Califf, Bradley. Hammill: Conceptualization, Lesley Curtis, Molly Hoffman, Bradley Hammill: Methodology, Data curation, Writing- Original draft preparation. Lesley Curtis, Molly Hoffman, Robert Califf, Bradley Hammill: Analysis, Writing- Reviewing and Editing.

Declaration of competing interest

LC, MN and BH declare no conflicts of interest.

RMC reports employment from Google Health and Verily Life Sciences and Corporate Board for Cytokinetics and Centessa with payment and stock.

Appendix A. Supplementary data

The following is the Supplementary data to this article: