Did Low Uninsurance Rates Win Obama a Second Term?

Almost 122 million Americans visited the polls on November 6th to cast their votes in arguably one of the most controversial election cycles in decades. After three days of early ballots and a full day of voting, President Barrack Obama was re-elected and offered a second term as President of the United States of America, winning 332 of the 538 Electoral College votes. Voting behavior has been historically (and consistently) associated with values and attitudes related to beliefs, views on race, and even church attendance and reading a newspaper. Yet, in the wake of the Supreme Court’s decision to uphold the Affordable Care Act (popularly known as Obamacare), the outcome of the 2012 elections weighed heavily on voter’s decision to keep or repeal it. Provisions included in the ACA are designed to increase access to preventive health services, giving access to health insurance to over 50 million uninsured Americans.

The purpose of this post is to identify, both visually and statistically, any relationships between voting outcome and proportions of Medicare enrollees, Medicaid enrollees and proportion of uninsured by state, then by county for 3 swing states.

  • US state and county-level health insurance information (Medicare, Medicaid and non-insurance total population percentages) were obtained from the US Census Bureau and three-year ACS estimates were plotted on base. At the time of this report, aggregate official sources for election data remained unavailable and the website politico.com was used as primary source for county-level voting outcome. County-level base maps were obtained using the TIGER GIS database while the US base map was provided by ESRI. Choropleth maps were created using the ArcMap 10 software suite, using the commonly recognizable color blue for democrat and red for republican. Three toss-up states (Florida, Ohio and Wisconsin) were included in the analysis to determine any specific differences between the two candidates. Percentages of Medicare and Medicaid enrollment, as well as percentages of uninsured were graphed at the country-level using graduated quantities, separating values by standard deviation, rounding up to the nearest integer. At the state level, only percentages of uninsured were included in the visual analysis.Data was then analyzed using Microsoft Excel 2007 Data Toolkit to identify any significant differences between states/counties that voted Republican or Democrat. Two-sample F-tests were performed to determine if samples had equal variances. Samples having an F-test showing a p-value ≤ 0.05 were considered having unequal variances while samples having an F-test with a p-value ≥ 0.05 had equal variances. Depending on the F-test outcome, a two-sample t test (assuming either equal or unequal variances) was performed to determine if there was any significant difference between the two samples. Tests with p-values ≤ 0.05 were considered statistically significant.
  • In terms of statistical significance, no difference is seen between republican and democrat voting states with regards to Medicare or Medicaid enrollment. Of course, it’s important to remember that the voting data and health care services enrollment data come from two complete different data sources and that this is not a breakdown of enrollment BY voter! On the contrary, the health care services data represents the entire state proportions, not just proportions among 2012 voters. What is startling is the clustering of high Medicaid enrollment among states.
  • At the country level, visually, there is somewhat a similar distribution of percent uninsured between Republican and Democrat-leaning states. Both types of states have similar levels of uninsured populations. There seem to be a higher number red states with high uninsured proportions but not much can be said visually.Yet when we run some quick numbers, there’s a significant difference between levels of uninsured among the two state-groups (two-tailed p=0.0004). Additionally, republican states had a consistently higher proportion of uninsured when compared to democrat states, as evidenced by the relatively low variance of republican states as compared to that of democrat states.

    Then comes the issue of the 3 swing states included in this analysis: Florida, Ohio and Wisconsin. While Ohio followed the national trend of higher uninsured in republican counties, Florida and Wisconsin go against the national trend. We can also observe this trend visually across the three states.

  • Country-level uninsured data indicated that democrat states have a high variation in the percentage of population uninsured when compared to their republican counterparts. Consequently, republican states had consistently higher rates of uninsured. This remarkable difference is in opposition to initial considerations and can be explained by a multitude of factors, ranging from population income levels, population employment levels, education, as well as gender and age breakdown – all of which were not accessible and were beyond the scope of this study. Continuing to state/county-level data, the variation in voting behavior reinforces the argument that voter behavior is likely explained by a foundational set of characteristics, potentially including insurance status. Because 2011 ACS estimates were used in this analysis, no solid conclusions can be drawn as the health data available is not directly related to voting behavior. Nevertheless, the strong correlation between voting outcome and percent uninsured poses a critical concern, not only to future presidential candidates, but also to current and future policy makers. Voter turn-out is highly dependent on income, where lower income individuals (those making under $50,000) generally vote 33% less than those making above $75,000 (Nonprofitvote.org, 2010). Additionally, income is correlated with a number of other variables that have the potential to influence both voter turn-out, as well as voter choice. Characteristics such as education and employment status will influence income, and income will influence insurance status.The lack of any significant relationship between Medicaid and Medicare insurance and voting outcome at the state level is potentially confounded by differences in proportions of high-age groups (for Medicare) and low-income groups (for Medicaid) at the county level, as well as total population size at the state level. While this study showed no significant difference in voting outcome (and by proxy, voting behavior) among Medicaid and Medicare enrollees, further analysis and the consideration of other confounding variables (such as age, education) could provide additional statistical insight. Additionally, the limitation of the data introduces a certain lack of study reliability, since health data was obtained from a different source, in a different year estimate (2011) than voter behavior data (2012). The analysis of voter-level demographic data, that was not available at the time of this study, has the potential to provide significant findings that this study does not and cannot.

Leave a Reply

Next ArticleA Look at Gun Ownership and Firearm Death Rates