![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
Abstract
I examine the relationship between demographics and vote choice using pre-election tracking polls, exit polls, and data on the American states. A number of important findings emerge: (1) there was a fair amount of preference stability in 2016; (2) Clinton underperformed Obama in 83% of demographic groups examined; (3) at the state level, the percent of whites with low levels of education and the size of the rural population had important effects on vote choice, and both of those variables were more strongly correlated with Democratic vote share in 2016 than in 2012; and (4) while state turnout rates were moderately correlated with Clinton’s vote share, changes in state turnout rates from 2012 to 2016 did not have a substantial impact on Clinton’s performance.
Notes
Notes
1 Most forecasts from political scientists predicted a Clinton victory, though one notable exception was Helmut Norpoth who predicted a Trump victory. Details on his forecasting model can be found here: http://primarymodel.com/2016-forecast-full/
3 http://white.pbs.org/newshour/updates/trump-campaign-has-ground-game-problem/
5 See this report for details on demographic changes from 2012 to 2016: http://white.pewresearch.org/fact-tank/2016/02/03/2016-electorate-will-be-the-most-diverse-in-u-s-history/
7 http://www.cnn.com/2016/11/09/politics/donald-trump-hillary-clinton-blue-wall/
8 See: http://www.cnn.com/2016/09/19/politics/election-2016-hillary-clinton-barack-obama-voters/and http://www.usnews.com/opinion/articles/2016-09-30/obama-coalition-may-not-turn-out-for-hillary-clinton
9 Data were gathered by the author from the following website: https://www.washingtonpost.com/wp-stat/polls/postpollarchive.html
10 The Washington Post-ABC News decided to conduct two separate polls in the month of September. This is the only month when they conducted two tracking polls. Note that the dates for the September polls are different (they represent two different samples).
11 I am limited to the demographic breakdowns provided by The Washington Post-ABC News. Thus, the specific categories (e.g., White versus Non-White) that I plot were determined by available data.
12 See, for example: http://www.washingtontimes.com/news/2017/jan/13/charlie-sykes-trump-critic-and-contrarian-conserva/
13 This quote is from the article “Donald Trump ‘hold your nose’ Presidency” by Chris Cillizza.
14 Exit polls, like any other types of polls, are imperfect. For an overview of how exit polls are conducted and potential issues with exit polling, see the following websites: http://www.edisonresearch.com/behind-numbers-2016-national-election-exit-poll/and http://www.pewresearch.org/fact-tank/2016/11/02/just-how-does-the-general-election-exit-poll-work-anyway/
15 Data compiled by the author from CNN and NYT exit poll pages for the 2016 and 2012 elections. See the following links for data: http://edition.cnn.com/election/results/exit-polls/national/president, http://www.cnn.com/election/2012/results/race/president/, https://www.nytimes.com/interactive/2016/11/08/us/politics/election-exit-polls.html?_r = 0, and http://www.nytimes.com/elections/2012/results/president/exit-polls.html
16 Data taken from Dave Leip’s Atlas of U.S. Presidential Elections: https://docs.google.com/spreadsheets/d/1D-edaVHTnZNhVU840EPUhz3Cgd7m39Urx7HM8Pq6Pus/edit#gid=29622862
17 Data on candidate vote share from Dave Leip’s Atlas of U.S. Presidential Elections (link provided above). Ideology is captured using the state-level measure from the American Ideology Project (http://www.americanideologyproject.com), Democratic partisan advantage is measured using 2015 Gallup data (http://www.gallup.com/poll/188969/red-states-outnumber-blue-first-time-gallup-tracking.aspx), state unemployment measure is from 2015 Bureau of Labor Statistics estimates, number of manufacturing jobs measure is from the 2015 Bureau of Labor Statistics estimates, Hispanic/Latino share of the CVAP (Citizen Voting Age Population) is calculated using the American Community Survey 2015 estimates, the percentage point change in the Hispanic/Latino CVAP calculated using the American Community Survey 2012 and 2015 estimates, the percentage of whites with less than a high school degree or a high school degree is from the American Community Survey 2015 estimates, as is the percentage of white men with less than a high school degree or a high school degree, the percent of the state population living in a rural area, and the percent of the rural population that is white.
18 When the unemployment measure is restricted to whites (unemployment rate among whites in 2015), there is still a weak relationship (r = .23, p = .11).
19 Note that the results are identical to this specification if one uses the change from 2012 to 2016 as the dependent variable and includes Obama 2012 vote share as an independent variable to capture the baseline from which change is occurring.
20 Although some of the independent variables are highly correlated (e.g., ideology, Obama vote share, and Democratic partisan advantage), the model results are quite similar when the ideology and Democratic partisanship advantage measures are omitted from the model (and the Obama 2012 vote share measure and the other independent variables are left in the model). When the full model (shown in ) is estimated, the mean VIF is 5.50. When the alternative specification is used (ideology and Democratic partisanship omitted), the mean VIF drops to 2.57 and the Obama measure remains significant at p < .05 (one-tailed), as does the percent of Whites with low levels of education measure and the rural population measure. The only changes are that the unemployment measure becomes significant at p < .05 (one-tailed) and the Hispanic population measure becomes significant at p < .10 (one-tailed). The other variables are not statistically significant at conventional levels. Full model results available from the author on request.
21 See: https://www.nytimes.com/2016/11/20/opinion/sunday/the-democrats-real-turnout-problem.html, https://fivethirtyeight.com/features/voter-turnout-fell-especially-in-states-that-clinton-won/, and https://fivethirtyeight.com/features/no-voter-turnout-wasnt-way-down-from-2012/
22 The article was published only a few days after the election (using preliminary data) and was updated on Tuesday, November 15, 2016.
25 Data on overall turnout rates in presidential elections can be found here: http://www.electproject.org/home/voter-turnout/voter-turnout-data
26 Turnout is measured as VEP (voting eligible population) turnout from the United States Elections Project: http://www.electproject.org/home/voter-turnout/voter-turnout-data
27 Osborn, McClurg, and Knoll (Citation2010), using county level data to compare the impact of changes in turnout from 2004 and 2008 on Democratic vote share, note that “Consistent with the hypothesis that the Obama campaign benefited in those places where turnout increased, we see a positive and statistically significant coefficient. The substantive effects are impressive, especially considering the range of the dependent variable. For every unit increase in turnout, there is approximately a 0.2-unit increase in voting Democratic. Thus, if turnout in 2008 was 10 points higher in a county than in 2004, Barack Obama’s campaign could expect to outgain John Kerry in the county by a little more than 2%” (p. 223). It would be useful to replicate Osborn et al.’s analysis to see how county turnout (and changes in county turnout) influenced changes in Democratic vote share from 2012 to 2016. Interestingly, as a robustness check of their county-level results, Osborn et al. estimated the effect of turnout change (from 2004 to 2008) on the change in Democratic vote share (also from 2004 to 2008) in the states. They note, “we estimate our models using state-level data. Though the small n in this data source renders the produced coefficients imprecise, the same pattern emerges in these results: Obama received more votes [than Kerry in 2004] but not enough votes to influence the outcome” (p. 224).
28 Only 40 percent of U.S. adults had a favorable assessment of Clinton, and a mere 35 percent viewed Trump favorably (Gallup 2016).
29 See, for example: http://time.com/4292683/hillary-clinton-black-voters-al-sharpton/