Understanding the 2016 vote: Democratic turnout

Reading Time:

Why did the election in 2016 turn out the way it did? One story is the that too many Democrats stayed home. Soon after the vote, articles like this one (“Don’t blame minority turnout for Clinton’s loss. White Democrats in key states just didn’t vote.”) went up.

Very few Obama voters have turned to Trump, but it seems plausible that many of them did not vote. Fashionable reasons in support of this narrative include:

  • Clinton was expected to win, Democrats could have taken a victory for granted.
  • The mythical "enthusiasm gap". The view is that Clinton was not an appealing candidate; the relatively low approval ratings are consistent with this story.

Besides, Hillary Clinton's book contains stories about people coming up to her after the election and apologizing for not having voted.

Where was Democratic turnout lower than it could have been, using the 2012 as a benchmark?

I will briefly summarize one well-known analysis, and one chart of my own to suggest that turnout did play a role.

Democratic turnout differences: A Dartmouth post-election alaysis

David Cottrell, Michael C. Herron, and Sean J. Westwood published dived into the data on county votes, mainly evaluating whether there was voter fraud, but also illuminating many voting patterns along the way. Worth highlighting is the difference between total Clinton votes and total Obama votes from 2012 as a percentage of the citizen voting age population that they look at in one of their regressions.

Compared to 2012, more votes were cast for the Democratic candidate in counties where:

  • The share of Whites without college degrees is large.
  • The median household income is low.
  • More people work in manufacturing jobs.

By contrast, more Democrats voted for Clinton than for Obama in countries with larger Hispanic or Islamic populations. County attributes pertaining to the shares of Blacks, Asians, and Jewish Americans were not correlated with changes in Democratic turnout.[1]

Turnout and implications for Democrats

I looked at all U.S. counties with available data and noticed that in those places where turnout increased more the share of the vote for Clinton was higher:

Note: The source of county-level data is D. Leip's Atlas of U.S. Presidential Elections. Only counties where turnout increased are displayed due to data transformation; Chart by Jan Zilinsky.

This relationship can be interpreted as mechanical to some degree[2] but it would confirm the claims of those observers who say that Democrats would have benefited if it had been easier to register and vote and in the United States.

Of the 3,112 counties for which there is county-level data, 2,728 shifted toward the GOP, 383 shifted Democratic, and 1—Barrow County, Georgia—stayed exactly the same.
The counties where Clinton gained on Obama occurred in much larger counties, where a median of 75,554 people showed up to vote in 2016. The median for the 2,728 Republican-gaining counties is 9,905.

  1. Cottrell and co-authors use state fixed effects. ↩︎

  2. Democrats do better in larger/urban locations, and if shocks to turnout rates were random, then one would expect places where turnout increased a lot to "look more Democratic." The general pattern in county-size differences is summarized the following passage from Zeke J. Miller and Chris Wilson: ↩︎

NAs in R: some warnings (and a worked example; calculating standard deviations)

This post shows why is.na and !is.na are not ideal approaches to “clean” a dataset with missing values when we want to compute summary...

Are we talking about anything other than the President?

Yesterday, many people were circulating this article: Trump challenges Tillerson to an ‘IQ test.’ The White House claims it was ‘a joke.’ - The Washington Post...