How did Donald Trump win the presidential election? In Part 1 of this four-part series, I showed how Trump performed poorly by the standards of challengers in post-incumbent elections of the past. (I define post-incumbent as elections with no incumbent on the ballot following the reelection of an incumbent.) In Part 2, I repeated the same analysis at the state-by-state level and found that Trump also didn’t do especially well in the battleground states, but he won because the handful of states he flipped narrowly from the Democrats were almost all large states.
But did Trump win — or did Hillary lose? Those aren’t separate questions, but to get a better handle on what really happened, we need to look beyond the two-party percentages to turnout.
In the sample of post-incumbent elections since 1868 that I’ve been discussing in this series, challenger candidates who weren’t facing an incumbent always increased their party’s raw vote total by at least 10 percent — until 2016. Trump got only 3.4 percent more votes than Mitt Romney did in 2012.
Trump’s 3.4 percent improvement over Romney looks pretty weak next to the historical average of 38.3 percent. It’s weak even compared with the improvement of recent candidates like George W. Bush (28.7 percent), Barack Obama (17.7 percent), and Michael Dukakis (11.3 percent, the previous low). But Hillary’s performance was also not strong: She joined John McCain, George H. W. Bush, and Hubert Humphrey in getting fewer total votes than her predecessor. (Bush won anyway, largely because Ronald Reagan left him with such a large margin of victory.) On the whole, before we adjust for population growth, the 6 percent increase in total turnout over 2012 is the third-lowest of these post-incumbent elections, ahead of only 1968 and 1988 (when the incumbent parties were running sitting vice presidents). Most of that modest net increase in turnout was soaked up by third-party candidates.
Traditionally, fresh new candidates bring a lot of new voters into the process, but Trump didn’t do that on a large enough scale to significantly offset the Republican voters he was losing. (More in Part 4 on that.) Instead, he was helped by the fact that the Democratic presidential candidates (Obama and Clinton), for two elections in a row, received fewer votes than their party’s nominee had in the previous election. In fact, in 2012 and 2016, Democrats got fewer total votes than Obama won in 2008; the only parties in power to suffer similar vote-total drops in consecutive elections were the Republicans in 1988 and 1992 (coming down from the Reagan landslide of 1984) and the Democrats in 1940, 1944, and 1948 (coming down from the FDR landslide of 1936). Judged by the overall popular vote, 2016 was much more a repudiation of the Obama Democrats than an embrace of Trump.
But raw vote totals are only half the story, because the number of eligible voters is always increasing, sometimes faster than at other times. Michael McDonald at Elect Project has compiled data going back to 1980 from which he estimates the number of eligible voters nationally and state by state, which he designates the “Voting-Eligible Population” or “VEP.” The rest of today’s installment is derived from McDonald’s VEP figures.
Once again, as in every election since 1980, the winner of the popular vote got substantially fewer votes than the number of eligible voters who didn’t show up at all. 2016 set a new record: More than 95 million eligible voters did not cast a presidential ballot, producing the widest spread between the nonvoters and the winner of the popular vote since 2000:
The number of people tuning out of the process spiked for the second election cycle in a row, while the third-party vote increased and the number of voters for the major parties stayed flat (this time, unlike 2012, without a multi-state natural disaster ten days before the election). This suggests a serious failure of either side to inspire much faith or enthusiasm.
But we can do more with the VEP data than measure the overall turnout. The votes received by each candidate as a percentage of the overall pool of eligible voters is a revealing test of how strong the candidates really were compared with past nominees. Of course, when looking at the turnout in each state, you have to bear in mind that different states have different traditions of higher or lower turnout, due to a variety of factors: demographics, voting laws, local traditions of civic engagement.
Let’s take a look at the national and state-by-state shares of eligible voters for Trump and Hillary compared with the previous four elections. There’s a lot to unpack in these charts, so I’ll go through them in a few ways. States are ranked here by how much the Republican share of eligible voters went up or down between 2012 and 2016:
States are color-coded here to reflect voter racial demographics, a sad fact of the racially polarized electorate of the past decade. The states in dark green are states where at least 20 percent of the population as of 2010 was African-American, the states in light green are at least 10 percent African-American. Shown in tan are the states where at least 80 percent of the population as of 2012 was non-Hispanic white. These are, of course, only rough proxies for the voting-eligible population as of each year. As I’ve discussed in some detail before, the states with the largest African-American populations were highly correlated with the largest Democratic turnout in 2012 relative to prior years. But where the top of the Obama 2012 turnout chart was a forest of green, the top of Trump’s 2012 chart shows a lot of tan, and we see the beating heart of Trump Country running from Western New York and Pennsylvania down through West Virginia and Kentucky, as well as New England states such as Maine and Rhode Island that have a lot of white working-class voters who don’t typically vote Republican.
The gold standard for turnout is a third of the state’s eligible voters, and I’ve bolded the states where one or both parties cleared that bar. (Only the Democrats in D.C. since 2008 ever clear half). But by that measurement, there are more Republicans who show up to vote in Minnesota (where the GOP has won at least a third of all eligible voters four elections in a row) than there are in Texas (where even favorite-son incumbent George W. Bush couldn’t quite break a third in 2004). Minnesota and Texas are at opposite ends of the turnout spectrum, which is how the Texas GOP can dominate its state with fewer voters than the Minnesota GOP perennially turns out in losing causes.
Looking at the chart above (“R &D Votes as a % of Eligible Voters, 2000–2016″), we can first see the national trend: Trump got the votes of 27.2 percent of all eligible voters, down from 27.4 percent for Romney, which was down from 28.1 percent for McCain, which was down from 30.5 percent for Bush in 2004. The last election in which Republicans increased their share of the eligible electorate was 2004, when turnout was up significantly on both sides from 2000. The 2000 election was held in a time of relative peace and prosperity, without a defining issue or large obvious differences between the candidates; this may explain why turnout was low. The opposite was true on all counts in 2008.
Democrats, by contrast, having increased their vote persistently up to 2008, have bled voters in the last two cycles, down from 32.6 percent of all eligible voters in 2008 to 29.6 percent in 2012 to 28.4 percent in 2016. On the whole, it should not be that surprising to see that Obama in 2008 was easily the strongest Democratic candidate of the last five elections, and Bush in 2004 was easily the strongest Republican candidate in the same time span. If you were reading this as a medical chart, you’d conclude that both parties were in pretty poor health right now.
Let’s cover the lower half of the same chart, below, and then we’ll break out the Republican and Democratic turnout trends separately by state:
Yes, it was hard to think anyone could drop that far off Mitt Romney’s 4.5 percent of eligible voters in D.C., but Trump got barely more than half of that, at 2.5 percent. Trump got the support of just 17.7 percent of all eligible voters in California, his lowest outside Hawaii and D.C., and dramatically down from Bush’s 26.1 percent in 2004. Here’s just the Trump 2016 turnout:
That is, as you’d expect, a lot of states with a lot of white people. And some surprises: New Hampshire and, as noted, Minnesota both had a lot of Trump voters in a losing cause, and Idaho had a lot of Trump voters even though he did unusually poorly there for a Republican (Idaho has a big Mormon population, and Mormons did not like Trump). New York had comparatively few Trump voters, even though Trump did much better there by this metric than Romney did. (New York’s 2012 turnout was undoubtedly affected by Hurricane Sandy. Many New Yorkers, myself included, got electricity back only the Sunday before Election Day.)
If we look at the past-five-election-year trends, with Trump 2016 as the baseline, we can get a clearer view of where Trump did well:
States above 100 percent show places where Trump won more votes than the GOP nominee had in a prior year: Thus, Trump got 141.5 percent of Bush’s 2000 vote in West Virginia, but just 88.1 percent of Bush’s 2004 vote in South Dakota. The “Avg” column shows the 2000–2012 unweighted average: Thus, Trump got 124.1 percent of the average GOP share of eligible voters in West Virginia over 2000–2012, but just 97.4 percent of the national GOP share of eligible voters over that period. “T+100” (for Trump +100) just tallies up how many of the past Republican candidates Trump beat — i.e., the number of the last four elections in which the Republican got a smaller share of the eligible vote than Trump did this year. There were six states where Trump got the largest share of eligible voters for a Republican candidate in the 2000–2016 period: West Virginia, Rhode Island, Kentucky, Arkansas, Pennsylvania, and Missouri. Only one of these, Pennsylvania, has been a swing state in that period.
At the other end of the scale, in terms of votes as a share of eligible voters, Trump did the worst of any Republican candidate in the 2000–2016 era in D.C. and eight states: Utah, California, Alaska, Washington, New Mexico, Idaho, Oregon, and Texas. These are also not swing states, but Texas stands out for being the biggest red state on the map. If you ranked the states by Hispanic population, you’d find that Trump did unusually poorly for a Republican in nearly every state with a lot of Hispanics except for Florida, Nevada, and New Jersey; thus, we see New Mexico, California, Texas, Arizona, Colorado, Illinois, and Georgia all on his list of states where he underperformed the Bush/Bush/McCain/Romney average. And Trump continued the deterioration of the GOP’s position in Virginia.
Notably, Trump actually underperformed both the Romney-Ryan ticket and Bush 2004 in Wisconsin; to understand his victory there, we need to look to Hillary.
This looks like a more typical blue/red chart. Hillary was a poor candidate in a lot of ways but fundamentally a conventional Democrat. Wisconsin was the only state she lost in which the turnout was above 32 percent of eligible voters. The only states she won in which the turnout was less than 28 percent of eligible voters were Nevada, New Mexico, and Hawaii — all states with large Hispanic/Asian/Native-American populations that tend to be low-turnout groups.
By recent historical standards, Hillary drew just 97.3 percent of the average Democratic turnout over the prior four elections, and that drops to 93.4 percent if you compare her to the last three elections. Clintonism was once a winning formula for Democrats, but they clearly turned out more voters for Obama and Kerry than they did for Hillary or Gore (or Bill Clinton, for that matter). Hillary did worse than Obama 2008 in every single state, and only in seven states did her turnout top Obama 2012, Kerry 2004, and Gore 2000: Georgia, Texas, Arizona, California, Massachusetts, Utah, and Illinois. Besides Utah, a result unique to Trump, that’s mostly a list of states with big Hispanic populations. She also did significantly better than Kerry and Gore, but worse than Obama, in most of the states with big black populations (Georgia, Virginia, North Carolina, D.C., Maryland, South Carolina, and Mississippi). Georgia was the only one of those states where she beat the Obama 2012 turnout. That suggests that black turnout was down in 2016, but it did not drop to pre-Obama levels — good news in the long run for Democrats.
Aside from Florida and North Carolina (which Obama won in 2008 but not 2012), you will notice that most of the swing states that Trump won are on the chart above. The Democratic vote was way down in the Midwest outside of Illinois; if you look just at the Democrats’ average vote as a share of eligible voters in the last three elections, Hillary drew 78.7 percent of the average in Iowa, 82.5 percent in Ohio, 83.1 percent in Wisconsin, 84.2 percent in Michigan, 84.5 percent in Minnesota, 84.8 percent in Indiana, and 92.6 percent in Pennsylvania. That’s where she lost the election, but working-class New Englanders were also unenthused: Hillary drew 85.4 percent of the Democrats’ 2004–2012 average in Maine, 86.5 percent in Vermont, 87.6 percent in Rhode Island, 90.5 percent in New Hampshire, and 94.1 percent in Connecticut.
#related#Hillary managed the worst Democratic performance as a share of eligible voters over the past five elections in 17 states, almost all of them states with above-average white populations: West Virginia, North and South Dakota, Wyoming, Tennessee, Missouri, Oklahoma, Iowa, Arkansas, Kentucky, Wisconsin, Michigan, Rhode Island, Alabama, Louisiana, Kansas, and New York.
On the whole, adjusting for the growth in the voting-eligible population, 2016 was much more a Hillary loss than a Trump win. But it was a good year for the GOP presidential ticket in key states such as Pennsylvania, Florida, Iowa, Michigan, Ohio, and Maine.
Would a different Republican have done better? In Part 4, I will look at where Trump ran ahead of, and behind, down-ticket Republicans.