He flipped big-enough battleground states.
In Part 1 of this four-part series on Donald Trump’s victory over Hillary Clinton, I examined the history of post-incumbent elections: elections following the reelection of an incumbent president, in particular the true post-incumbent elections where no incumbent was on the ballot. The trend in the national two-party vote in those elections has been overwhelmingly in favor of the party out of power, to the point that Trump’s gain of less than a point in the two-party vote against Hillary was decisively the worst showing by a challenger in 200 years. So did he win by doing unusually well in the battleground states? By historical standards, not really — except in one important way that made all the difference.
Flipping the Battleground States
As I detailed when I studied this in depth in 2014, swings in the national two-party vote against the party in power in post-incumbent elections are a meaningful measurement because, in all but one such election, the trend in the most hotly contested battleground states has been consistent with the national trend. (More on the one exception below).
For purposes of this analysis, I’ve defined “battleground” states in two ways. First, looking prospectively from the prior election, I included every state in which the incumbent won at least 45 percent but less than 60 percent of the two-party vote. Second, looking retrospectively to identify battleground states that didn’t look like battlegrounds four years earlier, I included states outside that range that flipped to or from the incumbent party in the Electoral College. I did not include states that flipped between the challenger party and a third party, unless they met the other criteria, although those were rare. I counted as partial “flips” the states that awarded some of their electoral votes by congressional district (Barack Obama flipped 20 percent of Nebraska’s electors in 2008; Trump flipped 25 percent of Maine’s in 2016) or elected the electors individually (Maryland had an idiosyncratic system in 1904 and 1908 that helped Democrats flip seven of eight electors in 1904; William Howard Taft flipped one in 1908).
Here’s how the battleground states swung in 2016:
If you knew no history and drew your view of the intransigence of red and blue states from the political media of the past few years, this would look pretty impressive for Trump: He outperformed Mitt Romney in the two-party vote in 18 of 23 battleground states, and Hillary underperformed Obama’s 2012 percentage of the overall popular vote in 21 out of 23. Hillary won a majority of the popular vote in just six of these states, compared with 18 for Obama. Trump flipped six battleground states and a congressional district in Maine, and he delivered an 8-point shift in the two-party vote in Iowa, almost 6 points in Ohio, and almost 5 in Michigan. Then there are the non-battleground states:
As in years past, the pattern in the non-battleground states was mostly similar to that in the battlegrounds: a few big swings to Trump, a lot of small swings to him, and a handful of swings to Hillary, including an enormous collapse in the GOP’s position in Utah (though not enough to change the outcome) and significant shifts in the largest blue and red states, California and Texas.
By historical standards, however, Trump did not do particularly well in the battleground states. (For the battleground states, I go back only to 1868, for a variety of reasons explained in the original study.) Let’s start with how many states he flipped. (The Flip-% column is states lost by the party in power divided by the battlegrounds they won in the last election. The Net % column is states lost by the party in power, minus states captured from the challenger party, divided by all battleground states.)
Six states flipped (6.25 if we are counting fractions) is the fewest in a post-incumbent election since 1928, or since 1944 if you count races with an incumbent on the ballot. Nobody else has captured the White House while flipping fewer than nine. On average, the party in power loses a staggering 61 percent of all battleground states it had won under the last incumbent; on balance, it sees 44 percent of all the battlegrounds flip to the challenger in the next election. Trump flipped fewer than a third of all the Obama-won battleground states, roughly half the historical average, and his overall pickup of 6.25 states was just over a quarter of the battleground states.
Shifting the Battleground Voters
If we look at the shift in votes in each state rather than just the states flipped, we can put Trump’s progress in the two-party vote in the battleground states in historical context. The first set of columns (V+, V- and Loss%) tracks the battleground states in which the incumbent party gained or lost share of the two-party vote. The second (V+3, V-3, and V-3%) tracks the battleground states in which the incumbent party gained or lost 3 or more points in the two-party vote. The third does the same for 4-point swings. The last two sets of columns track 7- and 10-point swings (there were none in favor of the party in power in 450 battleground state elections since 1868):
As the chart makes clear, Trump’s swing of 3 points in 39.1 percent of battleground states was about half the historic average; his swing of 4 points in 21.7 percent of battleground states was a third of the historic average; and his swing of 7 points in just 4.3 percent of battleground states was a tenth of the historic average. And in each case, it was even worse compared with the average in elections when the Democrats are in power. Michael Dukakis shifted six states by 7 points or more; Obama shifted seven states by that amount; George W. Bush, nine states; JFK, eleven states; Jimmy Carter, 24 states; Richard Nixon, 31 states.
Trump matched that only in Iowa. Sort the same chart by the number of states with 3- or 4-point swings in the two-party vote, and you can see Trump at the bottom of the pack in making a real multi-state dent in the opposition, trailed only by 1928 and the races with incumbents on the ballot:
I’ve mentioned 1928 a few times because it was an exceptional case. The Democrats decided to do something very unusual that year: Having found themselves reduced in 1924 to a regional party of the Deep South with no visible path to victory (just four years after Woodrow Wilson left office at the end of his second term), they nominated Al Smith, the Catholic, anti-Prohibition governor of New York. Smith brought the progressive Northern Democrats back into the fold (they’d bolted to a third party in 1924), greatly improved the Democrats’ showing in a lot of states where they got crushed in 1924, built the foundation for a major demographic shift to the Democrats in the 1930s, and even stole the nation’s two most heavily Catholic states, Massachusetts and Rhode Island, from the GOP coalition.
But in the short run, Smith was an Electoral College disaster for the Democrats: The party’s remaining Southern base bolted to Republican Herbert Hoover, who carried Tennessee, North Carolina, Oklahoma, Virginia, Florida, and Texas, all states the Democrats had won in 1924. By contrast, Smith lost 19 states where the two-party vote swung between 7 and 32 points in his favor; without actually winning the states, he ate into the huge margins that the GOP had there four years earlier.
The unconventional choice by the challenger party in 1928 severed the usual relationship between national popular-vote improvement and Electoral College winnings in the battleground states, and in September 2014, I didn’t think there was much chance that the GOP would try anything similar in 2016. As I wrote at the time:
It’s conceivable that Republicans could choose a nominee who brings in a lot of new voters while alienating a big chunk of their base — Rand Paul might excite a lot of libertarians while losing many national-security hawks and law-and-order voters, Brian Sandoval could win over Hispanic voters while alienating pro-lifers — but a Paul nomination is not all that likely, and Sandoval shows no sign of running. It’s more likely that the 2016 Republican nominee will be someone acceptable to Republican primary voters. That doesn’t leave much in the way of hopeful examples for the Democrats to emulate.
Obviously, I didn’t anticipate a nominee like Donald Trump who would (to varying degrees) alienate national-security hawks, pro-lifers, and other elements of the party base (supply-siders, free-traders, immigration moderates) while reaching out beyond traditional GOP voters. But that’s exactly what happened. And as we’ll see, the result was an uneven voter movement in place of what should have been a consistent shift to the GOP. It’s worth noting that many of the “fundamentals-based” political-science models predicted such a shift, too: Knowing nothing of Trump, they foresaw a Republican victory. Political scientist Alan Abramowitz even took the unusual step of explaining in August why we should not believe Abramowitz’s own, generally successful “Time for Change” model, which was predicting at the time that Trump would win.
The flip side of the states Trump won is the unusual spectacle of the challenger party losing ground in five battleground states. That had happened in only eight battleground states in total in the prior seven post-1948 post-incumbent elections (counting 1976) combined.
Traditionally, the few times this has happened in the past have been a combination of a state following a trend toward the party in power (note three examples here of Tennessee’s getting redder) and candidates who were demographically and culturally out of step with the state (Obama in Arkansas, Louisiana, West Virginia, and Tennessee; JFK in Tennessee and Oklahoma). And Humphrey’s improving on LBJ’s percentage of the two-party vote in Georgia in 1968 was a Pyrrhic victory, because George Wallace (running as an independent) won Georgia by taking slightly more Goldwater votes from Nixon than the LBJ votes he took from Humphrey.
Hillary’s gaining ground in Georgia and Arizona is the first example we have since 1948 of the party in power actually improving its position in a battleground state held by the challenger party in the previous election. That should be deeply worrying to Republicans concerned about the demographic drift of those states. But it also speaks to the fatal misallocation of resources by the Clinton campaign when it focused on poaching those two states instead of shoring up the Midwest. Which brings us to the one key battleground-state measurement by which Trump did historically well, rather than historically badly.
Winning the Big Ones
I’ve focused so far on shifts in the national popular vote and the number of battleground states with shifts in the vote and shifts in the result. In a normal year, these two indicators work in tandem and show us either a large or a small movement toward the challenger party. As a result, I haven’t weighted the battleground-state movements by the size of the battleground states (the national popular vote usually picks that up).
But what made 2016 unique was that the incumbent party’s losses were minimal both in the national popular vote and the number and margin of movements in the battlegrounds . . . and yet the election was decided by the size of the battleground states that Trump flipped. Trump won seven of the nation’s ten largest states, compared with two for McCain and three for Romney. He flipped only six battleground states, but he managed to flip four of the seven largest states. On average, Trump earned 16 electoral votes per state flipped (counting Maine as a quarter of a flip). That’s not quite as unusual as you might think, but it’s almost the highest average since 1868.
The average challenger in a post-incumbent election earns 11.1 electoral votes per state flipped, and net — including states lost and states that don’t change — adds 4.1 electoral votes per battleground in play. Only Tom Dewey in 1948 pulled in a higher average (16.2 electoral votes per flip), including 47 electoral votes in New York, 35 in Pennsylvania, 19 in Michigan, and 16 in New Jersey. But Dewey lost the election because he also gave back five states he’d won over FDR in 1944, including Ohio (25 electoral votes), Wisconsin (12), and Iowa (10). Trump’s 4.3 electoral votes per battleground in play is roughly even with the historical average but lower than that of losing candidates such as Dukakis and Wendell Willkie. Trump’s haul of 100 new electoral votes is less than those earned by Obama and George W. Bush and a small fraction of those won by Eisenhower, Nixon, JFK, and Carter.
By historic standards, Trump didn’t do particularly well in the battleground states. But he did just well enough in the biggest ones to make the difference. But was that a win for Trump, or a loss for Hillary? In Part 3, I’ll move on from the two-party vote to look at each side’s turnout.