Donald Trump trails Joe Biden in the RealClearPolitics polling average by a significant margin, as he has for the entire year. However, the rioting and shootings in cities, combined with the end of the Republican National Convention, have raised concerns in some quarters on the left that Trump may be getting a bounce. Remember, because he likely has a built-in edge in the Electoral College, a reasonably small bounce at the national level would make this a very competitive race.
The truth is, we probably have to wait another day or two for more post-convention polling to give us a better view of the picture. (An Emerson survey shows the contest tightening.) But there has been a growing argument, raised by the Biden campaign in his Pittsburgh speech Monday, that unrest in the streets should be attributed to Trump. Perhaps it will be, but this raises some larger questions about how elections work, so it is worth addressing.
Most of the electoral analysis you read, at least from quantitative analysts, rests upon an old theory that political scientists call “retrospective voting.” The idea is pretty straightforward: When voters go into the voting booth they aren’t making a choice between two candidates. Instead, they’re mostly passing judgment on the incumbent candidate’s performance. It is the political science equivalent of Ronald Reagan’s famous query: “Are you better off today than you were four years ago?”
So, when we model elections, we typically look at things like the state of the economy, or perhaps how wars abroad are going. If they go well, the party in power will perform well. If they are going poorly, the party in power will not perform well. This theory is well-tested and, at least to my mind, a strong one.
And yet, part of the problem that quantitative studies of elections face is that our dataset is limited. Presidential job approval numbers (which, we should keep in mind, are not actually a “fundamental”) date back to the 1940s, while relevant economic metrics usually don’t go back as far. In other words, we are looking at only about 18 or so cases when we model.
If we go back farther, there are instances that don’t fit the referendum model particularly well. One obvious case is 1820, where James Monroe was almost unanimously elected in the midst of one of the most severe depressions in American history. The 1876 election really had no business being as close as it was, and the incumbent party held the White House in 1908 despite the economy bottoming out in the second quarter of that year. If we include congressional elections we see years like 1890, when Republicans lost 90 seats in the House on the back of a very minor recession, or 1910, when Republicans lost control of the House despite a booming economy.
Put differently, most people remember Bill Clinton’s famous “it’s the economy, stupid” dictum from his 1992 campaign. What they don’t remember is that this was actually only one plank in a three plank talking point list, with the other two being “change vs. more of the same” and “don’t forget about health care.”
There are also elections that superficially fit the retrospective voting model, but when you interrogate the data a little more they fit much less well. One of those cases is 1948, which may shed light on the current election. In that election, Harry Truman famously came back from trailing badly in polling over the summertime to win the general election, including a popular vote win of almost five percentage points.
This fits with the retrospective voting theory to a degree. After all, the economy peaked in the fourth quarter of 1948, which is about when you would want the economy to peak as an incumbent. Under this telling, voters turned to Truman over the summer and fall because they decided that they liked his stewardship of the economy.
But this doesn’t jibe with the way Truman ran his campaign. Look at the following map, which shows the shift in the Democrats’ share of the two-party vote (that is, excluding third parties) from 1944 to 1948. Click to enlarge the image.
As you can see, Truman lost ground across much of the United States (Alaska and Hawaii were not yet states, and Truman didn’t appear on the ballot in Alabama). His losses in the South aren’t that surprising, given the presence of a Dixiecrat candidate, Strom Thurmond, but he lost several vote-rich states that Franklin Roosevelt had carried four years earlier, including Michigan, New Jersey, New York and Pennsylvania.
What saved him, in part, was his solid performance in farm states. Indeed all five states that flipped from Dewey in 1944 to Truman in 1948 were states with substantial farm economies: Ohio, Wisconsin, Wyoming, Iowa and Colorado. In fact, Truman was the first Democrat to carry Colorado and Iowa since FDR’s sweep in 1936.
This is where things get interesting: The farm states that swung toward Truman were not experiencing the same economy as the rest of the nation. As Zachary Karabell relates in his 2000 book, “The Last Campaign: How Harry Truman Won the 1948 Election,” these states had slipped into a recession before the rest of the country had. If they were engaging in retrospective voting and automatically attributing bad news to the Democrats, Truman likely would have lost these states.
Instead, Truman campaigned heavily there, and used the ailing economy to his advantage. He emphasized Republican opposition to farm support systems, and convinced voters in this area – again, many of whom had voted against FDR twice – that the Republican approach to farm policy threatened to make many of them worse off.
This, then, is how the issue of crime and rioting is potentially dangerous to Biden, even though it is happening on Trump’s watch. The Republican argument will be “the people rioting in Portland and Seattle and Kenosha are not Trump supporters, and he will be more likely to stop them.” It may not be fair (most antifa supporters don’t care for Biden either), and it certainly may not work. We’ll have to wait for more polling to know for certain, which, fortunately, we should get in the next few days. It would be a mistake, however, to assume that voters will blame this on the incumbent this time.