This morning the Bureau of Economic Analysis released its figures for 2nd quarter Gross Domestic Product for 2020. They're not good. According to these measures, the economy, measured in terms of real GDP, shrank almost ten percent just between the first and second quarters of 2020. Can we make a forecast about the presidential election based on this?
I make an economic forecast below, and I'll get into what it is and how I derived it. But let me just say at the outset that the forecast based on these economic figures is absurdly low. Like, probably not plausible. But the message we should be taking away from this is that the fundamentals are extraordinarily and historically bad for an incumbent president seeking reelection.
Now, does it even make sense to make a forecast based on the economy at all this year, given how bizarre and volatile economic behavior has been amidst the Coronavirus pandemic and how much things could still change by November?
Honestly, I'm not sure of the right answer here. I tend to take the view that voters will give Trump, as the incumbent president, credit or blame for the economy, and that this will be the main driver of the vote in November, as in previous presidential elections. It may be, as Jonathan Bernstein has suggested, that voters will cut Trump some slack since the economy was at least partially shut down intentionally to combat the pandemic. My assumption is that voters will assign credit or blame no matter what the reason behind the economic situation is, but yes, it's a weird year, so it's hard to be certain.
Also, I should note here that economic forecast models are useful and interesting but aren't substitutes for other knowledge. We might treat them as a baseline -- is the incumbent performing better or worse than we'd expect given the fundamentals? -- but not assume they are better predictors of the vote than, say, polls asking people how they're going to vote. (Although sometimes they are.)
Okay, what can we learn from GDP? Well, it's good to know how much economic activity occurred in the 2nd quarter of 2020, but we need to know what to compare it to. Should we measure growth between the 1st and 2nd quarter? Or maybe for the whole year between the 2nd quarter of 2019 and the 2nd quarter of this year? Or something else?
Here's a little table showing the R-squared (a rough measure of goodness-of-fit) for a variety of measures of economic growth predicting the vote share for the incumbent party between 1948 and 2016. More recent growth in real GDP (between Q1 and Q2 of the election year, or between Q4 and the previous year and Q2 of the election year) tends to do better as a predictor of the vote. Growth closer to the election may work even better, but since we won't know 3rd quarter growth until very close to the election, I'm ignoring that here.
But the best of these has an R-squared of .33, meaning that variations in real GDP early in an election year explain about 33% of the variation in the presidential vote. That's useful, but hardly complete. Throw in a few other simple explanatory variables -- the president's Labor Day Gallup approval rating, whether the incumbent president is running for reelection, whether the U.S. is at war -- and that R-squared jumps up to over .65.
So let's just go with that measure of GDP growth between Q1 and Q2 of the election year. Here's what that looks like as a predictor of vote share for the incumbent party's presidential candidate between 1948 and 2016:
Okay, that's kind of useful. When GDP growth is low or negative early in an election year, we know the president's party tends to lose.
Here's the thing: the figure for 2020, according to the latest release from the Bureau of Economic Analysis, is that real GDP shrank by 9.5 points between the 1st and 2nd quarters. That doesn't fit on the graph above. Not remotely.
If I plug that figure into regression equations accounting for GDP growth between quarters 1 and 2, assume the President will have a 40 percent Gallup approval rating on Labor Day, give him credit for being an incumbent and for the absence of a war, I get a predicted two-party vote share of 31 percent. If I use a similar regression equation that accounts for GDP growth between the 4th quarter of the previous year and the 2nd quarter of the election year, I get a predicted two-party vote share of 39 percent. Those models have R-squareds of .69 and .65, respectively.
I am highly skeptical of these forecasts. The better of these forecasts for Trump would be on par with the vote shares received by Barry Goldwater and George McGovern in 1964 and 1972, respectively. That's historic landslide territory, the sort of thing we haven't seen in a long time.
There will be future revisions of these numbers, and the third quarter will have interesting things to tell us, but probably the most useful thing we can conclude from this is that the fundamentals are really, really bad for President Trump. They are not an image problem. He is trying to sell a very bad product right now, and the fact that he's only losing by 8 or 9 points is kind of amazing.