GDP and forecasting the 2020 Election: What the Q2 numbers mean
OK, so this is it. Second quarter numbers are in, and that means it is time for a post.
After the 2016 election, I made the following pledge: I will not question the forecasting models again. In particular, in the Abramowitz Model I trust. That's the "Time For A Change" model. Contrary to what so many pundits have told you, the political science forecasting models got 2016 right. Forecasting models and polls are different. Polls are samples of the population, preferably close to the election, and we attempt to extrapolate from the sample when the sample is neither a random sample of the overall population nor necessarily reflective of the electorate. That can be difficult, and it was in 2016. That is not what forecasting models do. Forecasting models take variables that predict election results well in advance of the election, and plug them into parsimonious regression models to predict results without polls. The polls in 2016 were wrong. The forecasting models were right. The forecasting models, overall, indicated a generic-R advantage. Many of us ignored that because we couldn't fathom Trump actually winning. After he won, I made a pledge. I wouldn't disregard the forecasting models again.
Since then, whenever prompted by requests to forecast 2020, I have said, ask me again when I know what the Q2 2020 data are. Why Q2? That's the quarter that Abramowitz uses in his "Time for a Change" model. Once I know what the picture is in Q2, I can give you an idea of what we're seeing.
Well. Q2 data are out. And... um... what a mess. 32.9% drop, annualized. Gulp. OK, now keep in mind that's the quarter of the economic shutdown from COVID-19. Hence, it is anomalously horrific. What happens when you plug that number into the "Time For A Change" model? Along with Trump's below-water approval rating, the incumbent gets historically brutalized. It sort of blows up the model.
In 2008, the election wasn't even worth predicting with forecasting models because it was so easy to call. 32.9% drop in GDP? That would make 2008 look like a squeaker. Like I said, it blows up the model. And in fact, we're gonna have a mess on our hands in future years. This is just sort of a glitch in the math. In statistical terminology, this is going to be what we call an "outlier." That's... assuming any future for democracy, but that's another matter.
Substantively, though, 32.9% says... say hello to President Biden. And then President Whichever Woman He Picks As VP two months later after his eye explodes internally this time, causing a cerebrovascular event sorrynotsorry.
OK, now remember a couple of things. First, that 32.9% figure is an anomalous number. The economy shut down because of COVID. That's not an accurate assessment of the economy as it will be perceived in November. We use Q2 based on the observation that there is a lag between economic data and their perception among the public, but Q2 is the weird quarter here. Still, COVID is the big thing in 2020, and Trump is botching it in every way possible. Welcome to James Buchanan-territory in terms of presidential horribleness. A 32.9% GDP drop overstates Trump's disadvantage, but he has a yuge disadvantage. If you have to predict anything here using political science forecasting models, he's in big, big, big trouble. Yuge trouble.
To the point that the bigger question is what happens if he loses. Remember, he will not step down voluntarily. No, there is no margin of victory that is so big he'll be forced to recognize it. A landslide will just mean he says the magnitude of the Biden victory "proves" how much fraud there was.
If Trump actually wins with economic data like these, our models are really going to be screwed. Let's put it that way. Right now, if you're looking for a prediction, it's this: Biden wins, Trump refuses to vacate, chaos.
That's why Trump is putting out feelers for "postponing" the election. Just because it would be grossly illegal doesn't mean it is a zero-probability event. You have to stop thinking that way. The highest likelihood is the Biden wins-Trump refuses to vacate-chaos sequence, but don't rule out anything.
After the 2016 election, I made the following pledge: I will not question the forecasting models again. In particular, in the Abramowitz Model I trust. That's the "Time For A Change" model. Contrary to what so many pundits have told you, the political science forecasting models got 2016 right. Forecasting models and polls are different. Polls are samples of the population, preferably close to the election, and we attempt to extrapolate from the sample when the sample is neither a random sample of the overall population nor necessarily reflective of the electorate. That can be difficult, and it was in 2016. That is not what forecasting models do. Forecasting models take variables that predict election results well in advance of the election, and plug them into parsimonious regression models to predict results without polls. The polls in 2016 were wrong. The forecasting models were right. The forecasting models, overall, indicated a generic-R advantage. Many of us ignored that because we couldn't fathom Trump actually winning. After he won, I made a pledge. I wouldn't disregard the forecasting models again.
Since then, whenever prompted by requests to forecast 2020, I have said, ask me again when I know what the Q2 2020 data are. Why Q2? That's the quarter that Abramowitz uses in his "Time for a Change" model. Once I know what the picture is in Q2, I can give you an idea of what we're seeing.
Well. Q2 data are out. And... um... what a mess. 32.9% drop, annualized. Gulp. OK, now keep in mind that's the quarter of the economic shutdown from COVID-19. Hence, it is anomalously horrific. What happens when you plug that number into the "Time For A Change" model? Along with Trump's below-water approval rating, the incumbent gets historically brutalized. It sort of blows up the model.
In 2008, the election wasn't even worth predicting with forecasting models because it was so easy to call. 32.9% drop in GDP? That would make 2008 look like a squeaker. Like I said, it blows up the model. And in fact, we're gonna have a mess on our hands in future years. This is just sort of a glitch in the math. In statistical terminology, this is going to be what we call an "outlier." That's... assuming any future for democracy, but that's another matter.
Substantively, though, 32.9% says... say hello to President Biden. And then President Whichever Woman He Picks As VP two months later after his eye explodes internally this time, causing a cerebrovascular event sorrynotsorry.
OK, now remember a couple of things. First, that 32.9% figure is an anomalous number. The economy shut down because of COVID. That's not an accurate assessment of the economy as it will be perceived in November. We use Q2 based on the observation that there is a lag between economic data and their perception among the public, but Q2 is the weird quarter here. Still, COVID is the big thing in 2020, and Trump is botching it in every way possible. Welcome to James Buchanan-territory in terms of presidential horribleness. A 32.9% GDP drop overstates Trump's disadvantage, but he has a yuge disadvantage. If you have to predict anything here using political science forecasting models, he's in big, big, big trouble. Yuge trouble.
To the point that the bigger question is what happens if he loses. Remember, he will not step down voluntarily. No, there is no margin of victory that is so big he'll be forced to recognize it. A landslide will just mean he says the magnitude of the Biden victory "proves" how much fraud there was.
If Trump actually wins with economic data like these, our models are really going to be screwed. Let's put it that way. Right now, if you're looking for a prediction, it's this: Biden wins, Trump refuses to vacate, chaos.
That's why Trump is putting out feelers for "postponing" the election. Just because it would be grossly illegal doesn't mean it is a zero-probability event. You have to stop thinking that way. The highest likelihood is the Biden wins-Trump refuses to vacate-chaos sequence, but don't rule out anything.
Comments
Post a Comment