Don't pay attention to election forecasting models based only on economics
They might not capture our politics like they used to
The takeaway: Election forecasting models will never be perfect. Those based only on economics are more imperfect than others. They have higher errors and are built on a theory of political behavior that may not apply in an era of polarized politics. Therefore, we should probably not rely on them.
Thanks for reading my thoughts on this subject. And thanks for subscribing! Your membership adds up and makes all this newslettering possible (reminder: I do all this work independently). Please consider sharing online or with a friend; the more readers, the merrier. Remember that apart from getting special articles subscribers can also comment below each post and participate in exclusive threads.
As always, send me your tips about what you’d like to read about next. Or what you don’t want to read. Or your feedback otherwise. Also, cat pictures are nice, so please send them to me! I’m firstname.lastname@example.org, or @gelliottmorris on Twitter.
PS! I’m giving away free trials to this newsletter from now until November 1st. I’d love if you shared it with your friends. Click the button below or share this link: https://thecrosstab.substack.com/subfriends
Moody’s Analytics, a financial firm, predicted that Hillary Clinton would win the presidency in 2016. She would take the White House and 326 electoral college votes, Moody’s proclaimed.
Instead, Clinton won just 232 votes (before faithless electors) to Donald Trump’s 306. How could Moody’s be so wrong?
One possibility is that electoral outcomes no longer match economic “fundamentals” as well as they used to. I have written for The Economist that partisanship has blocked some voters’ ability to rationally assess the state of the economy. Today, it seems that our attitudes are “moderated” by whether or not we approve of the party in power. Notably, this is a huge departure from the historical trend. Here is the main graphic from the piece, “American voters don’t care about the economy”:
Such changes to voter psychology would render economic predictors—like GDP growth or unemployment—much less powerful in predicting presidential elections. That’s because the old theory went that voters punish presidents “responsible” for bad economies and reward those responsible for good ones. But if we all evaluate the economy through our partisan lenses, the relationship falls apart. This could also explain why economics-based forecasting models aside from Moody’s were also wrong in 2016.
It’s also possible that models that only use economic predictors just aren't that great at forecasting elections in general. According to PollyVote.com, an election forecast aggregator, economics-only forecasts have the second-largest historical error of any method. In other words, maybe 2016 wasn’t a problem of induction, rather economics-only models are often that bad. The following graphic depicts historical forecasting errors and comes from PollyVote:
Finally, there’s the possibility that Donald Trump is simply a unique case. He’s unpopular. He’s polarizing. He’s unconventional. Perhaps Trump simply throws off the models. For the record, I’m not to convinced by this argument. The president’s unpopularity has created pretty predictable consequences, such as historic mid-term losses. (Perhaps you could even argue that impeachment was more likely because Trump is so unpopular). Even if he is “special”, or perhaps “abnormal”, there seem to be some measurable ways we can quantify the president’s abnormality.
All this considered, I’d like to give some bold advice: you shouldn’t pay attention to forecasting models that are based only on economics. They seem to be prone to error, perhaps more so than other types of models, and are not theoretically sound.