How do markets and the economy affect the forecast?
An early look into the methodology behind our presidential model.
What’s going on?
In case you missed it, the July U.S. jobs report came in pretty rough today. The total number of new jobs was about 60,000 below initial estimates, a sudden reversal compared to past months where employment data was surprisingly strong. Additionally, unemployment ticked up to about 4.3%, from 4.1% in June.
Both domestic and global markets were quick to react, with the Dow down over 750 points (as of writing this article) and Japan’s stock markets experiencing their biggest plunge since the start of the COVID-19 pandemic. Many were frustrated with the Federal Reserve and Chair Jerome Powell for refusing to cut interest rates this month, opting instead to likely do so in September. (That rate cut may now be twice as large due to today’s numbers.)
But this isn’t an economy-oriented site. Although I may write about some basic economic topics at some point, it will probably be tied to politics most of the time, like it is today. The phrase “it’s the economy, stupid” has long been used in the political world, and voters continue to rate the economy (or prices/inflation) as their most important issue today.
The forecast
But as someone making a model of the election, I have to decide how to implement (if at all) economic data in the fundamentals/national-polling side of things. Some (smarter) forecasters like Nate Silver use actual economic metrics like stock markets and industrial performance indices to adjust their priors, and G. Elliot Morris at FiveThirtyEight (now apparently stylized as 538) use GDP, wage data, and inflation.
But I didn’t feel that using raw economic data was the best way to adjust projected vote shares for candidates for one key reason: sometimes, as much as it annoys me, voters are more vibe-based. Despite inflation cooling off significantly, despite relatively strong job numbers, the fact that prices are going up can be a huge frustration for voters, enough of a frustration that they may vote against the incumbent party. (in this case Kamala Harris) At the same time, a dip in the stock market probably isn’t going to sway many people’s votes. So instead of relying on the economics themselves, I use “consumer confidence” trackers, which effectively poll Americans on how they feel about the current state of the economy, or how confident they are in its position. For the forecast, I incorporate three major organizations that conduct these surveys:
Each of these current readings is converted to an adjustment (typically a few percentage points), and then the three adjustments are simply averaged out to get the final forecast economic adjustment (currently -2.3%). A positive value boosts Harris’ margin by that much, whereas a negative value boosts Trump’s. Here’s how that works with today’s preliminary national polling average:
Harris: 43.3% - (2.3%/2) = 42.2%
Trump: 43.0% + (2.3%/2) = 44.1%
Third parties remain unchanged.
It’s also key to note that this only affects 40% of each state’s forecast, at most, because that adjusted national polling average is used to get a fundamentals-based state-by-state expectation. By Election Day, however, this fundamentals projection will only be around 5-10% of the forecast in competitive states, so it will have very little effect. It’s still early, though, and we have barely any polling in most states, so it’s fair to say the economic adjustment is playing a role in depressing Harris’ chances.
But in terms of today’s events, they’ve had no effect on the model whatsoever. We only get new consumer confidence data around twice a month, at the end and near the middle, so if that drops around mid-August, it could change things. (Paid subscribers to HDR Analytics can view the economic adjustment on the forecast page now, but it will not be public for free viewers until Election Day)
So I just wanted to give a teaser of the full methodology post (coming later this weekend or next week) with regards to today’s economic events. Hopefully, it helps make the model more straightforward. More on methodology in the coming days, and check back here tonight for the daily forecast update.