
Illustrated by Emily Scherrer
Over the past few weeks we’ve been trying to figure out just how much ground the Republicans have regained in the race for control of Congress. There is none.
FiveThirtyEight’s deluxe prediction put the Republicans’ chances of winning the Senate at 34%. This is up from a low of 29% in mid-September.
The betting market has seen more drastic changes. In fact, the market is nearly even, and he has a 49% chance of a Republican controlling the Senate. This is up from a low of 33% in late August. The market has been fairly consistent with FiveThirtyEight’s forecast for most of the cycle. Not now.
Why? Well, as I wrote two weeks before him, I don’t think it’s strange for the market to worry about systematic poll bias. With Thursday morning’s worse-than-expected inflation report, it’s no wonder the market expected an impact on Democrats.
However, the gap between markets and models is large and growing. FiveThirtyEight’s models aren’t the only ones. Some other models are similar or show no Republican backlash at all.
I don’t care so much if stable Gap between market and model. If the market wants to take the predictable poll bias hypothesis more seriously than the model, or if it wants to give more weight to the long history of presidential party poorness in midterm elections, that’s fine with me.Various Aggregating hypotheses is generally the strength of the market.
Three Races Five Thirty-Eight Lowering Odds For Democrats To Hold The Senate
Markets and models are also trajectory But I think this leads to two potential ways political bets sometimes come to market. is not Its smart.
One of the weaknesses of these markets is that they tend to follow media narratives about racing rather than underlying evidence. Source for this claim: It really belongs to you. very Long time.
My impression of the past few weeks has been that media coverage has leaned towards talk of a Republican backlash. (No hypocrisy on this one, but we are too.) And in some ways, this coverage is justified. Republicans have clearly won support in the Senate elections in Wisconsin, Pennsylvania and Nevada, and these are important contests! The steadily improving political climate that the Democrats experienced over the summer is over.
But the Republican position hasn’t really improved in the Arizona Senate race, and Herschel Walker probably lost his seat in Georgia. Difficult to distinguish from noise. The popular vote and President Biden’s approval ratings haven’t changed much either. The House outlook remains in favor of Republicans, but there is not much change in either direction.
It’s confusing. People tend not to click on headlines like: The national trend is unclear, perhaps a hint of a Republican recovery, but it could just be noise ¯\_(ツ)_/¯.
Another weakness of these prediction markets is the lack of technical sophistication of traders. election predictionsI’m not saying it’s generally unsophisticated; Traders may know a lot about politics and may be familiar with market psychology.But I have some questions Through the process of actually building the model For example, determining how much election predictions should change in response to small but noisy changes in polls.
Most of these traders have not built election models themselves. Because I know people who bet on elections. Whether it’s his players in poker, his bettors in sports or his quantitatively oriented researcher, I have a lot in common with Nate World and I often meet them in my personal and professional environment. They say they’ll bet on elections, they’ll say they follow FiveThirtyEight and other predictions, The usual We are not building an electoral model from scratch.
Frankly, they shouldn’t be building them. To build a half-hearted electoral model, I estimate from my own experience that it takes months of hard work if there is a lot of support, and more if not. If elections are every other year and he only happens once, it’s not a great investment. You’d be better off working on a model where you trade options or bet on soccer, which has more liquidity and more frequent opportunities to make money.
As mentioned earlier, much of the value that models provide is all Opinion polls are often highly unrepresentative samples, not just those highlighted in the media.
The model also understands important facts about the medium term. No Turn on the penny, as is sometimes done in presidential elections. (Most “game changers,” even presidential elections, are false alarms.) In general, when you hear about a change in momentum in the midterm elections, you should be on your guard.
why? Well, voters aren’t paying that much attention. Yes, you are a FiveThirtyEight reader, so you are paying attention. But the general public pays less attention to midterm elections than presidential elections. For example, take a Google search for the word “polls”.

The surge in presidential elections is much larger, about 3-4 times. People certainly pick up on Big He news, such as inflation and the overthrow of the Roe v. Wade case, but they don’t follow the daily news like they did during the presidency. What happens on cable news tends to stay on cable news.
As such, we found that caution is needed when interpreting general voting changes. Your default is to assume that the public’s preferences for which party controls Congress are fairly stable. It takes time to move the needle.
Another difference between midterm elections and presidential elections is more obvious, but worth mentioning. Every state and district puts a different candidate on the ballot. A scandal involving Biden or former President Donald Trump would ripple through every state. Anything involving Herschel Walker will probably only make an impact in Georgia.Unlike presidential elections, very No You can safely infer how one state is behaving based on other states. Changes in Pennsylvania polls tell us very little about Arizona polls.
FiveThirtyEight’s site traffic is probably a bad thing to say this, but the final weeks of the mid-campaign may not be so climactic and action-packed. Sure, many of the individual races are very attractive, but an ‘October Surprise’ that affects one race may not affect others. There may be some changes in the big picture and top line numbers, but probably not by much. That means going into Election Day with a lot of uncertainty about which party will control Congress.