What's going to happen

Brace yourselves...

This post will focus on some of the forecasts of Tuesday’s results (which may take several weeks to fully determine, since we’ll likely have a runoff for the Senate seat in Georgia, and California, among other places, takes a long time to count all its votes. Oh, by the way, Brazil managed to count 120 million or so votes in about four hours last Sunday).

I’ll start with Nate Silver and 538. As of today, the main model gives Republicans a 54% chance of retaking the Senate and an 83% chance of retaking the House. As everybody knows, some of the polls-based forecasting models got the 2016 presidential election “wrong.” Silver, in particular, took a lot of grief for “predicting” Clinton would beat Trump. In reality, Silver’s model, which gave HRC a 70% chance of winning a majority of electoral college votes that year, was much more cautious than other polls-based models. Silver himself expressed bafflement that modelers like Princeton’s Sam Wang essentially thought Trump had no chance of winning. What Silver expressed, substantively, was the belief that Trump had a real shot, especially if the polls were slightly off in a way that he thought they could be slightly off. After all, something that is expected to happen three times in ten tries is a highly plausible outcome. Context matters for how we lay people assess probabilities. If you went to the doctor and they told you, based on an examination they’d just completed, that you had a 30% chance of developing a terminal illness, you probably wouldn’t breathe a sigh of relief that you had a 70% chance of being in the clear. Instead, you’d be freaking the F out. The common interpretation of Silver’s 2016 performance was that he said Clinton was going to win, yet she lost. But I think Silver is pretty careful and provides a reasonable approximation of the range of likely outcomes.

Turning to 2022, what do his model forecasts mean, in substantive terms? About the Senate, a 54-46 probability means control of the Senate is a coin flip. Or more accurately, a series of coin flips, particularly in Georgia, Pennsylvania, Arizona, Nevada, North Carolina and Ohio. You don’t need the 538 model to know this. But the point is that no one has a more credible guess than that about what is likely to happen. Senate races hinge to a meaningful degree on candidate quality and, partly thanks to Trump, the GOP nominated some doozies this year, including Dr. Oz in Pennsylvania, Herschel Walker in Georgia and J.D. Vance in Ohio. They might win all of these seats anyway (I don’t think they will, but they might), but less awful candidates would certainly have improved their chances of recapturing the Senate).

As for the House, the picture doesn’t look good for Democrats. It’s worth noting that things that are predicted to happen one time in six (17%) do indeed occur. Silver has noted that while polling error in recent cycles has tended to result in overstating Democratic support, it’s not a given that, if there’s meaningful polling error this time, it will necessary be in the same direction. So, for those of you who obsessively check in on these sorts of forecasts (and I have no idea what gave you the idea that *I* am one of those people), these models can’t, of course, actually tell us what is going to happen. They’re making guesses about possible outcomes based on their estimation of how the factors they deem relevant now compare to those that have yielded particular outcomes in the past.

There is another set of models that doesn’t rely upon polls, but instead on longer term “structural” factors, like the state of the economy and who controls the White House Indeed, if the only thing you know about a midterm election is the party of the occupant of the Oval Office, you can guess with a high degree of confidence which party is poised to do well in midterm elections - the other party. So, expectations for Democrats’ House chances this year were always going to be bad, based solely on the fact that Joe Biden won the presidency in 2020. The added uncertainty of and general dissatisfaction with the economy only adds to that negative picture. One of the political science models that uses these assumptions as a basis for its forecasts, developed by Charles Tien and Michael Lewis-Beck and with a good track record, says that Democrats can expect to lose 3-5 Senate seats this year and anywhere from 20-40 House seats.

In a recent interview, Lewis-Beck criticized pundits and observers (he’d include Silver in this group) who “assume the election at hand exists apart from past elections, with a de novo electorate moving into unchartered territory. That perspective may sell newspapers, but it tends to be false.”

A key reason for the uphill battle the “in” party faces in midterm elections is that, not surprisingly, voters are going to blame that party for whatever is going wrong at the time. In this time-honored view, voters cast ballots based on what officeholders have done for them lately, not what they think officeholders might do in the future. Whether Republicans have viable solutions for *any* of the issues voters seem most concerned about - inflation, crime, gas prices, etc - is mostly irrelevant. Americans are generally in a sour mood these days, and the party in charge is going to pay the price for that, according to this reckoning.

One wrinkle this year, however, is that it’s not normal for the party in power to have such a consequential issue decided against its interests. In political terms, that’s what the Dobbs decision represents. Democrats may control Congress and the White House, but they don’t control the Supreme Court, and the Court just made arguably its most politically consequential ruling in decades.

In that vein, some political scientists do think - contra Lewis-Beck - that there could be special factors this cycle that have the potential to upend normal expectations. Writing in the Guardian last week, Laurel Elder, (friend of the substack) Steve Greene and Mary-Kate Lizotte believe that abortion might help Democrats in ways that neither the structural approach nor the polls-based models can fully capture.

Elder, Greene and Lizotte note that one of the hardest problems to solve in election polling is figuring out who is going to turn out to vote. Pollsters make assumptions about who is a “likely” voter, and give those assumptions more weight as elections approach. They also assume, correctly, that younger voters will tend to turn out at lower rates than older voters.

But, Elder, Greene and Lizotte, say that assumption might be missing the depth of anger among young women, in particular, in the aftermath of the Dobbs decision:

A recent poll of Gen Z Americans in swing states supports this, providing empirical evidence that young people are energized to vote and continue to rank abortion as their top issue, even while the issue has slipped in importance for older Americans. Young people’s passion on issues has failed to translate into actual action in the voting booth in the past; however, if young pro-choice women actually do turn out in higher numbers than forecasting models are expecting, this could provide a multi-point bounce to Democratic candidates in key House and Senate races.

Indeed, earlier this summer, it looked like that issue might scramble the normal midterm calculus. While Democrats have never been favored in this cycle to retain the House, they started polling much better than the incumbent party typically does after the Dobbs decision. Furthermore, in August, Kansas voters decisively rejected a ballot measure that could have paved the way for a near total ban on. abortion, a position even many Republican voters don’t support.. In addition, the very substantial attention the January 6 committee received during the months of its hearings, from June to September, may have had an impact on the polling picture, including reminding many Americans that, though he is no longer president, Donald Trump remains a standing menace.

In sum, there were arguably two highly publicized issues that may have upended more straightforward assessments of who is in charge, who has power and, perhaps most important in our an era of negative partisanship, who voters should be most afraid of.

But as election day has approached, it appears that those anomalous factors are less salient to voters’ expressed preferences. The apparent deterioration in Democrats’ position in recent weeks suggests that the “structural” approach, which is not swayed by the ups and downs of the short-term news cycle or by polling, has a certain wisdom to it.

In sum, I don’t expect a good night Tuesday night and, yes, I’m sick about the fact that the Party of Trump, Vance, Oz, Lake, MTG, Cruz, Walker and so on - every one of whom is a demagogue and a menace - appears to be paying such a small price for its collective venality, dishonesty and malevolence. I’ll have more to say about that in my next post.