In trying to figure out who is likely to win an election, it’s important to remember that one is trying to predict human behavior about something that is certain to be affected by events that have yet to occur. At least in the old days, it was easier for pollsters to be confident that they had reached a reasonably fair cross section of the electorate; now that is not so clear either, despite plenty of methodological tweaking. It is with this in mind that I marvel at those who seem to be supremely confident that they know what will happen on Nov. 8, or Dec. 6 if a runoff is needed in Georgia’s Senate race.

It helps a bit when one party is clearly enjoying a political tailwind. Just a few months ago, Republicans were enjoying such a boost. Voters were very unhappy with President Biden and the Democratic Congress for a variety of reasons, from their handling of the economy, to crime and immigration, to the execution of the withdrawal from Afghanistan. For a year, Biden had either the lowest or second-lowest job-approval ratings of any elected president at that point of his tenure in 75 years.

Those headwinds have declined some but have not gone away. Biden’s still-very-bad numbers and the economy remain enormous challenges for Democrats. But now, Republicans are encountering their own headwinds, a result of the reversal of Roe v. Wade and former President Trump’s high (and increasingly unflattering) profile. The narrowly focused spotlight that had highlighted real problems for Democrats has now widened out to become a floodlight, which illuminates and emphasizes the problems facing Republicans as well. Now neither party has the benefit of a real tailwind; each has serious political exposure on important issues. A midterm that normally would simply be a referendum on the president and governing party has now become more of a choice election.

Democrats now have a lead of 0.3 percentage points in the RealClearPolitics generic ballot and 1.3 points in the FiveThirtyEight average. That’s not enough when you account for the fact that Democrats need to win the national House vote by about 2 points to have a strong chance of winning 218 seats. Republican votes are simply more efficiently allocated, because they tend to live in smaller towns and cities, while Democrats are concentrated in densely packed urban areas. Biden won 70 congressional districts by 70 percentage points or more.

Here’s another factor working against overly confident prognosticators: “Toss-Up” HouseSenate, and gubernatorial races have a tendency to break in a common direction rather than evenly.

Two years ago everyone was whipsawed in the last six weeks. Things seemed to collapse on Republicans after then-President Trump’s hideously bad performance in the Sept. 29 debate. Down-ballot Republican candidates’ numbers plunged as well, triggering speculation of an impending “blue wave.” But after several weeks of conversation about the implications of such a massive victory for Democrats, a last-minute gust shifted in the opposite direction, turning the blue wave into a Dead Sea as independents decided to hedge their bets. In the House, every single race rated as a toss-up going into Election Day went Republican.

But this is not only a House or a 2020 phenomenon. Over the last 12 elections, an average of 77 percent of the Senate races rated in the toss-up column going into Election Day broke in the same direction. Democrats won eight out of nine toss-ups in 2006, Republicans eight out of nine in 2014. In 1998, Democrats won six out of seven toss-ups. Democrats won eight out of 12 in both 2008 and 2012, and seven out of nine in 2000. The trend is even a bit streaky: Toss-ups broke in favor of Republicans in the last four elections (2014-2020), but for Democrats in the previous four (2006-2012).

Margins are still close enough in too many races to be certain about the outcome. Just using the margins in the RealClearPolitics averages in Senate races, two Democratic seats have margins of less than 2 points—Sen. Raphael Warnock leads Republican Herschel Walker by a statistically insignificant three-tenths of a point in Georgia, and GOP challenger Adam Laxalt has a lead over Sen. Catherine Cortez Masto of 1.7 percentage points. Among Republican-held seats, Sen. Ron Johnson has a 1.5-point lead over Democrat Mandela Barnes. In North Carolina, Rep. Ted Budd has a 2-point lead over Democrat Cheri Beasley. In Ohio, Republican J.D. Vance’s 2.2-point lead over Rep. Tim Ryan is hardly any wider, and for that matter, Sen. Marco Rubio’s 2.8-point lead over Rep. Val Demings isn’t, either. Keeping in mind that these averages are of polls that could easily be a week or two or three old, in such a volatile year with dangerous crosscurrents, putting too much confidence in very small leads doesn’t seem terribly prudent to me.

The unprecedented situation of this midterm election and the volatility we have seen over the last few months demand a bit of humility employed in this year’s prognostication. It isn’t a bad thing any year, but this year in particular.

The article was originally published for the National Journal on September 26, 2022.

More from the Cook Political Report