In my post from Friday, I suggested the following variables/categories of variables to explain electoral outcomes, in descending order of significance.

1. Partisanship
2. Prevailing national conditions
3. Candidate quality
4. Other factors, such as campaign spending

To reiterate: the single most useful variable in understanding an outcome of a contest is the known partisan lean of a given district. If I gave you a map with the historical averages of electoral outcomes in the fifty states, or of 435 House districts for a ten-year period, that information would be more useful in predicting the next election, for example, than a map showing you campaign expenditures.

One could make a reasonable prediction of the 2028 Electoral College map right now, without knowing the candidates or how much money will be spent, by using the history of past votes. I am not saying that one could nail such a map without some luck, but just that that bit of information would get a person quite far in getting the map right. I feel pretty confident, for example, in boldly predicting that California will go Democratic in the 2028 presidential election (and 2032!) and that Alabama will vote Republican.

Indeed, other variables on that list co-vary with partisanship. One reason, for example, that the candidate quality in a given contest is bad is because the partisan lean of a given district is so skewed that candidates of the minority party know they are going to lose; hence, those elections end up with poor-quality challengers. This is clearly not universally the case, but the basic pattern is clear.

Money, too, can co-vary with partisanship, insofar as lower-competition contests that attract poor-quality challengers also tend to attract less money. Indeed, competitiveness is a major driver of spending, which can be fueled by varible #1 or, as we are seeing this cycle, variable #2.

But the one thing that campaigns cannot control is the prevailing national conditions. And that leads me to the G. Elliott Morris piece that took me down this pathway to begin with, Poll: Low-information voters have turned hardest against Trump and the GOP.

This poll illustrates the power of the #2 variable. He returns to a poll I noted a while back, which looked at low-information voters, which he defines as follows:

In our October Strength In Numbers/Verasight poll, conducted Oct. 13–19 among 2,028 U.S. adults, we asked respondents two factual questions about politics: which party has the most members in the House, and which has the most members in the Senate. I classified anyone who got at least one of these questions wrong as “low-knowledge.” About 30% of adults miss at least one question, including 25% of 2024 Trump voters, 22% of Harris voters, and more than half of people who didn’t vote in 2024.

I would suggest dwelling on those numbers for a moment. Just shy of a third of adults don’t know which party controls both chambers of Congress. And a similar percentage of voters for both Trump and Harris fell into that camp. This is a stark reminder that a whole lot of people pay only passing attention to politics. It also underscores that a lot of arguments about messaging are pointless. And to beat a dead horse, probably the main reason a significant chunk of those 25% of Trump voters voted Trump and those 22% of Harris voters voted Harris was (you guess it!) partisanship.

To return to the model I am presenting here: most voters come to elections pre-set with a partisan preference, and here we see that a chunk is responding primarily to prevailing national conditions. Variables #1 and #2 explain the lion’s share of outcomes in the aggregate.

Back to Elliott.

First, a chart:

This is an almost mirror image of what the evidence showed for 2024:

Low-knowledge 2024 voters backed Trump by 8 points. Today they favor Democrats for the House by 12, a 19-point swing (using unrounded figures). Voters who got both questions right backed Harris by 1 point and now favor Democrats by 7, a 7-point swing.

He explains:

It is well documented in political science that the least-informed voters are the most likely to switch parties from one presidential election to the next. People with more political information have the partisan filters to resist news that cuts against their side, so their views stay put. They see politics through the lens of highly calibrated ideological labels and partisan filter bubbles, cognitive heuristics that are often unavailable to voters with less information about politics. But people with less political knowledge can have their views shaped by broader economic and political conditions.

This reinforces my point about variable #1 v. #2 (emphasis mine).

He concludes:

If the polls are right, in just ten days, Democrats will be swept into power in the House and may even capture the Senate. If they do control Congress, it will be due to gains they’ve made with these elastic, reactive, largely apolitical low-information voters. They should be careful not to misunderstand what this victory would stem from: frustration at the incumbent and a demand for change.

If Democrats can’t deliver in 2026 — and especially after 2028, when Democrats will have a shot to win the presidency and therefore be unconstrained by President Trump’s veto power — it’s likely these voters will just swing back to the right again.

Let me conclude by noting that I think a key reason that there are often disagreements/points of contention over these kinds of posts is this: I am here trying to explain how voters behave, not how I want them to behave. While I have normative assessments about how I might like them to behave, that is not what this analysis is about. The empirical evidence is strong: most voters are driven by partisan identity, which (as Eliott notes) shapes the way they process and prioritize information. The empirical evidence further demonstrates that the kind of voters who are more likely to resist partisan filters and be affected by prevailing conditions, like the economy, are voters who don’t know a lot about politics.

Such observations are continually being reinforced the more data we have. It may not be the way we would like people to behave, but reality is what it is. Indeed, one of the reasons (and certainly not the only one) that I think we would be better off with structural conditions that foster more parties is because the rigid dichotomy we observe here is not healthy and does not create openings for solving national problems since two-way politics is definitionally zero-sum. Zero-sum politics makes compromise, and therefore policy-making, incredibly difficult, especially in a system that gives the minority more power than the majority.

I note the above to try and help connect all of this to a broad conversation.

I will further note the following.

Do I think it is a failing that many people pay so little attention to politics that they don’t even know which parties control Congress? Yes, I do.

Do I think it is a failing that even a lot of people who can identify who controls both chambers still don’t understand how the chambers function? Yes, I do.

Do I think it is a failing that a lot of voters are willing to vote for terrible people because of the letters in their names? Yes, I do. (Although even there, it gets complicated. How awful a person would you, dear reader, be willing to vote for in the Senate this year in your state if you thought said awful person was the bulwark against Trump nominating a fourth Justice to SCOTUS?)

But here’s the deal on all of this. Regardless of my value-laden preferences, the empirical reality is, and continues to be, that a lot of people can’t name who controls Congress, don’t know how things work, and will vote for terrible people because of the letter by their name.

Again, that’s the science. These are confirmable, replicable phenomena that we can empirically study.

Now, as I frequently state, none of this absolves an individual of their moral choices.

All of this is why I am such a fan of the quip: democracy is the worst form of government, except for all the others that have been tried.

And it is why I keep arguing for reforms, large and small, to try and improve our democracy because any number of key flaws in the system are why we are where we are.

To return to structure, I will note the following chain of events again:

1. Trump only won a plurality of the primary voters in 2016, but the combination of a crowded field and the basic fact of our weak parties led to his nomination. 55.1% of GOP primary voters wanted someone else.
2. Trump lost the popular vote by almost 3 million votes, but won the White House via the EC.
3. This then placed him in the leadership of the Republican Party and allowed his faction to become the dominant one in the party (which would lead to his re-nominations in 2020 and 2024).
4. In 2024, the combination of variable #1 and #2 led to a narrow victory and a return to power, and the opportunity for an even more significant MAGA takeover of the party.

I would note that 1 through 3 are situations in which the structure of our system was more important than voters. And you can’t get to #4 without 1, 2, and 3. Plus #3 (and what happens in #4) is a direct result of presidentialism as an institutional feature–it makes the person who holds the presidency especially influential over their party.

All of the above is just a series of empirical observations.

If we return to the values question, yes, people like Mitch McConnell and Kevin McCarthy in particular should have behaved differently in early 2021 and thereafter. But the reality is that power-seekers gonna seek power. No political analyst is going to go broke assuming that politicians are less likely to be guided by what other people think is the moral thing to do when they can make choices that they think will maintain or enhance their power.

Hence the reason I argue for reform: we need to do everything we can, big and small, to increase the reliance of politicians on voters rather than on other factors, but also to break up the partisan duopoly so that voters and politicians have options that would reduce the zero-sum nature of most of our political interactions.