Strength In Numbers Q&A for June 2025, part 1

Are right-wing pollsters biasing Trump approval averages? Should America switch to ranked-choice voting? And what's up with the Democratic Party brand?

Welcome to the inaugural Q&A post for Strength In Numbers! I've wanted to do a monthly mailbag for a while now, since a column like this lets us cover good questions that don’t merit a full article and keeps the newsletter a two-way conversation. SIN has recently reached the scale necessary to gather a critical mass of questions, so now seems like a good time to start this feature!

For this issue, I have selected a baker's dozen of questions readers submitted to my email inbox and the Strength In Numbers reader chat on Substack. If your question didn't get picked this time, that means either (a) I didn't have anything to say about it, (b) I’m saving it for part 2 of this Q&A (more on that below), or (c) I'm saving it for the July mailbag.

On the menu this month: Questions about recent special elections, what's going on with Trump's approval rating, and some thoughts on ranked-choice voting.

Before we get started, two programming notes: First, I have split this Q&A into two parts, because the questions were good and my responses went long — too long for this to fit in your email inbox, according to Substack. This gives me more room to address everyone’s good questions!

And second, this article will get paywalled about halfway down, since I promised this in part as a benefit to paying subscribers. If you learn something from this post (or otherwise just enjoy it), please join the Strength In Numbers community as a paying member. Your support helps us do lots of impactful data-driven journalism and makes this a sustainable career for me.

Let's go!

Send any questions you have to me for the next issue here! (I’ll remind you again closer to publication day.)

---

Elections

Gayle asks: How are Democrats overperforming or at least doing well in special elections when our national brand is in the toilet? Is this just anti-incumbency flipped?

First, some numbers for the uninitiated: According to an analysis of special election results by the folks at The Downballot, Democrats in special elections for state legislative and congressional seats are running about 16 percentage points ahead of Kamala Harris's vote margin in the 2024 presidential election. That is a significant shift, and about 6 points larger than the swing in specials from 2016 to 2018, before Democrats won the House popular vote by 7-8 percentage points.

Yet, the Democratic Party brand is still terrible. In YouGov's tracking, just 38% of U.S. adults have a favorable view of the Democrats, while 58% view it unfavorably. This is actually considerably lower than the party's rating ahead of the 2024 election, and worse than the rating for the GOP.

So what gives? How is the party doing so well with a bad rating? I talked a bit more about this in Saturday's interview with Paul Krugman and The Downballot's David Nir, but the answer boils down to two big factors:

• Differential turnout: Read my article on specials from May, then read the follow-up from two weeks ago. I find that a little over half of the nationwide increase in Democratic vote margin in these special elections can be attributed to Democratic voters showing up and Republican voters staying home. This is not surprising; it's what we should expect as educated voters become more Democratic-leaning, since educated voters vote more often.
  But notably, that leaves about half of the variance in these contests that is explained by other factors. That's stuff like fundraising, local party and candidate brands, and just plain old-fashioned campaigning. And on that note:

• Backlash against the incumbent: The other big factor solving this puzzle is that a party (when it is out of power) does not necessarily need to be popular if the incumbent is more unpopular. When conditions aren't good or a leader is underwater, you should think about voters as searching for mechanisms to express their displeasure, both within the system (elections) and outside of it (protests). There are only a certain number of levers to pull in politics to express that displeasure, so special elections draw a lot of attention.

Looking forward, we do not know how much the Democrats’ overperformance in special elections will translate to future electoral success. That’s because when the overarching question about politics turns away from "is the incumbent doing well?" to "which party do I support," voters are accessing a different part of their brains with different information and considerations for their behavior.

Jonathan shares this post and asks: Ranked choice voting has many positive qualities, but most places it is used do not apply a counting method that finds the Condorcet winner, which sometimes leads to undesirable results. What are your thoughts on the debate over counting methods?

Not the first mention of Condorcet at Strength In Numbers!

Jonathan raises a popular criticism of ranked-choice voting, an alternative counting method for ballots that is supposed to deliver majoritarian results more often. Specifically, here we are talking about instant-runoff voting, which successively reallocates votes for candidates in last place until one candidate has a majority of ballots remaining. Read more about how IRV works for a particularly notable race: the Democratic primary for mayor of New York City, happening tonight). And for more context on the question, a "Condorcet winner" of an election is the candidate that people would vote for in a hypothetical head-to-head matchup against all other opponents. It is the ideal, most majoritarian candidate.

Some more context: The criticism of IRV is that it can sometimes elect the "wrong" winner — that is, a candidate that is not preferred to all other alternatives. This happened in the 2009 election for mayor of Burlington, Vermont, when the candidate who won the head-to-head matchups against his opponents was eliminated in the third round of the tallying process before the more ideologically extreme Republican and Progressive candidates, who had higher support in rounds one and two. Here's a great visual walkthrough of what's going on.

This criticism is usually dismissed on the grounds that RCV with IRV gets us much closer to a world where 100% of elections deliver a Condorcet winner than the plurality-winner system used in most U.S. elections, which mostly does not. That is true, and why I generally support RCV.

However, there is a fix to the problems with IRV. You can use a different counting method called a bottom-two runoff, which takes the bottom two vote-getters and eliminates the one who would lose in a head-to-head matchup, which the algorithm decides based on voters' rankings. This is all nice and well, but I think it suffers from being significantly harder for voters to understand. And after all that is a key part of the process: If voters don’t understand how their votes are being counted, how can we expect them to support the democratic process?

I'm not an expert in IRV/RCV/etc, but here are my thoughts, since Jonathan asked: Recent studies have found no evidence for the other supposed benefits of RCV, which include things like ideological moderation and policy responsiveness. These problems have even caused perhaps the leading proponent of RCV reforms to all but abandon the movement entirely.

Here, we should acknowledge that other alternative voting methods exist and have a place in the conversation. I personally really like open-list proportional representation, for example, and wrote about it years ago. I would like to see a new version of the Fair Representation Act that uses OLPR.

But one thing to say is that we don't have to have the perfect solution first; we can take small steps toward the ideal method one step at a time. The most important thing is that we move away from the plurality-winner single-member system as fast as possible, and IRV takes us in the right direction. Given how undesirable the current system is, “don't let perfect be the enemy of good” is a persuasive argument.

Polling

Micah asks (on June 13): Trump got several really bad polls in the last few days which are well below his current average. But from a quick glance at some of the different pollsters it seems like these aren't actually all that much worse for him than their previous polls, many of which seem to have come out around his 100 day mark when he bottomed out. If you just look at individual pollsters' trends, does that indicate that his approval rating may not have changed as much over the past month as the averages suggest? How much of it was real movement, as opposed to compositional effects? (Presumably, you account for this in your average.)

Brandon asks a similar question, but in the opposite direction: What’s going on with the InsiderAdvantage polls? I know they’re a right-leaning pollster, but releasing Trump +10 net approval seems sort of ridiculous. Why are polling aggregators even including these results? They are major outliers and skew the averages upward.

These are two great questions about so-called "house effects" in polling data, or a tendency for results from certain pollsters to be much more favorable (or unfavorable) for Trump than the average poll. Go ahead and look at this table of polls in our average and see if you can spot some of the trends and culprits:

Two things stand out:

First, polls of all adults tend to show lower approval and higher disapproval numbers for Trump. That's to be expected; throughout Biden's presidency, all-adult polls also showed lower approval for him. We think this is due to a tendency both (1) for more apolitical Americans to say they "don't know" how they feel at higher rates, and (2) to view the incumbent party more negatively because of poor economic conditions or political strife, all else being equal.

The second thing that stands out is that pollsters typically aligned with Republican candidates or causes, or that put out a lot of GOP-friendly commentary, are significantly more positive for Trump than the average. Their residual is even higher than the one for all-adult surveys. Note the InsiderAdvantage poll on June 16 that gave Trump a +10 net approval rating, for example, or the Trafalgar Group's +9 rating on June 20. These surveys are between 15 and 20 points more positive for Trump than the average poll (on a net approval basis).

These are firms that have very strong GOP house effects over a long time period. So it’s plausible that whatever data-generating process they are using is producing a distorted picture of Trump's approval rating. That would significantly impact a simple polling average that doesn’t take them into account, such as the RealClearPolitics average (whose methodology is suspect for a variety of reasons).

Let me show you this graphically, with a few custom averages and charts.

---

The rest of this post is paywalled for members of Strength In Numbers. Paid subscriptions help pay the bills and grow our business, so we can engage in more ambitious projects and bring on freelancers and eventually staff members!

Become a member of our community for independent, data-driven journalism today, and get access to the rest of this article and other perks.

---