I do not like to do a lot of tweet myth-busting a lot for this newsletter, but something this week warrants it. Will Stancil, an urban policy researcher who posts a lot about the polls, published the following Tweet on Monday, August 29. In it, he claims that Joe Biden’s approval rating has turned around not because of inflation flattening out, gas prices falling, or the Dobbs decision, but because Democrats announced their deal for the Inflation Reduction Act on July 27th.
I admire the effort at trend-fitting here (it does look like something happened on July 27, after all) but there are a few reasons why I think this explanation (and the conclusions in his thread) are wrong. Before I get there, though, let me explain that I am not writing this post just to #dunk on Will, or otherwise attack people who like the IRA (it’s a good bill!). Instead, I think it is important that we properly understand why things changed in the past so we can make better forecasts for the future.
The past
So here is my argument against Stancil’s tweet.
First, consider that the US Senate did not pass the IRA until August 7th, 11 days after the inflection point Stancil points out. While some Biden-skeptical Democrats may have been excited by the news of its potential passage beforehand, it wasn’t an actual accomplishment until nearly two weeks after the slope of the trend Biden’s approval ratings turned positive.
And besides, people are not so attuned to the news that we would expect shifts in their opinions within a few days, and especially not on the very day of an event. Nate Silver thinks convention bounces, for example, take about a week to reach their peak — and that’s if you’re dealing with a highly-watched campaign event. Effects for the mere announcement of a deal on a bill in one chamber of Congress surely takes longer for the public to price in.
Now consider alternative explanations for Biden’s bounce. We can do so by looking first at evolutions in the average of generic ballot polls. They began shifting much earlier, bottoming out for Democrats at a -2.6 percentage point margin on June 16th and rising to -1.7 a month later (a week before Joe Biden’s approval-ratings trend reversed), then +0.5 by August 16th.
This suggests that other causes may have moved the political environment. The Supreme Court’s Dobbs decision overturning its past precedent on abortion rights is the likely culprit. How can we know that empirically? Well, for one, polls since the decision have shown an increase in enthusiasm for voting in November among Democrats, women, and people who say abortion is important to them. Results from special elections where Democrats have campaigned on the issue have also been bluer than expectations.
Then, there is the economy. Presidents usually benefit from growth (and are punished for inflation and recession), warranting an investigation of the economic data to identify any changes in recent trends that may be helping Biden. Expectations for inflation over the next year began to fall back to earth in late June.
At the same time, inflation in the price of a gallon of regular unleaded gasoline also peaked in late June — June 20th, to be precise:
And since changes in gas prices are inversely linked to changes in presidential approval, we should expect that to benefit Biden, too.
The future
As I mentioned, it is crucial to get our study of the past correct so we can better predict the future. Predictions models are almost all trained on historical data, after all.
But it is also important to understand the past so we can make better decisions about how to shape the future.
Take the implication of Stancil’s thread. He says Joe Biden has gotten more popular because he has passed a big, progressive bill that will deliver left-leaning priorities and cut inflation. Conditioning on the White House believing that, the natural conclusion of the West Wing would be to push for even more liberal laws before the midterms. Under this model so-called “deliverism” would rescue unpopular parties from assured defeat in November (at least in years not divisible by four).
In reality, we see that this is not a winning prescription! There is a relatively strong negative correlation between the number of net liberal laws changed by a Congress and the change in the Democratic Party’s share of the two-party vote in the subsequent mid-term election. This chart from TheEconomistillustrates the dynamic:
This stands in striking opposition to the normal political science explanation of why parties rise and fall: voters punish them when economic conditions are bad and reward them when conditions are good (all else being equal). If you look at the recent correlation between inflation expectations and gas prices and Biden’s approval rating, then you come away with the conclusion that beating inflation and making goods cheaper will buy a lot of votes in November. And, indeed, polls show that these issues are very important to voters today.
That gives Democrats and the White House a set of actions that are more empirically defensible, so long as their goal is to actually win in November.
And as for Stancil’s tweet, it is also a reminder that the simplest explanations are not always the best, Occam’s razor be damned.
"It's the economy, Stupid!" - James Carville to Bill Clinton.