Did the campaign to "Defund The Police" really cost Democrats votes in 2020?📊 March 28, 2021

A popular explanation for their relative losses, especially with Hispanics, might not be so cut-and-dry

Thanks as ever for reading my weekly data-driven newsletter on politics, polling, and democracy. I invite you to drop me a line or leave a comment below with your thoughts. Remember to click the ❤️ under the headline if you like what you’re reading. If you want more content, I publish subscribers-only blog posts twice a week. 


Did the campaign to "Defund The Police" really cost Democrats votes in 2020?

In America’s current era of hyper-competitive elections, every vote counts — a lot. The past two presidential contests have been decided by fewer than 100,000 votes out of 150 million. In terms of electoral handicapping, it makes sense for analysts and activists to put a premium on candidates who are more palatable to the median voter, relative to extremists. There, amongst the ideological middle of the masses, they can win more votes.

In 2020, this intuition combined with the ideological centrism of the press to increase the base rate of circulation for takes that the Democrats’ high-profile shift to the left on policing probably costs them votes. Eric Levitz’s interview with David Shor in New York Magazine last month was even tweeted out by Obama.

But I think the issue isn’t as black-and-white as pundits, who have largely adopted the thesis as conventional wisdom, make it seem. It could very well be the case that Defund The Police and BLM hurt Democrats among some voters. But public polling and county-level electoral results suggest competing thesis. At the best, the #Defund movement is one big explanation among many. But it might not even be that big.

First, consider the results of the last two presidential elections. The following map shows the swings toward and away from Democrats between 2016 and 2020 in each county. An area shaded red is one that got more Republican. Blue is more Democratic:

This map shows the county-level change in per-capita disposable income between 2017 and 2019, with green counties being ones that got richer and orange being counties that got poorer.

Now, I have talked a little about this on Twitter, but if you control for the standard demographic variables in each county — such as the share of whites, Hispanics, African Americans, and educated voters that live there, as well as the results of the 2016 election — the above percent growth in income is residually predictive of whether a county swung to the left to the right. This is especially true in counties with more Hispanic voters. I don’t usually drop regression coefficients in this newsletter, but in case you like that sort of thing, here you go:

To me, this makes a lot of intuitive sense! Hispanic voters are more split along ideological lines than Black voters are (see the next chart), and also less polarized by education than whites. It would stand to reason that they might be less moved by the ideological issues that dominated the 2016 campaign and more prone to reward the incumbent party when the economy was doing well, as it definitely was before the lockdowns imposed on Americans to help stem the circulation of covid-19.

On the other hand, we don’t have a whole lot of evidence of this pattern in past elections, and I can’t find much scholarship on racial differences in the effects of economic growth on incumbent vote shares. It would make sense that poorer groups would also react more to modest percentage increases in their income, but there is a difference between thing x makes intuitive sense and has modest evidence for its existence and thing x is true across data sources and model specifications. So I don’t think we have disproved any hypothesis about Defund The Poice here, but the county-level results and economic data are an enticing alternative.

Now, consider the polling. The table below (also from my Twitter) shows the 2016 and 2020 voting breakdowns of the 3 most populous racial groups and a lumped category for the rest, each further split up into thirds by voters’ self-described ideological affiliations. This is according to self-described voting behavior in YouGov’s online panel. Their data shows that moderate Hispanics swung 10 points to the right in 2020 relative to 2016, and that conservative Hispanics swung by 5 points.

Before I say anything, I want to issue a caveat: I doubt these numbers are exactly right. The actual county-level swing in highly Hispanic counties suggests the group shifted closer to 20 points (on margin) toward Republicans, which isn’t reconcilable with this polling. Other data show something similar, especially among conservative Hispanics.

The discrepancy could be due to online polls having problems reaching the “right” Hispanics. But the point is not the precision of the Hispanic estimate; it’s that Democrats made gains with white liberals and moderates, who make up about 41% of the electorate according to YouGov. If that’s true, it throws a lot of water on the theory that Democrats are hurting themselves by moving left on policing; considering the electorate as a whole, it may have even helped them.

Other online polling data also show (a) increasingly left-leaning views on the police and (b) stable voting behavior of Hispanics throughout the 2020 election cycle. See the trend in approval toward Black Lives matter over the past couple of years, per Civiqs…

…or the highly negative trend in favorability the police between summer 2019 and summer 2020 in Nationscape data, this courtesy of my friend Rob Griffin:

Rob has also posted this picture of Hispanic loyalty toward Trump increasing a lot from 2019 through 2020, with the Goerge Floyd protests and rise of the chant “Defund the Police” by Black activists not causing a significant change in the trend:

Now, maybe all this internet polling is wrong. It could be the case that Hispanics responding to online polls weren’t the ones that were put off by the #Defund movement, and so they don’t show up in the trend. But if that’s the case, I haven’t seen any alternative evidence for it. I’d really like to hear from a private pollster, like Shor, who talked to hard-to-reach Latino voters and found evidence that BLM turned them away from Biden in huge numbers.

One added wrinkle is that the Hispanic voters who immigrated from Venezuela and Cuba were much more likely to swing from Clinton to Trump, again according to Nationscape data:

To me, this speaks to larger patterns of ideological sorting than it does to a specific effect from the popularity of defunding the police. It is reasonable to expect that Donald Trump’s campaign against “socialism” hurt Democrats with groups that had the most experience in poor socialist countries.

In fact, the table of YouGov data above supports the idea that Hispanics simply became more ideological voters in 2020, “sorting” themselves into the Democratic and Republican camps by their ideologies. This makes sense; in 2016, motivated by racial issues and Donald Trump’s crusade against Mexican immigrants, Hispanics would have been more likely to see their place in the electorate delineated along racial lines than in 2020 when immigration was not so much of an issue.

According to my friend Alexander Agadjanian, this is very likely a huge part of the explanation. He analyzed the recently-released polling data from the 2020 Congressional Election Study and found that “The relationship between police attitudes and 2020 Trump vote switching is not unique to Hispanics” and “Police attitudes, compared to non-police attitudes, are not consistently stronger predictors of vote switching among Hispanics (or any racial group).” He posts the following two charts on his blog, which shows that attitudes toward policing — especially toward defunding the police — are not any more predictive of Clinton voters switching to Donald Trump in 2020 than, say, their positions on Abortion or general ideological position.

Now, Agadjanian’s analysis has the same asterisk next to it that Griffin’s and mine do; the data we’re relying on might miss hard-to-reach Hispanic voters. But if there really is data out there that shows Defund The Police is particularly damaging to Democrats, and especially among Hispanics, I really would like someone to show it to me.

Finally, one thing that (especially white) election analysts often miss when talking about police attitudes is how much the issue matters substantively. #Defund and BLM are not just tools to win votes, but real social activism campaigns that change minds and might improve conditions for Black and other marginalized Americans. How would the aftermath of the killings of George Floyd and Breonna Taylor (or even Michael Brown) look different without mass mobilization of voters? Even if Defund The Police did cost Democrats more votes than it earned them — and I haven’t seen any raw evidence suggesting it did — we can’t ignore that it may have helped a lot of people in different ways, too.


Posts for subscribers

What I Wrote

I wrote three pieces last week:

  • First, on how the ultra-racist white terrorism film The Birth of a Nation increased racial violence and KKK membership in the places it was shown.

  • Second, on the pandemic-fueled rise in hate crimes toward Asian Americans.

  • And third, on how Americans are opposed to restricting voting access but also surprisingly hesitant to expand it.

What I’m Reading

  • Mill Town: Reckoning with What Remains by Kerri Arsenault. I missed this when it came out last fall — something else was eating up my time… But I’m happy to be getting around to it finally.


That’s it for this week. Thanks so much for reading. If you have any feedback, please send it to me at this address — or just respond directly to this email. I’d love to hear from you.

If you want more content, you can sign up for subscribers-only posts below. I’ll send you one or two extra blog post-type articles each week. As a reminder, I have cheaper subscriptions for students.

In the meantime, follow me on Twitter for related musings.