The presidential election polls were significantly off in 2020 with President Trump garnering many more votes than pollsters anticipated. Data analyst David Shor argued that the polls were wrong because Democratic voters became more politically engaged than Republicans during the lockdowns and answered more surveys. Rather than a shy Trump supporter phenomenon, there was a loquacious Biden supporter phenomenon. This created a pre‐election blue mirage whereby Democratic support appeared to be much greater than it really was.
But that wasn’t the only polling shock. Exit polls showed a relatively big shift toward Republicans among Hispanic voters. According to exit polls, Trump garnered 32 percent to 35 percent of the Hispanic vote in 2020, up from 2016. Trump’s gains appeared to be especially big in Miami‐Dade County Florida and in some small counties in Texas along the border.
However, the same polling problems that led to bad pre‐election polls also afflict the exit polls. About 100 million voters voted early, 65 million by mail and 35 million in person. Fewer than 50 million voted in person on election day. The in‐person election day voters were more likely to be Republicans while those who voted early were more Democratic due to different party messaging. Thus, in‐person election day voters who were polled skew Republican regardless of their other demographic or economic characteristics. Exit pollsters do try to call early voters, but that doesn’t give me much confidence given how Democrats and Republicans had different levels of survey engagement in pre‐election surveys. Wouldn’t they also have different levels of survey engagement after the election? Thus, I suspect that exit polling of those who voted early is as unreliable as the pre‐election polls if not worse. Major differences in the two major exit polls only reinforce my suspicions.
There’s some truth hidden in the exit polls. Trump did win some counties in the Rio Grande Valley of Texas that he could only have won with a significant share of the Hispanic vote and he probably got a higher share of the Hispanic vote in 2020 than he did in 2016. But, by that measure of success, the pre‐election polls were great too because they correctly predicted the outcome of the election. Apologies to Milton Friedman, but the predictive power of a model isn’t the only way to judge the explanatory power of a model … even in polling.
Of course, many people who are quick to rightly dismiss the pre‐election polls after November 3rd have embraced the similarly flawed exit polls. Some commentators and politicians were so excited about Trump’s improvement with Hispanic voters in the exit polls that they predicted a bright future for a populist and multi‐ethnic worker‐centric GOP. That’s an odd way for them to view an election where Trump got 1 Hispanic vote for every 1.8 Hispanic votes that Biden earned – at best. It’s especially odd to be happy about that because George W. Bush did better with Hispanics, winning 40 percent to 44 percent of their votes on a pro‐immigrant platform in 2004.
The untrustworthy 2020 exit polls bring up the issue of political assimilation for immigrants and their descendants. Hispanics, Asians, and immigrant voters are more Democratic than the rest of the country, but there are different theories that explain why that is true. One recently dominant theory is that they are Democratic because the Democratic Party is pro‐immigration relative to the Republican Party. This theory is consistent with the evidence that immigrants agree with Americans on just about every policy issue except immigration, so they pick the Democratic Party for that reason. It’s also consistent with the timing of the change in how Hispanic voters in California went from evenly supporting Republicans and Democrats in state‐level gubernatorial elections prior to 1994 to becoming overwhelming Democratic in that year. Hispanic and immigrant affiliation with the Democratic Party could be reinforced from the broad social science finding that most people choose their political party first based on identity and then later say they support the policies pushed by their party. As their identities change over time and generations, their party membership will converge with that of the American norm.
But that theory has limits too. There hasn’t been a similarly large shift in Arizona Hispanic voting shares when it enacted two harsh immigration enforcement laws in 2007 and 2010 nor has it happened nationwide in reaction to President Trump’s anti‐immigration administration. In California, Proposition 187 may have helped turn white voters more Democratic in 1994 while the same thing may have happened nationwide in 2020, but the 2020 exit polls hint at the possibility that Hispanics are more pro‐Trump now than in 2016. And Hispanics and other immigrant voters are not single‐issue voters focused on the issue of immigration and never have been.
The 2020 exit poll data raise more questions than they’ll ever be able to answer. But I have confidence in these conclusions: The safe bet is that Hispanics and immigrant voters would be more Republican if there was a pro‐immigration Republican like George W. Bush at the top of the ticket, Hispanics and immigrants will become more Republican as they continue to successfully assimilate, and that the lessons from Proposition 187 are probably a case study not generalizable outside of California due to peculiar historical circumstances in the Golden State. The exit poll data are being used by political pundits to produce partisan talking points about immigrant and Hispanic assimilation and voting patterns — don’t believe them.