NEWSWIRE: 11/9/2020
- As returns came in for the 2020 race, pundits everywhere began asking: Is political polling dead? Pre-election day polls showed Democrats rolling easily into the White House, regaining the Senate, and expanding their majority in the House. (The New York Times)
- NH: Once again, wolves are circling around the pollsters. An Atlantic writer calls the 2020 election a "catastrophe" and "disaster" for the polling industry. "Polling seems to be irrevocably broken," says another columnist. "The polling industry is a wreck, and should be blown up," says another.
- So how bad was it, really?
- That depends, I suppose, on which election you're looking at and how accurate you expect the polls to be. The leading pollsters keep telling us that the margin of error for the national vote is 2 to 3 percent. By that standard, it is likely that Biden's final presidential vote margin will fall within that margin.
- The RCP national average on election eve was +7.2% for Biden. Currently, Biden's vote margin stands at +3.0% (50.6% to 47.6%). But by the time the final count is in, that margin is expected to grow to around +4.5%. So that's within a 3-point margin--albeit barely.
- Why does this margin keep growing? Simple. There are untold millions of heavily Democratic mail-in ballots yet to be counted, especially in giant blue states like California and New York. (The futures market give better than 50-50 odds that the final margin will be over +4.5%, so if you're certain Biden's margin will be smaller, you can go ahead and bet your conviction).
- A popular vote margin of 4+ percentage points, BTW, is not small. It is the largest since Obama in 2008, and before that, Bill Clinton in 1996. It is the largest vote margin running against an incumbent since Bill Clinton beat George Bush Senior in 1992, and before that Ronald Reagan beat Jimmy Carter in 1980.
- But let's get back to Biden's late-count rally. That brings us to the first reason so many Americans feel betrayed by the polls. Due to the pandemic and to the weird, politically-inspired layering of votes--day-of votes counted first and most early votes counted later--there was an initial surge for Trump followed by a long and late (and still ongoing) surge for Biden. Everyone anticipated this would happen in most states: an early "red mirage" followed by a "blue shift." Indeed, the White House clearly intended the vote to unfold this way--so that the President could rally supporters against the "fraud" of Democrats. Still, everyone was surprised.
- What about the polls in the individual swing states? Overall, they too underestimated Trump. And, as we might expect, their margin of error was more variable. But again, take a look at these RCP election-eve poll averages.
- Georgia: 1.0% for Trump. MISS
- Ohio: 1.0% for Trump. HIT
- North Carolina: 0.2% for Trump. HIT
- Wisconsin: 0.7% for Biden. HIT
- Florida: 0.9% for Biden. MISS
- Arizona: 0.9% for Biden. HIT
- Pennsylvania: 1.2% for Biden. HIT
- Michigan: 2.7% for Biden. HIT
- All 8 of these swing state poll averages (except maybe Michigan) were small, essentially pointing to toss-up elections. And the polls got 6 right and 2 wrong. Overall, Biden had the edge in these polls, and Biden ended up winning 6 of them. So again: Nothing shocking here. It's true that Trump enjoyed a late polling surge in several of them, especially in the last 48 hours in Pennsylvania and Michigan. But that's not a pollster error. Rather, it's why pro-Biden prognosticators like myself (see "What Trump Needs to Win") kept pointing to the size of Biden's margin. Trump had a similar late surge in 2016, but this time he was too far behind for it to tip the balance.
- In the Senate elections, the pollsters' critics have a much better case. Here, no question about it, the pollsters sizably and systematically underestimated Republican strength. Yes, the Democrats netted a gain of at least one seat. But they came in way below expectations. They lost brutally in Maine, which we thought "likely" to switch to the Democrats. And they lost in all four toss-up elections (NC, IA, MT, and GA). What's more, they often lost by wide margins. In Iowa, for example, we thought Greenfield (D) was an even shot to beat Ernst (R). Instead, Ernst won by 6.6 points.
- Ditto the for the House elections. The Democrats were widely expected to gain seats, maybe 5 to 10 net, and those expectations were rising late in the season. Instead, the Democrats flipped 3, the Republicans flipped 8, and the GOP may pick up a surprising 5 seats net. The Democrats similarly underperformed in state elections. They picked up majorities in 2 more state senates and 4 state legislatures, but lost one governor. They were certainly hoping for better in advance of next year's redistricting.
- As I pointed out last week, maybe the most unexpected lesson of the 2020 election is that the traditional Republican party brand may be able to beat the populist Trump brand after all.
- So let's shift gears. Let's now ask why the pollsters once again significantly overstated Democratic strength. Even if they did get the ultimate presidential decision right, why couldn't they improve their performance over 2016?
- The answer isn't yet entirely clear, but let me offer some suggestions.
- First, the pandemic introduced a clear asymmetry in voting--since the Dem leadership wanted to encourage "distance" campaigning and voting, while Trump defiantly made in-person politics almost a signal of partisan loyalty. The Democrats reached out to supporters largely on-line, and Biden campaigned largely by Zoom. But the GOP organized house-to-house visits, and Trump held nonstop crowded rallies in swing states. Democrats called this foolish. But Trump supporters called it bravery, and they turned out in higher-than-expected numbers. For getting people to make a commitment, in person beats on line every time.
- Second, many moderate Democrats in the House are complaining that Biden did not sufficiently distance the party from its social and cultural left--especially on issues of race ("systemic racism") and crime ("defund the police"). There's no question such issues motivated voters. According to A.P. VoteCast/NORC polling, protests over police violence were an "important" factor for 67% of voters and "the single most important factor" for 19%. These voters split roughly 50-50 for Trump or Biden, but given the distribution it seems clear that more of these were moderate Dems who switched to the GOP than the other way around. This didn't hurt Biden so much in the choice against Trump, but it may have done greater damage down ticket.
- What's more, however the Democratic Party branded itself this time around, it didn't much help them with their demographic home base--the young, the educated, and minorities. Contrary to expectations, Democrats didn't do any better among <30 or among college or post-grad voters than they did in 2016 or 2012.
- As for minorities, this was a big fail. Despite the Democrats' extravagant outreach, race-baiting Trump did just as well among blacks and Asians in 2020 as he did in 2016--and better than Romney in 2012. With Hispanics, he did better than ever (35% in 2020 versus 29% in 2016), especially among Cubans and South Americans and in red-zone states like Texas. (See "Trump's Favorability Rises Among Hispanics.")
- A bellwether of Democratic miscommunication on race was signaled by California's Proposition 16, an attempt to revise the state constitution to allow discrimination in hiring and education to favor disadvantaged races/ethnicities. (In 1996, under GOP Governor Pete Wilson, the constitution was amended to bar any such discrimination.) Democratic progressives in Sacramento simply assumed that their overwhelmingly blue, majority-minority electorate would have no problem saying yes to this revision. They vastly outspend their opponents. But Prop 16 failed: 44% yes, 56% no. Whites and Asians voted heavily against it--as did an estimated one-third of Latinos.
- Finally, it's time we revisit the "shy voter" theory for why pollsters may have gotten it wrong. There may be some merit to it. Arie Kapteyn, who heads a team of USC voters, said before the election that if you ask people a "social circle" question ("who do you think your friends and neighbors will vote for?"), you get a 5 to 6 point lead for Biden--a lot smaller than asking people themselves who they will vote for. This, as it turned out, was a more accurate estimate.
- True, most researchers have failed to validate the "shy Trump voter" hypothesis. But that's because it needs to be reframed as a "shy Republican voter" hypothesis, meaning that Republican voters in general--especially moderate and educated Republicans--may be even shyer than Trump voters about disclosing their voting preferences.
- This may explain why many of worst pollster errors in 2020 came with down-ticket Senate or House races involving moderate Republicans. Consider Senator Susan Collins of Maine, about as mainstream and respectable as they come. Polls underestimated her support by at least 10 points. Shy voter theorists think it's all about the identity of candidate. In fact, it may be more about the identity of the voter.
- Is there any evidence supporting my idea? Well, that's grist for another NewsWire. But I will mention two studies. One, by Morning Consult, confirms that the biggest gap between in-person and on-line responses by Trump voters (a common way to estimate shyness) is "driven by adults with higher levels of education." Another is a series of surveys by the Cato Institute. They show, to begin with, that conservatives and moderates are much more likely than liberals to agree that "the political climate these days prevents me from saying things I believe because others might find it offensive." They also show that highly educated adults are much more worried than adults with less education about possible damage to their careers "if your political opinions became known."
- So to recap. Yes, the pollsters got it wrong again. But their error on the presidential election was within the allowable margin. When all votes are counted, Biden will gain a substantial victory, though it will seem a lot less impressive in the electoral college than in the popular vote. In the Senate, House, and state races, on the other hand, the pollsters' overestimate of Democratic strength was more substantial. And it is already triggering heated debate among Democratic leaders over what exactly they got wrong.
