It didn’t take a genius to know that California would vote big against Donald Trump, and that’s just what every pre-election poll showed.
But the fact those surveys also got Democrat Joseph Biden’s landslide California margin very close to correct showed that pollsters learned something from their 2016 debacle.
Just not enough. In fact, the surveys had already made adjustments in time for the 2018 midterm election, successfully pinpointing the seven California congressional districts Democrats would flip away from Republicans in this state.
Here’s the lesson the polls learned: They could no longer settle for surveying stratified random samples of registered voters, but from now on would have to do the extra work of figuring out who was really certain to vote and who was not, and then pretty much exclude others from their surveys.
There was also the matter of “secret” or “shy” Republican voters who would go for Republican Trump in the privacy of the voting booth or their kitchen table, while telling pollsters they were either undecided or planning to cast ballots for Trump’s opponent.
It turns out many Republican websites and other Trump supporters were dead right when they loudly contended the number of those hidden Trump supporters was greater than even new polling techniques used in many major surveys detected.
Some surveys mixed computerized responses and answers to robocall questions with results from live telephone interviewers, trying to find out who was who.
The results indicate even those tactics were not nearly good enough.
Yes, the polls were correct in a few tightly contested states like Iowa and Arizona, but in Pennsylvania, Michigan and Wisconsin — the three states that figured, correctly, as it again turns out — to decide the election, they were off significantly, almost all picking Biden to win all three by anywhere from 2% to 5%.
Most major polls, including those touted by the New York Times and the Los Angeles Times, counted on a supposed factor cited by Sean Trende, the wonderfully named senior elections analyst at the RealClearPolitics poll-aggregating website.
Said Trende in a pre-election interview: “In Internet polling, you click on your keyboard and no one really knows how you answered. When you get a robo poll, it’s just an automated voice and you’re pressing buttons.” So what’s to be shy about? Apparently something.
With Election Night returns largely in and large numbers of mail ballots yet to be counted in key states, it was apparent the outcome would not be known for days or weeks.
Which means Democrats who doubted the (for them) positive poll prospects were right to be worried.
Many felt like football fans whose team held a lead after three quarters, but had a history of blowing big advantages.
Biden might yet sustain his early lead and the final results may show he actually padded it in the last pre-election weeks by campaigning repeatedly in battleground states from Wisconsin to Arizona, Nevada to Michigan.
But Trump demonstrated that hard work and myriad rallies in key states have power even when a candidate is down in the polls and many big donors have deserted.
Biden’s war chest rose to about $500 million in the pre-election weeks, while Trump’s fell below $300 million, one result of Trump’s early spending, but Trump’s hard work helped in states like Florida, where almost all polls gave Biden pre-election leads, but Trump won handily.
Another reason polls proved less reliable this year than in 2018’s mid-term election is that voters appeared more volatile than expected, many changing their minds moments before they voted. Four years ago, every poll showed a narrow election, with the lead fluctuating back and forth and the likely margin projected at 2% for Democrat Hillary Clinton. This year, Biden had poll-average leads of 5% or more since he essentially clinched the Democratic nomination on Super Tuesday in early March.
The bottom line: Polls were far from perfect this fall, about as unreliable as four years ago, especially in state-level polls outside the West.
Email Thomas Elias at [email protected].