November 14, 2020 3:15:28 pm
In this Nov. 7, 2020, file photo, Vice President-elect Kamala Harris holds hands with President-elect Joe Biden as they celebrate in Wilmington, Del. (AP Photo / Andrew Harnik, file)
Almost the day after the US election, pollsters and election forecasters readily admitted that their models and polls appeared to be wrong once again.
Even though the votes are still being counted and the data is still being scrutinized, American analysts have begun to reflect on the entire election forecasting industry, which predicted a much bigger victory for President-elect Joe Biden than what we saw. last week.
How do US statisticians create their electoral prediction models?
The models combine two types of numbers. The first are the “fundamentals” – the factors that shape voters’ choices. For example, how the state of the economy affects the chances of occupation or the fact that a party that wins three times in a row has only happened once in the last 70 years.
Andrew Gelman and Merlin Heidemanns of Columbia University, who created a poll aggregation model for The Economist, wrote: “Like most forecasts, our model … applies past patterns of voter behavior to new circumstances … “How often have previous candidates in similar positions left? to win? If those historical relationships break down, our forecast will fail. “
The researchers then look at the surveys (responses from representative samples). The model averages the surveys, weights each one according to the sample size, and then corrects for any biases. Nate Silver, a front man in the election forecasting community and editor of the established data outlet FiveThirtyEight, distinguishes himself specifically from a pollster, stating that his organization’s job is to understand how wrong polls could be to create probabilistic forecasts.
The final model combines the fundamentals with the survey averages. With these two types of information in place, the researchers run simulations a large number of times to find how many times a candidate receives more than 270 electoral votes. In 1,000 simulations, if Biden wins 500 times, he has a 50 percent chance of winning. As Election Day approaches, researchers are giving polls more weight on fundamentals.
What happened in 2016?
The mathematical oracles of the American elections had surely prophesied a victory for Hillary Clinton. Respected mainstream surveyors gave Clinton a four-point lead. He ended up leading by 2.1 percentage points in the popular vote. FiveThirtyEight faced pressure for predicting that Hillary Clinton had a 70% chance of winning the White House. Silver said people were taking the election poll results out of context.
The Economist wrote: “Trump’s unlikely win in 2016 left many quantitative election forecasters looking foolish. Princeton professor Sam Wang vowed to eat a bug if Trump, who he said had only a 1% chance of winning in November 2016, was even close to winning. (He chose a cricket) “. 📣 Click to follow Express Explained on Telegram
Autopsies from institutions like the Association of Public Opinion Research found that polls had underestimated the weight of voters without college degrees. The New York Times Upshot found that the lack of weights for educational status miscalculated Trump’s support by four points, matching the error. In many ways, it was a simple understatement of how many voters were white and did not have a college degree. In another mistake, late decision-makers ended up voting for Trump more than anticipated, and Trump’s overall voter turnout exceeded expectations.
Claiming that they had fixed the bugs, the statisticians claimed that they had learned the lessons of 2016.
What happened in 2020?
“There is no doubt that the polls were lost (again). But we won’t know how much until all the votes are counted (including the rejected ballot estimates). Then we will reevaluate. But I think it’s fair to say now that in many ways, including political polls, Trump is sui generis, ”Monmouth Poll editor Patrick Murray tweeted the day after the election.
Polls showed Biden leading at least eight percentage points down the stretch of the campaign season. Most likely it will end with a victory of four to five percentage points. Even both campaigns’ own private polls underestimated the Republican candidates.
At the state level, the predictions were even more wrong. RealClearPolitics and FiveThirtyEight forecast too much for Biden in every changing state except Arizona. Florida, in particular, was out of line; With nearly four points, Trump took the state that polls had predicted on average for Biden by three points. The New York Times and the Washington Post had Biden leading 17 and 11 points at Wisconsin. So far, it has a difference of one percentage point. The races for Congress were even worse, with Democrats surprised by their losses.
“Polls (especially at the district level) have rarely led us further astray and it’s going to take a long time to unravel,” Cook Political Report editor Dave Wasserman tweeted the day after the election.
What went wrong?
It’s too early to tell, but theories have started to leak out. One theory from Zeynep Tufekci is that there is not enough past data to accurately create the fundamentals because the factors in the choices change so substantially each time.
Other possible answers could be in the final participation data. Nate Cohn of the New York Times says that 2020 presented a new set of problems or that the problems of 2016 may never have been fixed. It leans toward the former, mainly because the weighting of education did not change the predictions. Polls found that white voters without a college degree would vote for Biden at higher rates than Clinton, but the final results showed that they did not change as predicted. Another error was in the calculations of high-level voters, who were predicted to vote for Biden by 23 points more than Trump. But in reality, older people didn’t vote for Biden at higher rates.
Cohn points out that these are not faults in estimating the size of the groups, but even more so about their attitudes. This is related to claims on the right of a “silent majority” who vote for Trump but hide their political beliefs. After the failures of 2016, the polls lost credibility and perhaps fewer Trump supporters were willing to answer the poll questions.
One obvious potential problem in the numbers was the pandemic. Polls prior to the pandemic (between October 2019 and March 2020) were more accurate than as the elections neared. One theory suggests that Democrats were more likely to be blocked during this time and were more likely to respond to polls than Republicans. Responses increased in that time, and the hot spots began to show more support for Biden. In other words, this was not a major support for Biden; this was an increase in the likelihood that a Biden supporter would respond.
Is it a matter of substance or presentation problems?
Some political experts say that the problem is the presentation of the numbers to a mass audience, rather than a problem of numbers. For example, if Biden is given a 65 percent chance of winning the election, that means he has almost a one in three chance of losing. However, most voters who hear a 65 percent probability imagine a high probability.
The contenders argue that political oracles have created margins for errors and warnings so large that they can say they were right regardless of the outcome, effectively rendering them useless. Silver has vigorously rebuked the narrative that the polls were wrong, writing that his organization had rightly predicted that Biden could survive a normal or even slightly larger poll error and still win. “Voters and the media need to recalibrate their expectations around polls, not necessarily because something has changed, but because those expectations demanded an unrealistic level of precision, while resisting the temptation to ‘drop all polls. ‘. … If you want to be certain about the election results, the polls won’t give you that, at least not most of the time. “
How are institutions and people responding?
Although analysts prefer to present their arguments in opposition to each other, the general reflection across the board seems to be relatively consistent: decrease the obsession with polls.
“Much of American democracy depends on being able to understand what our fellow citizens think. That has become a more challenging task as Americans sort into ideological bubbles … Public opinion polls were one of the last ways we had to understand what other Americans really believe. If the polls don’t work, we are flying blind, ”wrote David Graham of the Atlantic.
Some have problematized the entire numbers game, not just 2016 versus 2020 weights. Silver rose to fame predicting baseball games, “but unlike baseball… this game doesn’t always have a predictable set of rules. that all players abide by. There is a lot more noise in the signal that can interfere with an algorithm, ”said Slate policy editor Joshua Keating.
“We should take the money we spend on pollsters and we should dedicate ourselves to organizing on the ground. I understand Trump had people on the ground for a year in Florida. I would like to see us relying less on polls because it is becoming less and less perfect to get what we want, ”Congresswoman Pramila Jayapal said in a webinar the day after the election.
Similarly, news organizations are investing more in Internet trend analysis and local news coverage to make up for poll errors.
Tufekci said: “Instead of refreshing the page to update the predictions, people should have done the only thing that really affects the outcome: vote, donate and organize. As we discovered, everything else is within the margin of error. “
.
