Are the pollsters all in the bag for Biden?
President Trump loves Fox News, except when their polls come out.
But it’s not just Fox News that has Trump and his supporters exercised; they think all pollsters are in for a rude surprise on Election Day 2020 — just like in 2016. That still leaves the question as to whether current polls are an accurate reflection of voter intention.
National polls from June have sported a wide range of results from a 3-point Trump deficit in the IBD/TIPP poll (45 percent to 42 percent) at the beginning of the month to a 14-point rout in both the most recent Siena/New York Times (50 percent to 36 percent) and CNN polls (55 percent to 41 percent). More recently, the gap has been widening with the average Biden advantage at nearly 9 points.
Not a single poll has Trump ahead; only two polls (IBD/TIPP and Hill/Harris) have Trump within 5 points.
But are those polls an accurate representation of the eventual November electorate?
The Census Bureau provides an estimate of the demographic makeup of the presidential electorate by race, age, and gender. From this data one can make a rough guess as to what voter demographics will look like in November.
In 2016, the gender split was 53.6 percent female to 46.4 percent male. The age breakdown was 38.2 percent for ages 18-44, 37.6 percent for ages 45-64 and 24.2 percent for 65 and older. Both age and gender have changed little from election to election.
The ethnic profile has been changing with the share of the white vote (73.3 percent in 2016) falling an average of 1.8 percent per election since 1996. Meanwhile, the Hispanic vote share (9.2 percent) has been rising by an average of 0.9 percent and the “other” category (5.6 percent) by 0.6 percent. The African American vote has fluctuated, reaching a high of 12.9 percent in 2012. Not surprisingly, the share crashed to 11.9 percent in 2016 — what you get when you go from a charismatic black candidate to a sour, old white woman.
Given current trends, we should roughly expect gender and age to change little and the ethnic profile to be (roughly) 71.3 percent white, 12.4 percent black, 10.1 percent Hispanic and 6.2 percent “other.”
Are polling samples biased?
Two prominent pollsters (YouGov and Siena College/NYT) make detailed crosstab data available, (the only way to check for biased sampling). If the demographics of each survey differ from the predicted profile of the electorate, one can re-calculate the surveys and determine any changes. In other words, if a survey includes a lower proportion of white voters, the final survey can simply be adjusted. (Note: The Siena demographics for age, ethnicity, and partisanship do not add up to 100 percent. The analysis below adjusts for that omission.)
YouGov gives Biden a 49 percent to 41 percent lead, close to the RealClearPolitics average, while Siena/New York Times gives Biden a larger 50 percent to 36 percent lead. In both cases there may be some structural bias — but not enough to turn Trump from a laggard to a leader.
A structurally biased poll against Trump would under-sample seniors, white voters and males while oversampling younger voters, minorities (particularly African Americans) and females. Both polls actually oversample men by about 1 percent, advantaging Trump — but by a barely perceptible degree.
YouGov under-samples 65-plus voters by 1.5 percent and oversamples young voters by about three-quarters of a percent. Adjusting those demographics to the above predicted profile results in a Biden lead of 48.7 percent to 40.7 percent — essentially within mathematical rounding. Siena also oversamples the youth vote — by 2.4 percent, but also oversamples seniors by just under one percent. Adjusting those demographics results in a 50.1 percent to 36.9 percent Biden lead.
Siena does have a problem with its ethnic mix, projecting just under 63 percent of the electorate to be white, while overestimating Hispanic voters by just over two percentage points and assigning “other” twice the expected proportion. Yet, adjusting to a more accurate profile barely moves the needle for Trump. Biden still leads 49.9 percent to 36.5 percent when white voters are added and “other” voters are reduced. The YouGov poll provides no good news for Trump, slightly oversampling white voters and under-sampling African American voters.
Trump vote: Underreported, but not a lot
Neither YouGov nor Siena show major demographic biases, although Siena’s ethnic breakdown is difficult to square with past elections. Shifting age and ethnic mix might narrow the gap for Trump about one percent, but not much more.
There is one significant problem within the Siena poll: independent voters break for Biden 47 percent to 29 percent, an 18-poiunt margin that which is out of line with the YouGov results (an 8-point margin: 43 percent to 35 percent favoring Biden) and Quinnipiac (a 3-point margin: 43 percent to 40 percent favoring Biden).
Siena and The New York Times may have failed to account for Trump voter suspicion of polling and the mainstream media and a possibly higher propensity to refuse to take their poll. This is not a problem for the Republican quota of participants as the pollster just has to keep dialing until the quota is reached (and GOP numbers for Trump remain high). The problem is with independents. If the refusal rate for independent Trump voters is high, then a poll would fill up its independent quota with more Biden and truly undecided voters.
If independents in the Siena/NYT poll are assigned the same Biden-Trump numbers as the YouGov poll, Biden’s advantage would shrink to 50 percent to 39 percent. If the same as Quinnipiac, Biden would lead 50 percent to 40.8 percent.
Bottom line: Trump is trailing
No matter how you cut it, Trump is behind. Some polls may be undercounting his strongest demographics, but not by enough to make a big difference.
Some Trump voters might be lying to pollsters, but the effect appears little, if any.
It is far more likely that there is a small reserve of people too embarrassed to admit they are voting for Trump. But likely only a percent or two.
When all is said and done, Trump’s complaints about the polls are worthless and fruitless. Unless he directs his energy toward addressing the country’s problems, his polling numbers will keep falling — and with it, his presidency.
Keith Naughton, Ph.D., co-founder of Silent Majority Strategies, is a public affairs consultant who specialized in Pennsylvania judicial elections. Follow him on Twitter @KNaughton711.