NYT/Siena Recent Poll May Have Underestimated Black Vote in Michigan, Pennsylvania

On  Monday, Nov. 5th, the New York Times released an extensive poll of 6 swing states won by President Trump in 2016, matching 3 major Democratic contenders against Trump, Joe Biden, Elizabeth Warren, and Bernie Sanders. The states are Arizona, Michigan, Pennsylvania, North Carolina, Wisconsin, and Iowa. The results show that President Trump remains intensely competitive in these states against any of these contenders, meaning he could once again lose the national popular vote and win the Presidency by winning a majority of these states through the Electoral College. The Democrats must win three of these states that Trump won in 2016 to win the Electoral College, assuming that the results of all the other States stays the same.

Nonetheless, my own look at these polls leads me to believe that the Times/Siena Polls may have underestimated the Black vote in Michigan and Pennsylvania sufficiently to improve the chances that the Democratic nominee will win those two States.

I am not disputing the overall takeaway of the Times/Siena Poll, that the election in these 6 states  will be close, and close in Michigan and Pennsylvania. I will explain my analysis after this outline of the national polls and the NYT/Siena poll results.

Here are the matchups in the six states from NYT/Siena Poll:

Recent national polls stand in sharp contrast to the swing state polls. Here are the matchups with the same three contenders in the major national polls:

The 3 Democratic contenders are running strongly against Trump in these polls. Biden has margins of 9-17 points in 4 polls. Warren has margins of 5-15 points in 4 polls, and Sanders has margins of 7-14 points in 3 polls.

Can the national polls and the swing state polls both be right? Well, of course. We saw what happened in 2016. But a detailed look at the swing state poll in Michigan and Pennsylvania suggests the Democratic candidate may run stronger in those two states.

The reason that the swing state polls in Michigan and Pennsylvania may underestimate the Black vote is that the proportion of the Black vote in these states was reported higher in the national Exit Polls in the 2016 Presidential Election than in the recent swing state poll. National Presidential Exit Polls involve a consortium of all the major media, who hire a firm called Edison Research to interview about 27,000 people in a range of states chosen for their importance, either because of size or how competitive they are. The Exit Polls help each of the major news outlets call the states and predict the winner of the election.

Here are the results of the NYT/Siena Poll in Michigan, and the ethnic composition of the electorate in the sample. I used the results for Trump vs. Biden to perform the comparison; you can see the results for Sanders and Warren earlier in this article. The size of the sample was 501 registered Michigan voters. Trump beats Biden here by 1 point.

Here are the Michigan 2016 Exit Polls, which included 2812 respondents.

In 2016, the Michigan Exit Poll reported that Blacks were 15% of the electorate and 92% voted for Hilary Clinton. By contrast, the NYT/Siena Poll reports Blacks as 12% of the electorate, and they would vote for Joe Biden 74-15% (with 5% of this group saying they won’t vote). Since the size of the sample in NYT/Siena was 501, the number of Black respondents at 12% would be 60 people, an extremely tiny number from which to extrapolate results for the entire population.

If the 2016 Election poll numbers for Michigan for Blacks were used to substitute for the result in the NYT/Siena Michigan Poll, the Democrats would win 13.8-1.2 among the Black vote, for a 12.6% margin, instead of a 9.5-2.0 win in the swing state poll, a margin of seven points. That would be a gain of 5.6 points, meaning Biden would lead by 4.6, not lose by 1. Sanders would lead by 6, and Warren would be in a dead heat. The U.S. Census reports Michigan 14.1 % Black in 2018.

Here’s the NYT/Siena poll results for Pennsylvania and its electorate composition.

Biden leads Trump among the sample of 661 registered voters 47-44, and the poll reports Blacks as 8% of the electorate ( with 7% of Blacks saying they won’t vote), meaning the Black vote is about 7 1/2% of the electorate. 91% of Blacks say they will vote for Biden. The sample of Blacks, 8% of 661 voters, is small.

Here are the 2016 Pennsylvania Exit Polls:

In the Pennsylvania Exit Poll, the Black vote is 10% of the electorate, and 92% report they voted for Clinton. So there is no difference in the swing state poll on the percentage voting Democratic, just the difference between 10% of the electorate in the Exit Poll vs. 7 1/2 % of the Electorate in the NYT/Siena Poll. But an extra 2 1/2 points for the Democratic candidate would have brought Biden’s lead to 5 1/2 points, not 3. Sanders’ lead would be 3 1/2 points, not 1, and Warren’s lead 2 1/2. Sanders and Warren would still be in the margin of error in the poll. But the Democrats become much more competitive using the 2016 Exit Polls. The U.S. Census reports Pennsylvania was 12.0% Black in 2018.

The Exit Polls are not perfect, and the main point of the swing state poll remains: Trump is competitive in these states, and even if my analysis is correct for Michigan and Pennsylvania, he is still competitive there, and all-out efforts are required.