Caveats about polling

Yesterday, I called attention to a new Quinnipiac poll of three battle ground states -- Florida, Ohio and Pennsylvania -- that showed a tight race between former Secretary of State Hillary Clinton and Donald Trump for the presidency. Last night, the poll garnered a great deal of analysis and commentary on the airwaves. And, Nate Silver, widely viewed as the best number-cruncher in the business, went into a Twitter rant about how useless such polls are this far out and with so little data. What to make of this?

In discussing polls, pundits almost always note that a poll registers attitudes "if the election were held today." This was, in part, Silver's objection: There is a campaign to be run between now and November and the shape of the race can change depending on how the candidates perform. The next big moment for both candidates will be their selection of a vice presidential candidate, a choice that will be viewed as their first "presidential" decision. I don't think these choices usually help or hurt very much, except at the margins, but elections are often won at the margins.

The bigger caveat about polling, one that pundits less likely note, is this: These polling results register attitudes of voters if the electorate looks like the polling sample. Demographics will be especially important this year because the Obama coalition of minorities, young voters and unmarried women does not necessarily transfer completely to Clinton, and because Trump believes, with some evidence, that he is bringing new, mostly white male, into the Republican ranks. The Quinnipiac pollsters apparently believe him: Their samples in the three swing states registered three to five percent more white voters in their poll than cast ballots in the 2012 election.

Why? No one knows for sure. Pollsters guard their decision making and algorithms for determining the demographics of a race. For example, look at this analysis from David Wasserman at the Cook Political Report, the gold standard of non-partisan political analysis. Lots of really useful information, but check out item number 5. There we read this analysis of the fact that Latino voters are not concentrated in many swing states and so their influence is diluted: "Nationally, Latinos were 10 percent of all voters in 2012. But Latinos averaged just 6.9 percent of voters in the 12 heavily contested swing states. Even if Romney had won a 10 percent higher share of Latinos in every single state, he would have won only one additional state: Florida." Only Florida? Ask Al Gore about that. Everything Wasserman says about the numbers is correct, but the significance of those numbers is always something to be debated.

Usually, people who will be voting this year are the same people who voted last time. The first thing to look for in any polling data is whether it measures registered voters or likely voters. Lots of registered voters have not actually voted in years. But, one candidate can catch fire and bring people to the polls who never voted before. This is the claim of Sen. Bernie Sanders but, unfortunately for his argument, the number of people voting in the Democratic primaries this year is well short of the number who voted in 2008. It is also Trump's argument, and I think all pundits have come to recognize that predications regarding the Donald are of limited value.

I do not believe that demographics is destiny, but it is a big part of destiny. For example, at Five Thirty Eight, the same David Wasserman notes that demographics have been tilting the electorate for Democrats for 30 years:

There's no question that recent demographic trends have aided Democrats enormously. In 1980, Ronald Reagan won 56 percent of all white voters and won election in a 44-state landslide. In 2012, GOP nominee Mitt Romney carried 59 percent of all white voters yet lost decisively. What happened? African-Americans, Latinos, Asians and other non-whites -- all overwhelmingly Democratic-leaning groups -- rose from 12 percent of voters in 1980 to 28 percent in 2012.

He goes on to note, however, that educational attainment has become an even more determinate factor in electoral patterns:

In both 2008 and 2012, Republicans' best group by far -- of the five we examined -- was white voters without college degrees. The GOP carried that group by 14 percentage points in 2008 and a whopping 26 points in 2012. However, these voters -- who skew older and more rural -- decline 3 percentage points every four years as a share of the overall electorate. In contrast, white degree-holders -- who still lean Republican but are much likelier to support Democrats than whites without a degree -- rise a percentage point every four years.

And, it is precisely this demographic, non-college whites that constitute the core of Trump's support. The ratio of non-college white to the rest of the electorate is also significantly higher in those large industrial states in the Midwest -- Michigan, Ohio and Pennsylvania -- that Trump argues he can flip from blue to red in November. The last time a Republican won Michigan or Pennsylvania was in 1988. No Republican has ever won the presidency without Ohio. And, the Latino population, one of the fastest growing demographics nationwide, remains relatively small in all three states. Conversely, all three states have a significant black population and it is anyone's guess whether or not Clinton can perform as well among African-American voters as President Obama did and, even more of an open question, how many African-American voters will turn out to vote in November compared to the previous two cycles.

I agree with Nate Silver that we need to wait for more polling data before we start drawing any conclusions about what will, and what won't happen in November. And, before the different demographic slices decide for whom they will vote, the campaigns decide which demographic they will appeal to. Will Trump try and cut into Clinton's lead with black voters? Will Clinton focus on winning the votes of moderate Republican women or make a play for the white working class voters who have backed Sanders? Voters do not only choose candidates. Campaigns choose which voters to target and go after.

One thing is clear. This year will be like none other.

[Michael Sean Winters is a Visiting Fellow at Catholic University's Institute for Policy Research and Catholic Studies.]