Let Me Add A Fourth Theory

Mark Blumenthal and Ariel Edwards-Levy have an article up over at Huffpost Pollster wherein they discuss three theories for why polls underestimated the GOP. It’s worthy of reading.

You had to know a “however” was coming, did you not?

I have another theory to throw into the mix, and it is a theory I feel really good about believing is true. Without further ado, let me throw it out there.

The theory is that polls, at any point in time, are not measurements of what will happen, but rather are measurements of what the parties have to work with.

The parties spend a good amount of money on “get out the vote” (GOTV) activities. Some of these activities (voter contacts to spur interest, drives to register voters and/or request absentee ballots) start weeks before the election, and to some extent will get reflected in polls as they are ongoing. However, the late pushes to ensure the absentee ballots yet to be returned, and the Election Day efforts to get people to the polls, happen after most polls have been conducted.

It seems obvious to me that these operations must be capable of "moving the needle", or else the campaigns would not spend time, effort, and money on them.

Further, there is nothing that says that the parties will perform equally well in their respective efforts. In fact, it seems to me to be patently obvious that there will be times where, through skill of leadership and especially through advances in information technology, one party will improve from one election to another more than the other, that at times a party will develop an edge in this regard that can last a cycle or two, and that when that has happened it is inevitable that the other party will eventually catch up.

That dynamic, if my theory is correct, would play out such that in some years, we would see the Democrats outperform the polls, being better at converting their potential voters into actual voters at the end of the campaign than the likely voter screens in polls ‘assumed’ due to the methodology involved. In other years, the same would be true with Republicans. We might even see, for a few years in a row, the same party outperforming the polls, until such a time as the other “ups their game” - and if pollsters had adjusted their models and/or methodologies to try and rectify widespread misses, a party just drawing even would subsequently outperform the polls in a widespread fashion.

We might even see it look like it does in the chart Nate Silver presented here.

We also might see the effects be really big in some states, where field operations had changed more from election to election than in other states. As happened this year.

That the Democrats had developed an edge in micro-targeting and GOTV over several election cycles is broadly assumed to be true. Republicans spent a lot on ORCA, a software system designed to close that gap, but it failed to function on Election Day in 2012. This year, there were instead stories about how the GOP had developed a much improved system.

As Silver wrote in the article linked above, “The polls may be biased again in 2016; we just won’t know much about the direction of it until votes have been cast and counted.” I agree wholeheartedly. However, I suspect this has a lot to do with the fact that polls are measuring attitudes and opinions of people, and can never measure the relative efficiency of the respective ground games at the end. That plays out, to a large extent, after the polling is done. If all of the effects of the relative strengths of voter mobilization efforts will be captured in pre-election polling, then my theory is bunk. However, it seems clear to me, since those efforts continue until the balloting stops, and are actually quite different in mechanics at the very end, that pre-election polls cannot ever truly measure it.