Polls, Projections and Politics

Right From the Start

This is boring stuff. Boring to the point of making your eyes roll back in your head. Boring to the point that you might rather watch the Kardashians on television. But it’s information that, unless you are knee deep in politics, you probably have never come across. So sit down, get comfortable and let’s slog through this.

As the political season enters into high gear we are being inundated with polling information. I don’t think my wife and I have been to a social gathering in the last month that somebody doesn’t comment on a recent presidential poll. Unfortunately the comments generally reflect a lack of understanding of the value of polls, how they are conducted and what they really mean. Because the reference to polls will only increase between now and November 6, 2012, and the poll numbers will be used to try to influence you, it is probably important that we discuss the use and abuse of polling data. In this instance let’s confine our discussion to the “presidential polls” although the comments are equally applicable to all other political polls.

Unless you have been forced to endure an education in Portland Public Schools you will already know that polling is done by a sampling of a target group and, thereafter applying a statistical analysis to that sampling. The validity of a sample often times depends on the demographic information that is gathered and applied to the analysis.

Let’s begin with the target group for a presidential election. If the target group is all voting age people the results of the poll will be interesting but about as accurate as People Magazin identifying Michelle Obama as one of the Most Beautiful Women in the World (as if placing her ahead of Halle Berry wasn’t sufficient to question the accuracy.) The reason simply is that not everyone of voting age is registered to vote and, therefore, will not vote. Unfortunately a large number of the polls taken and broadcast over the summer use this sampling. It is a poll designed to bias not to inform.

A poll based on all registered voters while more accurate is still highly inaccurate because it reflects the views of a large number of people who are highly unlikely to vote. For instance, in Oregon, in 2008, eighty-five percent of registered voters actually cast votes while in 2010 only seventy-two percent actually voted – nearly 600,000 registered voters did not vote. While the views of those registered but not voting might be interesting, they are unimportant in projecting the likely outcome of the election. Going back fifty years indicates that the years in which presidential elections are held usually results in about ten percent more registered voters actually voting than in non-presidential years. Even the advent of “vote-by-mail” has not materially changed either the differential or the total turnout.

A poll based on likely voters (four of four and three of four in the previous four general elections) provides a higher degree of accuracy. Those who have voted routinely in previous general elections are more likely to continue that practice by voting in the next general election. And the key here is voting in the general elections as opposed to primary elections. If the poll doesn’t begin with this targeted group, it is unlikely to provide useful information for the outcome of the election.

But even at that, polls can be and are skewed based on the demographics of those most likely to vote. For instance, in the 2010 election there were 863,300 registered Democrats, 664,100 registered Republicans and 423,800 independents. That breaks down as 41.7 % Democrat, 32.1% Republican, 20.5% Independent, and 5.7% minor parties. But let us assume that only seventy-five percent of the Democrats voted, eighty-five percent of the Republicans, ninety percent of the Independents and 60 percent of the minor parties. If your sample included registered voters you would over sampled likely Democrat voters by three percentage points under sampled Republicans by almost two percentage points under sampled Independents by almost two and one-half percentage points and under sampled other parties by one and one-half percentage points. There have been a number of national polls favoring President Barack Obama cited recently that, upon examination, were shown to have over sampled Democrat voters by as much as eleven percent. The point here is that the accuracy of the poll is dependent on the accuracy of the demographic information as to who is most likely to vote.

To make matters worse, the national media routinely fail to disclose the demographic information necessary to allow us to make informed judgments as to the value of the results. The bias of the national print media, NBC, MSNBC, CNN and to a lesser extent ABC and CBS has been pretty well documented. On the other side of the coin the bias of Talk Radio and the opinion portion of Fox News is also well-known. The news has been democratized and so it is incumbent on all of us to ask the critical questions concerning the polls in order to test their validity.

And finally, the Constitution of the United States does not care about national polls because the president is not elected by a popular vote. Any number of Democrats still haven’t gotten over the fact that while Al Gore won the popular vote, George W. Bush one the electoral vote – that is the popular vote in sufficient states to capture a majority of the electoral votes. Whether, for instance, if Mr. Obama wins a plurality of votes in Oregon, or all of the votes in Oregon, he will still only get seven electoral votes from Oregon. Therefore, national polls, even of registered voters, even of registered voters likely to vote, are of little value when compared to the polls focused on registered voters likely to vote in the so-called swing states – the states where polling indicates that the presidential race remains within the range of the statistical error.

The campaign professionals now refer to these as the “battleground states” and include in their ranks, Nevada, Colorado, Iowa, Wisconsin, Florida, Ohio, Virginia, and New Hampshire. As the election proceeds it is entirely possible that Minnesota, Pennsylvania, and New Mexico could be added to that list. The remaining states are pretty much assumed to have locked in their presidential preference and are unlikely to change.

So while the “spin may stop” with Bill O’Reilly, it starts with those who tout polls without disclosing the requisite demographic information.