# Math Geek Mom: Predictions

Polling and accuracy.

## By

June 9, 2016

In statistics, we often talk about the difference between “correlation” and “causation.” Just saying that two things are correlated does not mean that one causes the other to change. As one statistics text noted, saying that a rooster crows and then the sun rises does not mean that the rooster crowing caused the sun to rise. I found myself thinking of this recently when I learned that both of the major political parties now have presumptive nominees for President of the United States, and the realization hit me that the next five months will be ones in which many, especially those in Ohio, will be trying to predict the outcome of what will happen in our country on November 8th.

The most common way to predict the outcome of an election is to take a poll of likely voters. This is especially important in Ohio, since we are the ultimate “bell weather state.” This status comes from past elections in which we have been the deciding factor in determining who won, but also because Ohio demographically looks a lot like the rest of the United States. In a sense, we are a “representative sample” of the entire country, with the same proportion of residents who are of different age groups, married states, races and employment statuses as the rest of the country as a whole. As you can guess, pollsters love us, and therefore disturb our dinner hours on a regular basis.

While taking polls of potential voters if the most common way to predict the outcome of an election, this may not always yield accurate results. One reason for this may be that people who have the time and desire to answer such polls may or may not look like the people who are going to actually vote. For example, I always feel a responsibility to answer polls, mostly because I make a living studying data, and feel that I should do my part to make sure that people collecting such data are able to do their jobs. However, I may or may not look a lot like the other voters in my area; I may have different political opinions and some very different life experiences than do my neighbors.

So what are we to do in predicting the outcome of the upcoming election? Over the years I have become aware of other ways to predict the outcome of presidential election, some more orthodox than others. Many of them look for correlations that seem to persist over the years, and then go on to (possibly incorrectly) make the statement that such correlations are rooted in an actual cause and effect relationship.

My favorite approach to looking at the outcome of a presidential election is seeing it through the lens of the “political business cycle.” This is an idea that says that voters will react to “fundamentals” in where the direction that the economy is currently moving, and vote accordingly. An improving economy, it notes, is likely to favor an incumbent president or party.

A little more creative approach, loved by economists, involves “prediction markets.” These involve people betting on the outcomes of markets, using all of the information that they have available to them. The fact that they are placing money behind their assessments makes these “markets” more reliable predictors than polls, as it is no longer just a matter of giving someone on the other end of a phone your opinions; now money is involved, so people will be less likely to signal incorrect information. In recent elections, such markets did an excellent job of predicting the outcome of national races.

Perhaps the most unusual approach has been one noting a correlation between the winners of presidential elections and the winners of the World Series. This has led some to notice that there is also a correlation between the winner of presidential election and the winner of the NBA playoffs. According to what I have heard, there should be many “blue” people out there this week joining almost everyone in Cleveland, especially my daughter and my family, in holding out one last shred of hope and cheering “Go Cavs.”

+ -