By Bruce A. Love
When I was in grad school, a professor explained, “there are three types of liars: liars, damned liars, and statisticians.” It was a variation of a quote often attributed to Mark Twain. The odd thing was, the professor was not teaching a course on ethics, philosophy, or literature; she was teaching a statistics class. I quickly realized that Dr. Szabat was poking a little fun at her profession. One thing I learned in that course was that although numbers never lie, interpretation of the numbers, and the way in which they are collected, can be deceiving. That point was made clear this week as exit polls were erroneously interpreted to suggest that Senator Kerry would become the next President of the United States.
Businesses, politicians, and military strategists all use “intelligence” to help them determine the best course of action to achieve desired outcomes. Military commanders use information collected from a variety of sources to assess emerging situations. This enables them to effectively engage or avoid the enemy while minimizing risk to their own human and material assets. Politicians use telephone surveys, and focus groups to help them understand the concerns of citizens, and to assess the impact of points made during speeches and debates. By studying survey data and analyzing responses of focus groups, politicians can make adjustments to their campaigns and improve their chances for election.
Businesses also use surveys and focus groups. They also use questionnaires, mall-intercept interviews, and numerous other research methods, to understand the wants and needs of potential customers. All of these activities are considered functions of marketing. Marketing research is conducted to maximize profits and minimize the financial risk that would be greater if assumptions were made about markets. Companies do not want to introduce new products unless they are relatively sure they will receive consumer acceptance.
This week, we have heard a lot about the voter exit polls, and how they were flawed or used inappropriately. There is a lot of finger-pointing and accusations regarding who is at fault for the flawed projections that were based upon the exit polls. The major media outlets, which all use the same source for the information, blame the polling companies and bloggers (people who post information and opinions on Internet web logs) for the confusion. The research firms that conducted the exit polls claim that some members of the mainstream media jumped on early reports when they were specifically warned against drawing conclusions from such data. They have also suggested that some political operatives posing as pollsters leaked the early data to bloggers who were used as conduits for political purposes.
Having watched some of the blogs on Election Day, I can honestly say that there were some very interesting exit polls being posted. This turned into an ecstatic party atmosphere on liberal blog sites, and as word propagated to conservative sites, it triggered panic and depression. It was truly remarkable to witness the hysteria that results from not understanding the nature and limitations of polling and statistics. Given the rocky ride on the stock market that day, I suspect traders were listening to bloggers reacting to exit polls.
There are several other plausible explanations for the flawed exit polls. Some have suggested that the morning sampling of voters resulted in a disproportionate number of polled women – a group that presumably leaned toward Democratic candidates. My theory is that a statistically significant number of conservatives (Republicans) prefer to keep their opinions private and are therefore not represented accurately in surveys and polls.
The desire to satisfy the insatiable appetite of Americans for fast and accurate information has driven pollsters and news services to reach beyond the limits of statistical modeling in extremely close elections. With such a tight elections, it is understandable that pollsters were not 100% correct. To get a more accurate picture would require a much bigger sample of voters. In the end, when the votes were counted, that is exactly what we got. The only numbers that mattered were not statistics, but the voice of individual voters.
