President Obama was recently inaugurated for his second term after a tough race, with pundits on the left saying the President would prevail and those on the right predicting a clear Romney victory. News viewers were no doubt on the edge of their seats waiting to see who would ultimately come out on top.
Unless they read Nate Silver’s blog, that is, and knew with 90.9% certainty that Obama would win.
But, how could one person successfully predict such a dynamic and complex outcome? Furthermore, how could he predict which candidate would win each of the fifty states with perfect accuracy? Maybe he just got lucky, right? Nope …
Silver runs what is considered a “poll aggregator,” a process that combines and weights the results of all polls to arrive at a conclusion, rather than standing on the result of a single poll. The problem with single polls is that they are subject to considerable error; you get a much more reliable result if you combine your observed poll data with all other prior poll data.
Silver uses Bayesian statistical methods to combine and weight polls. The methods, used as a way to update one’s beliefs based on evidence, were originally developed by a minister from the 1700s named Thomas Bayes.
What can these methods teach us in the Virtual Job Tryout and larger employee selection business? Turns out, a lot! At a conceptual level, these ideas help us to realize that each data point (e.g., the way a candidate answers an interview question) should be balanced with other competency evidence. In other words, we should not over-interpret something we observe, especially if it conflicts with other prior evidence. By taking into account historical data, we can ultimately make more accurate decisions in hiring, and also in life.
Stay tuned! In the next few weeks, we’ll be examining Bayesian ideas in more depth, and seeing how they can add accuracy to our decisions, in both employee selection and life in general.