Presidential Polling Lessons

While the race to the White House is long since over, there is still a lot we can say about the election process, especially with regards to presidential polling. One of the best places for this type of analysis is at the FiveThirtyEight blog at the New York Times. The blog looks at the results of every published poll from every well-known (and less well-known) polling company on both a state and national level, and uses a complex algorithm to figure out the state of the election at the time, including probabilities of victory for each of the candidates.

The most important thing they emphasize, though, is that they're just probabilities. When one candidate has a 66% chance of winning and another has a 33% chance of winning, that doesn't mean the 66% candidate is going to win. The authors emphasize that the probabilities in their model indicate that based on the information available at the time, one candidate has X chance of winning, while the other has 100-X chance of winning.

There is a lot to learn from this blog – and presidential polling – with regards to survey research and analysis. Let's look at some of the most interesting issues as they relate to survey methodology. There Are Always Unknown Variables First and foremost, it's important to remember that there are always variables that are unknown.

That's why even if the polls show a clear leader, it's possible that the other candidate could win. For example, voter turnout may be different than voter preferences. Polling locations could affect voter apathy.

While these don't play a role specifically with survey research, it does show the same issue – that there are issues that you may not know about that affect the quality of your results. Different Polls Have Different Biases Another interesting thing that has been found in the polls is that several polls show a "lean" for reasons that are not quite clear.

For example, Rasmussen Reports has always showed a slight Republican lean, for reasons that not even Rasmussen is entirely clear on. Sometimes these things happen in survey research as well, which is why comparing your results to the results of someone else's survey isn't going to tell the whole story. We'll continue with other things we can learn from presidential polling in the next article.