More Polling Lessons

This is a continuation of a previous article. We're looking at the things that presidential polling analysis can teach those interested in survey research about their analysis. It's not always as simple as analyzing the results, nor are the results to be taken at face value, since presidential polling is rarely accurate enough to call a race as it's decided.

Before we went over two points that are important to remember: There are always unknown variables. Polling can introduce different biases. There are many other things to learn as well about the field of survey research from polling analysis.

Let's look at some of the others. There Are Way Too Many Bad Surveys Another issue with polling is that many of the companies that engage in polling do it wrong or poorly. Many of them.

It's hard to believe how many polling firms get results that are all over the map and nowhere near the actual number they should be. At FiveThirtyEight, Nate Silver gives surveys weights based on how accurate they appear to be, and the vast majority of pollsters receive a very poor rating. People do surveys wrong all the time.

Measuring the Individual Factors is Important When measuring outcomes for presidential races, nationwide polls are interesting, but the electorate is all that really matters. That's why looking at polls in each state is important as well, and often differs from the numbers in the national race. This matters with your research as well.

You can analyze how people view your company on a companywide level, but really it's each individual part that matters. Regress Outliers Finally, make sure that you don't start to assume a huge shift in the data is accurate. The reality is that outliers happen all the time in polling.

Even the most accurate pollsters occasionally get a result that makes very little sense. Polling, like surveys, is working within confidence levels, but even a survey at 95% confidence has a 5% chance of being way off the mark. Weird numbers may be inaccurate, especially if there are potential survey problems.

Polling and survey research are different. But there are many things you can learn about survey research from the way polls are now being collected and analyzed, and the more you know, the more valuable your research will be.