In Defense of Pollsters

Pollsters have been blamed for predicting wrongly that Clinton will win the presidential elections. I find these accusations unfair and, as a marketing researcher and data cruncher, I would like to lend a helping hand to my fellow political pollsters.

I averaged the results from all general election polls monitored by Real Clear Politics in the month of November. These polls were conducted between November 1 and November 8, 2016. The unweighted average Clinton advantage from these 74 polls was 1.95%. As of November 23, 2016 Clinton has received 62,829,832 votes versus 61,488,190 for Trump. This is 50.5% vs. 49.5% or 1% advantage for Clinton. Therefore, the poll average overestimated Clinton's advantage by less than 1%. Given that the margin of error for most of these polls is at least 3%, you can say that the pollsters actually got it right. I know that this will contradict all you have read in the press but this is the story as told by the numbers. The numbers don't lie: the pollsters were not that wrong.

The ones who got it totally wrong were not the pollsters but the forecasters. When the general election polls predicted less than 2% advantage for Clinton and the state polls in key states like Pennsylvania, Ohio, and Florida started to lean towards Trump, forecasters kept predicting a Clinton victory with high degree of confidence. On November 8 Nate Silver predicted that the probability of Clinton winning the presidency was 71.4%. How you get from 1.9% advantage in the polls to 71.4% probability of winning is a bit of a mystery. I believe that this grave error was a result of too much data crunching. If you torture the data long enough, it will tell you any story you want.

Finally, I would use this post to pat myself on the back. I predicted the election accurately and I have a proof to show. As a member of the forecasting Good Judgment Project, I have a record of my forecasts that you can see if you follow the link below. On September 7, 2016 I gave Trump 55% chance of winning. The main reason I gave Trump an advantage is because I used findings from some qualitative research to complement the data from the surveys. Now, in light of the election results, many researchers talk about not relying on surveys alone to make predictions and complementing quantitative research with qualitative studies for forecasting purposes. I have always believed that this is a very productive approach.

https://www.gjopen.com/memberships/28315/forecast_history

Happy Thanksgiving!