Pizza and Polling

Ever since word spread that the best place for pizza in town was a food truck, lunch time at Tony's Pizza Truck was complete chaos.

When Tony started his pizza business, he paid meticulous attention to every pie, checking every detail. It started with Tony hand picking the tomatoes for the sauce; carefully kneading the dough and paying personal attention to getting just the right mixture of cheeses. The end product of all this careful attention was an amazing slice of pizza.

As word spread about Tony's, large crowds started to show up on a daily basis. The lines that stretched down the block for the pizza truck were incredible. The demand for pizza was through the roof.

Although he was elated over the success of his business, Tony found he has to rush to get his pizzas out. Suddenly, the attention to the little details just wasn't the same because of the pressure he felt. No more handpicked tomatoes. He succumbed to using prepackaged dough and didn't always get quite the precise mixture of cheeses as he had before.

Not surprising, the quality suffered. Each slice wasn't quite as good as the original. Over time, Tony found his business suffered as well.

In the last week, more than 40 Presidential polls were released characterizing the state of the race at both the state and national levels. A seemingly insatiable demand for up to date polling results from the news media, the politicos, and the public at large, leads to a daily flurry of headlines, news reports, and updated averages. But, in the midst of this mad dash, sometimes it is easy to overlook the underlying metrics that drive different poll results. For example:

  • Likely vs Registered Voters: One of the underreported differences in polls after Labor Day is many organizations begin to focus on likely voters rather than registered voters. Polling organizations often relying on a combination of direct questions about how likely someone is to vote and turnout models based on previous elections to determine who is included in these poll results. This difference in who is included can change polling results based on whose supporters are estimated to be more "energized" and other demographics.

  • Margin of error: In the haste to report on latest polls, especially interesting results that go against the grain, many news organizations fail to report this key measure of statistical precision or relegate it to the fine print. But, failing to be aware of margin of error is misleading because without considering it, you don't know whether differences in the support of each candidate are meaningful from a statistical perspective or whether we can't be certain of the results.
  • Past performance of the poll: As polls start to proliferate, less reputable or less well-known polls begin to gain traction. This phenomenon can be especially true at the state level. Past performance of a given poll can be lost in an effort to glean the latest news from a key battleground state or a state where limited polling has occurred.
  • Just as Tony's Pizza Truck had a short lived run at the top of the pizza universe once its focus on quality was lost, voters who fail to look a bit closer at the quality metrics underlying the polls could find them quiet surprised by the final results on Election Day.