10 Polling

Polling is a powerful tool. This chapter is being written during the early build up towards the Democratic Primary of 2020. They;re used to determine who gets into debates. Politicans are dropping out of hte race months before anyone votes based on their poll numbers. The media gets exited everytime they’re released. Yet, tehy’re each only based on the opinions of typically between 1000 and 1500 people. Why do we give those people so much power to determine the national conversation? That’s the topic of this chapter.

10.1 Building a Poll

The strength of a poll is determined before anyone is asked any question. A poll only has value to explain the views of the group that it is representative of. When the nws media reports on a poll, the assumption is that it is representative of the population. What is a population? The group that is being studied. Becuase one can’t ask everyone in the population who they will vote for (that’s for elections, and they’re quite expensive) we form a sample. A sample is a subset of the population that is representative of the whole population.

But the term population here doesn’t refer to everyone in the country (or even a state). Population again refers to the entire body of people that are of interest. It could be the entire country, or residents of New Mexico, or Independent voters, or any other group. Whatever group is being studied is the population. For a typical poll before a Presidential election, the assumption is that the poll is representative of all likely voters.

That’s why polls run during MSNBC or Fox News programming have always seemed odd to me. Polls of MSNBC viewers are representative of MSNBC viewers, not all voters, and not even all democrats. So what do we learn when we find out 87% of MSNBC voters want to see universal healthcare enacted? Nothing, except that 87% of respondents to that poll want to see universal healthcare enacted.

When a poll is not representative, it’s results are meaingless. This fact has led to several high profile failtures of pollsters (more highprofile than the deaily missuse of polling by cable news). Many pollsters incorrectly identified who would vote in 2016, leading to wide divergences between predictions and the outcome of the election. Figuring out who will vote of all eligble and registered voters in the United States is really hard work. The (NYTimes)[https://www.nytimes.com/interactive/2016/09/20/upshot/the-error-the-polling-world-rarely-talks-about.html?mtrref=www.bing.com&gwh=CA51C3A312313E0A3646DF9AEF3ABBA0&gwt=pay&assetType=REGIWALL] gave a set of pollsters the same polling data a month before the 2016 data and told to determine for themselves how to weight the data to match their expectationsfor who would vote in 2016. The outcome?

The accuracy comes from from getting a representative sample.

2016 wasn’t the greatest failing of predictive polling in US Presidential History.

Not to give away spoilers, but Dewey did not defeat Truman. Truman was expected to lose the 1948 election based on polling. The Tribune, which did not like Truman, had to begin printing before many states reported their results, but early returns showed Dewey leading. Truman would go on to win the electoral college 303 to 189 (39 votes went to Dixiecrat Strom Thurmond).

10.2 Plus or Minus 3 Points