I can hear the sighs of cynics across the land. Yes, 2016 was not a political predictor’s shining year. Yes, they said Hillary Clinton was ahead. Yes, President Trump won. But, no, polls still matter, whether correct or not, they matter. They are still key points of information which, like journalism in general, should be critiqued, questioned, and evaluated. They are signals and signs, not set in stone values.
What can polls teach us?
- You can learn which electorates matter. In 2020, the ‘Rustbelt’ states are getting a lot of attention after flipping from the Democrats in 2016, lots of polling has been done on these. As we have moved through this election cycle: Texas, Georgia, and North Carolina have seen a surprising amount of attention across all races, Presidential, Senate, and Congressional. This can teach us where the parties are also sending money and resources. Parties will commission expensive polls in states they want to target, watch out for which states get the most attention in the final days. Polling suggests the Democrats are trying to expand their potential electoral map and investing in more analysis of these Southern states.
- They teach us more than purely who may get the most votes. Polls reveal how voters feel about the candidates as people and politicians. Favorability scores are vital, Clinton had poor favorability ratings, Biden is above water in many important states – suggesting a stronger footing regardless of candidate preference.
- You can learn about the differences anticipated between Conservative and Liberal groups. Many pollsters are hired by partisan groups. Trafalgar and Rassmussen are very Conservative and their polls show Trump in a stronger position than most non-partisan polls. Yet, even these polls show close races and sometimes Biden winning nationwide – the fact that even conservative social scientists are occasionally predicting this shows us that Biden is doing stronger than expected.
Key concepts to look for this year:
Sample size & weighting: The larger the sample size the better (as it can be more representative) but depending on the pollster, some sample populations may not be weighted correctly. In other words, they may not be fully representative of various demographics (such as race, age, location). This is what went wrong in 2016, many pollsters didn’t weigh for education level, leaving some white non-college-educated voters out of the results, or not increasing their weight of importance in the results.
Registered voters vs Likely voters: Always look at WHO is being considered in a poll. Likely voters show people who are more motivated to vote. Remember, on average only around 59% of eligible Americans actually vote, so the opinions we need to value are those of who will really head to the ballot box.
Undecided voters: In my opinion, this is the most important metric to look at. In 2016, many were undecided up to election day, suggesting malleability in results. Undecided voters broke for Trump, which in turn impacted the end results. This year, undecided numbers are very low – suggesting that there is less flexibility in poll numbers.
Margin of Error: The first thing you should read in a poll is the margin of error. This shows you how much the results can deviate from the number given. If Biden is at 51% to Trump’s 49% in Pennsylvania (for example) with a 3.5% M.O.E. (3.5-4.5 is a standard margin of error), then the results of that poll can really vary from Biden at 47.5% and Trump at 52.5% all the way to Biden at 54.5% to Trump at 45.5. So the difference between the two candidates can really vary from a 9% Biden lead, to a 5% Trump lead (theoretically speaking, of course, this is unlikely). The margin of error is most interesting in states such as Wisconsin and Michigan, where the New York Times, for example, has Biden up by 8%, with a M.O.E. of 4 points. This means Biden’s lead is almost certainly between 4 and 12%, providing a level of reassurance to democrats.
Things to Remember
Polls aren’t meant to predict: They are measurements. As in school, when the only grade that matters is the one you receive, you can spend the semester guessing and doing the math to figure out your percentages and probabilities of meeting a grade level, but that may not matter once the final result comes in. This math isn’t pointless, it’s about positioning and understanding what work needs to be done. Polls help us do that work, we can figure out where to spend our money as a voter – for example, if you’re a democrat, sending money to Amy McGrath in Kentucky is not as helpful as Theresa Greenfield in Iowa. While both are seen as strong candidates, only one (Theresa) is level in the polls and may actually flip a seat. Polls should be used to understand what messages are winning and where your support should be focused on.
Polls need to get things wrong: Failure is important, it’s humbling and it creates change. We needed The New York Times to mess up and stress the electorate out with their infamous electoral needle. They have refined this idea to predict election outcomes, publications and pollsters will not get everything right, in order to make the science better and more accurate, we need to allow these institutions to learn from their mistakes. It can be frustrating, but as political observers, it’s also up to us to always question the information we are being given.
Our most basic duty should be to vote in this upcoming election. Regardless of where you reside, national and statewide polls can teach us about the country and how it’s feeling these days. Informing oneself shouldn’t just include journalism and online research, but also gauging the mood of the country with all its diversity. This year especially, we must learn the lessons of 2016 and keep our eyes open to the temperature of the nation.