Polls, polls, polls! We have polling on just about everything. And now we have the latest fad in political punditry – to dismiss polls because Trump won. Which like most Conservative commentary completely misses the point of polls.
In 2012, Nate Silver shocked everyone by predicting the Presidential election. That impressive feat spurred a surge of polling reporting as gospel. But even in 2012, Silver’s predictions allowed for error in possible outcomes. Because, and this is the most important point, polling is not a prediction. Polls ask a sample of the population questions and then extrapolate their answers to the entire population. So you should ask if the sample collected represents the population as a whole. Landline polls tilt more conservatively. A small sample may skew either way, requiring a large margin of error. Finally these polls assess opinions, not actions and voting is undoubtedly an action.
Fast forward to 2016. If you remember the Primaries, then-candidate Trump routinely lauded the polls showing his lead over the field. Day after day, Republicans polled favored Trump and he ultimately won. Then in the General, the polls routinely showed Clinton with the lead and suddenly Trump hated polls. September through October the difference between the two candidates widened and narrowed. At the end of October, the polls were widening after Trump’s erratic behavior, terrible debates, and a audio of him admitting to sexually assaulting women because he is famous. But with a Week to go, Comey sent a classified letter to Congress about some Clinton emails on another computer which was promptly released by the Republicans. The national polls narrowed to near the margin of error. Election Day would be tight with an advantage to Clinton.
![](https://politicalpunch.politics.blog/wp-content/uploads/2019/07/a70f90a6-08c8-4c6d-b226-ef807cdaae03-4034-000004015bc14c6d.png?w=1100)
But alas, the President isn’t elected by the nation. He or she is elected by states. States are generally harder to poll due to costs, responses and other constraints. The national poll reflects the overall mood, with a 3% advantage to Clinton. Again this is based on opinions not actions. That 3% has to be translated into votes. If people don’t vote, it doesn’t matter what their opinions are. But people did vote. Clinton received the second most votes in history, only behind Obama. And her 3 million votes more than Trump was a difference of 2.5%, just as the national polls predicted. Unfortunately, the votes were not evenly distributed and about 80,000 votes in 3 states accounted for the Electoral difference from the popular vote. But let’s state it again more loudly, THE POLLS WERE RIGHT! Clinton had a 3% advantage and received roughly 3% more votes. The problem with 2016 wasn’t the polls, but the pundits.
Professional predictors eager to repeat Nate Silver’s success jumped on Clinton’s polling advantage to predict that she would become the first female President. At one point, someone gave her a 90% chance to win. Nate Silver by the way forecast a 66% chance of a President Clinton. Now, conservatives pounce on these prognostications as proof that polls don’t matter. But even here they are wrong again. Just as the polls were right about a 3% difference, the pundits were still correct. Clinton had a better chance to win. The prediction all still gave Trump a 10-33% chance to win. It was still possible and their predictions showed that. Pundits took these predictions and declared a landslide in an election that was already over. These pronouncements no doubt affected the outcome as well. Voters that preferred Clinton over Trump but still didn’t like her were given license to either not vote or throw a vote to a third party. Add the #nevertrump Republicans voting for the Republican anyway and what was an unlikely outcome, but still possible, happens. Again, the predictions were still correct in giving Trump a smaller, but possible path to victory.
So now people like to say that they don’t trust polls. Americans really suck at statistics and probabilities. And that applies to 100% of poll-haters. Now I don’t hate polls, but I recognize a major issue about them – the reporting. Reporters that don’t understand stats attempt to explain them poorly. But my biggest peeve has to be reports about what Americans believe. If the polls asks people if they believe in God. Sure, the report will say 75% of Americans express belief In God, 15% don’t believe in God and 10% didn’t answer/don’t know. I’m fine with that. But I hate, loathe, despise polls that report beliefs about facts. It doesn’t happen all the time. I read a report that only 30% of Americans can name all three branches of government – executive, legislative, and um, the other one, oh yeah judicial. The report clearly that people got the question wrong. But ask a question about climate change? Forty percent of Americans don’t believe in climate change? No 40% of Americans are wrong. It’s a scientific fact whose understanding is expressed in terms of belief. It might not be their fault that the don’t understand climate change, but it doesn’t change the fact that they are wrong. The worst offenders are polls about Republicans. Over 50% of Republicans believe Obama was not born in this country. No, over 50% of Republicans are morons that can’t answer a basic question about the former President. It’s not a one off. Polls of Republicans routinely demonstrate what Republicans don’t know – either legitimately don’t know or actively “don’t know” to own the lives – reported as matters of belief instead of knowledge. That’s the problem. Facts are real things that don’t care about your political persuasion. Reporting falsehoods as beliefs contributes to this idea of alternative realities that Trump takes advantage of. In the real world you get things right or wrong.
As the 2020 primaries kick off and we approach the 2020 election, remember this about polls – it’s just a number. Not good, not bad. Just a number. The media will hype the numbers, those ahead in the polls with praise them, those behind will attack them. Look for polls with large sample sizes, broad samples, and the types of questions asked. With any luck we will get through this together.