Why polls can be wrong

• [Instructor] In previous lessons, we've talked about how polls and surveys are used to measure public opinion, but the important thing to recognize is that they are estimates of public opinion. Ideally, they're done as scientifically as possible, as statistically robust as possible, but even then, they might not give an accurate picture.

And perhaps one of the most famous recent examples of that is the 2016 election. In this chart, which I got from Real Clear Politics, you have the results of many of the polls of Monday, November 7, 2016. You might recognize that as the day before the 2016 presidential election.

And if you were to just look at this chart from the day before the presidential election, who would you think would win the presidency? As you can see, most of these polls have Hillary Clinton having more support among likely voters than Donald Trump, but we know what happened on election day. Donald Trump won the election.

Why did that happen? Well, there's a lot of potential theories, and political pundits continue to debate why this happened. One idea is that polls, when they randomly sample people, they're trying to randomly sample likely voters.

And some people theorize that there might have been a group that voted in this election that the pollsters did not view as likely voters but they voted nonetheless, and amongst that group, they voted disproportionately Trump. Another idea is that maybe someone else about the sampling techniques wasn't completely random, that for some reason, it might have skewed in favor of people who leaned towards Clinton instead of people who leaned towards Trump.

Another idea is that these are national poll results, while we elect our president through the electoral college. So it might be more interesting to look at especially some of the swing states what were happening, but even there, it was a surprise for most political pundits in terms of who won many of those states.

So the big picture here is that polls and surveys can be valuable. They can start to paint a picture of where the public's views on things are, but you should not view them as indisputable truth. They are samples from the population, and it is hard to do a truly random, unbiased sample.

And even when you do that, you're not even sure if people are going to tell you who they're really going to vote for. And even if they do tell the truth of who they're thinking for voting for at the moment, we don't know. Maybe their minds change by the time they actually go to vote.