Transcript for Be Wary Of Exit Polls This Year (Well, And All Years) | FiveThirtyEight
I'm me. In preparation for Election Day it's important to understand that the first data that we yet is not actually. Results from any states or precincts or counties. First exit poll. Add there's a lot of excitement here deep fear of those exit polls for some sort of indication. How the election will come. But there are important hot shots in interpreting that data as we get to our election night before the results cut him. So accurately to talk about what to make a collection via our elections analyst at Bankrate it's in an aerial. And ultimately that's its quantitative editor of war art at war and still. Exit polling date is not necessarily reliable in any year but there are particular reasons that we should think carefully and our heart we use it. Because of the way that it's been conducted during that and so that they can you kick us off by just talking about. What the usual cot yachts are when it comes to exit poll. Yes definitely. So it's important for people who know that exit polls are not intended to be used as a tool for actually. Saying who won or what the margin is instead they're just a way to kind of tell how different demographic groups voted after the fact. And therefore you really camps. You know use to exit polls to say oh. Trump is leading 5149%. And therefore he's one Pennsylvania for example. On the data is weighted to the actual results after the factor here is a probably. Familiar with the concept of waiting polls. I'm so those data are gonna shift over the course of the night. Accurate data as the bulls. Attack. Anyway so there are particular reasons we should be you know thought Fuller or keep in mind. Dowayne that this is being done differently because. And directly bore out what's different this year. Yeah usually do way to exit polls are done is that people. The people taking them. Goes to the locations where people and they interview people as they need the case that's and so this has. A huge advantage over rated vehicles. Because even you pretty sure that the people you're interviewing him actually vote and as as of now no one of the big difficulties great vehicles is trying to figure out. Whether the prison interviewing it's gonna vote ends a different cultures ethnic likely voter models are different ways of trying to figure that out. The examples actually know that and so that that's a huge. Advantage that they happen that's one things that. Makes them somewhat more reliable. And pre election polls he usually. The problem is in you know in a world in which. Many people are voting early or by mail. The people that are coming out of polling locations on election night are not necessarily represented example. All of the people that are voting and and in fact in this in this year we had reason to believe that people voting on Election Day are probably more likely to be Republicans. Given partisan differences in how what's being said given that the people that are being interviewed or not a representative sample exit. Examples and obviously know this and so what they've done is they've supplemented. The regular exit poll. When it. Two different things one is a full loophole. People that have voted absentee or by mail. Current year. And in some states what they also then additionally is early exit so going to early vote locations in interviewing people. That are voting there are two Robinson pat one is for the phone poll part like for the people that voted by mail absentee. That you don't exactly know again whether they actually turned out there are people that will say that they turned out but they actually happens that your back in that. In that situation of two and a half tanks of gas who's who's actually turned out. And the only problem is that yes she don't know the composition of all these three. Things there are some people they're going to be interviewing. In there you don't locations their other people they going to be interviewing on election being greedy here exit pull. Style and then there's there's this phone pole and so it's going to be up to the pollster to sort of determine the composition. Of those three things and and that's going to be some amount of guesswork until we have actual results to eight. Trying to figure out how much of the electorate is on Election Day or vote by mail verses earlier worst act pollsters are going to have to try to meet assumptions in order to figure our how many people in the American electorate fall into each group. Now we at 538. Are gonna and what are we until the election is over the exit polls have been. We needed to be actual results to try to meet sensible act. However on election night itself decision das are going to be using exit polls to try to beat their projections of racists and they're gonna have to use this imperfect leader to some extent. Two you know projected the reasons are all watching on TV getting out of the party do that. Home wins with data that we described as not necessary. Or why apple. Right so I think that protection task worth its salt is only gonna make a projection based on exit polls when the margin is something super clear outside of the margin of error so like they might call New York because that exit polls showing that Biden is meeting their victory points or some time now. But you're always have to wait for real results before you call state like Florida are canceling. Edited just do give. Our audience outside RB going to be talking about the exit polls at all on the wire walk are we gonna friends were people. Auto auction back. Well I think we're not going to be talking about the exit polls. On election night itself I do think that after the fact once we have real results. And exit polls it can be we be using those results we can use them too meek. Two draw inferences about Saddam how different groups have voted. You know women rents it to men have voted that kind of thing. But that's in that's gonna required. You know the the waiting using actual results because otherwise we're just not sure that you know the numbers in the bowl. There in the country's nascent area if. Greta I think people should think about exit polls on election night has basically. Being not that much different from one of the many pre election polls that we debt you know essentially that's what it is because these exit polls are going to have to be calling a significant part of their sample. Com and using a likely voter model the same way that most polls. Soon. Essentially. Consider it but to the same expect you'd consider any other individual poll is not enough that your. Think that's safe to say and of course as we know any one individual poll should be put into the average. How so realistically should be looking at the 530 forecasts. This is seeking new and Nathaniel has been saying on and on that Kenya is that a big skeptic and exit costs in general. I am a skeptic there fine but there are you know a lot of people believe that day you know when they say that. Women voted for Clinton 6040 that is the literal gospel truth of what happened and that's just after. Write edit it could take months or even like a year after the election for us to get. Verified voter survey data that may have very different numbers. And give us some different sense of how. Women vs men voted or college educated people purses not college educated people voted at Saturn Saturn we're gonna have multiple data sources to try to Rick all. Right personally I would use some of these more sophisticated. Surveys that take longer time like the C yes. Or various academic studies. Are. So you've heard it from our exit poll skeptic in chief Nathaniel rich rich as you hear people rookie ethical principles exit polls. Throughout the evening keep all of this in mind. But more importantly go to the 538 Y walk. So you can get some real time analysis that is hopefully put it at all in contacts. And not relying too much. Poll. But states are much for an hour then on. He's. All right so adamant and handed over to hear that I pretty quiet block. To get all of our. Real time Roberts added 10 o'clock in the morning. On November 3 quarter warrants are for you also on the YouTube channel so make sure to subscribe to 538. You. War.
This transcript has been automatically generated and may not be 100% accurate.