Transcript for Model Talk: How the 2020 presidential forecast works | FiveThirtyEight
I'm dealing drink. I came Nate Silver. And address. Is model talk. Where are you really you really tight on synchronized married to him. I N I felt a little spunky person. Hello and welcome to the identity politics pod cast. I'm Gayle injury I'm Nate Silver. And and this is model Arnold saw it. All right America so. If you're listening to this pod cast it means that we have officially launched the 20/20 presidential forecast model. And you can go check it out at 530 dot com. It has been. A while it has been. Four years since we've had a presidential forecast model but it is finally here thank you for being patient all of our listeners and New York. Very eager to take a look at it there is lots of information to explore on this new version of the forecast. But I will give you all the top line number as of Tuesday August 11. It's that Biden is favored to win the election but he 71%. To check its. So Nate. You know who else innocently when percent chance of winning election year read my mind so those odds probably. Sound familiar to our listeners they definitely sound familiar to me and at a lot of questions for you but I'm just let it go straight about most obvious one. Which is that. Those were basically Hillary Clinton's odds on Election Day 2016. We just discussed on the podcast on Monday. How I didn't has been in a stronger position than Clinton was and so far this cycle so why isn't the model more bullish on Biden at. 'cause it's August nothing is. Really happened yet I mean a lot of things have happened a 150000 people have died. There than most. Significant. Protests against racial inequities since you know 1960s. A lot of stuff has happened right. But we're still. Fairly early in the campaign. And in a time when the world is pretty uncertain. And it's on very colloquial like there are ways you can actually. Account for that indirectly including the model we've had the most. Kind of violently swinging economic data. Basically in American history right there's more news and every yes you can quantify these things. We haven't had the BP piccolo maybe maybe we'll Kennedy could pick by the time you hear this site and I'd be very nice of them to it Tynan with the 538 model. We haven't had the debates we haven't had the conventions right. We haven't seen a whole time that advertising. Yet. You know it's still it's still pretty early and other Biden has a pretty big lead now. There are reasons to think the election could tighten if the election does tighten the and the Electoral College probably still favors from as it did. It when he sixteen. So you know. Looked to me. In August to say candidate has a seven year 72. Or whatever percent chance of winning is like that is. You know. It is fairly high in some sense right it's much lower than other models but you know I think other models have not been designed. As. Carefully to think about it and certainty in the world. So we were gonna get in. A lot of that but. I think it's important to distinguish upfront. Is this forecast model substantively different then the forecast model we put out in 2016. And if so why. So it's different because of 26 scene. We think are modeled treated well and when he sixteen. It gave trump a much better chance than other models did than the conventional wisdom is that prediction markets to its that you've been following our model you would have. That a lot of money on trumpet fewer Clinton that kind of thing and mean a lot of money. So we didn't think there was anything fundamentally wrong about. The math behind or model we thought carefully to how it's presented however we have this thing called. Starters COV to write code written nineteen. And we have. Like I said I'm economic data likes which we haven't seen. Maybe ever in this country serve and that's it's a Great Depression for some of this like GDP decline in the second quarter for example. So we do a lot of thinking around. Around that right around an additional uncertainties that coded. Create spending more time on the economic index that we've used. And the thing about how the mechanics of voting could be altered. You know so yat I mean if this we're. A boring ordinary reelection campaign which I don't know it's possible minute in the era of trump right let's say it was you know. Mitt Romney running for re election. And there is a mild recession in. And in the polls are released able Romney was eight points behind Joseph Biden or something and there is no Kobe pandemic. Van and okay they maybe could have. Fairly high confidence that that. That Biden would win right. Although still not I wouldn't think ninety year 95% confidence it's too early. But sought the world that we're right we're in a world where. There's unprecedented about of news we're we're voting is going to be very different than it was before we're even accounting for you know what I call just elect. Extracurricular shenanigans by. President from pressure constitutional right where he might you know. We're trying to account for it prism will dramatically about any when trying to. Steal the election right it's going to be on the scope of the model even assuming the election goes on as normal than. It's still. Pretty early and if you go back and looked at OK look at elections which we have data rate here's your bust polling data. For state level data going back to 1970 to their national polls back to 1936. Rate. You know having an eight point lead and really it's more like it the quilt and elect a six point lead Electoral College having an eight point lead. In August is not. Necessarily secure that their elections when the polls swing by. More than eight points. Between now and and November. So from what I understand in looking over to forecast model and reading your methodology. Is that. By including more data from history that includes big events like the stock market crashing or pass pandemic in 1918. War you don't other big events going back to the nineteenth century it in Jack's more uncertainty into the forecast. So I guess I'm curious. Can you try to quantify how much more uncertainty this current forecast model includes as opposed to. Past ones like the one and 2016. I'm not sure that it intrinsically includes more uncertainty right we're not just kind of I mean you know what we're saying is like there are now Maurer. Factors that we use to measure uncertainty. A couple of which are high because of coded right there once and high because of co winner of this factory look at. How much due to economic variables that we use and economic index change the changed a lot. And then when based on and literally how much front page news is they're so how often does in your terms of like. Full wit headline. And typically it in my in my happen ten times all year right. Disregarded have like thirty of the goes right. But they decoded in the protests of president comes impeachment. So you know so it. In 26 team there was also. Decently high and certainty but for for different reasons right. Increase Eckstein were a lot of undecided voters between sixteen the polls fluctuated more than they have this year. So yeah we're not even saying like. That this year's is incredibly especially on certain. We're saying it's probably about average relative to the entire dias at that we use. And the entire data set includes case like 1988 where Michael Dukakis was ahead. At this point and lost in a landslide onion and right includes cases like X 92 where you know there are giant. Ten or twenty point swings back and forth 1980 right we're. Rick and was way ahead in Carter made a big surge that Reagan won a landslide right. People are used to like the scan a period of my 2042 when you wait for me twelve where the polls were were fairly stable and those that's actually. More the exception in the rule. Yeah I mean sir you said this forecast model is a bit different from the one in 2016 I'm curious if you plugged in the data from 2016. Incidents forecast model would be much different from what we actually forecasted. Now there Witten because again we think like 2016 a model that great pretty much. And I think if you're like. Let's listen click but they collect this right if you design a model that's holding driven. That would have predicted a Donald Trump win in 2016 then you made terrible model and you're fighting the last war you. Obviously manipulated. The way you structure that model and may use that you're fighting last word you probably don't know what you're doing right. So I don't really care. What. Someone's model would have said. Between sixteen and if you didn't exist. In 26 teaching right. It doesn't matter because like oftentimes are building a model then you can over fit to past eight and it's not necessarily a good indication of a model. That it retro fits well right prediction and retrofitting are related up two point. The first and 80% of model building is saying okay Howard you have optimize production on past elections but the 20%. Is where people actually. Differentiate themselves right in the 20% is thinking about like what are things would lead me to it. To overestimate how certain item about future data and overestimate how much you know about a known data. Give in this relatively limited sample size a past data. So I've never impressed when someone tells me in early care. You know when some tells me. With their mop with their forecasts would have said ray and Carole little bit like about edged cases we are few it's a very strange. Election. We're something really weird happen under a weird state right out here a little bit more about that I think it's usually more important. For a model to give reasonable answers quarter quote. To his many cases it's possible the man in to give exact answers if that makes sense. But you know. But yet I think kind of what with the model lesson to me sixteen I think is it really overrated question right because I can optimize the model a way that would give an institute would like that question. But it witness it would make them island better this year. Right I guess the question is more getting act is this forecast model one that just has a more open mind about how. Ree months away from an election we don't actually know what's gonna happen is there just more uncertainty in general. And it sounds like in some cases in this election yes but others now. I mean there are her. Indicators pointed low and certainty like to a number of undecided voters but Peters Peters Covert and there's you know. And there's 10% or whatever is now unemployment right and there's all the news. And incorrectly when you have means kind of you know. Polls change in response to news now. There are fewer swing voters and their once where rice you have a lot of stimulus are relatively few people but but still and even you know even even with disabled wholesome mean. You know the polls have gone from Biden being ahead by like four points ahead by more like. Nine points or maybe now more like eight points right mean you know going from 49 is a decently large. Shift at four points and you kind of get more into twin sixteen on Election Day territory we're just a slight polling this combined with the trump Electoral College. Advantage could be enough to it. Two you know allow him to win a second term. If the current poll Lang and fundamental data held. On November 2 the day before Election Day what do you think the forecast Boortz show. Or Biden and try. You know how much of what we're seeing is the result of being months away from auction. Soul yet most of it is right I think if you. Tell our model the election is today. And beak is Biden's Republican nine the year 92% chance of winning. The Electoral College. Yeah I mean look. An eight point lead nationally units and probably more like six or something in the swing states six or seven right that is a pretty big lead it's not our other and the question that like you could. Have a pulling here that large could happen might that that would be a genuine upset. You know but now it's just the questions most of like things to change a lot between now and in November. You've been working on mass. For months at this point kind of ago combing through historical data all kinds of stuff working colleagues. And Hannity and you're actually altogether on the forecast and you get a numbers. Yourself that you're seeing for the first time great. Were you surprised at all when you saw that number. I EU is lower than expected for Biden you know I mean happens other models out there a kind of assume that they were being very careful about picnic. All of these things after Corey sixteen and I think they were being careful so the kind of is anchored my priors in different place in the division. And to be fair like you know kudos other people for having other models like. Bird strike that like. But now I was surprised. That it wasn't like by Nate 85% or something. I want to talk a little bit more now. Fundamentals. Go in tears forecast model for. New listeners or for people who want refresh since the last time we've gotten into the new secret. To break it down for folks what are the key components that go into a four model like S pulled fundamentals. What we mean when we talk about fundamentals and what kinds of polls. So. We used. Pretty much every poll. But state polls are more important in the forecast. National polls had an important role in tact doing. What color trend line adjustment rights of national polls can tell you OK well Biden's gained two points since last week. We haven't had a poll of Nevada recently so therefore we can. You know nudge upward a poll of Nevada to account for for the shift in the race overall. But for the most part. It's an election we're obviously the vote is tallied up state by state. And having it pulled a state is usually better than trying to make inferences about the sitting in that state of all you can do blend. And so he notes during by state polls. We also have. Fundamental forecast the call under a prior. Which is based on an index of economic conditions but also counts were incumbency and polarization. You know perhaps surprising to people that prior actually. Says it trump is only a narrow underdog. Their reasons being that number Y and actually not all the economic data is all that bad so income for example where. Because of the cares act the government has given it tons of money to people. On them to be running out now that's a big. Concern if I were trumped right. But the government gave a lot of money to people that Meehan disposable income actually went way up. And it's one of the six indicators that we use meanwhile inflation has been low the stock market. Has you know. Bizarrely against it has been pretty good as the other indicators that we use and so you know. Between that and the fact that it's a highly polarized climate so maybe things don't have as much effect anyway. And the fact that trump is an elected incumbent. The prior only has and losing by. A couple of points on you know he's behind by eight points or something. In the polling efforts of actually expects election to. To tighten a bit. And how much does the economic data and weighing on the forecast for example. If congress doesn't come to some sort of agreement on a relief bill and that extra 600 dollars in unemployment benefits goes away. How much could that potentially change the fort. So. It would depend. We used. Different ways of forecasting the it's a one image imagine rate. We actually are forecasting what can look like in November on him because we're an academic workers are cells we use. Surveys of professional forecasters who use the Wall Street Journal. Economic. Forecasting panel we also use the stock market which even those will annoy some people actually. Does have some predictive power in predicting macroeconomic. Conditions. So yeah we're currently. Assuming based on those forecasting device is an economy will improve. Pretty meaningfully between now and November. If you don't have further stimulus passed and I think that question that's a more questionable assumption right and see probably initially see that may be in. The stock market going down her forecasters predicting. Less growth in the third quarter. Men and eventually kind of effect that it. The variables themselves that we track potentially you know you can imagine. Dispose Lincoln going weighed down this month for example would have a negative effect on on the economy and therefore comes chances. But in a but the tigers that we had that much right at this point. It's you know Sydney ivory percent. Base in the polls and when you're 25% based on on the prior in the prior rents down to zero by election. We mentioned that. We're accounting for Covert nineteen in this forecast models of how exactly Donna's before incorporate act deed in. So. Currently in some ways that are. Not that important to the overall model design. So there's not like any assumption and the model felt like. Because trumpet screwing up on code today he'll lose whatever right. You know in fact I think the reason. To think trump will lose is not because of the polls but it is because of light payload he screwed up on coded people. I'll have their lives kind of ruined for the past it. Six months because of coated and they won't get that much better before Election Day and so therefore he'll lose or that's a perfectly coherent hypothesis. It's not something which is something that would be the result of the polls are Siskel model right. So in sums of I think it's like I think it. The model should be cautious and everyone else. Should maybe being cautious right instead you know I'm not been demean other models but trump is not. But only used more than 10% chance of winning the Electoral College apparent armed but OK how is Kobe data used. In the model in a couple of ways. One is that we run as we elect in cancer alternative to the polls he run. He regression analysis that says okay which factors predict. State by state polls he can use that prediction. That model to estimate we Colledge to feel in states we have little polling or no polling right. And won the factors it tries to assess is. Is Anita pulling driven by Covert cases where seats we have. The worst of it outbreak is that affecting trump standing there positively or negatively for that matter. Also we allow. Error to be correlated based on Caylee cases writes it could be that like. Trip would underperform his polls and a bunch of states with. Cove it. Or that could affect the mechanics of voting C have. Error in ways that would be correlated with Covert. But those things are like so tactically yet co we did he gets used for its night. We're not making it cookery strung. Assumptions about it. We're just allowing like Kenneth coming to be like. Vector if you will. That affects how the votes to be it state by state bricks are. In the way they're in 2016. Of the upper midwest. Be tied correlated. Met that we gave trump a better chance than other forecasts. In this case. States we have high levels of covered cases are now also correlated. For example I think new York and Arizona couple pat pat outbreaks so while they may not seem obviously correlated from an elect oral standpoint. Doubts however correlated and that's model is that correct. That's right yeah I'm there like. You know literally. Dozens of different correlations that we account for were reaching for him you know how old the states are turned to their you know the average citizen how oldest person is still like. What reach in the country there in ten of latitude Marta to it you know racial and religious categories on both added cove waiting for that matter also. How much the State's vote we expect to be cast by mail all. To those things which are potential sources of correlated here. What considerations do you make how different states are voting because in a lot of states people are going to be able to do. Cast a ballot by mail if they wish because of coated in some states including states like Texas that's not the case not anybody who wants to will be able to cast a ballot by mail so how much does that changed the kind of fundamental assumptions about that. So we use. Something here useful that a group of academics put together called but it habits of the cost of voting index that. That looks at how easy to a disability each state. And when their changed in that can change the composition of the electorate in ways that are. Fairly predictable meaning that when you make it easier to though. Democrats. Tend to benefit. When you raise barriers to voting Republican send a benefit. So those academics graciously provided an update. Of their index for 20/20 to presume to be published you're corny sixteen. I'm not entirely sure. How much they're doing to account for Covert or not. But it but Texas a state where. It's become harder to vote according there index. And so that actually makes the model. More bearish. For Biden. Member bullish for trump in Texas and then other models might be. And that's actually been a point of contention which we've talked about on this podcast whether or not making it easier to pro life youth vote by mail. For longer periods for earlier voting. Actually boost democrats' odds it sounds like you're saying that they do you. While we've heard from other academics that they don't that it doesn't really shape the outcome of an election so that we can view he said to eat on the what should we suit I think conflicting messages I think. The voting by mail piece of it is something where people meet stronger claims and I think that's partly because like. Partly because voting by mail something that both parties really did take pretty good advantage variety and there are things like you naps in the ballot it helps to have your older units ended in it. Get a stable household right soap. So voting by mail maybe one of the exceptions actually weird. Where it did have a neutral effect or maybe who knows even help Republicans right but overall like look. There is a reason. That Republicans. Spending a lot of time trying to make it harder for people to vote which is that Republicans benefit. When fewer people especially people who are poor who are minorities. The benefit when those people have a tougher time voting right it's Perry plain and Democrats and a lot of time fighting back against that. And so. If any academics are are. Saying that overall oh it just kind of totally neutral then number one I don't think that's what the data actually shows and number two you're saying that all these. Apparatus that exist for both parties to spend huge amounts of legal fees and time energy. On in the GOP's case making hard people who though Democrats case making it easier recent all these people are idiots for doing that I mean. You know again the narrative about mail voting right I haven't looked at that I could reasonably be academics are right about that. But now look I mean news flash right Republicans don't want it like black people to vote. Basically right that's kind of one of the things yet include about the American political system I'm glad we're finally incorporating. Some of them logic logic and remodel because it's incorrectly shows up. Pretty strongly. What other news data are we considering this time around. You've we have the Kobe did we have how easy it is to vote. What else is now. I mean that's really most of it we have someone Maurer. Sophisticated. Sees me handling around. Events like debates. Where. What can happen around debates is that. Or conventions of a costs example a candidate will. Get a bounce in the polls and that will dissipate over time on separate as the model had like a convention bounce adjustment. But now there's a second strategy which is that when it's a major event. It will kind of take the snapshot. Of the polls before that event happened. And then hold on to some of that snapshot. For a couple of weeks until we see if any changes in the polls is permanent right does that make sense. So we'll hedge against. So instead of you know someone how to get current debate and the forecast immediately. Seeks access but let's wait and see if the enthused yeah it'll say candidate is to say. Let's give you have credit for the apparent in the polls right now and after weaker to them or going to be for credit basically. It's a pretty minor Justin but it it does mean that like. You know the model wouldn't have. Balance as much toward Mitt Romney for example in after the first debate 2012. Or just were Clinton after the first and second debates in 26 teams so it'll. Be a little bit more conservative about. About flopping around as a result of a minute of it and and an event like that we've also expanded our data thread. Yes so that economic index this now constructed. On elections when all it back to 2000. Sees me to 1880. It's a big problem we have a small sample size. In these fundamentals models and so. Going back to 1880 lets us capture. The great depression and other very kind of strange economic conditions right. And so I think that makes or model. More were route bus. Also to go back to 1880 it's not as clear that the economy is so determined basically. Predictive. Of kind of which party. Wins the election. Since when the reasons why. Why you like we don't we the fundamentals that much is if you actually expand the sample size. Today that you couldn't kind of test the model on originally well it turns out may be the economic fundamentals are. Are not as fundamental you know certainly obviously Hoover did not do very well right if you have a really severe. Depression by it but yet and finally I would take to like. You know even. If you're kind of using an economic prior. It's really hoping that voters are blaming trump for an economy is a pro readings on economy are actually. Are actually break even their fairly decent. And so if you rigged up a model that. That's as though you know its economy it's all about the economy no it's it's all about Kobe hit it here but I mean that's if you actually look at the polling data is what it says and so. Maybe coincidentally your kind and indirectly. Having two wrongs make a right right. The economy gets wrecked because of coded. And therefore use economic variable and therefore you say allegedly went term because the economy right. But yet again are in our model it's kind of it it more reversed worded Pryor of of zero or I think you know trump losing by two points or something to be very close electoral college and so. It just kind of saying hey things might tighten down the stretch for an here. All right so let's total rapid fire before we close out this episode of model talk. I don't say that in future episodes of model block we roll answer questions from listeners out. Prisoners if you have any questions you can to greeted us you can also email us act podcasts at 530 dot com. We're going to be during this through the elections are one of a lot of time to answer your questions but are here's a matter of irony erratic. Moot. At this point according to the model what is the tipping point state. I don't know actually there isn't one tipping point state ahead of time. To frequencies determined. Only after the fact. But the two seats it thinks are most it's like it tie between Florida and Pennsylvania. So. Between the three kind of mid western whether you can debate whether it's having his midwestern not right but that it Pennsylvania Wisconsin Michigan group. You know the one that has pulled the best for Biden. It's Michigan. That would has a lease electoral votes. Is Wisconsin. Right so you could afford to lose Wisconsin. Pennsylvania certain soup super important and in Florida has an electoral votes and also like. If Biden wins Florida than he can lose some midwestern states and still with the elections so so Florida Pennsylvania. Potter. The most important states according to the model not the sexy here. Wisconsin and Arizona people talk about as much Florida has 29. Electoral votes. Pennsylvania has twenty if you wind between nine electoral votes in the becomes theory hard for trying to win next question is. What is little likelihood of an Electoral College popular vote split according to a model. It's in the neighborhood. 10%. So I think we have. Biden it in the seven years MET percent to win on the electoral college. And eighty or 82% to win. Popular. There is also liked. Very outside chance at the polls are really just about violated and that Biden wins. The popular vote. On a Bible is elector college it was popular that's like a one in 500 she answers of and it's quite low on but the chance at trump. Does we just we sixteen are you know there's a 10% chance of a repeat of that roughly. And that is key here right and her yeah if we had. If we were not his Electoral College then. Then we say up gaelic it Biden's currently in the polls and even though might tighten nets would translate to an eighty. Something percent chance of winning so look for college as an important factor here. And lights be things. With events should listeners expect a senate or house forecast the cycle. Yes. But don't you think about it right now who can act that's just yet. I thought I thought I. All right Ari fair enough well let's leave it there name I know that you deserve a break a lot of work. Make yet. Great thank you. And a couple housekeeping notes before we close out this podcast we're going to be doing night we podcast during the conventions so for next week. The Democratic Convention and the beat after the Republican Convention. You can expect near daily podcasts are Monday through Thursday but it also means that our regular Monday podcast will come out a little bit. Leader than usual we're gonna be recording our park that's on Monday after all of the speeches I've concluded. At the Democratic Convention and that's anger following week so you get near daily podcast after we eat a little longer for the Monday podcast. Also on a repeated note. The next humans are going to be very exciting and busy here 538 so. We are hiring a producer in a temporary roll through inauguration day 20/20 one to help out on the politics podcast to work two to three days a week so check out that posting. I'm fine 38 dot com or my Twitter feed where it's also posted. If you're interested and it usually deathly plot or if you think you know somebody might be interested let them know. My name is due Hendrick Tony child is in the virtual control room you can get in touch by emailing us at podcast I've pretty calm. The house of course leaving us with questions or comments. If you're a fan of the show beavers are reading or you in apple podcast or or tell someone about us. Thanks for listening and wheels. Yeah.
This transcript has been automatically generated and may not be 100% accurate.