Model Talk: National and district-level polls disagree

In this installment of "Model Talk," Nate Silver talks to Galen Druke about why state-level polls and national polls are showing different degrees of competitiveness.
57:17 | 10/23/20

Coming up in the next {{countdown}} {{countdownlbl}}

Coming up next:



Skip to this video now

Now Playing:


Related Extras
Related Videos
Video Transcript
Transcript for Model Talk: National and district-level polls disagree
But later grimace over it. Not. Might just have fun. Okay. Hello and welcome to the identity politics pod cast. I'm deal injury I'm Nate Silver and and this is model saw. Neatly Edward boisterous models like interest network it was a subdued. We have maybe it's because we were up late last name and or party back to the grind podcasting. A gag I have to say I didn't really get any of that rest and relaxation. And a as a runs or youngest questions that we talked about podcasts met yet I think in that room this morning ruthless like. Well union and Iron Man I also wanted to mention that this is likely are. Anti pen ultimate. Model talk of the year. It's you can ultimately set in third to last I did tonight I didn't hear that. So we have today. We have the Friday before the election and then probably once we get results will a model talk roundup but. Yeah it's really things are calming down to the wire. Yeah I'm. I don't get excited and repetitious. Article let's. Different strokes a picture are forecast looks likes went eleven days ago. Until Election Day. Not that much has changed so the split for the presidential race is still 87% chance that Biden wins twelve or 13% chance that trump wins. Democrats 174%. Chance of gaining control the senate. And in 96% chance of maintaining control. The house. So and he let's focus on that steadiness to Starr wears. What are they dynamics at play here because we've set that as we get closer to Election Day. If the polling remains the same Democrats God's word in group. And it. The forecast has been relatively steady for almost two weeks now. So does that mean that the race is tightening a little bit and that's why we're seeing those odds remain around the same what's. The the overall scheme booklet. I don't know I'd say it's an instant steady if her me a week or something but look. The national polls have tightened a hair. Them for castle sneezes state polls it's a bit less clear those have tightened however. The state polls were never really consistent with Joseph Biden having it's like eleven point lead it looked more elected. A nine point race. So basically you know the model is projecting the popular go in. Hire him. Eight points for Joseph Biden that got out. Eight and hats now eight point one right. So it's tight knit tiny bit. Meanwhile time has ran off the clock. That is who's titans win probability you know I think it's like if you're 87 that's. Pretty hot right spot a hundred it's very importantly not a hundred Wright is like that much more. Room to go right Salina where Biden again in the exit 93 year 94 or something it holds by Election Day in and it's still pretty. Glacial pace try to gain and a half appointed day at the polls are steady there's been maybe teeny turn negative. Tightening. Com. You have a theory as to why the national polls. And and then national picture has heated IXT level holds. Arc different. The high quality pollsters which are often networks and academic pollsters. They kind of all tend to think a little bit alike into the kind of pull the same states here poll nationally simpler times right. Post debate post quoting diagnosis. We had like a lot of high quality. Holsters. Polling national polls high quality light color polls and shall we say the staples were a bit more mixed there were more kind of Rasmussen intra faltered and public policy polling molecular matter fear pollsters right. And you know when he sought higher quality polls in this state think it seemed kind of more consistent with like a larger. Finally they beat me beat. Nine and half points or something so it's partly an exit pollsters right it could be also like. The can be a little bit of hurting right especially the lower quality pollsters. Tend to want to. Come in pretty close to whatever the 530 Gator got the politics average was in this state right. Becomes probably more presents later in the year so that can create. Some funny stuff some times. You know one of the things we don't get very many polls. Not swing states right you know now available at like at West Virginia poll or at Kentucky poll or something that shows. Joseph Biden doing. Surprisingly well there. Right. Or Kansas right. Her North Dakota. Led by those polls I'm that your and it's sort of plausible right now mean you had basically. You know if you have a lot of kind of white older rural voters now you finally have a a white guide at vote for her it instead of Clinton or Obama write him active or John Kerry was a white guy bit. I'm kind of series this mrs. Kennedy liberal right so maybe by this event a states that gab and it could be that like. Biden is making up ground of these really red states that doesn't really matter tipping point states and their theory that any at a Tennessee Tennessee even out. You know. Which so neat you mention holidays. We often times and up on this topic somewhere around Election Day are we now on hurting alert our re seeing behavior from pollsters that should be testing that some of them are hurting. And and sorry I should say for listeners hurting as essentially when pollsters published results that converge on the conventional wisdom of what the outcome of the race will be seeking out Qaeda. A peephole that represents the expected result before Election Day so that date what good it was that reputation but oftentimes what actually happens is that as pollsters hurt polls become less accurate because they are not open to you. Different twists and turns that the race could take so that's why we're on the lookout for her rhetoric are at. I mean here different types of workers right. What happened pulled it hurts is probably may be a poll on liner IDR there are some very good online polls and they're so let's get online polls right. I think some. Online polls that are less good. In part pretty strong assumptions. Into kind of how they wait and massage their data. Because you're not getting it. Random sample is simply fear greatly do you with a phone poll. It seems like sometimes those decisions are. Influenced a little bit by the other pollsters are saying you're polling averages are it's kind of one category. And at a category is. Sometimes these. Quads bipartisan poll news like this big shadow area right of polls are not officially partisan. The way we define which is actually pretty strict definition but in practice. It's clear who this fox those polls are rooting for to win and they. Mammals that writer kind of framed in a particular way in the trying to influence the averages are boost morale right those lists nearly occurred. Toward the exact average at 538 that they like. You're very aware of kind of their position in the landscape in it kind of seems like. Incentive. A poll. Beginning. Polling input they don't doubt that they're actually calling. People. But just extremely strong friars and your basic skating like. Somebody's opinion the third type of herder is. Something you see more in higher quality polls which is that high prestige polls sometimes it. Last fall they published the Pendleton police publish right they know it's going to be judged in the major us. Your little more careful right you know if they have a now mired they would find some. Way to the public theater where work. It applied for turnout model right. And get a little creative write. That test actually Ian of the race. Sylvia their multi consumers. I sought Connor sent through is promised Bloomberg right. I'm Twitter quote my favorite new wildcat polling theory. Is now that a 530 average is seen as the Wall Street consensus you have so much potential polls are hurting. That the real polls art house district polls where polling averages don't exist. There's a lot on taxpayer by the we've. Establish what the general meeting of hurting as an and whether or not pollsters are doing it. Do you get the sense that the house district level polling is showing a different picture of their race vent some of these other polls that we may be able to. Coverage its interest sitting right. So first of all EB. Sarah and I should assign a stewardess and minutes after the at about like actually rounding up. All these polls and looking at which ones happily presidential race but certainly anecdotally at least when people on people concluded it seems like these polls show. Very good results for Joseph Biden even once you count for the fact that. A fair number of them are partisan polls right. You know they might be consistent with the race where. Biden is winning by twelve or thirteen points or something right these house to house yet and Connors theory is that because there's no I pretty average and because you know frankly. Now with him give a pollster. Too much. If they have. Trump losing a district by seven points. And than four months later. After we kind of crunch the numbers like congressional district it turns out that he won it by one point your bracket it was gonna remember that earlier so maybe it's a more kind of honest take on their race it's an interesting theory I don't know right it's clear that like kind of I mean the national polls show. A period of Biden lead right the state polls in the chose these smaller quietly in the district polls show. The script actually buy it's a little weird right. It's a little weird I mean I asked think. One thing they think about a little bit. I think pollsters probably feel comfortable with. In state polls with various kind of single digit Biden needs you now. It seems like you don't actually see except with the rest of the world. Very many polls actually show trump ahead even though it might is actually had backed by Catholics in the state you're going to buy random chance alone. Get your occasional light truck plus one writer current post to its appeal cedar and it goes right it also seems like you don't see hearing people through going to publish. Double digit margins for Guyton even though. Even though if he's plus I do get occasional plus ten apostle or something right. So I don't know right I mean I am not be very articulate here experts think it's like. If it pollster showing Biden ahead anyway. There incentives might run pretty strongly against. Showing a larger margin if that makes sense I don't know. So essentially we should be on poles are hurting watch bite. We won't really be able to say what is happening until we know the results but keeping in buying that we are seeing differences between. House districts state local and national. Should make us question that there that there is something going on that is seeping these different result hurting could be one of prospects. Is that it Ellison's equator. Yeah and you know what I think I think two is like the sometimes the Vieri. The last polls because of herding wind up being. Less accurate elected polls here for days before and it shouldn't be that way but you know it's possible I mean in 28. Sixteen. People get it was a little bit of a tick back toward Hillary Clinton in the final like forty hours right. Turco got have to like 35% in our forecast men trapped between 9% in the forecast affect different like but sometimes that last little blip can reflect. Hurting. And may actually not. Direction only kind of when you in the right place. The other thing haven't paid attention Q this week it is turner and it looks like fart 38 has also updated its turnout projections. Upward revised upward arc and our projections because it looks like turnout in the selection could huge so. What led to that and what kind of turnout re expecting. The turn I should say like the turner remembered a forecast. Doesn't. The fact that. Margin percentages right it just kind of like a side product. And Italy the the ratio will turn out in different states affects our. Popular route estimate Reich is a way that works for the national popular road as we have it. Forecast in each state and we at an upbeat need to know kind of what turnout is in each state a percentage basis right but Texas makes up say 7% to turn out nationwide. Nordic and awaited estimate the national popular business. Before we go any forever. Let's put some numbers address and our projection is now that the total runner in the presidential race is going to be a 154 million. Within eightieth percentile range at 344. Million a 165. Million. Into dark and sixteen were 137. Why or where was that before we meet the repairs and how much did we organize our ports. Own. So we had been at around when 43. In so add another ten million actually what most of that was is. We had been using estimates. How much population had grown since when he sixteen that. Apparently were too conservative. Pastor Michael McDonald. Has actual estimates of the voting eligible population. We kind of an indirectly trying to ask me that the population growth and sure that there are more eligible voters in some states then. That we estimated by the way the fact that in some states you have removed. Some degree a felon disenfranchisement. Can also increase the voters eligible population. So using a more contemporary. Up to date. I asked him how many people are actually eligible to vote. But also our model those look at polls. Ask how enthusiastic players are those are off the church right. We also made an adjusted based on looking at. Voting laws of whether they've been liberalize or. Or made more conservative Brian if your podcast with. Emilia and tell your day right in general actually there are a lot of laws that. Make voting easier this year. Some of them are permanent some of them are temporary because of coded right. But you would expect that if there are more ways to go than ever which in many states there are. That that would boost turnout. Just a little bit more people who would like to cast about will wind up having a week to cast a ballot. So we wind it now at a rent 154. Million. Do you were saying that that doesn't actually change are rejected national popular. Mark. Can you explain why that is because we're not. Using turn me we rely on pollsters predict turnout right. And the pulls have to wrestle with but they're gonna turn out obviously right that we don't like. Adjust our margin up or down based record projections right we Louis averaged the polls as an average the kind of pollsters estimate of what turnout will be right implicitly. Pollsters don't explicitly say that all we're gonna predict. 70% turnout of the whale population right. We have to kind of inferred that. You know that our opinion. All of what turnout will be year model wicker will be doesn't affect whether Biden is up. By nine point three points that are now in practice subjectively. My guess is that some pollsters are being a little. Too conservative with their likely voter models and maybe that. Hurts Democrats will now over the past week though symbols were like Democrats gave electric motor Messina kind of seem to be price an in the polls. An assumption that if there is differential turnout it would help Democrats it's not surely was naturally case earlier in the cycle. If that one theory about Biden's growth in the polls could just be actually that like he actually comes from the fact that. Now you're counting for early voting and it's becoming more apparent the Democrats have a high turnout they Republicans will also they probably will right. By Ian how Poulter. Wrestling with early voting. Is I think one of the big. Important question. We've actually gotten some specific questions about that's why we're seeing significantly different results in some polls surrounding registered voters likely voters but that also. The high turnout scenarios and low turn our scenarios and one point particular that showed us with a senate pull out of Iowa so what is going on here. So that was at Monmouth poll. And the Monmouth poll found actually that. Meehan stimulus right. Democrats would prefer a lower turnout Vick is. They are voters have already voted and simplifying here right. And therefore you don't want more turnout has the remaining people who have left to go to Republicans. So you know Monica's an interest in pollster and then I believe they used. Probable stick waits for the likely voter model which a an absolutely the correct way to do it. Like I don't understand why every pollster doesn't do this it's kind of idiotic not to a minute Finland pollsters by doing that right like van in like. I love you pollsters. But use these problems to models it's clearly like it's you can't defend he's in his heart and doesn't make any sense at all Kenny explains it probably problem ballistic. Turnout model as. Yeah you take whatever inputs whether. Whether it's you telling me how what the or though some pollsters asked for how much good in the past they transferred knowledge of like. Where's your voting location right. I mean the tone of the times have done good work on this with Siena College in if you kind of check all the boxes. And yet the highest propensity score right you're still in a 100% to vote may be years. 90% or something right. And if he'd say you're not gonna vote in donuts is out question to you still might note excellent you 20% chance or something incorrectly when you go back to check the voter file later. And see what it's when voters right. So. What most pollsters they have a hard cutoff right. Whether you and your 100% essentially your zero if you're about suppression blows Drexel. Which doesn't necessarily make a lot of sense right. Like what if there is a long table. Low propensity. White working class voters right. Each of whom have a 30% chance to vote individually but collectively. Friction that will vote right and they may be their temp voters' right earlier hurts when it. Latino voters and black voters Wright who haven't voted before your little skeptical all right we really have this nineteen Internet search it may be blown. And maybe two thirds of people who used. Filter out what one of voting but he went there are well if there really democratic vote in some states like Arizona that you can little dog Democrats there. So. Yes there are questions of light. If you have people who have already voted. Should you treat him as a 100%. Right and in people who are likely vote he should treat is. Something well below 100% right I don't know if that's what month is doing exactly this argument for doing something. Right. On the flip side ebbing our base to be like okay. If you have democratic like others have already voted. In about the likely voters will probably vote but maybe well right that could be a reason why. Why likely voter polls would underestimate how well Democrats who do you give your military law and right. The flip side is let's say that you're also now excluding some of these unlikely. Republican voters that will vote maybe kind of all balances out in the end right but their various options you can make about. And how you treat. A likely voter verse and already voted. And how that affects year. By the office of polls that don't. Even ask about whether you already voted or not at all right it may be he would ask the question you should keep a pollster it maybe you already voted. But the poll kicks you out because likely voters stream that he would ask you or you noted already you know. So there are various ways this could create. Diocese in the polls. And it's an issue additional sources here. Are so hopefully there are sufficiently. Into weeds of polling methodology for. Our listeners. I ward something manner. So let's move on and focus on some listener questions from Sierra now as we mentioned and and those late night podcast last night. We're getting down two you the wire theaters a week and a half laughed about in this campaign. And posters are going to be conducting their final holes relatively. Soon. So our first question comes from idea and the question is. How would the model reacts to a last minute surprise story it is too late to be properly polls. Yeah I mean that's just. You know part of the uncertainty. Him come to the fact that you don't actually. Incorporate. Every last news development write a Khamese stuff that happens on. Election morning in theory. I keep in mind though that this is a case where there early voting does matter align right something like two thirds of the vote is expected to be in. On election morning a minority 57 million votes. So you know something that happens like people who vote very early. Tend to be pretty big partisans Republican in. The same way no matter why right. But you know a lot of kind of more regular eat an undecided voters may vote. In the handful of days before the election and so it's pretty hard for. The sun last making teacher to have the full effect that it might have an election where. 80% of people are voting on Election Day mean that is a pretty change. But I mean look certainly if there's some major event. And we don't know direction which affect it would happen which can have to say okay well you know. The model kind of stakes in some uncertainty in this might be case where it's not making enough uncertainty. Because there's this event and we knew that happened but we can't quantify it the flip side of that might be that if there is anti climactic finish the race. Then I mean that model has too much uncertainty right consistently that hey that's when added. Eight times records crazy news development. That we underpriced rate but now and then you know give a very smooth finish than it might make things less likely to have a polling here. So somewhat repeatedly we dot. A question that asks what Disney mean when he says that the 538 model is somewhat conservative. I think man. We make assumptions about how much air there isn't polls that we assumed mayors more error. Other people who build. Models do. You know for example we assume. There are three ways in particular that we assume the final polls might be. But were wrong ban models. Its number wine is that. Most of the model is calibrated. Based on polling data from 1972. Awkward. The cassettes we start to get some what robust. Polling withers those state and national polling. And polling throughout the year not just the right. RS though kind of how much the final polls went beyond. Also incorporate. Data from back to 1936. Which is the first you're Gallup did final national polls. In galloped national polls and the thirties and forties. Our at a pretty checkered track record. But that it's accounted forward. In our estimate how the final polls might be right we calibrate that based on vehicle and ethnic and 36 and so like literally that. Do we vs Truman polling here. As a kind of where all right let's and one. Never too. Because. We think that turnout is going to be hard to predict. Because of all this new mail in early voting. Because we saw in the primaries that turnout was kind of all over the place relative to expectations. We also add. Some additional air for that so we state polls are going to be 20% more airplane. That it would be otherwise. Finally we have. What we call that detail here. Which is because the sample size of elections isn't all that large. We use. It's a faculty students T distribution and sit at a normal distribution. Which affects the weight on the tails of the chart right it means that you know you know the models reluctant to do about. 98%. Gets about buying up a to put thirteen point lead said that ten points or something right. You get these tales that that. Like gay noble distribution darker Rick it's very very game. Details. Are thicker and took it's harder to get to 9899. Point five. And whatever else to go beyond that and so those are three things like you can argue we don't. Have to do right if the rest the most opportunity to be too awkward them wire we. All this and using. In 1936 data for this last part of it right is that really pertinent accurate polls are now we have a lot more polls maybe maybe not right. You know be effect helping is the correct practicing it's kind of. An ambiguous. And we we've talked about Arafat fails before better listeners should be from the air but this is adjusted for kind of turnout as a you know. A little bit lower. A little bit lower debatable I think right sort of set rate I think we're making I don't know I mean act kind of hate the fact that like I don't connect people perceive. I thirty eight's forecast I kind of hate that like if you kind of take a system like avatar. Of certainty I mean Biden is ten points ahead. In national polls and there are eleven days to go on the campaign attend were taping this right trump is kind of out of money they're nowhere debates. We have what raging pandemic it's getting worse right. Everything he tries kind of seems to not be very popular. And yet he has a book tickets for him. Will yeah I mean I think that's that's why people are curious. You know. How we got to that place that seems relatively conservative you did mention that we have no more debates last it's only been it's been. Less than half a day since we published last podcast we didn't actually get some polling and that the how the candidates did in the debate whether or not Americans thought Robert Martin one. Yeah or so repulsed. Out. Of that show that might be better although the margin from his first debate. You don't typically if you had yet he if you lead who wins the debate polls like fifteen points that you might typically expect. That translate collective further one point bite gains. In that top line polls. I'm not sure you'll see that here because. By Missouri so far ahead. Like Tammy and went in custody. Who wanted to be equal to fifteen points and or treatment in point is very different. You know our poll with. If so let's show although voters thought that might do better. There would is essentially no change in the top line people's voting intention there. So. I don't think I'd expect to be. Change. If it shifted another point toward Biden. I would be shocked if particular point for Trout to be because the debate because like. They seem to be teeny tiny bit of tightening going already. And maybe he's kind of short and it's hard to lose an election by. By ten points. In the United States can a couple of residents Tuesday that is pretty hurt you. So maybe it would be a little bit of tightening there is because of the debate. Again we should say almost talking about it's all. It's all based on pre debate. Oh right slope. If you believe that because the debate. Biden is now in either. Remain steady in the polls or increase appointment we've probably mean for the steady. Then by the time we get host of people in six days from now right. Then apparently be about 80% he said percent to. Right and you mean post debate polling as the top line polling here in the groceries and are. Unit desperate are people pollsters are going to be asking candidates in the eight. Artists arson caused from Leanne and she asks is there significantly more state polling in the battleground states in 20/20. That in 26 teams and of course we got a lot of questions along these lines so let's answer that question. But I also more broadly want to ask what does the polling the landscape look like right now are we getting a lot of high quality polls. To what extent are they disagreeing towards an army getting low quality poles and two and might that be. Messing around we have our averages but let's answer about the battleground state questions person and brought it out. I think we will have. Enough polling. In the -- urge. Majority. Of swing states I'm a little worried about now that which is always a state that a lot of pollsters avoid. Missed it ripples of disagreed quite a bit actually. IA unit captures him and gets a little bit ignoring its advocates not back those this year in Nebraska second district won't get pulled very much but like. A kind of bigger higher population swing states. I would expect you Gannett. Not be lacking for happily pulls to park has. You know people are editions particular obviously that you're tied Siena College. Doing all those polls in. Keys races as is very valuable and there one of the best pollsters around right. You know mine that is pretty prolific. And good. ABC news is doing. Polls in. In states this year. Which they haven't done that much for several cycles of our colleagues if you watched those. Those are by locals to have. So you're not really lacking for her for high quality polls have to go. In its streak of a week or two where some steak it's ignored Revver little light on Arizona polls recently for example I wouldn't mind. Pollsters revisiting Minnesota. Which got all the literally in the cycle in comparison to 2006 team this is a lot more. We state polls. You know I think the high year. Echelon of online polls like you've got a what not a pretty good right there's a lower Echelon. Where they seem to herd alive right there's an Echelon of like pollsters from like. The UK and Canada in places like that that maybe could maybe not but don't have a track record of polling here. And then there are. The shadow partisan pollsters talked about the word right like I think that like. You know almost seems like a photo group. Is just trying to poll often enough and they're always in the real clear politics average you know I mean and always show like the same numbers always show. Biden plus two Pennsylvania and minus two admissions always saying it's very weird experience. Why do we even include those kinds of polls being that horror average and in our forecast. Because we are a rule driven society Galen and we have to have actual systematic rules of that. So that I don't have to make a whole bunch of subjective decisions. Right. You know do you use a what do we do this year's we actually kind of house effect adjustment in or model which is that you look at where poll lines up relative to other polls and is consistent bias in the model just weren't right. The house after disinfection like a little bit more were modest and has been in some past years not a mostly contained within. When state rights of perform a group. Has pulled Michigan six times OK you don't suspect you can just that right if it ain't going to poll. Florida for the first time I think apple portable to the Pope for the first time it wouldn't carry over that much. The house effect from other states right. That was a change we need this year we found that actually it's a good change of terms the accuracy of about his what you don't wanna do is you don't want to basically just kind of ignore the local. Pollsters that only pull one state that I do think it like there are. Some pollsters that like. Kind of abuse it a little bit right and kind of like. There are are. I mean good faith is a strange. Thing gate level I am sure that whatever the true faltered group guy's name is Robert something I'm sure he believes account it went down doubt that right. But like. It certainly seems like. You know it real clear politics has like. Five holes in state and went and mr. Volcker. It's just 15 that dudes opinion you know what I mean when they describe their polling process where tobacco use seven different. Ways to collect eight and look at you know the it's a patchwork well when we weave together their. Their electorate right I mean you know you're supposed to it. Infer electorate looks like from. The responses you get at least in part right I don't know. How much of our average and therefore our forecast is based on the use as you call it equals a look at it's it's not that much has never wanted to keep that much weight. And there are a lot of good polls it's probably. You know five to 10%. In the swing states actually a little bit more in the swing states in the national polls is an ordinance races much. Due to get beat it without a gift if you did take the time to subjective Rico through these different polls like this garbage. Do you think it would make our average any matter or just looking throughout history including everything is what kept you close all. I am less concerned. About our average having a role bite. Bias. Because we do include these. The span he polls which this year pretty Republican right the fact that you can't. Predict. The direction of polling biased in your year right meaning that when your computer Democrats when you could be Republicans or computer what the secret to your hero but it's not predictable. That part reflects like the marketplace for. Her for pollsters right. And the fact that trim. Beat his polls went sixteen. Creates more of a market for mr. Volcker groups. And their Rasmussen Reports and the insider advantage is right. You know that kind of have very Republican leaning results as well what about when a sixteen also the kind of mainstream polls were wrong right. So that's part of the market reacting to it you know and I mean. We technically sixteenth so maybe American Iraqi receipt is. I think these polls sock. However they're also if you pulls out here. That still are waiting my education that's probably fairly big mistake you have electorates that are are we to college educated they're probably mean we're democratic nation right. And with full sucked him. You have like balancing stickiness on both sides there. I would rather kind of lit. Trying to manually kick out these polls right and it could be right I mean it could be immediately the thing that keeps these. Error prone holes on both sides alive right is that. You know when every three or four years there is going to be a systematic pulling years in a certain direction right. So rest to secure live off the fat a when he sixteen. And. Other years where we're going to speak their polls and the eight people ignore the fact that they're generic Gallup poll was like ten points and 28 scene right. So you know so sometimes kind of I know it's a marketplace and put it like that right and I trust the kind of overall process for. For how we wait and incorporate poll. It's. Sounds like a proper capitalized. So let's get to use them rapid fire questions from listeners. Here's the first one. From day one figures on the keyboard programming to creating in the final version. The forecast model how long. Teach the presidential models actually quite a bit of work this year's trouble like. Let me it's hard to say it is like also don't like these new polling averages right and kind of entered a process that let me say like. Twelve weeks of work. And where an average of and other things to right where an average of doing forty hours. A week. So probably you know 500 hours of worker in the presidential models hear something like that. Probably another. 100 on the mid term model which is mostly just. They didn't change much it was just kind of collecting this data which takes forever as a tentative and uses. But you know but putting in. 500 hours. On a model where in principle things are very similar in the past we've made some. Changes are unanswered Unix or what I mean it's a lot of its a lot of work. Next question. Is there a reason and at 538 doesn't forecast gubernatorial elections the cycle. And yeah I because it's not a rarity. It's generally not the large population states that have gubernatorial elections in. Presidential years our policies to North Carolina and Missouri and Indiana and so forth. Yeah I just like you know there is a me your costing you might think. Just a maintaining. All these different models right again that it. The house model this in the Gresham analysts say we're. We literally made it like to retreat changes right it's still a lot of work to update all the different sources of information that we need. To debug. Any edge chase races echo wrongly convict a long time but Georgia's special senate election or something right. So. It's just a matter of of prioritizing. Next question what are the odds that a runoff in the door to us partial election determines which party controls. I have looked at specifically and that could be to run ups right injury to actually. I mean we have the senate to keep. Them. So started to show called watch me if you map why. Somewhere in the near ahead of me. Twelve to 14% chance that the senate to come down to it once seat. If if it's two seats which could be then it could be double that represents a twenty or 25%. But you know you need to apple it carrying it a certain share of the vote in the George regular election. Or Oslo where Purdue could get 50% so it's maybe a toll. 15%. Yet fairly high. This is it. A question that requires listeners to imagine the forecast when you go and look on the page you see. That there are all of these balls that get distributed across different probabilities different Electoral College part comes and then they show. Maps. But different possible outcomes. And oftentimes those get screened shot people share them wit us and and we how does this map happen. And really this is just are a very random process and not representative of what we actually think happen necessarily under a fight and win war its crop went. But we got a question how are the maps in the balls warm. I guess we're calling that distribution a balls form. Charles. We choose maps. Such that the overall number of times that a state is a certain color. Matches the forecast right so if Biden wins Texas conditional. Line. Winning the Electoral College. Pitched in a time. Then if Egyptian of the Biden winning maps should have Texas colored blue or whatever. I think that we implement that can sometimes lead to like. Making it hit a bit more likely that some of weird looking mansard. Chosen. The same time you know we delivered the the think that weird looking things can happen right if you can go back and look at you know how accurate are. The forecasts in the sample. I mean we're back testing right. Especially in states that like our poll that often meaning answer is you can get. The occasional stay where your forecast is. Ten points off right. So randomly biting winds South Dakota in a blow out or something where there's been no polls right I mean I think people like people have weird. Attractive it forecast rivalry stuff. Human trait. Place some people at a certain. Rival model have extremely strange ideas. About how certain you can be. About one not weird events occurring we have it's like tiny sample size of data we actually kind of back cast your model. You do come up with fairly weird errors in states now and then especially states are polled very much right. I'm. So you know you're going to have. Weird stuff happens sometimes. Happens less let me ask you can get a close election in the air reduces so they kind of chaos in areas. Palmer happened less office of the maps are lost we are and they were in August. But like you know. You gonna have weeks to happen. All right so before we wrap up I wanna come back to topic from last night actually. We are talking about it to be we got on the topic. The editorial section of newspaper verses. The newsroom I ask you. Actually didn't ask you if you were drunk because you thought that art listeners didn't know the difference. I ask you you're drunk because of how quickly you progressed consider it like three different topics over the course to send it does something like that. And then we started talking about whether or not our listeners know the difference between. The editorial report the news room. I am I didn't clicker poll on Twitter recordings and a whole and these are people who follow me on lessons of this podcasts. Presumably lessons as podcasts 85% of people said they know the difference. In this and Boris you miss him said they did when you really interactive news right. Have really young but I wanted to but I did I do want to come back and say that regardless. Pole we'd stop. Several questions asked gates if we could go back to that topic. And talk about. You know the difference between an editorial report at a newsroom and why it is important in the context. Of that Wall Street Journal story I am sorry me for cutting you off and now hours or opportunity to. Equipment it I don't think people realize I didn't realize this before I went and worked for the New York Times. Is that. They're run by separate people right I thought he was meant to enter an administrative Bill Keller I didn't know that Bill Keller doesn't. Also have domain over the op Ed page to your times. That is a separate products that like that. Dared to bosses basically maybe the publisher. Missiles worker plan I guess they're kind of overseeing things they do have some joint authority. But yeah and a lot of new lutherans. The opinion section and the newsroom. A air firewall and B probably. Compete with another and see maybe kind of hate when other's guts to some extent right. You know certainly and you see a lot of because it's very public now right you see a lot of concerns. For. New York Times number public now they were before words about you look at your time skilled right about like. Stuff that appears on the op Ed page that might seem to validate you know like racist when severe or whatever right. And I you know that can create the conflicts right. You know the fact that I mean. They are very expects to restore rant about that Wall Street Journal. Story and to clarify for listeners what is the story and talking about in the wall street journal editorial pages. I mean it's kind of confusing scandal right. So the alleged story is that sent emails surfaced on a laptop in a Delaware computer repair Stewart. Where Hunter Biden having talking with various associates. About a deal involving China. What do those emails seem to be trying to get. Joseph Biden involved is an you know the big guy whatever his call we get a 10% cut in referred to Joseph Biden I mean right. Later on when the people in that ring. What's his name. Allegedly leaked letter or letters. Right in front of it it is wait for a Tony Bobble in ski now obelisk is meant alleging that like he actually took the additional step of meeting with Joseph Biden. After the election I mean you know what issue here is like. Even if you kind of take everything at face value in even Kimberly council's opinion writer the journal he was taking it seriously. Even errors reported like Biden says kind of capital and no joke might. Right. So. It's a little bit weird that a home. Like it doesn't seem to paint so Biden in that negative. A lot right but then but. The overarching theme here right is that this didn't pass Muster Weir's. People who were actual reporters. It passed Muster waved people who published opinions. I mean both based time. What we saw it yesterday where the Wall Street Journal newsroom publish your story with the headline hunter Biden's X this is partner elect is bothered you about venture. And man. Assault that is former masters this is you know involvement. Court records rebate both journal showed no role for divide between the kind of batting that story down right and they're very specs to resettle here. About. When the truck Fultz were excited. In the kind of tipping opportunity big story. In the Wall Street Journal that they may be thought that would be a story covered by the newsroom. And that game. The Wall Street Journal's newsroom would have a higher degree of Christie's. And office that your post right but it's a little her the totem poles farmers. Credibility with the mainstream media if the Wall Street Journal entertainment's. Like a question of like. Covering the conference would actually kind of covering that allegations right. Apparently. Whenever they saw. The news or was not satisfied kind of wound meeting chapter to another. Sensitive sources that were Lowell portable one which is the Wall Street Journal op Ed page right which has. Lower editorial standards. It's strange to me by the way why newsrooms by newspapers' actions saying that newsroom for newspapers. Have these two entities with totally different sets of standards it doesn't make any sense of I think that if you if I ran I would never really Erica and DoubleTree Austin computers knows when that happens. If I view. Op Ed page. Work and you pioneer any. Many tell people luckily it's like. Newspapers I would make sure they get awesome jobs writing to New York Times in your times newsroom. Remember you're such an awesome great. But these infections which is unsigned. Opinions story is he's like endorsements like why are you endorsing candidates right like. What are you doing and why are indicative publishing stuff Wall Street Journal op Ed page you're newsroom rejected. I mean what are you doing. Right names we do not always agree but this is definitely a place where read dear I'd like I've been saying this for our quarterly agrees that the opinion sections of newspapers. Should not exist I think once upon a time when people who really didn't had any older. Platt former way to interact with the news or rates or express an argument. Were bright in suited newspapers and that's how. They can get published and their ideas could be put out into the public square. Now we have media on the and social media and all kinds of different ways for people who want to serve our opinions and arguments that are not. That'd and driven by evidence. They can do that in other places I think it frustrates. Americans view. The press and the role of newspapers too much by putting it in. The same place today and for me the difference here is really in a newsroom you start we have. Questions. And you go where the evidence takes you in order to find answers and it should be you know of course we all have our piracy is not. Obviously the case. But ot it's a rigorous process of allowing evidence to help you paint a picture what the world looks like for your audience. Oftentimes the opinion section is starting with answers in your job is more to convince the world that the weight use yet is the correct way to see it. And the evidence is not always quite as important. And again I will reiterate what gates had that there are very talented analysts who work in some of these opinion sections. And if their work as rigorous enough you should absolutely just put them in the regular newspaper. Right and I'm not calling person 1950s style I mean we try to. We do have our own mr. Galen. Region I would iris house. And we had her own style that's now style that you know it's not it's that needed carrying us for being honest it's kind of like. Just a style that all the editors and all the writers have I think we have evolved. There are plenty of pieces that I 38. That are argumentative. Right. They're funny piece is that kind it. Taking a point of view. Meters and lots and lots and I think really sharp writing at 538. But second of the voices. From nowhere sty all right. But you know what extent possible because it can be blurry line some times you're starting with the evidence first. And that's. We come to your conclusions right and you have to demonstrate things improve things that things are back checked air check right. And you're not trying to just. Commit similar except maybe some time leasing okay here's a doubles you know next it'll probably right. Some devils out. The case for. What term could win. 4% chance right. That very clearly say. That this is a devil's advocate case right so in clear labeling can can. Can matter. And and her. Just like all. Track notes I mean it's it's a real struggle right it's a real struggle too real struggle obviously you know sometimes to Canada and social media. But this is true for lots of journalists write your credit expected to lean and also there's a payoff for its helicopters and twitters is huge tractor. I mean a lot of journalists. Built huge columns and Twitter drive traffic to their newsrooms. I get jobs because they're good and Twitter right. And I'm look. Other thoughts on Twitter pop I think. We aren't you talked about. The opinion section vs the newsroom as our as the one terminals that topic for that. We will get to Twitter someone time maybe after the election is over but. I hope we've answers sports questions about. Where we regardless conversation last night again I'm very sorry I interrupted and Claire what's with the humidity in amusing drive time. If it better anyway. That is it for now so thank you. Thank you do. And I'll also say before we reviewed the fire at 38 store is up and running. If you want your order to be received between November 23 and November 30 you should please that order by November 5. Essentially you know it's not that shipping schedules so depending on when he ordered changes when your stock. All that is to say that we've got lots of new stuff on the factory store. Two coats sweat shirts with the new designs particularly featuring art fox. And finally caught 80 are. By the facts are IV the fact. It's fighting fox I guess I did a fox sounds more like guy. Like a qualitative judgment moved Friday. You know there's. Near the mid century modern house. Hears you road cult fox road. And there is a copy number 55. Three so maybe it's Revis. That's probably where he lives and when we eventually create a feature film for fighting a fox and we pitch it to our over reports that EDC area and the Walt Disney company that is exactly where Friday fox five fox record. But until then go checked out all of his stuff past 538 dot com slash store. My name is due to enter Tony chow is in the virtual control room cleared editor Curtis is on audio editing you can get in touch by emailing us at podcast at 530 dot com. You can also of course greeted us with any questions or comments if you're a fan of the show leave as a rating review in the couple podcasts to work or tell someone about us. And remember to subscribe to fart duty on YouTube. Base for listening and policies. And yeah.

This transcript has been automatically generated and may not be 100% accurate.

{"duration":"57:17","description":"In this installment of \"Model Talk,\" Nate Silver talks to Galen Druke about why state-level polls and national polls are showing different degrees of competitiveness.","mediaType":"default","section":"ABCNews/fivethirtyeight","id":"73796659","title":"Model Talk: National and district-level polls disagree ","url":"/fivethirtyeight/video/model-talk-national-district-level-polls-disagree-73796659"}