Who hello and welcome to the five thirty politics podcast, I'm due drink, I'm Nate silver and this is model dark, then you have to resign are now praised into the model and we ve gotten some limited polling also, yesterday Thursday, the possibility that nobody winds and outright majority of pledge delegates in the democratic primary briefly overtook Sanders to become the likeliest outcome in the model currently Sanders and nobody are tied at eight thirty, I've been percent likelihood Biden is down to a thirteen percent chance of winning a majority. Bloomberg is up to eight percent. Buddha judge is at four percent, and Warren is at two percent so neat what happened? Why has the likelihood that no one wins a majority increased over the past few days? Well, you kind of have a interesting result in New Hampshire which reflect

also- and I think the poles who seen since New Hampshire, on the one hand, just getting away from like, like, I think, narratively in terms of the way the election was covered. I think Bernie should have gotten the credit as the headline for winning New Hampshire right, but you peel back a little bit deeper. You say: okay, Bernie got twenty five. Seven percent of the vote. I think it was actually a little lower than we expected based on his poles in a state that, should have been one of his stronger states, a Bernie wine, but brain doesn't look particularly strong. On the other hand, everything behind Bernie is a total mess. The two candidates New Hampshire. Apart from burning Navy people, a Chechen Amy Club Bashar, are far behind the pace in poles. Meanwhile, Elizabeth Warren, who had the opportunity to build a broad coalition and Joe Biden, wound up in a single digits in New Hampshire meanwhile, you have Michael Bloomberg, continued to rise in pole.

It is actually leading now samples of a couple of Super Tuesday states around fifteen percent and national poles, he seems likely to get at least a fairly decent chunk of delegates on Super Tuesday and so and so that I would make a contested convention or no majority more likely, it's going really well from the standpoint of of chaos. I guess- and we should say like you know, Bernie has a very, very plausible case to win the nomination by himself with the majority there's another thirty seven percent chance and he is doing spectacularly well. Like doing way there than anyone else near lot of ways. These could break in his favour. Cutting voters said you know what I don't want to test the convention. I want I'd rather take Bernie. It is not my first choice, but still I mean this is like this is on them: the chaotic side and by the way, one more thing to nobody on the major Canada has dropped out yet right, typically, you might expect for there to be dropouts after this. You know you might expect

Frankly, you know, Biden or Warren might have dropped out of there. So disappointing in New Hampshire or Club Bashar after finishing fifth place, A neighbouring state for her might have dropped out and said she stated from a single good. I stated now. I have their place little bit of a mental, so you ve had The minor can drop out, but none of the major Canada dropped out yet an everyday that occurs like I'm running the model, if assume every candidate states in including through Super Tuesday. Then no one chance would go up. I think because it makes it really hard we can t get it Georgie of delegates later on. I think you may have partially answered this question, but I did Bernie Sanders chances of winning the majority of delegates fall from the forties to the thirties after winning in New Hampshire. Was it simply under performing his poles because the model is smart and the model did Bernie a win New Hampshire, two or three times and unexpected that,

those times you one New Hampshire. He would often win by a clear margin. Then he won by so yea actually managed to win the state well underperforming his poles Abed, which had been. You sure you shall, if you, when you can ever perform your pulse, that's a basic rules, either there are several things. Pseudo rub: murmured like again the lack of people. The lack of major can is dropping out, hurt senders a little bit some of the minor candied Stroppy may actually also hurt him weirdly, because it means that you actually have it makes it easy. Now for someone to get fifteen percent if there are non Bernie can't eight right if you're not having Michael Beneke is one percent or Andrew Yang at his four percent or whatever else it makes it easier for people to Chechnya and fifteen percent in some future state which what you need to get delegates for example? So it's not a fairly settles. The also like the hand

of post new Hampshire polls that we ve seen and there really early three, as this recording there's one national Paul one in Florida and one in Georgia. That's it. Nepal has not been particularly strong for sanders. The national did show him gaining, but the Florida and Georgia all showed him kind of mired. In the end you know in the mid, teens or lower. I second Florida he was at ten o clock is a weird, safer centres, because there are a lot of all people and, despite Bernie Sanders being old old people do not like Bernie. Young people are Bernie Sanders, so you can imagine for it is not a great state for him, but still like it does not look like he's getting this big, accelerating bounds, despite winning kind of one and a half. If you will the first two states you mentioned that floor, Nepal were actually Bloomberg is leading and right now Bloomberg is on the cusp of overtaking bide and in our national pulling average. So how has boom

rise affected, the model, maybe other than increasing the chances of no one. You know where has he taken support from otherwise I mean it pretty clearly does seem to be from Biden. If you look at the state, where Bloomberg as strong a kind of actually is among older voters. It is african american voters. To some extent it is a little bit in the north Is probably the model think so to get my polling to the northeast? Was the states are pretty back loaded in the calendar, so it seems like a fairly direct flow of Biden. Voters saying I have doubts about Joe Biden, but instead of going to Bernie, which you know you might not expect, but they both do well among working class white voters, for example, but instead of going to Boerne or Buddha statues. One. These states are club a shower alive and is going to to bloom bearing- and I should say this is something that, from my point of view, where the models point of view is something that,

is weird and unusual and kind of is not supposed to happen. Corner quote: well you're not supposed to be able to get a balance in an early stage if you're not competing in that state, but somehow the teams have happened for Bloomberg, and so the model has gone from very sceptical a Bloomberg. To not so sceptical really has such an unconventional campaign Limburg is waging created complications for the model, given that the model is obviously based off of historical data that doesn't have any historical president's like Bloomberg. Yeah finish that really I mean I've got back and forth. Trying to scrutinise. Ok, what worthier assumptions about Bloomberg and- and I think people makes them assumptions- baby are wrong about Bloomberg right. These well he's rising the pull so far. So therefore for continue to arise right, I think people believe you're not may act slightly like believe,

Even with his Bloomberg data to people may slightly overrate. The importance of money in the primary Limburg has gradually been gaining from, like four percent to eight or nine percent right, but this shift from eight to nine percent to nine fifteen or sixteen percent. Is something different and that has to do with media attention. Around Bloomberg. That has to do with, I think Democrats not being very comfortable with other catch in the field. I'm sure the fact like he's anemones mine, because until the other time helps a lot to write, but this isn't like a slow, gradual, build up based on agitate, it's a slow, gradual, build up, and then eight rather sudden bound and our model says. Ok, we'll is always some chance of anyone bouncing some Cisco chance one. Whatever right, but you're not supposed to bounce after not competing in states underhand. We don't really have any data on what happens when a cat aid is not competing in states. In fact, when you're not competing in a state,

Usually it's because, like you know, you're gonna lose, and so therefore you try to lower expectations. That's not a case like Berger, like literally, he got into the re so late. That, although there were some discretion like South Carolina, I think, but like literally gotten the re so late that, like he missed filing it's a kind of has a excuse not to compete. So no, seventy case, where it's kind of violating the assumptions the data we have in the minors? I could get pressed him for it with with all that said, you know if you have something is being quite unlikely to her, then that you, because there are multiple barriers at a canada- has to break through or an event has to break through first, for it to happen I saw Bloomberg. On the one hand, clearly he is much more likely to be many than the model would have thought as of how many weeks or months ago, at the same time, he still has a pre difficult road ahead of him. He

probably gonna becoming from behind. After Super Tuesday, it's maybe they are for me, get a majority. Still we have a one in fifteen chance of that getting the mild. I was fully kind of caught up to two Bloomberg right. If you go to confess to mention it most likely nominee. I dont know right. We have a lot of endorsement from democratic lawmakers for Bloomberg. If you get out of the realm of things at the motto considers you know, Bloomberg has gotten mostly positive, attention because in number one parsing he's adds number two he has kind of competing quote unquote. So we had really kind of taken the scrutiny that, if reminder, might get number three kind of cable news talking heads love him because he confirms their prior is about a technocratic, moderate clay, but that's gonna change, you're gonna have Limburg compete, maybe in one or more debates. You gonna have scrutiny of his record brown issues involving policing and race,

formerly being a Republican. You have a lot of opposition research. Dropping on him, you gonna see is able to clearly compelling speaker. I mean one thing that we do know from talking about a mile off on the primary is like the more expectations rise, the more risk you have if you fail to meet them right, numbered all. The sudden now has a merely pretty good super Tuesday. Let's say that he gets seventeen percent in average. Trust You stay states, maybe winds, one state alot of second and third, and if you fourth place, finishes right, it's actually perfectly fine and I put him on a path to win. Some delegates is kind of intermodal x like that might not be seen as a disappointing performance before we wasn't aggressions. I do want to ask about Buddha, judge, of course, keys performed well in the first two states, winning the most delegates legally in Iowa booted edges only at four percent. So I assume that debate, of poor

showing so far in the pulling in South Carolina Nevada and national hme. Do expect that to change much. Is there any sense that he's gaining traction? I mean you know he's gotten up to eleven percent or so in national poles, which is fine, but he's probably going to be helped by. We know a million about Nevada, which is of a state where it, when assumes where he's gonna do well, but no one really knows right, but yeah I mean look coming. The model is based basically on pulling. It does account for actual results. Obviously, in terms of geography went so far in terms of trying to calibrate sit, we don't have a lot of polling, but like But you know Buddha church is an eleven percent nationally. You probably expect both Nevada and South, I had to be below average states for him. We only have one posting Hampshire pole. If you see much opposed this week or next week on where we can have met, fifteen percent National Olympics are pretty decent,

size, difference right that could easily double his eyelids were more our model and by the way we say it's actually five percent. This update, not four percent. That's jets! Many a majority. He has a eight percent chance hitting a plurality. Which is higher and obviously when you have a thirty seven percent chance of nobody, winning a majority does. All mean that, like literally we wind up in a work zone where there's no nominee, someone gets nominee It often will be Buddha, chatter, Bloomberg or Sanders or binding Warren or Club Bashar addition, the chance you have some one from outside, come in and when God and also closure is still not really on the board. She has about a point. One percent chance warrant at two percent chance will see if any of them now given day or point. Let me let me remind people that the model is making assumptions about the bounds based on very little polling data right, so some sounds like the times when people are most interested in the model. Rapturously is when the model is kind of weakest. Because

release. Data work with that's relevant, foregoing, right we just had a lot of data. I'm New Hampshire and now we actually have the result. So that's all settled, but we don't have any data, basically for Nevada, so we're in voting pattern a little bit and we will. I should mention prisoners. We're gonna record are regular weekly pie cast on Tuesday of next week because of the holiday. So hopefully we will have some data by that. Hopefully, hopefully, hopefully I think we should get into round Tuesday, but there's all kinds of ways that pulling Nevada is tricky, and maybe that's also part of the reason that posters haven't jumped on it coming out of New Hampshire. Is there may be a little scared of power Nevada D, dipping their toes and that potentially difficult spot to Paul. I want to get too listener questions, but first today's podcast is brought to you by Super Creetur. Hiring is challenging, but

one place you can go, we're hiring is simple fast and smart and growing businesses connect to qualified candidates, zip recruiter, quotable, cofounder, Gretchen he'd never experienced how challenging hiring can be after unsuccessfully searching for a new game artists to grow her education.

A company, but don't you switch to zip recruiter and saw an immediate difference, and you can't do by signing up for free, zip, recruiter, dot com, slash five three eight zipper critter doesn't depend on candidates, finding you it finds them for you and by using zip recruiters screening questions to filter candidates. Gretchen found it easier to focus on the best ones, then find the right one with results like that. It's no wonder: four out of five employers who post Unzip recruiter, get equality candidate within the first day, see right. Zipper. Critter is effective for businesses of all sizes. Try zip recruiter for free at this web. Address, zip, recruiter, dot com, slash five, three, eight again: that's zip recruiter, dot com, slash five, three: eight zero critter, the smartest way to hire today's podcast is also brought you by light stream.

How much money? Are you paying an interest on your credit cards every month too much, then? Why not consolidate your credit cards into just one payment at a lower fixed rate and start saving money? It's easy with a credit card consolidation loan from light stream rates or as Lois five point. Nine five percent, a pr with ATO pay, get alone from five thousand to a hundred thousand with absolutely no fees, and you can even get your money as soon as the day you apply. Light stream believes that people with good credit deserve a better loan experience and that's exactly what they deliver. Just for listeners apply now to get a special interest rate discount and save even more. The only way to get that discount is to go to light stream dot com, slash five, three eight, that's ally G Hd asked he are you a doctor five hundred and thirty eight subject to credit approval rate include two point: five percent, auto pay discount terms and conditions apply and offers are subject to change without notice visit, lightstream dot com, five thousand three hundred and eighty four more information.

We're back and we got a lot of questions from westerners and will start with Katrina. She asked: how does the but all respond to dropouts, notably Yang and Bennet, dropped out on Tuesday night. So what's the model do with that information? it takes you chances down to zero or our technical. You can actually accumulate delicate still right of Bernie Sanders dropped out now he might still in some delegates have stay in Vermont, for example, but you know there are fairly subtle, knock off effects if you drop out Why is that is always a residual chance that someone would have had a bounce later on that can be eliminated. To is like it affects the vote, shares rather candidate. So I guess I met you before you have to get fifteen percent and a state or district to qualify for delegates. If you'd know of sudden have have gang getting three or four percent now he's out and Bennet

Patrick combined for one or one and a half per cent and pulls actually can help a little bit to have more vote to go around among the still living. If you will candidates now, if you have what we consider major candied strap out, which mean someone who is pulling above five percent nationally at the time the drop out, then the model makes more interesting assumptions where it tries to guess where that support will go so, for example, if worn were to drop out, then the model assumes that probably looking at her support, probably That goes to Sanders and Buddha chat because actually, among the college dictated by they both dwell but a more and meanwhile among liberals, obviously Bernie

or undue well, so more and dropping out might help Sanders immunity at the expense of Say, Biden and Bloomberg, for example, but the model is trying to forecast the chance that people drop out and when you see for it people in South Carolina, I think the model currently Bernie favoured in South Carolina. That's probably has expired, might drop out before softer China, where is it thinks he's very little chance of burning, will drop out. So additional on Biden competing South Carolina, then its potentially much more competitive. There thought it next question. How does the model account for undecided voters which, at this point around fifteen percent? It doesn't do a term with that that there are some assumptions built in about like when they feel is more unsettled, theirs,

relativity in the model and one of the ways that it accounts for how settled the field is to look at the number of undecided voters I'm simplifying here, a kind of looks like you know, takes all the can. Its vote shares hacked something called diffusion index, and that gives you an indication of how much. Might change so right now, both the fact that you have a forerunner of undecided and the fact that, more importantly, that, like nobody, is above like twenty three percent national poles means that things are like like going to be fairly wilder. Essentially when you get into a two way raise that can be a lot steadier. Some of that fifty five, it comes at forty five. Those can be holding patterns but like in this race, you still have. The potential for big swings in the polls is, to some extent I suppose, work So the next question is from Alex who ass a pretty black question and it is what's the point of a model that takes such wild swings. Obviously, the model have shifted a lot since Iowa and the question is what information?

can I glean from today if I expect that it could swing as wildly say tomorrow These things are very while they think this wings are saying that I was really important and that whatever happens in Iowa, has fairly profound effects in the rest of the race. You know we're not talking about a case where you have information revealed incrementally right. You have certain times when Key information is revealed. I owe Tuesday New Hampshire, occasionally when you're looking for poles after those states poles by ship things a lot you know. So it's like this is resembles the actual event. It's not gradual, incremental anything p, Have like weird assumptions about like how model should behave. It does not linear, a good model or show nonlinear effects. When major news happens so basically Alex you can learn some life lessons flourish in life. This model loss of life, lessons that things kind of don't shift in it gradual away. Sometimes big news happens right,

you get fired from work. You know your wife comes in her husband says it's good earth right. You know you get sick girl, I mean in oh, yes, you mean a lottery mean big events happened that shift your life in a hurry. And that one day where that thing happens hey matter more than the next five hundred days combined. I that's an interesting way of putting it. I like lifeless next rush it. How did you test the model on past elections? How well did it perform, but I guess the first growl you tested on past election. Obviously any model is based on past data and we look How well calibrated was so meeting other can set at ten percent chance of winning. Did they have a ten percent chance in the backed has an answer

yes, it was well calibrated at the same time like this is a real protest for people who are working on models. People overrate the importance of be careful about this. The goal is not to back test well, the goal is to predict well, and those things are highly correlated most the time to the extent there not correlated, then you shouldn't sweat too much like how well a model back tests, because the extent you're saying ok, this may have predicted well the past, but it's probably a coincidence or whatever, or this head structure Let's see the mile or a kind of violates our theoretical assumptions. Right, like you know, model should back test pretty well or else. What are you doing right otherwise, like United and using like data from the past in for a mere year priors, but a model that back test super well verses?

backed has pretty well. I'm actually not sure that you want the one that back tests very works. It could be what's called over fit, meaning that, like it using, we passed it too rigidly and it won't handle new cases very well, so I think you gotta embrace uncertainty going forward. Well, you also have to make certain structural assumptions with the model. Like a mean thing in Bio, again way. I did say price. We last budget that will get catch a flight here in a second. What I think tricky about this model is like, because it s sequential process you have voters and party actors, kind of acting to information in a sequential way. You kind of almost crying like anticipate group behavior. You know what I mean, and so I might think it's ridiculous that everybody is talking about Bloomberg as the Saviour. If Biden collapses right,

and the model might say why talking about Bloomberg, ease at seven percent he's not competing in Iowa Hampshire, New Hampshire Right, but the fact that people are talking about a kind of ask me you'll, like it could re. If I itself, you know what I mean So the economics of Lubeck, weird kind of one year using a model that kind of check the conventional wisdom, but the commission wisdom actually can affect what happens a little better. Right you're like ok. Well, why would call Bashar get about after finishing in third place after finishing in fifth place? say right next door to her right. Well, people talk about that than even though about the kind of inconsistent that could still happen. Potentially such that interesting nature, remodel like this one, so don't under rate? the media narrative, whether or not it alliance with reality. Well, I think I think is our. Besides underrate the media narrative, I mean first Alameda Nervous. I was stupid, but you can like you, can use

more emanates, a real effects. I never effects life, obviously right, you're, saying sort, and we are trying to forecast immediately We're saying that if you want to state the especially for the unexpected winner, then you get about connection. Some of the balanced budgets have been pretty decent. So far quite decent. The one major Kelly's he other clearly in the model so far is is Bloomberg in its you're Bloomberg, getting as big bounds despite he was on the ballot in Ireland, got like open one percent of the vote, or something like that right. That does not happen. The model thought could could happen right. It's ok when there's a early state voting. There is volatility for all became too are competing, but not for Africana who isn't incomplete, You know what I mean an down, so that's been. I think that one assumption so far that like clearly is wrong and clearly like and leave it probable that, when the nomination machine syrup but like even if he didn't denomination, even if he's Taiwan or Mark, even then

they go back and say. Ok now we have some data on what happens when a when a country is not Putting in a state- and they can still be in affected- not infected, so be affected by the outcome in it. State. So now we have some data on that for twenty twenty four, I suppose, but sorry, if you're Michael Bloomberg fan now I liked what model lot better. It was definitely under way to him before our raw Imagine you drive to catch a flight where you have after you share with listeners going The cargo, the NBA all star game there are well enjoy, have a great weekend and was the on Tuesday. Thank you. Ok, my name is Galen droop Tony Chow is in the control room. You can get in touch by e mailing us at pod concept: five, thirty, eight dot com. You can also, of course, tweeted us with any questions or comments. If you're a fan of the show, leave us a rating or review in the apple pie, a store or tell someone about us thanks for listening and will see us,