Atop Democratic primary polls, Bernie Sanders and Joe Biden are re-igniting a debate about whether moderates are more electable. Are voters pushing the candidates to the extremes or just looking for moderate alternatives? Andrew Hall finds that moderate candidates are more likely to win general elections, but that they are running for office less often than extremists. The benefits of office are declining and the costs are increasing, especially for potential moderates. But Stephen Utych finds that moderates are far less advantaged in general elections over extremists than they used to be. Partisan polarization means voters increasingly treat politicians in each party as interchangeable, lowering the costs of nominating extremists. Either way, voters are not the main cause of polarization.

Studies:Who Wants to Run” and “Man Bites Blue Dog

Interviews: Andrew Hall, Stanford University; Stephen Utych, Boise State University


Grossmann: This week on The Science of Politics, if moderates are electable, why are we electing ideologues. From the Niskanen Center, I’m Matt Grossmann. Atop Democratic primary polls Bernie Sanders and Joe Biden are reigniting a debate about whether moderates are more electable. Are voters pushing the candidates to the extremes, or just looking for moderate alternatives that they never see nominated? We have a long history of Congressional elections that might help adjudicate the debate. Today I talk to Andrew Hall of Stanford University about his new Chicago book, Who Wants to Run? He finds that moderate candidates are more likely to win general elections, but that they are running for office less often than extremists. The benefits of office are declining and the costs are increasing, especially for potential moderate candidates. Voters are left with only extreme options.

But I also talk to Steven Utych of Boise State University about his forthcoming Journal of Politics article “Man Bites Blue Dog.” He finds that moderates are far less advantaged in general elections over extremists than they used to be. Partisan polarization means voters increasingly treat politicians in each party as interchangeable. Hall’s book finds polarization is mostly due to the changing pools of candidates, not voters.

Hall: The book is about a very basic question, which is we know that the polarization in our legislatures, that is the ideological gulf between Democrats and Republicans, has gone up over time. We’re not really sure, and we might want to know why so that we could think about how to reduce if if we think it’s something we want to reduce. The main finding and argument in my book is that political science has been very focused on voters, and that makes sense because voters play an important role in this process, but we’ve missed the fact that actually a lot of the action in polarization is coming from the candidates. What I mean by that is over time, what I show in the book is, the options that voters have to elect to Congress, the Democrats and Republicans who run for Congress, are more polarized than they used to be. That lack of moderate options seems to have a lot to do with why the legislature has polarized.

Grossmann: He says lots of factors matter for polarization, but we place too much emphasis on voters.

Hall: Even if voters weren’t becoming more polarized, we might still see the legislature polarizing because of all these other factors that are dissuading moderate people from becoming candidates for office. I think of it as yes side by side with these arguments about everything that’s going on in the voter’s side. If you look at what are the arguments out there it’s that people themselves are becoming more polarized, the primary elections are favoring the views of the more polarized, that things like gerrymandering are changing the way voter’s preferences are mapped into election outcomes. Those probably all have something to do with what’s going on, but my point is even if you put those all to one side we’d still have this other thing going on, which is that moderate people don’t want to run for the House or for other political offices.

Grossmann: Even if voters choose the most moderate candidate across all races that wouldn’t have helped reduce polarization much.

Hall: You take the mix of campaign contributions that every candidate has received, that allows you to measure something about their ideological positions and to identify who’s more moderate and who’s more extreme. Now you have that for every district and every year. You know all the people who ran and which ones seem to be more moderate and which ones seem to be more extreme. Then you can ask the following question. You can observe who actually won the race and served in the legislature, and you can compute in the real world how much polarization was there in the legislature in each cycle. Then you can compare that to a hypothetical, which is since we know the ideology of all the losing candidates in every race we can now ask what would have happened, what would the legislature have looked like if instead of electing the person they actually elected the voters in each district instead elected the most moderate person running in each district in each year.

Hall: When you do that what you find is that actually polarization in the legislature would look mostly just like it looks in the real world, even in this best case hypothetical where the voters always elect the more moderate candidate. That tells us I think, it’s very suggestive at least, that there’s more going on than just voters preferring more extreme candidates. There’s actually not that many other choices voters could have made to make the legislature less polarized than it is.

Grossmann: Hall’s work shows that voters still prefer moderates. He takes advantage of close primary winners to obtain a causal estimate.

Hall: A central challenge in this literature is figuring out whether more moderate candidates do better or worse in the general election. There’s people with very strong views on both sides of that debate. One of the things that makes it very hard to study is that of course who runs for office and who’s nominated to run in the general election is not a random process at all. So you end up with this very strong selection bias where in the places where more extreme people run for office they do better electorally because they’ve survived some process that favored that type of candidate. You need to do something to account for this selection bias in terms of the underlying types of districts that have more extreme candidates versus the types of districts that end up with more moderate candidates. There’s many different ways to do this, and in the book I actually explore a variety of empirical methods and all of them arrive at similar conclusions, but certainly yeah the one I’d use in my job market paper and the one I prefer in some ways is this regression discontinuity design.

The idea here is what we would really like is for there to be an experiment where some districts randomly got a more extreme candidate and other districts randomly got a more moderate candidate. That would break this selection bias. In this experiment the districts that have more extreme candidate are otherwise, at least on average, just like the districts that instead got a more moderate candidate. That’s supposed to be the central advantage of running experiments. We can’t run that experiment, but we are able to take advantage of a natural feature of the electoral process in America, which is primary elections, to do something like this. What I did was I looked at a set of very close U.S. House primary elections that were between one more moderate candidate and one more extreme candidate. Then the idea is basically because these elections are so close it’s almost like an experiment whether the more extreme candidate won the nomination or the more moderate candidate won the nomination.

Then now you have a set of general elections that are kind of all comparable in some sense. The districts that have fielded the more candidates are a lot like the districts that have fielded the more extreme candidates. Then you just compare performance in the general election and you see that the more extreme nominees do a lot worse. That was sort of a key piece of evidence for me that suggested there really is some kind of advantage for more moderate people, at least in some context.

Grossmann: But Utych thinks the conventional wisdom about moderate advantage might be outdated.

Utych: I think a lot of the popular and scholarly conventional wisdom lines up quite nicely saying that moderate candidates are probably more likely to win election, the ideological moderates, than extreme candidates based on things like matching up with voter preferences and things like that. What I see is there’s a reason that conventional wisdom existed because early on in the 1980s and probably before, though I don’t have data on that, it was right. Moderate candidates were incredibly more likely to be elected to Congress than extreme candidates. But kind of where I challenge that is seeing that in more recent years extreme candidates are doing better and better. As they’re doing better they’re becoming more likely to win election, and as a result of that moderates are becoming slightly less likely to be elected.

Grossmann: He entered a new field to assess that idea after hearing pundits talk a lot about it.

Utych: This is really different for me when I started it. I’m a political psychologist who studies the effect of language in politics and things like identity and out groups, so this was really out of my wheelhouse when I started it. Kind of one thing that interested me was my personal direct experience of hearing mostly pundits talk about oh this candidate can’t win because they’re too liberal or too conservative. Then I see well why are so many now reasonable candidates for major office who are too liberal or too conservative even though they’ve been elected before. I kind of thought maybe this conventional wisdom doesn’t hold up anymore. I was like someone should study that. Then I convinced myself I was somebody and I could be the one to do it. I decided to just kind of take a look at what was out there. I was aware of Adam Bonica’s data on candidate ideology. I thought maybe I could use that and see what happens.

Honestly I found really big effects over time and seeing big changes in how people were likely to get elected. Honestly probably a lot bigger than what I originally expected. I was thinking of this originally as like a Republican tea party type thing that had changed things, but I found that it’s happening for Democrats too. It surprised me how big the effects were, and maybe that was part of me being kind of new to this area.

Grossmann: He finds that ideologically extreme candidates are more likely to win over time.

Utych: In this paper it was basically using data on ideology of Congressional candidates and interacting that with years, so basically seeing if the ideology of Congressional candidates changes their likelihood of wining over time. What I found was, and this data was from 1980 to 2012, what I found is that as time progresses, as we get to more modern times, strongly ideological candidates, that is candidates who are very liberal or very conservative, are becoming increasingly likely to be elected to Congress, and moderate candidates are becoming decreasingly likely to be elected to Congress.

Grossmann: Hall agrees that the moderate advantage might be going down, but doesn’t think it’s been eliminated.

Hall: Definitely think it’s plausible and possible that this relationship is going down over time, but I guess where I would push back a little bit on the other finding is that I don’t think it’s zero today. You can see that either by running a primary election RD like I do, and Dan Thompson and I recently published a paper that did this for pretty recent elections and continue to find big effects, but you can also see it using some of these other observational methods that I used in my book. Actually just recently, prompted by an email from a friend, I ran a similar observational approach year by year. What you see is some evidence for a decline over time, so I did it through 2012 and it seemed like it had gone down particularly in 2010 and 2012, but nevertheless it was still, the advantage for more moderate candidates even in this observational setup, was still not very close to zero even in 2012. I think it’s plausible that there’s been a decline over time, but I don’t think it’s zero today is what I would say.

Grossmann: Utych’s analysis is a simple comparison of relative candidate extremism with a linear time trend.

Utych: My evidence is relatively straightforward. I used the data, as I alluded to before, from Adam Bonica’s database on ideology and money in politics and where he estimates ideology of both incumbent politicians and challengers based on their campaign contributions. He estimates these ideology scores going back to 1980. What I do is I simply use the relative extremism there. I kind of collapse that to say it doesn’t matter if they’re liberal or conservative, it just matters if they’re extreme or moderate. I use that and I create an interaction term with year. Basically that allows me to say does the effect of ideology on things like winning an election and electoral margins, or the vote percentage, does that change based on time. What I find there is that it does change based on time.

It’s a very simple and straightforward model just using just some linear methods, or in the case of probability of winning a logit model. Very simple and straightforward, just a quick way to kind of test that main hypothesis of do ideological candidates have a better chance of winning over time as the year gets more approximate.

Grossmann: He finds that the trend applies to both parties in both their chance of winning and the margin.

Utych: There was no difference there. They were increasingly likely to win and increasingly likely to have a greater electoral margin. Interestingly there was no difference between Democrats and Republicans. The effect was stunningly similar for both parties.

Grossmann: He does find a year to year variation, but says the main trend is decreasing moderate advantage.

Utych: I’ve looked at the data in terms of some quick scatter plots and things like that, and we do see that we are seeing more extreme candidates over time. We’re seeing not necessarily that they’re more extreme, but just that there’s greater numbers of extreme candidates. This, probably part of it, is that if there are more extremists then they have a better chance of being elected, but the extremists who did run in the early study period they weren’t winning really. There’s one notable exception I found looking through exactly who the candidates were and it was Ron Paul. He was doing well in the 1980s and he was consistently identified as an extremist, but most of the other extremists weren’t really doing well at all.

I kind of tried to look at this year to year and I kind of, when I went through I was looking at year splits to see if things happened. I was thinking especially like 2010. All of those choices seemed very arbitrary to me and they didn’t seem like necessarily the best and cleanest way to present the data. It’s not necessarily a completely linear relationship, but I think looking at the data it’s a roughly linear relationship and one that I could defend that choice of model pretty well with.

Grossmann: He acknowledges that Hall’s estimate is better for causal inference, but says you’d still find an advantage over the full study time period regardless of methods.

Utych: I think the big explanation for the difference between those findings, and I know his work is obviously great, it’s very methodologically sophisticated and I don’t really think there is actually a huge disjoint between those two findings in the sense of what he’s finding over his study period pretty much falls in line with what I might expect. It’s the idea that there is an advantage to moderates over that period from 1980 to 2012 in the whole period. His design kind of forces him to look at all those years together. I think looking at those years separately gives us kind of a more nuanced view of it, a way to say well they’re not all the same and if it’s changing over time it makes things a little more difficult to actually do with those designs, which I certainly think has a lot more causal purchase than the design I have in my paper.

Grossmann: They both use Adam Bonica’s campaign finance based measure of candidate ideology. It’s not perfect, but Hall tries to validate it in several ways.

Hall: The drawbacks that worry me the most are one, the possibility that it creates a mechanical association between appearing to be more moderate and being expected to perform better electorally. Imagine you’re a relatively strategic donor, you’re a pact for a corporation for example, you want to give to winners because you want to have access to people who are in office. We have lots of other research to suggests that true. That will mean that candidates expected to do well will garner more money from these strategic donors, and because the donors are not themselves acting in a particularly partisan manner it will make the candidates look quite bipartisan and therefore moderate and it will create this artificial relationship that doesn’t reflect that true moderate candidates do any better electorally.

I put a lot of time into making sure to short circuit that particular risk. The way I did, there’s a couple ways. One is I had worked on a version of this scaling with Jim Snyder. This is slightly different from Adam Bonica’s approach, although they don’t end up being all that different. What we did was we characterized your ideology as a candidate by asking who did you raise money from in the primary you ran in before you were an incumbent. That we hoped would short circuit some or perhaps all of that relationship because you wouldn’t know if the person was a winner yet or not.

The other thing that you might be able to do, and it depends on what you’re studying, which I was able to do for my job market paper in the book, is use a completely independent measure of ideology, which is the roll call votes you cast in a state legislature if you’re a state legislator. Something like 30% of House candidates are former state legislators. In the book and in my job market paper I confirmed that there appears to be a similar penalty for more extreme state legislators even when you don’t use their campaign finance to scale them. That was an important reassurance for me.

The other thing to note, and this is Adam Bonica’s big point about this risk, is that actually most donations come from individuals who don’t seem particularly strategic in the way you donate, and so the scalings are actually not that weighted towards the somewhat unusual strategic pacts. But nevertheless if you wanted to you could form a version of the scaling that only uses individuals, try to again get rid of the strategic donating another way. When you do that you end up with scalings that are extremely highly correlated with the normal CF scores that Adam uses. Those are all I think important things about avoiding this mechanical link.

The other thing I think you raised, which is equally important, is like is this comparable over time and does how much does it reflect ideology versus how much does it reflect some notion of partisanship. It could be that the way candidates have raised money and the way that donors behave has changed over time, and the scalings are really picking that up rather than any change in candidate platforms themselves. I rely on a couple things to try to address that.

Hall: The number one thing, which is something Adam did more recently, in a followup paper in the AGPS Adam Bonica released a set of scalings that aren’t so much about characterizing the variation in donor behavior, but instead they’re specifically intended to predict your roll call votes as a member of the House if you win office. That very clearly I think gets rid of the core partisanship concern because you’re entirely trying to predict how they’re going to vote on different roll call votes when they’re in office. When you do this, or when Adam does this, you see some reassuring patterns in the sense that the donors for example within the Democratic party that most predict whether you’ll be a more moderate or a more extreme member of the House is like the Blue Dog Democrat Coalition. It really seems like it’s picking up these within party ideological divides. One of the things I do in the book is use that alternative scaling to replicate all of my results.

Grossmann: Hall also has an explanation for why moderates do better. Extremists lead to counter mobilization, stimulating turnout on the other side.

Hall: You have surveys of voters that suggest most voters aren’t super informed about the specifics of particular policies or the positions of their candidates, but on the other hand in the aggregate it seems like more moderate candidates do better. Connecting those has proven somewhat difficult, and so our idea in that paper was both things could simultaneously be true. It could be that voters are not voting entirely on the basis of candidate positions, and yet more moderate candidates are advantaged if something is going on with mobilization. Again studying the House what we find is that when you nominate these more extreme candidates an important part of why they seem to do worse in the general election is that their party goes on to have a lower share of the turnout in the general election, in particular voters of the other party seem to turn out at a higher rate in response to an extremist nomination. At the same time the extremist nominee does not produce the bump in own party turnout that people sometimes predict.

There’s this idea of firing up your own base, and we just don’t see evidence that that works for these U.S. House races that we look at. Now, maybe Presidential primaries are very different. Maybe when you nominate a more extreme Presidential nominee they fire up the base in the general election in a way that we don’t see at the House level, but I would kind of think that the House would be an easier test for that theory just because it’s a smaller context in which a small number of dedicated voters could make a bigger difference, but I don’t know.

Grossmann: Utych agrees, but says cross-partisan swing voters might also help moderates.

Utych: Turnout versus vote choice is challenging. It’s one of those, it’s a weird situation, especially in the context of the current Presidential primary where Bernie Sanders last I read seems to get a fair amount of support from people who don’t necessarily identify as Democrats anyway, though Biden could obviously bring in more ideological moderate independents and things like that.

I wonder about those effects. I’m going to defer to Andrew on that because I think he’s studied it in a really systematic way and it’s compelling, and I’m going to believe him on that. I think it does matter in terms of mobilization maybe more so than trying to convince people, though I have found in another paper through an experimental study that really where ideology tends to matter is moderates tend to do really well among voters of the other party, voters from the other party like moderates more than extremists who are completely opposite in their ideology. That might be a beneficial thing for, especially in this election where Donald Trump there are the never Trumpers out there who might be maybe convinced to vote for Biden, but maybe not for someone like Bernie Sanders.

Grossmann: Utych also agrees that the changing candidate pool is important.

Utych: I think there could be a candidate explanation here of a way that sometimes voters have forced choices between two extremist candidates. They don’t always have that opportunity to pick a moderate. They might have to pick an extremist from one party or an extremist from another party. Certainly that would mean in an election with two extremists obviously one of them is going to win.

I also think there might be something there on candidate quality that he’s getting at, is the idea that if moderates are less likely to run it’s possible that in previous times the moderates who were running were better candidates than the moderates who are running now. It’s also possible that today the extremists who are running are simply better candidates than the extremists who have run in the past. I think the availability and choices of candidates matter a lot for this. I think it’s a reasonable explanation.

Grossmann: Hall says moderates are deterred from running by campaign finance and legislative partisanship.

Hall: I think the number one cost in my mind is campaign finance and the requirements of what it takes to be a viable candidate today in America. You have to be willing to spend really most of your time, most of your waking hours, and a lot of people don’t want to do that. Again, like a lot of parts of the book, I’m really building on other people’s work and other people’s arguments. I think the best evidence that it is campaign finance specifically that’s deterring people are qualitative accounts of people who did run and people who didn’t run for office. I’m thinking particularly of some work by Jennifer Lawless, but others as well that basically just, it’s just logical. I mean you are expected if you want to run for the House to spend something like four to six hours a day calling people and asking for money. There’s just not a lot of people who want to do that as a full time job.

Bear in mind you’re also, in saying that they need to do that, you’re asking them to stop whatever other job they might have. Again, that’s something not a lot of people want to do. Most of my evidence in the book is at a more aggregate level. It’s that indirectly there appear to be these costs and it appears to be leading to this divergence in the ideology of the people willing to run, but qualitatively it is my hypothesis that a lot of it has to do with campaign finance and what a burden it’s become for everyone.

The other cost I think is really big, and here I’m building very directly off of Danielle Thompson’s recent book, is that once you get to office you don’t get to do as much as you used to, particularly if you’re a moderate. The more the legislature has become organized around party leadership led policy initiatives, the less scope there is for individual members of Congress to have an impact. If part of your goal in becoming a politician is to influence the policy process, right now is not the most appealing time to run for the House. You can call that a cost, you can call that a decline in benefits, but one way or another I think policy opportunities and the partisan organization of the legislature are probably also taking a toll on who wants to run.

Grossmann: He finds that moderate state legislators run for Congress only when election timing means they don’t have to give up their current job.

Hall: Because state legislators are such a common source of House candidates, is I study the decision for state legislators to run for office. Again, I’m building here on Danielle Thompson’s work. As a function both of how extreme they are in the state legislature, but also as a function of how costly it is for them to run. What I think is interesting about the analysis is you can take advantage of the fact that sometimes state legislators can run for the House without giving up their current seat. That’s kind of a low cost way to run. If they lose, they still get to be a legislator. Whereas other times the elections are synced and they have to give up their state legislative seat in order to be on the ballot for the House. That’s a big cost if you like your current job.

What I show, or at least what I suggest because this evidence is not as strong I think as the rest of the book, it’s just underpowered in a statistical sense, is state legislators who compiled a moderate roll call record are quite sensitive to whether or not they have to give up their seat. They are much more likely to run for the House when they have these opportunities to do it without paying the cost of giving up their seat. State legislators who have compiled a more extreme record don’t display the same sensitivity. They seem to be mostly willing to run whether or not they have to give up their seat. It’s sort of consistent with this theoretical story that the ideological motivations of the people in the tails are different than the people in the middle. As we raise the cost of running for office, we’re going to dissuade people in the middle from running, but we’re not going to dissuade people in the extremes from running.

Grossmann: And states can actually increase moderate candidates and depolarize their legislators by increasing legislative pay.

Hall: A lot of state legislatures don’t pay very well, and a lot of them have actually altered their pay in one direction or another over time. So you get pretty dramatic variation in how much legislators get paid in different states over time. Again, if my argument holds some water, we might think that when a legislator pays its legislators more that increases the benefits of being in office, it makes it a job that a broader set of people can realistically consider holding as a full time vocation, and it should change who runs for office therefore and make more moderate people more likely to run. That’s actually exactly what I find.

I do two things. One is I do a case study in Alaska because Alaska had a particularly dramatic salary reform where they doubled the salary of state legislators in Alaska. It appears to have led to both a more moderate set of people running for office, and also a more moderate set of legislators. It actually seemed to have reduced legislative polarization. Then I do an aggregate statistical analysis, which comes to the same conclusion. I basically compare over time within states when a state increases its salary who runs and who wins, and I see evidence that more moderate people run and polarization in the legislature goes down.

Grossmann: But voters may be having the same problem as a the scholars trying to study them. It’s becoming harder to distinguish partisanship from ideology.

Hall: It’s incredibly hard to know whether voters are voting on partisan bases or ideological bases. What is the main evidence electorally for this partisanship story, it’s this well known pattern that House district Democratic vote share for Congress is more and more correlated with Democratic vote in that district for President. The two offices, if you the district votes 55% for Hillary, it also votes 55% for the Democratic candidate for Congress in that district. That correlation is held up as evidence that everything is partisan.

The problem with that story, it may well be true, I don’t know, it’s entirely, that pattern of evidence is just as consistent with the story where voters vote on an ideological basis and they’re not given any variation in an ideological menu of candidates anymore. So all the Democrats look one way, all the Republicans look another way, and so if ideologically you’re on the left, you vote for the left candidate because it’s your only choice and vice versa. People are really running away with these correlations between President and Congress vote share and saying everything is partisan. I think it’s entirely plausible, but it is not empirically distinguished from an ideological story.

Grossmann: Utych says that might be one reason extreme candidates can now do better.

Utych: There are some reasons here. One is that some people discuss the idea of redistricting, creating safe districts allowing the average voter in that district to be more ideologically extreme. That’s compelling in some ways. I think that we’re seeing extreme candidates from all types of districts sometimes, so maybe not the only explanation. There’s another, there’s kind of two explanations here that I kind of see to explain these answers. One is that ideology still matters and it’s through polarization that voters are simply preferring stronger ideological candidates to more moderate candidates. I don’t think in research I’ve kind of done since then that’s still in progress that that’s necessarily the best explanation.

I think the explanation is actually that ideology is mattering less and that things like a voter’s partisanship, their ideology and their presidential approval are increasingly impacting how they vote. If partisanship starts to matter more and more, things about voters, if Democrats are always going to vote for Democrats, Republicans are almost always going to vote for Republicans, then we see a situation where the candidates specific traits, like the candidates’ relative ideological extremism, should matter less and less. I think I come down on the side where I’d say that’s probably the most important explanation.

Grossmann: They are both hesitant to extrapolate their findings to the Presidential race. Hall says it may not apply in executive elections, but at least the current discussion does focus on the candidates.

Hall: It may be that the trade offs are different in executive offices than they are in legislative offices. My evidence is pretty exclusive to legislative offices. The one thing though that, there’s a piece of logic, it doesn’t really speak to the electoral penalty, but there’s a piece of logic inherent in the legislative work that I’ve done that I think is relevant to the public discourse about these things, which is I really do think the right way to think about ideological positioning is through the nomination of different types of candidates in primaries. You hear a lot, the way you phrase it I guess is exactly how I would phrase it is they face a choice, the Democratic primary voters face this choice between these candidates and they need to decide whether to nominate a more extreme or a more moderate person. Whereas a lot of times in the public discourse you hear things like the Democratic party needs to move left or right.

I think sometimes people have in mind sort of almost like a European style thing where the Democratic parties are going to write a platform and it’s going to matter for something. It just doesn’t as far as I can tell in the U.S. A party moves in the U.S. when its primary voters nominate someone different. I think it’s exactly the right question for us to think about is who should each party nominate if they want to win the election. My personal view, I mean I would guess that more moderate candidates have an advantage, but I don’t think my evidence speaks to the Presidential race very cleanly.

Grossmann: Utych says Presidential candidate effects might be larger because they’re so well known.

Utych: I think with state legislative races we’re probably likely to know even less about the candidates as voters than we would in a House race, which means increased reliance on things like partisanship and I know some work out there, the Hopkins Rogers doing things on state legislative and Presidential approval, showing that that has a big effect. I think this is probably happening at the state legislative level too if I had to guess because they’re less known. If people’s partisanship matters more, these candidate centric factors probably matter less.

The Presidential elections are a lot, they’re a lot tougher to get at because I actually think candidate centric factors in Presidential elections probably matter a lot more. You get into that Biden versus Bernie debate and you have two candidates who are incredibly well known, one of whom is much more of a moderate and one of whom is very liberal, and you kind of see those things and how does ideology impact that. The question is in a primary I actually have no real idea what’s going on. I tried to look at this just the other day on a whim at primary elections and tried the same type of interaction with time and likelihood of winning, and I found pretty broad null effects on that.

It doesn’t seem like ideology is doing anything different in primaries than it used to, with a caveat that that is incredibly preliminary. It’s something like two days ago I decided to look at, and then I shrugged my shoulders and was like well all right that’s happening. But in terms of that in a Presidential election with it being so much more candidate centric that I think these candidate traits probably will matter somewhat, and they’ll probably matter considerably more than they matter in House elections.

Grossmann: He says electability arguments do matter, especially in close races, so he does not expect an advantage for ideologues anywhere.

Utych: I don’t know that I think it will reverse, but I do think those ideas about, I have some experimental evidence in another paper that’s again the one that’s still in the process, that paper kind of shows that it actually is these electability arguments do matter. When a candidate is moderate, people tend to believe they’re more likely to be elected than a candidate who’s ideologically extreme. That does have some impact on how they’ll vote. It’s these ideas of yeah I think this candidate will win. I think that’s there, and I think as much as I kind of say oh the big takeaway from this paper is that moderates are no longer advantaged versus ideologically extreme candidates, I don’t necessarily believe that that’s true in every instance.

You could imagine a situation where we have a relatively Republican House district and we have a Democratic candidate, probably makes sense for that candidate to be more moderate than ideological in a situation like that. I don’t think moderates are necessarily disadvantaged everywhere, and that electability argument is true in some circumstances. The question is how true is it in every circumstance. That’s something I don’t really have a great answer to. I think it matters maybe more on the margins than it used to, but it probably still does have some effect. Democrats, a lot of Democrats certainly seem to be buying into this with Biden as the electable candidate.

But in terms of the trends changing, I think we will see some kind of change. The explanation that I favor on this is that it’s not that people like ideological candidates a lot more, it’s that ideology doesn’t really matter as much. If ideology doesn’t really matter, we’d kind of see a flattening where in the extreme where no one cares about a candidate’s ideology, they’re just voting based on their partisanship. It never really gets to a point in the explanations I find compelling that ideologues are going to be advantaged more than moderates.

Grossmann: And they’re still might be a moderate advantage in marginal districts where it matters most.

Utych: I think it absolutely could be those marginal races where a moderate is advantaged. Kind of taking some findings I have from experimental data is that really big focus on it’s the out party likes moderates. Republicans like moderate Democrats a lot more than they like liberal Democrats. If you’re in the type of district where you know you need to get a lot of Republicans to vote for you to have any chance at winning, it might not benefit someone to be ideologically extreme. It might in fact benefit them to be a moderate. I think that’s maybe the future for moderate candidates is running in these marginally districts. That’s not to say that being a moderate isn’t important then because those districts are huge. It makes a lot of sense that those are the ones you need to win.

It doesn’t matter who’s running in New York City. We know generic Democrat, whoever wins the nomination is going to win that election barring something truly absurd. But in districts where you see, like in Nebraska, Kansas, Iowa, places like that where we see these, we saw in 2018 a lot of Democrats winning districts that are Republican leaning districts, and we saw in 2010 Republicans winning some districts that were Democratic leaning districts. I think it helps maybe to have a more moderate candidate in those instances.

Grossmann: Hall and Utych also agree that voters might be playing a role in polarization, but don’t warrant full responsibility. Hall says we can’t conclude voters wanted polarized politics.

Hall: One of the reasons I think both parties have polarized to some extent is that voters have just not been given a lot of opportunities to elect different types of candidates. Now that could have to do with the primary system and how its changed over time. It could have to do with voter preferences, so primary voter preferences for example. But it doesn’t seem, I don’t think we can conclude just from the fact that the parties are polarizing that it’s what the voters wanted.

Grossmann: And Utych says voters are reacting to changing parties.

Utych: Falling on the lines of the polarization debate, I fall in the camp of what I’ve observed and what I believe is that it’s happening. Maybe not as extremely in the public as it is in Congress or within the parties. That would suggest that both of these things are happening. Voters might become a little bit more ideological, and Republican and Democratic voters might becoming more different, but certainly they’re faced with a choice between two parties who are obviously very different. They obviously present very different policies, very different ideologies, and they’re kind of left with a choice. It’s an idea of do I pick the one that’s closest to me even if they’re kind of far away, the answer is usually yes. I think that’s a big driver of it.

Grossmann: Hall’s worry is that we’re now in a cycle that’s hard to escape. People dislike politics and don’t jump in, leaving it to the extremists.

Hall: I worry a lot about our government falling into sort of a pessimism trap where people don’t have faith in our politics, and that leads a lot of good people of whatever stripe to not want to become politicians. That actually creates a feedback loop where then because good people don’t want to become politicians people think politicians are not good people. You can get into these very difficult to escape from traps once that’s true. I think in some sense we’re already there. I think that’s why for example Americans prudently in my view don’t support paying our politicians more, don’t see being a member of a Congress as a particularly prestigious job, don’t see Congress as an effective institution. The problem is that once you’re into that set of views it’s actually very hard to get out of them because you need to jump to a different situation where we like our members of Congress, and therefore we make it a prestigious job that we’re willing to pay people to do, pay people well to do.

I worry that in some sense I think what we’ve been seeing over the last two to four years is one last effort to encourage a bunch of different new people to run for office. If it doesn’t succeed in changing the state of affairs I worry we could fall into a more permanent deficit in terms of the people we’re able to convince to be politicians. It remains to be seen what will happen.

Grossmann: Where do we go from here? Utych’s next step is to look at why ideology matters less relative to partisanship.

Utych: What’s next for me in this agenda is I’m trying to, as I’ve alluded to, another working paper I have trying to focus on why it’s happening, giving kind of an explanation from a voter centric angle saying that voters are making decision based on their personal traits, things like their ideology, their partisanship, how they feel about the President. This gives less room for candidate qualities like the candidate’s ideology to really matter. I test that through survey data and through experimental data, and find some evidence that that’s happening. I also find evidence that using Adam Bonica’s ideology data to predict someone’s vote choice is increasingly has less of an impact over time as well. Kind of looking at it on voters in particular rather than broad electoral outcomes.

Grossmann: Hall’s next step is to turn away from polarization to examine the candidate pool.

Hall: Right now we’re massively over valuing studies of partisanship and ideological polarization. I say that as someone working in that literature. Not a day passes where I don’t see some kind of hyperbolic claim about how affective polarization or ideological polarization explains everything that’s going wrong with our politics. I fully expect the next turn to be a turn away from that topic and towards other pathologies in our politics. What I’m really interested in is sort of what I was hinting at in the last answer is how do we get “good” people to run for office. Of course it’s super hard to study because what does it mean for someone to be a good politician. It means a lot of different things to different people. But I’m interested in developing techniques for measuring who are candidates are beyond their ideology, and sort of how effective would they be as legislators, would they work hard, would they craft policies that are popular or more efficient, more effective. I think if we can solve some of those measurement challenge we can answer some fundamental questions about our elections that go beyond just ideology.

Grossmann: There’s a lot more to learn. The Science of Politics is available biweekly for the Niskanen Center. I’m your host Matt Grossmann. Thanks to Steven Utych and Andrew Hall for joining me. Please check out Who Wants to Run? and “Man Bites Blue Dog,” and then listen in next time.

Photo credit: Gage Skidmore under CC by SA 2.0.