Category Archives: Commentary

Who Will Win The Senate? A Primer on Midterm Forecasts

With the nominating phase of the congressional campaign just about over and the midterm elections less than four months away, you are going to see an increasing number of predictions, prognostications and more than a few statistically-driven forecast models purporting to tell us how the Republicans and Democrats are going to do in the House and Senate. Accordingly, I thought it might be useful to present a short primer on the types of forecasts you are likely to see, so that you can makes sense of the predictions.

Generally, you will encounter three types of forecasts. The first type are individual race-specific predictions made by the veteran handicappers like Charlie Cook, Stu Rothenberg and their associates. These predictions use a combination of on-the-ground reports, opinion polls and other bits of evidence to divide the field into safe, leaning and tossup (or their equivalent) races. With their descriptively detailed updates focused on the competitive races, these types of horse-race forecasts are in many respects the most interesting to follow, particular when they look at the high-profile Senate races. Right now, for example, Cook is predicting that the Republicans will gain between 4 and 6 Senate seats. Rothenberg puts the number at between 4 and 8. The Republicans, you will recall, need a minimum of 6 to regain a Senate majority so both handicappers see a Republican takeover as well within the realm of possibility. The implicit assumption in these models is that individual races can turn on factors idiosyncratic to that particular race, and thus the most accurate prediction depends on understanding these myriad influences. In short, if we want to know which party is going to control the Senate, you need to build from the bottom up by aggregating the results of the individual races.

The second type of predictions are those produced by the structural forecast models developed by political scientists. In contrast to the handicappers like Cook and Rothenberg, these models eschew any interest in local detail in favor of macro-level factors, such as national economic growth, the president’s approval ratings and the number of exposed seats, to generate a prediction which is usually measured in terms of how many seats will be lost by the president’s party. The assumption built into such models is that fundamental national tides affect all races and thus all forecasters need to do is to measure those tides to generate an accurate prediction. To the best of my knowledge, Edward Tufte constructed the first such midterm House forecast model back in 1974 that was predicated on only two measures: the president’s approval and the annual growth in real disposal personal income per capita. In effect, he was modeling the outcome of the House midterm races – specifically, the share of the national popular vote the president’s party received – as referendum on the President’s performance.

Since Tufte’s pioneering effort political scientists have generated dozens of such forecast models, with most of them trying to predict the distribution of House seats between the two parties, rather than the overall vote share. Some early models focused solely on economic indicators. But most are predicated on variants of Tufte’s model, with measures for presidential approval included. More recent versions include additional variables. These might include “seat exposure” (which party has more seats on the line); a “surge and decline” variable (the idea is to capture the effect of the decline in midterm turnout based in part on how the parties did in the previous presidential election); and a time in office variable to capture the waning influence of a party that has held the presidency for a long time. The most recent innovation to these models – and one that I will discuss in a moment – is to include a generic vote variable based on national surveys that ask respondents which party’s candidate they plan to vote for in the midterm election.

In assessing these structural forecasts, you should keep two considerations in mind. First, most of the models are based on midterm elections occurring during the post-World War II era. So they are predicated on a very limited numbers of cases – 2014 is only the 17th midterm election in that period – which means that even the most accurate predictions have a very wide margin for error. In assessing the prediction of a particular model, you should always look to see if a prediction interval is provided in addition to the predicted seat outcome. Moreover, as Ben Lauderdale and Drew Linzer have cautioned in their critique of presidential forecast models, there is a tendency with a sample this small for modelers to over-fit their predictions by basing them too closely to the particular elections studied. Nonetheless these models are theoretically the most interesting because, when done well, the modelers are very explicit in explaining why midterms turn out the way they do. So we learn the most from these efforts in terms of understanding the dynamics driving election outcomes. Note that these are one-shot deals – once the numbers are plugged in, a forecast is generated and that is that. There’s no updating based on new data.

The final set of forecast are what might be called mixed models. Typically, these start out with a variant of a structural mode (a prior, to use Bayesian terminology), but then the prediction generated by that model is updated based on race-specific polling data. By the time Election Day rolls around, most of these mixed models will be mostly poll-based, which means (as Drew Linzer demonstrated so effectively in the 2012 presidential election) they are likely to be very accurate. These are the models featured at the Washington Post’s MonkeyCage’s Election Lab or the New York Times’ Upshot site. The basic idea behind these efforts is that if you want to know how people are likely to vote in the 2014 midterm, you should probably ask them and incorporate their response into your forecast.

You might think these mixed models would all generate basically the same forecast. As of today, however, they are not. The MonkeyCage, for example, is giving Republicans an 86% chance of retaking the Senate. The Time’s Upshot, on the other hand, gives the Republicans only about a 59% chance of taking the Senate. Why the difference? As the MonkeyCage’s John Sides explains here, it is partly because as of today the Upshot is likely weighting the polling data, which is a bit more favorable in some states to the Democrats, more heavily than is the MonkeyCage. So, which forecast is more accurate? I don’t know, and neither do they! But that is probably beside the point since it is likely that the two models’ predictions will converge as we get closer to Election Day.

The more important point is that this combination of structural models and polling data is likely to produce a more accurate forecast than either approach alone. (For the more technically-minded among you, Simon Jackman explains why here.) The use of mixed forecasting is more prevalent now because of advances in computing power and the greater ease of access to polling data. Still, the forecasts are not foolproof – pay attention to that confidence interval when evaluating predictions! – which means that in a very close election cycle, as this one appears to be, none of these approaches may be precise enough to nail down who will control the Senate with perfect accuracy. (Control of the House appears not to be in play at all this cycle.)

Of one thing I am sure. In the face of forecast uncertainty many of you will cherry pick the model whose outcome you like the best, while damning those you disagree with for their bad data, faulty assumptions and poor methodology. If you are one of those who believe in advocating the equivalent of “unskewing the polls”, I apologize in advance for injecting a dose of reality, no matter how unpleasant, into your political fantasy world in the coming months. I promise to do so as gently as possible.

Now let the forecasting begin!

UPDATE 5 p.m. Sam Wang has waded in with his own Senate forecast that puts the projected final Democratic seat total at either 49 or 50. He makes the important point that a small swing in the partisan share of the vote is going to determine which party controls the Senate – it’s that close! See his post on the topic here.

Sunday Shorts: Balz, Polarization and a New Logo

The Washington Post’s Dan Balz posted this piece yesterday in which he tries to make sense of what appears to be contradictory evidence regarding whether Americans are becoming more polarized or not. Balz notes, correctly, that in a recent survey Gallup recorded a record high number of Americans – 42% – self-identify as independents. The number of self-identified Democrats and particularly Republicans, meanwhile, is on a downward trend.

But then he notes a seeming counter trend: “Voting behavior tells a different story. In recent elections, at least nine of every 10 people who identify themselves as Republicans or Democrats — or who say they are independents but lean toward one party or the other — vote for the candidate of their party down the ballot. In 2012, only about 11 percent of voters said they cast split tickets. The percentage of true independents may be only about 10 percent of the electorate.”

By now, regular readers should recognize that this seeming contradiction is not a contradiction at all – instead, the data present two sides of the same party sorting coin. Rather than becoming increasingly polarized along partisan lines, Americans are still mostly clustered in the ideological middle, as reflected in the Gallup results showing 4-in-10 self-identifying as independents. However, as I’ve discussed in several previous posts, the parties are also better sorted ideologically. So when moderate voters who may lean in one direction or the other are forced to choose from among two partisan candidates, those choices are increasingly going to break down along party lines as more people view the parties as either purely liberal or purely conservative. Again, and at the risk of kicking a neighing nag, people aren’t more polarized – they are just better sorted into either party.

Balz misses this point and instead spends much of the rest of his column drawing on the Pew data to buttress his claim that “The trend toward polarized politics is well documented.” As you know, of course, it is not – indeed, the evidence suggests precisely the opposite conclusion. Balz, as did many of those who reported on the Pew findings, mistakes signs of growing ideological consistency among some American as proof that they possess more polarized political views. By now I trust you recognize the difference, but see here if you want a refresher.

Near the end of his column, however, Balz does make one very important observation, drawing on input from political scientist Gary Jacobson, that addresses a question I am often asked by students. If so many Americans are clustered near the ideological center, why don’t candidates build on that fact to form a centrist-leaning party? One answer is that moderates tend to be less politically engaged. Moreover, as Jacobson notes, centrist-leaning voters often do not have ideologically uniform views on which to construct a party platform. Instead, they are much more likely to shift their views, and to swing from one party to the other, depending on political context and available choices and candidates. We saw that in 2008, when independents went solidly for Obama, only to turn against the Democratic party in the 2010 midterms.

Perhaps the best way to see that Americans remain solidly clustered in the ideological middle is to look at their views on specific policy issues. As we shall see, when asked to choose among policy options, most Americans on most issues stake out rather centrist views. I’ll take on that challenge tomorrow.

Meanwhile, a quick word of thanks to all of you who offered great ideas regarding the Christian Science Monitor/Presidential Power logo. Some of the better suggestions included: Air Force One; Mt. Rushmore (one version had it painted green to represent Vermont); the Oval Office desk (adorned with a “the buck stops here” plaque); the Presidential Seal; and a silhouette of a famous president. (Here I’m partial to one of FDR, with his cigarette holder held at a jaunty angle!) In the end, the CMS editor like the idea proposed by Olivier Knox: a graphic of Article II of the Constitution which, as you know, outlines the executive powers. It will look something like this, although the CMS graphic team is working on the final version.

Olivier wins the “It’s the Fundamentals, Stupid” t-shirt. Tradition dictates that he is required to send us a photo of him wearing the shirt, preferably in some presidentially-themed context, (which shouldn’t be hard for Olivier.) Here is some inspiration, if Olivier needs it:

Thanks again for all your suggestions, and have a great Sunday!

The Declaration of Independence

Home Is Where The Political Heart Is – Or Is It?

Jack Goodman, consumer of all things political and purveyor of great blog post ideas, recently sent me the link to this entertaining exercise hosted by the Washingtonian magazine that attempts to determine where you should live based on your politics. If you click the link, you’ll come across a series of eight agree/disagree questions pertaining to a variety of lifestyle choice and beliefs. Based on your responses, the program’s algorithm purports to tell you where you should live, within a particular state, according to your political views.

The exercise brings to mind Bill Bishop and Robert Cushing’s widely publicized 2008 book The Big Sort. In it the authors argue that during the previous three decades Americans were increasingly sorting themselves into politically like-minded communities. They did so not on the basis of overt partisan calculations, but due to life-style choices that produced, as a byproduct, more politically homogenous communities. As evidence, the authors note that in 1976 only about a quarter of American voters lived in a county in which a presidential candidate won by a “landslide” margin, that is, with 60% or more of the vote – an indication of a dominant political perspective. By 2004, however, the number of landslide counties had swelled to nearly half of all counties. The trend toward a more uniform political outlook within communities, they believe, has contributed to the growth in political polarization that has again become such a hot topic thanks in part to the recently released Pew report I’ve discussed in previous posts.

Somewhat puckishly, I immediately emailed Jack with a challenge: to name one individual who is actually living where the website said she should be living based on her political views. My challenge was based in part on political science research that has cast doubt on the Big Sort thesis that, in effect, home is where our political heart is. As it turns out, other indicators suggest Americans are not sorting into like-minded communities. Thus, professors Morris Fiorina and Sam Abrams show that if you examine party registration levels in counties, instead of vote choice in presidential elections, the trend is quite different than what the authors of the Big Sort would have one believe. Based on this alternative measure, they find a decrease in Democratic “landslide” counties, but an uptick in counties dominated by independents and Republicans. More generally, in their words, “If we define landslide counties according to their voter registration rather than their presidential vote, the proportion of the American population living in landslide counties has fallen significantly, from about 50 to 15 percent.”

Fiorina and Abrams do not claim to have the last word on this topic. But they do point to the need for a more fine-grained analysis that digs deeper than county-level analyses. In this vein, other research has shown that in 2008, the Democratic share of the presidential vote in most precincts was close to 50%, suggesting that by this measure at least our communities are more politically competitive than the Big Sort suggests. (This graph is by Yale political scientist Eitan Hersh via the MonkeyCage blog):

More importantly, questions like those in the Washingtonian exercise, or in the recently released Pew survey on political polarization, that ask where we prefer to live based, in part, on political preferences, aren’t very good at telling us where we actually live. That is because as Clayton Nall and Jonathan Mummolo show in this paper people’s residential choices are constrained by more fundamental concerns with factors such as crime rates, the quality of the schools and proximity to one’s job. In this regard, people may express a partisan preference on surveys in terms of where they would like to live, but that preference is rarely going to determine their actual choice of a home. So we should be wary of using respondents’ answers to survey questions regarding where they would prefer to live as evidence of increasing political polarization.

Note also that survey results based on dichotomous choices, such as the agree/disagree option in the Washingtonian exercise, don’t do a very good job at capturing the complexity of individuals’ political views. Thus, asking whether one agrees or disagrees with the statement “Abortion should be legal and accessible to all women” won’t come close to capturing what most Americans think about this issue, based on other survey data that gives respondents more options.

For all these reasons, I’m willing to buy Jack lunch if the Washingtonian Capital Comment algorithm actually places more than, say, 5% of those Vermonters who respond into the community in which they actually live. (Full disclosure – Jack has bought the last 23 lunches we have enjoyed together so this is a low-risk wager.)

And we can start with me – the algorithm didn’t come close to getting my residence location correct. And that is because I live here in God’s Green Hills not because of any affinity with my neighbors’ political views – indeed, I have very few neighbors in my very rural community to bother me. Instead, I have an abundance of swimming holes, hiking trails, woodchucks and, not least, stones. And stone walls, after all, make good neighbors. And woodchucks never question my political views. They just eat my garden.

UPDATE 12:47: The Fix’s Chris Cillizza chimes in, citing some of the same research: http://www.washingtonpost.com/blogs/the-fix/wp/2014/06/20/no-polarization-isnt-causing-us-to-change-where-we-live/

No, That’s Not Why Cantor Lost, and That’s Not What It Signifies

No, that’s probably not why Eric Cantor lost, and no, that’s not what we should conclude from his loss.

To all of you who were lurking in the twitterverse last night, I apologize if I seemed to take a bit too much pleasure in pushing back on the instant analysis issued by everyone from Chuck Todd to Chuck Wagon. But let me ask you: since no name-brand pundit that I know of saw Cantor’s loss coming (on this point, see Jaime Fuller’s wonderful “Holy Crap” summary!), why should you believe them when they then try to explain what it means? The answer? In the absence of actual data regarding who voted (more on that in a moment), you probably shouldn’t. Of course, that’s not going to stop pundits from trying to glean the national implications of Cantor’s loss.

So, at the risk of piling on, let me explain in a bit more detail why you should view most of the Cantor post-mortems with a great deal of skepticism. Let me begin by addressing some of the more popular but empirically vacuous bits of punditry.

1. Cantor’s loss means immigration reform is dead.

This was the initial reaction from pundits like Todd, and it is being repeated today by the Washington Post’s Chris Cilliza and others. The logic seems to be that since Cantor expressed a willingness to discuss amnesty for the children of illegal immigrants as part of an overall immigration package – a position his opponent David Brat attacked – then Cantor’s loss shows that immigration reform is a non-starter. There’s a couple of problems with this interpretation. First, it presumes that immigration reform wasn’t already dead, or at least on life support, even with Cantor in the House. Second, it’s not clear how much opposition to immigration reform in Cantor’s district had to do with Cantor’s loss. In fact, a Public Policy poll indicates that 72% of those surveyed in Cantor’s district strongly or somewhat support the elements of a bipartisan immigration reform bill. That support includes 70% of Republicans who responded to the survey, and 73% of independents.

2. Cantor’s loss is good news for Hillary Clinton/bad news for Marco Rubio/Jeb Bush/fill in the name of moderate Republican presidential candidate.

Ezra Klein, among many others, is pushing this line as one of his 11 lessons to draw from Cantor’s defeat. (Note: as I tweeted at length last night, my view is that at least 5 of Klein’s lessons are of dubious empirical validity.) But it is extremely misleading to draw national implications from one House primary race. We might just as well conclude that Lindsey (I support immigration reform) Graham’s Senate primary win suggests moderate Republicans are poised to do well in 2016. The fact is that you shouldn’t draw any implications regarding the 2016 presidential race from an outcome based on about 12% turnout in a single House district.

3. Cantor’s loss shows money can’t buy elections.

It is true that Cantor vastly outraised and outspent Brat by some 5-to1. But about 39% of Cantor’s money came from PACs, which is not unusual for someone occupying a leadership position, and only 2% (about $95,000) from small contributors. In contrast, Brat received no PAC money, but drew 33% ($65,000) of his contributions from small donors. As my colleague Bert Johnson is fond of pointing out, small contributors tend to be activists with strong partisan preferences. In addition, David Levinthal, using financial disclosure forms, indicates that only 12% of Cantor’s money came from within his district. So, the effective disparity in campaign contributions may not be as great as the gross spending numbers suggest.

4. Cantor’s loss means all Republican establishment candidates, particular in leadership positions, are vulnerable to Tea Party challengers.

As Klein asserts, “These losses mean no Republican is safe.” Presumably the pundits are referring to people like Senate minority leader Mitch McConnell, or Senator Lindsay (I support immigration reform) Graham, who both easily beat back Tea Party challenges? Keep in mind as well that two weeks ago these same pundits were explaining how the Republican establishment had figured out how to beat Tea Party-backed candidates!

I could go on, but I hope you see my point. What all these instant analyses have in common is a desire to draw national implications from a local race that, in the absence of data to the contrary, likely turned primarily on local constituent concerns. This tendency to draw sweeping conclusions from limited data is an unfortunate characteristic of today’s social-media driven punditry. That strategy may increase readership, but often at the expense of getting the story right.

So why did Cantor lose? We know from the PPP poll I cited above that he was not very popular in his district; only 43% of Republicans and 23% of independents approved of him, compared to 49% and 66% expressing disapproval, respectively. It is also the case that the House Republican leadership was not very popular (41% approval among Republicans, and only 16% among independents). Finally, we know turnout was up from the Republican primary in 2012 (a presidential election year) by about 20,000 voters according to Martina Berger, in an open-primary state in which there was no Democratic primary in Cantor’s district, so it is likely that independents participated in the Republican primary to a greater degree than might be expected. I hesitate to say much more about the composition of yesterday’s electorate without more data, but it wouldn’t surprise me if Cantor lost primarily because many voters viewed him as too concerned with leadership issues and thus out of touch with local district concerns. That’s not very earthshattering, and it is disappointing to those seeking some deeper meaning in Cantor’s defeat. But sometimes the simplest explanations are the best. In the absence of evidence to the contrary, and until more data comes in, that’s my story and I’m sticking to it.

Update 1:16. The WashingtonPost is doubling down on the immigration angle here and dismissing the polling data I cited by saying that the poll doesn’t survey only Republicans. Of course, it does have results for Republicans but to find them you have to actually read the poll which, evidently, the WaPo writers find too time consuming. And, of course, the Virginia primary is open to anyone regardless of partisan affiliation, so it’s useful to know the opinions of non-Republicans as well. Now, it may be, as the writer would have us believe, that turnout was dominated by the 23% who opposed immigration reform rather than the 70+% who supported it. But you can’t simply assert that this was the case without supporting evidence.

Update 1:56. I haven’t said much about whether crossover voting by Democrats along with independents contributed to Cantor’s defeat (but see this!) However, an initial analysis by Michael McDonald and by Scott Clement suggests the evidence doesn’t support the crossover voting thesis, although in the absence of exit polls it is hard to tell conclusively.