Tuesday, July 20, 2010

The Economy, Elections, and Political Science

I urge everyone to read John Sides over at the Monkey Cage on elections and the economy.  He's responding to David Paul Kuhn, at Real Clear Politics, who has a piece claiming to debunk what he calls "economic fatalism."

First, I think this is a very healthy debate to be having.  To the extent that people are looking, generally, at systematic reasons that elections turn out as they do -- and what is left to explain beyond that -- everyone is better off.  I think Kuhn underestimates the extent to which it's a major achievement to have the Washington Post publish an article devoted to the evidence on that sort of thing.  If Ezra Klein oversimplified things in his piece, or if there's more debate than he let on, by all means it's a good thing to take issue with it.  But make no mistake; what Klein produced is far, far ahead of what John cites -- the awful NYT op-ed page "How to Fix Obama" feature from this past Sunday.

So, yes, the economy isn't the only thing that matters in elections.  In presidential elections, it turns out that it's not really a good idea to subject your nation to an endless, high-casualty war, especially one that you're not winning.  There's also evidence that there's a general reaction against keeping the same party in office indefinitely, so it's a plus if your party his been out of the White House for three or more terms.  And, yes, it's not a good idea to select someone far from the ideological mainstream -- that's really only mattered significantly with Goldwater and McGovern (and even then, only to the margin of defeat), but it probably has made a bit of difference in other elections.  In Congressional elections, candidates matter to some extent.  Open seats are more difficult to defend, so an incumbent party hit with a wave of retirements will tend to be hurt in November.  Challenger quality matters, so a party that does a good job of recruiting a solid crop of candidates (as the Democrats did in 1974 or the GOP in 1994) will be better off than one that doesn't (such as the Dems in 2002). 

In other words, yes, there are systematic things that matter in elections in addition to the economy.  The point is that when we talk about elections (or, perhaps, presidential popularity) to look to those things first.

Beyond them?  Yes, there's also some margin of error, so we can try to explain that by factors specific to particular elections.  The complaint of the political scientists is that this should be done, and usually isn't done, in the context of the systematic factors.  So, yes, perhaps if Barack Obama gave a few more better speeches about better subjects he might have nudged his approval ratings up a point or two.  But the overall context of those approval ratings is going to be the big, systematic factors.  And, in fact, Obama is basically more or less where one would expect given those factors. Similarly, the Democrats should expect to lose seats this November because of the big, systematic factors -- the biggest and most obvious of which is just that they've done so well in the House recently that they're defending lots of marginal seats, and have very few marginal seat targets. 

There's a very strong, and really understandable, urge for us to believe that the day-to-day stuff matters to election outcomes: the gaffes, the debates, the ads, the strong speeches, the policy proposals.  And sometimes they do!  Mostly, though, they don't, or they matter just on the margins.  As I've said many times, that doesn't mean that the ephemera of campaigns shouldn't be reported (and it may in fact be important to what pols do once they're elected, even if it doesn't sway voters).  It just should be reported in context.  Yes, I'm repeating myself, but it's for emphasis: marginal things should be reported as if they were marginal things, which requires keeping the context in the forefront.  Or, reporters can simply avoid claiming or implying important effects for things that are unlikely to have such effects.  In other words, I understand that it's impractical for reporters to constantly include reminders that really the economy and other systematic factors matter much more than X; that's not necessary if reporters would stop making claims about how X is likely to affect election results. 

At any rate, what's terrific about Ezra Klein's piece is that it provides that a large piece of that context, in a clear and readable way, in a major newspaper. 


  1. Operating analogy:

    "Once upon a time there was a magnet, and in its close neighborhood lived some steel filings. One day two or three filings felt a sudden desire to go and visit the magnet, and they began to talk of what a pleasant thing it would be to do. Other filings nearby overheard their conversation, and they, too, became infected with the same desire. Still others joined them, till at last all the filings began to discuss the matter, and more and more their vague desire grew into an impulse. "Why not go today?" said some of them; but others were of the opinion that it would be better to wait until tomorrow. Meanwhile, without their having noticed it, they had been involuntarily moving nearer to the magnet, which lay there quite still, apparently taking no heed of them. And so they went on discussing, all the time insensibly drawing nearer to their neighbor; and the more they talked, the more they felt the impulse growing stronger, till the more impatient ones declared that they would go that day, whatever the rest did. Some were heard to say that it was their duty to visit the magnet, and that they ought to have gone long ago. And, while they talked, they moved always nearer and nearer, without realizing they had moved. Then, at last, the impatient ones prevailed, and, with one irresistible impulse, the whole body cried out, "There is no use waiting. We will go today. We will go now. We will go at once." And then in one unanimous mass they swept along, and in another moment were clinging fast to the magnet on every side. Then the magnet smiled--for the steel filings had no doubt at all but that they were paying that visit on their own free will."

    --Oscar Wilde
    (I thought it was Chesterton, but I guess I'm wrong.)

    The economy drives the results in such a way that it allows candidate X or Y to believe that she won because of her charisma and savvy selection of issues, that he lost because of bad luck and a dirty smear campaign. It doesn't decide EVERYTHING, but it decides a lot.

  2. I agree that fundamentals are what matter in an election, but what if the most important fundamental is no longer the economy? Let me put it this way. We know that some people's voting behavior is not affected by changes in the economy. The easiest example would be black voters. Black voters overwhelmingly support the Democratic party no matter what the economy does. The level of their support doesn't track at all with changes in the economy. The economy apparently doesn't matter at all to black voters (at least since 1964). In theory, something must have happened that made the economy less salient for black voters. I'll leave it as an exercise for the reader to figure what that might have been.

    The real question is: could something happen that would make the economy less salient for other voters? I think the answer to that is yes and that something was a deliberate strategy of the Republican party. It's gone by various names (the Southern strategy, the culture wars, etc.) but the Republican party and its paraparty apparatus embarked on a deliberate strategy to neutralize the salience of economic performance in partisan voting. Much to their detriment, the strategy worked. They made the mistake of throwing their lot with the wrong demographic groups.

    This first started working itself out in the 90s. If you look at clustering of issues important to various types of voters, you see a wholesale change from the '80s to 2000. Since the 2000 election, the difference in the outcome of national elections has been much more about the changing demographics of the electorate than the economy. The only way to really see this is to look closely at the demographic breakdown of voters in the three presidential elections. Unlike past elections, the differences between 2000, 2004, and 2008 do not track evenly across various demographic groups (as you would expect if the differences were driven by economic performance).

    The Republican party is well on its way to making Hispanic voters as predictable as black voters. That's pretty bad news for Republicans because they will likely make Texas a battleground state by 2016 or 2020. Even worse is that all voters born since 1978 prefer the Democrats 2-1. As this cohort reaches their prime voting age, this is really problematic for the Republicans. This is the echo baby boom. While the older voters who are dying off are split roughly 50-50 between the parties, new voters split 66-33 (roughly). Unless something dramatic turns that around, the Republicans are facing a very difficult prospect ahead. Looking farther out, when my generation (I'm 50) starts dying off, that means the exiting voters will be 55R=45D. The Republicans will have to really turn things around or face extinction.

  3. Sadly, while William's comments are truthful, the models remain peculiarly resistant to this notion. This is mostly because the "new" is swamped by the "old" in the data. The 1946-1988 data are locked in; it takes until 2030 for there to be an equal amount of data from the era that he identifies.
    As a field, though, we are starting to have enough data to employ some econometrics to suss out the answer to this. I know that Gary Jacobson gets a "post 1994" dummy variable to pick up significance (in favor of the GOP). I've never been a particularly big fan of that approach, though, because I think it does some violence to the underlying causal claim, which is gradual. I want to say that Erikson, MacKuen, & Stimson have been playing around with the gradual change argument, but their approach has generally left me unsatisfied.

    I'm at a point where I'm uncomfy with both the "stock" polisci answer of "it's the economy, stupid" and perspective of partisan hardening/demographics. I buy the partisan hardening argument, but the economic determinism approach doesn't seem to be missing the mark in its predictions. (For my money, one of the funniest panels I've ever seen was at APSA 2008 where everyone under the sun was trying to adjust their model to factor in Obama, while Abramowitz remained true to his three variables (economy, popularity, time in office) and 3 months later, I think Abramowitz's model came out pretty good).

  4. It isn't just that the economy is a big factor in how people vote generally, but that in this cycle, so far, it doesn't have much competition.

    We're fighting two wars, but both (relatively!) low casualty and low profile. There have been no big scandals to capture attention. The Gulf oil spill has dominated the news for three months, but my sense is that it has merely fed into the mood of frustration resulting from the bad economy. (Also, something I've hardly seen discussed is that the region most affected is the red-state oil patch.)

    It is remarkable in a way that Dems are not in worse shape, given that their medicine has not made the patient feel better. If I were a Republican I'd be worried that the operative analogy is not 1994 but 1982. Even a fairly glacial recovery will probably mean that things are substantially better in 2012 than they are now, with advantage to Obama and the Dems - and threatening to lock in the demographic transitions William O. mentions.

  5. Does it really make any sense to treat elections from the pre "one man one vote" era as relevant to today's elections? Has anyone ever looked at the difference in the variance of number of voters per congressional district between, say, 1948 and 2008?

  6. The 1980 Presidential election is, behind the 1984 election, the modern template for a landslide. However, the 1980 race was a statistical tie as late as 1 week prior to election day, when Carter faced Reagan in an ill-fated debate that pretty much sealed President Carter's (otherwise optimistic) fate.

    Carter v. Reagan refutes the "its the economy, stupid" meme, for while the economy was surely problematic in 1980, it almost certainly was that second debate that torpedoed Carter; as he was right in the race until that unfortunate night.

    I certainly agree that the economy matters, it always matters, but I think its wrong to assume that the economy is not routinely overriden by other, case-specific factors (e.g. Reagan v. Carter, debate II, 1980). These situational factors can swing the race by several points, turning a close race into a blowout.

    A contemporary example: this blog has been discussing Sarah Palin's chances in 2012. I believe there is a peculiar situational factor that will greatly hinder Palin, probably moreso than any other candidate, such that a 5-6 point lead (due to economic problems/other Obama gaffes) may evaporate prior to election day.

    That factor? Palin's GOTV effort is bound to be historically awful.

    Palinites will have no desire to man the phones the weekend before the election, they will not pound the pavement the day of the election, knowing that in either case they will encounter numerous fellow citizens questioning their sanity. Palinites will no more man the phone banks than Tom Cruise will give the keynote speech at the National Psychiatry Association convention. It just won't happen. And it will cost her a ton of votes.

    She may have built up a big enough lead to nevertheless win on Election Night - but whatever your standard assumptions about the impact of the economy, they have to be severely curtailed in the face of what should be an awfully thin operational support team for Palin around election day. Just one example to make the point that the economy matters as background, but other factors can overwhelm it in the particular.

  7. @CSH: You are wrong about the 1980 election, but your mistake is one that nearly everyone makes. You assume that the polls more than one week before an election are accurate and that is pretty clearly not the case. Here is a great example from today's news. There is a report today of a SurveyUSA poll that show Tom Perriello trailing by 38 points in the VA-05 congressional district race. Perriello might well lose, but he will not lose by 38 points. Likewise, Al Gore was never going to win in 2000 by 16 points (which one poll showed). Nor was John McCain actually tied with or ahead of Obama right after the Republican convention. Carter was doomed in 1980 long before that debate.

    One of the good things that Nate Silver has contributed to this whole process is a better understanding of how things are related. If Perriello was really done by 28 points, the Republicans would be set to take 100(+/-20) seats. To quote a famous man: "Ain't gonna happen"

  8. I'm not convinced on the other issue -- I think, as Matt does, that the data push the other way -- but I absolutely agree with William comment about 1980, and generally about horse race polls.

    To me, the classic example is the 1984 election, and the myth that Mondale had "caught up" until Reagan destroyed him with one good line in a debate.

  9. Carter was doomed in 1980 long before that debate.

    William, I'm not sure what you're getting at here. Are you suggesting that polls indicating a tie a week before the election were invalid (i.e. poorly constructed)? Or are you suggesting that they were unreliable (i.e. not reflective of voters' true intentions)? You must be arguing one (or both), since the only alternative is that something changed the last week of the campaign.

    Perhaps you brought up Silver to suggest that pollsters pre-Silver used invalid poll techniques? I enjoy Silver's writing as much as anyone, but the industry pre-dates him by quite a bit. Or perhaps the electorate is driven by animal spirits, causing them to inevitably wish to vote for Reagan even as they report their intention to vote Carter, a mere 7 days earlier.

    The '80 election is arguably an interesting case study in the distinction between these broad trends and the idea that case-specific stuff happens in a campaign that decides an election.

    Just one example that folks often forget: on Inauguration Day 1981, the fundamentalist Islamic regime in Iran broke with their universal anti-American sentiment by celebrating a Presidential Inauguration in releasing the American hostages. Given Reagan's relative hawkishness, it remains quite odd that the Iranians would celebrate his victory. Of course, there remain rumors that VP-candidate (and former CIA director) Bush arranged the inaugural release in October 1980 in Paris.

    I don't know if the Bush story is true...but if elections are determined by "animal forces" (e.g. the economy, or whatever), and if as you say Carter was doomed long before the homestretch of the election, why on earth would Bush - or whoever - arrange such a deal with the Iranians?

    Or maybe the mullahs were just celebrating Reagan's inauguration?

  10. Jonathan,

    I agree that there's a lot of data up until 1988 that shows that the economy had a huge influence on Presidential elections (in particular). And the economy definitely still has an impact on presidential approval. The problem I have is that it will take 40 years after this stops being true before the poli sci community will notice (if all they look at is the top line results).

    It really is impossible to come up with a plausible story about how the economy made 2008 different from 2004 if you look at the demographic breakdowns of from the exit polls. Is it really possible that everyone who is my age or older (50+) is completely unaffected by economic variation? That's not how it used to be. Why is it that way now?

    Just take a look at a graphic Andrew Gellman put together a year ago:


    What happened that caused the congressional results to start to track more closely with the presidential results? What happened to dampen the variation in presidential election results?
    Why are black voters immune to the effects of economic variation? Although the data is mixed, there is a real possibility that Hispanic, Asian, and "other" voters are also immune. If that turns out to be true, the impact of economic variation will lessen over time as the demographic composition of the electorate changes. Moreover, there is no reason to suppose that white voters will always behave the same way they did in the past.

    You put a lot of faith in a regression analysis that was significantly off in 1996 and 2000. It seemed to "come back" in 2004 and 2008, but that doesn't hold up if you look at the way changes were distributed across subgroups.

  11. CSH,

    Both, really. Invalid because they ask a nonsense question (If the election were held today...) and unreliable because it is hard to know who "likely" voters are until very close to an election. I brought up Silver to point out that the polls often show things that we know are clearly contradictory. Silver described one way to figure out which polls are most nonsensical. In 1980, Carter and Reagan being tied late in the election cycle was inconsistent with literally everything else that went on during that year.


Note: Only a member of this blog may post a comment.

Who links to my website?