Obamacare: The roll call and the damage done

It's become an article of faith that Democrats in Congress paid a price for having passed the Patient Protection and Affordable Care Act (or ACA, or Obamacare) in 2010. I'm pleased to report the publication of a new paper that not only demonstrates that this price was real, but also shows just how it happened. The paper, "One Vote Out of Step: The Effects of Salient Roll Call Votes in the 2010 Elections" (ungated version here), just published in American Politics Research, is the product of some earlier research by Steve Greene and me and by Brendan Nyhan, Eric McGhee, and John Sides. We decided to join forces as an Asia-like supergroup to produce this new paper.

Consistent with the earlier research, we found that those House Democrats who voted in favor of ACA ran around six points behind those Democrats who voted against it in the 2010 midterms. We conducted thousands of simulations and found that, in the majority of simulations, Democrats retained at least 25 additional seats if they had all voted against ACA. That's enough for them to have held the majority.

Politicians and political observers often talk about the demands of party ("Sometimes party loyalty asks too much," said John Kennedy), but it's rare that we see such an explicit tradeoff. Nancy Pelosi actively pushed to pass this bill, achieving a goal that the Democratic Party had been pursuing for decades. And the price of the goal was that she lost her speakership and a few dozen of her colleagues lost their jobs.

Now, how exactly did this happen? Why did a vote for ACA cause a Democratic House members' voters to turn against her? We examine this at the level of the individual voter using a CCES survey. The findings suggest that voting for health reform caused voters to perceive a member as being more liberal, even controlling for the members' overall voting record. In the graph below, respondents were asked to evaluate the ideology of their member of Congress. The solid line represents voters in the districts of Democratic House members who voted against ACA; the dotted line represents those in the districts of ACA supporters:

That one yes vote had an enormous effect on voters, causing them to perceive their representative as being substantially further to the left. And as numerous studies have shown, being ideologically extreme tends to reduce one's vote share.

A final question, which goes somewhat beyond the paper, is whether the effect of ACA goes beyond the 2010 election. Will it hurt Democrats again this year? Some early polling suggests it might, although Bernstein is skeptical. John Sides gets into this issue in a recent op/ed. One reason ACA might not have much of an effect in Congressional elections this year is that all the vulnerable Democrats were kicked out in 2010. The ACA supporters who remain have relatively safe districts. It may be that a few senators who were not up in 2010 could still pay a price (this may be part of Ben Nelson's reason for retiring), but that's much harder to say. 

My co-authors are blogging on this issue, and I'll link as those posts become available. Here's what Sides had to say this morning, and here's a post from Nyhan on this topic. Greene had this to add. We got a bit of pushback from the likes of Jonathan Chait, so here's Sides' excellent followup.

Fast analysis, deep insight

So it seems like journalists are increasingly turning to political scientists for answers to political questions.  Thank you! In return, political scientists are trying to speed up their pace of analysis, so that reporters don't have to wait 25 years to find out, say, the impact of Abscam on public opinion.

In keeping with this, be sure to check out the midterm postmortem in the Boston Review by Brendan Nyhan, John Sides, and Eric McGhee.  No doubt political scientists will be learning more about the 2010 election cycle with time, but this is a pretty impressive compendium of all that we know so far about things like the power of the Tea Party, the impact of campaign spending, and the effects of roll call votes cast by incumbents over the past two years.

Also, if you're teaching about any of the 2010 races in the spring, be sure to pre-order your copy of Pendulum Swing, edited by Larry Sabato.  It will be available for spring classes.  I mention this because it's affordable and because I have a chapter in there about the Colorado senatorial and gubernatorial races.

It's not just the economy

Are the swings in midterm elections getting bigger? Look at Brendan Nyhan's graph below. 1994 had the biggest loss for the president's party in half a century.  That was followed by two midterms in which the president's party actually gained seats -- also unheard of in the previous half century.  Then in 2010, we saw a larger loss for the president's party than we even saw in 1994.  As I mentioned previously, these extremes are not well explained by the economy.  The four biggest outliers in midterm elections relative to economic growth occurred in just the last five midterms cycles. 

If the swings are getting bigger, what might account for this? I don't really have the hardcore evidence to prove this, but it strikes me that it's not a coincidence that the swings are growing at a time when parties are becoming increasingly strong and unified.  

Greg Koger, Matt Lebo, and Jamie Carson put out an article earlier this year showing that members of Congress get punished for voting too much with their parties. This is consistent my finding with Steve Greene that the health care vote cost supporters roughly 5 percentage points in the election and with Eric McGhee's finding that the stimulus vote and the cap-and-trade vote also took a few percentage points off the vote shares of their supporters.

Why do members get punished for voting with their parties?  Because parties are not interested in pushing through popular legislation.  Parties have longstanding priorities (health care reform, tax reductions for the wealthy, etc.) that are molded and favored by the most active and passionate leaders within the parties.  These goals are priorities for the parties over many decades and do not wax or wane with public sentiment.  Indeed, in most cases, these priorities will run against public opinion.  After all, if everyone favored something, it would probably already be law -- it wouldn't take a whole lot of energy by a unified party to press for it.

So, yes, the Democrats did suffer this year because of the poor economy.  But they also suffered because they actually used their majority to do something.  No doubt a number of Democratic members of Congress -- particularly those who lost last week -- did not appreciate the pressure put on them to cast party-line votes on things like health care and the stimulus and would rather still be employed and still be in the majority.  On the other hand, if most Democrats were offered the chance to still have the majority in exchange for not having any of their legislative accomplishments over the past two years, what would they choose?

Tea Party endorsements

Whom did various Tea Party groups endorse in the general election?  The Washington Post has a really nice page devoted to this topic.

Generational Polarization

Were poor youth turnout and high elderly turnout responsible for the Democratic slaughter on Tuesday?  William Galston is skeptical:

The conventional wisdom before November 2 was that seniors enraged or terrified by changes in Medicare would turn out in droves to punish those who voted for health reform while young people disillusioned by Obama’s failure to create the New Jerusalem would abstain. That did happen, but only to a modest degree. Voters of ages 18-29 constituted 12 percent of the electorate in 2006; 11 percent in 2010. Voters over 65 were 19 percent of the total in 2006; 23 percent in 2010—noticeable but hardly decisive.

Okay, but Galston ignores an important trend: the electorate has polarized by age group considerably since 2006.  In 2006, 60% of 18-29 year olds voted Democratic, compared to 49% of 65+ voters.  That's an 11-point difference.  In 2010, 57% of 18-29 year olds voted Democratic, compared to 38% of 65+ voters.  That's a 19-point difference.  The increasing Democratic tendency of young voters relative to older voters made their failure to vote much costlier for Democrats than it would have been a few years ago.

The most damaging roll call vote

Back in September, Steven Greene and I did some preliminary research finding that Democratic House members who had voted for health care reform were running about three points behind those who voted against it.  Well, now that the election is over, we figured we'd check to see how those candidates made out.

Using the same analysis we did last time, only predicting vote share rather than polling results, we find that Democratic House members who voted for health care reform did an average of 5.2 percentage points worse than those who voted against it.  This effect is statistically significant.  This is controlling for district partisanship and the members' ideal points (an estimation of their overall voting record).

Suffice it to say this is a huge effect.  As it turns out, of the 41 House Democrats we examined, nine eight of them who supported health reform ended up losing by less than 5.2 points: Carney (PA), Kilpatrick (AZ), Klein (FL), Mollohan (WV), Perriello (VA), Pomeroy (ND), Salazar (CO), Spratt (SC), Wilson (OH).  That is, the analysis suggests that had those folks voted against health reform, they'd still have jobs in Congress.  Now, this is a pretty simplistic analysis -- we don't know whether Democrats would have fared better or worse overall if health reform had failed, for example -- but it's still a pretty astonishing effect for one vote.

Conversely, we find that five Democratic members who did get reelected would be out of a job today if they'd voted yes on health reform.  These include Altmire (PA), Chandler (KY), Matheson (UT), McIntyre (NC), and Shuler (NC).

We ran the same analysis for two other controversial votes, the stimulus and cap-and-trade.  We found no statistically significant effect for either of those votes.  (This is somewhat different from Eric McGhee's findings, so Eric, Steve, and I will need to hammer this all out over beers in Chicago next April.)

Some caveats: This does not include the entire Democratic caucus -- just the 41 members from the 50 most conservative Democratic-held districts who were running for reelection.  We'll expand our analysis on this soon.  It also doesn't control for spending, although most analyses I've seen on that suggests the effect was kind of a wash.

I've done some previous reflection on how to interpret the fact that Democrats finally following through on a longstanding party commitment appears to have hurt them dearly.  For more on this, I'd suggest reading Jonathan Bernstein, who suggests that delivering on this goal was worth losing an election over. I still maintain that health reform will be popular and untouchable, along the lines of Social Security, in the coming decades, but we're not there now.  It's currently unpopular.  If it were popular, it would have happened a long time ago.  Instead, what it took was a determined and (relatively) unified party with sizable majorities in both chambers.  Going against public opinion is a big part of what strong parties do, and it's not surprising that there's occasionally a large price to be paid.

Midterms 2010: What Happened?

As loyal readers of this blog are no doubt aware, my forecast of the midterm election was pretty wide of the mark.  We don't know the exact number of House seats the Republicans have gained yet, but it looks to be around 65 seats.  (I've updated the scatterplot above using this figure.)  I had predicted 40 seats, with a 28-seat margin of error.  So, while 65 seats is within that margin, it's really out in the tails.  And as the scatterplot above shows, it's a huge outlier.  I am gratified somewhat that most other forecasters missed it by a lot, too, but the question remains: why did we miss this?

John Sides offers a bit of speculation.  One thing he touches on is the importance of candidate recruitment -- the favorable political conditions for Republicans made it easier to recruit high-quality candidates this year.  Contrast this with 1994.  Yes, Republicans recruited heavily that year, as it was clear that conditions were running against the Democrats, but few believed that the GOP would really take the House that year until right before it happened.  Democrats had held the House for forty years; a Democratic House was believed to a permanent part of the political environment.  Today, we know that the House can flip back and forth.  It was clear by mid-2009 that the economy wouldn't be expanding robustly any time soon, that unemployment was going to remain high, and that Obama's popularity wasn't likely to surge.  Add to that the knowledge that a decent set of GOP candidates could actually flip the House, plus a nascent Tea Party movement that was producing potential candidates, and you have a great recruiting environment.

I'm not saying that's the only reason the GOP won so many seats, but it might contribute to it.  I'm still wondering about other reasons.  I can't help noticing, for example, that the four biggest outliers in the above scatterplot are from the last five midterm elections.  Perhaps large swings are a feature of stronger partisanship, or nationalized elections, or something else.  But the swings are getting larger.

Meanwhile, a spot of good news for my forecasts: I predicted that Republicans would take over 16 legislative chambers.  It looks like they've taken 18, although there are still a few that are up in the air (including the Colorado House).

The morning after

Final predictions

I have a bet with my colleague Peter Hanson over whose freshman seminar students can produce the best midterm forecasts.  Here are the mean predictions for how many seats the Democrats will control in the House and Senate after tonight:

                      House      Senate
My class          204          50.5
Peter's class     217          51.25
Me                   217           52

I'm guessing Peter has more liberals in his class than I do.

Obama-midterm-loss-blame-narrative Bingo!

Via Brendan Nyhan:

Freakonomics: Midterm Edition

The NYT's Freakonomics blog has a forum out today on the congressional midterms featuring yours truly. Yeah, I wanted to toot my horn a bit, but I also wanted to point you toward Justin Wolfers' forecast, which is brilliant.  Basically, Wolfers is arguing that it is far worse for a pundit to be uninteresting than wrong, so he proceeds to predict big wins by Democrats next month:

And if I’m wrong? We both know there won’t be any real consequences. I’ll be sure to sell some clever story. You know, there was weather on election day (hot or cold, wet or dry — it all works!) and this messed with turnout. Or perhaps, This Time Was Different, and my excellent forecast was knocked off course by our first black president, by rising cellphone penetration or a candidate who may not be a witch. I’ll remind you how I nailed previous elections.... I’ll bluster and use long words like sociotropic, or perhaps heteroskedastic. And I’ll remind you that my first name is Professor, and I went to a prestigious school. More to the point, if I’m wrong, I’m sure we’ll all have forgotten by the time the 2012 election rolls around. Shhhh… I won’t tell if you won’t.

Definitely worth the read.

Breaking: Tea Party candidates not all freaks

Did Tea Party influence cause Republicans to nominate a large number of unqualified candidates?  Via John Sides, Brendan Nyhan looks at the backgrounds of the current Republican candidates for the U.S. House and finds that about half of them have previously held elective office.  (Political scientists usually consider previous elective office experience to be a useful proxy for candidate quality, which is otherwise very difficult to measure.)  Not only is this figure similar to those of previous election years, but it's considerably higher than the figure for current Democratic candidates.

Yes, there are a bunch of inexperienced candidates with Tea Party backing out there, but as Nyhan notes, they are largely concentrated in uncompetitive districts, so they're not really hurting the Republicans this year.  In more competitive races, the Tea Party has chosen to back experienced politicians.

All this suggests that the Tea Party, to the extent we can define it as a unified entity, is much more pragmatic than the media usually portray it to be.  Remember that Tea Party members enthusiastically backed Scott Brown for the Massachusetts Senate seat despite his very moderate credentials.  That is, he stood for basically nothing that they stood for, but they recognized the importance of depriving Democrats of their filibuster-proof majority, so they sucked it up.

Can early voting numbers tell us anything?

Last week, I cited some early voting numbers collected by Michael McDonald that suggested a Democratic advantage for the midterm elections.  This post at National Review's Online, however, compares these early voting numbers with those of 2008 and finds the Democrats wanting this year.  That is, even though Democrats still outnumber Republicans among early voters, their advantage has waned significantly:

The average of these states show that early voting has shifted from a D+16.6 partisan split to a D+1.7 partisan split for a Republican gain of +14.9% since 2008.

Over at the Washington Post, however, Karen Tumulty reports that the partisan balance among early voters is almost identical to that of 2006, when Democrats took over both chambers.

So are the early voting numbers a good sign for Democrats or Republicans?  What's the proper basis of comparison -- 2008 or 2006?

Probably the most important question here, though, is just how predictive are early voting statistics of actual election outcomes?  I've not seen a serious study of this (please let me know if you have), but my guess is they're not terribly helpful.  For one thing, widespread early voting is a relatively recent phenomenon, so we don't have a whole lot of data here.  For another, as McDonald reminds us, early voting doesn't tell us how people voted.  All we know is their party affiliation.  Yes, any registered partisan who bothers to vote early is extremely likely to vote her party registration.  But a) this doesn't tell us the percentages of registered partisans who will ultimately vote; and b) this doesn't tell us a thing about the preferences of independents, who comprise somewhere around a fifth of the early electorate thus far.

At this point, only around 5 to 10 percent of those who are casting a vote in 2010 have voted, and I feel confident in saying that that 5 to 10 percent is poorly representative of the ultimate electorate.

Covering my assumptions

Nate Silver recently noted just how uncertain most midterm election forecasts are.  He's expecting the GOP to pick up 50 seats in the House, although the actual outcome, he suggests, could vary by 20 or 30 seats each way.

This seems like a good point to point out that my forecast of 40 GOP pickups (which now seems relatively conservative) had a standard error of 14 seats.  Which means that I can state with 95% confidence that the Republicans will pick up between 12 and 68 seats.  That's not terribly useful, I know.  The best guess is still 40, but there's a large range there.  As a professor of mine once noted, when a 5-year old aims a bow and arrow, the most likely place for the arrow to land is at the target, but everyone still runs for cover.

When making forecasts previously, I have sometimes been led astray by late information and by a reliance on my "gut," which seems to have a bias toward my preferred political party.  So I'm sticking with my original forecast.

Evidence for a wave

I'm sticking with my predictions about the midterm elections, although there are a bunch of anecdotes that would tempt me to move into either the Dems-will-hold-the-House camp or into the Dems-will-lose-8,000-seats camp.

In the latter category, I am hearing rumors that Colorado Democrats are worried about several of their Denver statehouse districts, including HD3, held by Rep. Daniel Kagan, who was appointed to the post last year.  The latest voter registration stats show that HD3 is 40% Democratic and 26% Republican, with 32% unaffiliated.  Is it possible that a Democrat could lose a district like that?  Suffice it to say that if Democrats can't hold onto statehouse seats in Denver, they're not holding onto much else.

Women Connecting in Connecticut

In last Wednesday's New York Times, Raymond Hernandez noted a recent CNN poll (PDF) on the Connecticut Senate race between Republican Linda McMahon and Democrat Richard Blumenthal.  Hernandez notes the substantial gender gap in the race: there's a 13-point gap between the sexes in their evaluations of the candidates.  Blumenthal is basically tied among men (48 Blumenthal, 47 McMahon), while women overwhelmingly prefer him to McMahon (61-34).  Hernandez's explanation for this: "Many female voters are turned off by her campaign."  She is allegedly failing to "connect" with female voters due to her previous work with professional wrestling and her current tough campaign style.

You know what would have been helpful to mention in this article?  There's a substantial gender gap in virtually all partisan races, regardless of the sex of the candidates.  In fact, there was a 13-point gender gap (identical to the current one!) in Connecticut in the 2008 presidential race.  Indeed, we've had a measurable gender gap in presidential elections since the 60s -- one that seems unaffected by the presence of female vice presidential candidates on the ticket -- and it's been in the 10-point range since the 90s.

Now, one thing that's kind of interesting from 2008 is that the gender gap in Connecticut voting was higher than it was nationally.  Is the gap in the Blumenthal-McMahon race higher than in other races?  That would be useful to know.  But all the Times story provides us with is some sex-based rationalizations for how people probably would have voted anyway.

Dems doing well in early voting

As of today, over 700,000 Americans have already voted in the 2010 congressional midterm elections. Michael McDonald is once again reporting on early voting statistics as they come in.  Interestingly, some states provide the party affiliation of early voters, which allows us to make a halfway decent guess as to how they actually voted.  It turns out that the Democrats are doing reasonably well so far.  In Iowa, for example, the breakdown of early voters is 42% Democrats, 29% Republicans, and 29% independents.

Doesn't this violate everything we've heard so far about the likely makeup of the 2010 electorate and the enthusiasm gap?  Well, to the extent this trend holds up, it suggests an interesting divergence in party strategies.  Republicans seem to be focusing on winning the air war, while Democrats are apparently pushing mobilization.  I say this extremely tentatively, but this may just be one of those situations where everyone thinks the Obama folks are losing because they're not winning the daily skirmishes, but they're actually husbanding their resources for when it actually counts.

The growth in early voting is really an interesting and under-developed area for campaign scholars.  Researchers have noted, for example, that when someone of the president's stature comes to down town for a rally, it can give him or some other candidate a short-term boost in the polls.  That boost will soon disappear, so it doesn't matter too much.  But if the visit gets thousands of people to walk over to an early polling station and vote, it can matter quite a bit.

The curious case of 1982

Several months ago, Brendan Nyhan took issue with political commentators who were urging Obama to campaign more like Reagan.  Jon Judis, in particular, had argued that economic models in 1982 had predicted Republican House losses of around 50 seats, but Reagan's communications skills limited Republican losses to only 26.  Brendan argued that this was silly -- that midterm elections tend to turn on the fundamentals rather than the president's political skills.

Of course, I'm in sync with Brendan on this point, but when I look at my midterm data, I find that 1982 really was a significant outlier.  Below is a graph showing how the president's party did in every midterm election since 1950 relative to my forecast (based on growth in real disposable income and the president's popularity).  Each data point is the actual gain in seats minus the expected seat gain.  So positive numbers means the president's party did better than expected, even if they still lost seats.

Interestingly, 1982 is the biggest positive outlier in the whole series.  My model forecasts the Republicans losing 47 seats that year, but they only lost 26.  How did they pull that off?

Some possible explanations:

1. My model isn't very good.  In Brendan's post, he cites Alan Abramowitz' somewhat more complex model that pretty much nails Republican seat losses for that year.  So maybe my bias toward parsimony has some down sides. 
2. Related to point 1, Republicans held only 192 seats going into the 1982.  It's just hard to lose 47 seats when you only control 192 of them -- you're pretty much cutting bone at that point.  By contrast, the Democrats today control 257 seats, which necessarily includes some districts that aren't generally friendly to Democrats.  A modest Republican wave can take out a bunch of those. (Notably, when I include the number of seats held by the president's party, the forecast for 1982 drops from 47 to 33 -- much closer to reality.)
3. Maybe Reagan and the Republicans really did have a better campaign and communications strategy than Obama and the Democrats do this year.  Again, though, I doubt this -- both presidents were quite aggressive in trying to pin the economic failures on the previous administration.  What's more, Obama actually enjoys slightly higher approval ratings than Reagan did at this point in his first term.

Better state legislatures variable

In my dataset used for forecasting midterm elections, I have changed the variable regarding state legislatures.  It is now the total number of state legislative chambers won by the incumbent president's party.  Thanks to Hans Noel for calculating this.

For what it's worth, my model predicts that the Republicans will take over 16 state legislative chambers this year.

How much did the health care vote hurt?

There was much speculation at the time of the passage of the Affordable Health Care for America Act that a yes vote might put Democratic House members from moderate districts in greater danger of electoral defeat.  Voters are known to punish members for being ideologically out of step, but would one particular roll call vote matter?  (Steven Greene suggested this idea to me -- blame him for this particular act of inception.)

To see how House members were doing, I used 538.com's brilliant election forecasting site.*  I decided to look at the members from the 50 most conservative congressional districts (as measured by the 2008 vote for McCain) that are represented by Democrats.  These range from Florida's 22nd (represented by Ron Klein), where McCain got 48 percent of the vote, to Mississippi's 4th (represented by Gene Taylor), where McCain pulled 68 percent.

Out of the original 50 districts, only 41 had members who cast a vote on health care reform and are running for reelection.  If we just divide these members based on their health care votes, those who voted for health reform are running 2.7 percentage points behind those who voted against it.  But, of course, we should control for other things, especially district conservatism, since those from the more conservative districts voted almost uniformly against reform.  I also included the members' DW-NOMINATE scores to distinguish the health care vote from the members' overall voting records.

What I found was that Democratic supporters of health care reform are running 3.2 percentage points behind Democratic opponents.  (This is statistically significant at the p≤.05 level.) That's a three percentage-point penalty resulting from a single roll call vote.  I would describe that number as large.  Most members of Congress win by much greater margins than that, of course, but for Democratic incumbents from conservative districts in a distinctly anti-Democratic year, three points is serious business.  Indeed, of the 41 Democrats I examined, only six are currently forecast to win by more than three points (and none of those voted for health care reform).


*The 538 forecasting site is far from perfect, and I question some of their results.  For example, they suggest that Rep. John Salazar in Colorado's 3rd will only get 48 percent of the vote, which I think understates his support significantly.  But they're depending on polling, which is pretty scant in most congressional races.  Regardless, 538's site is far and away the best thing going on out there in terms of House election forecasting.  Thanks to Wesley Hussey for the tip.

Update: The data are available here.