More on Those Declining Turnout Rates
Like Henry’s post below, I was going to start my post on Michael McDonald’s article by quoting the same delicious snark — albeit without Henry’s knowledgeable-and-therefore-dorky asides about Battlestar Galactica. Oh, and I wanted to add a graph.
And, like Henry, I wanted to express some frustrations with the commentariat. Not so much because they fixate on certain explanations for “declining turnout” but because they can’t even correctly describe trends in turnout in the first place. Here is Evan Thomas of Newsweek in early January:
Aside from an uptick in the 2004 presidential election, voter turnout has drifted downward since its modern peak in 1960 (from 63 percent to the low 50s), despite much easier rules on voter registration and expensive efforts to get out voters, writes Thomas Patterson, a professor at Harvard’s Kennedy School of Government and the author of “The Vanishing Voter.”
Even more depressing, from the perspective of numeracy, are characterizations of the “decline” as “steady” (e.g., here.)
It is a crass simplification to characterize trends in turnout since 1948 (or 1960) as a “decline” and certainly as a “steady” decline. (Ditto the “decline” in political trust. Look at the graph, people! But I digress.) There was a decline between 1960 and 1976. But after that, the trend has bounced around.1 In any case, recent elections have seen participation at levels close to the Edenic days of yore. Oh, except for 1948, which Evan Thomas no doubt doesn’t want to discuss. (And I’m not even getting into elections before 1948, which also complicate the story.) These are the niggling details that get ignored in service of a good woe-is-us narrative.
This little tirade is my way of saying “Amen, brother!” to Michael McDonald. The dirges commonly sung about turnout need to be re-tuned. And now, the tougher business begins: what explains the trends in turnout, including the recent increase?
McDonald notes four factors regarding 2008, which may also be applicable over time:
- The closeness of the election. Turnout was higher in the 10 states the candidates visited in the last two weeks of the campaign (66%) as compared to the other states (61%).
- Mobilization. Some prominent research by Steven Rosenstone and Mark Hansen explained the earlier decline in terms of declining mobilization efforts. This seems to be a promising explanation for the increase, as presidential campaigns have rediscovered old-school shoe-leather campaigning and political scientists have shown that voter contact, when done right, can increase turnout by several percentage points (see this repository of findings). At this stage, the only 2008 data we have is exit poll respondents’ self-reported contact from the campaigns, which isn’t sufficient to prove that mobilization mattered. And the data we have is also somewhat equivocal: the self-reported rate of contact was slightly lower in 2008 than 2004, as McDonald notes.
- Enthusiasm. Did Obama’s “enthusiasm gap” matter? This could help explain the increase as well as the small declines in some solidly Republican states. But, ultimately, we need better data. Comparing turnout rates across groups will give us a much better picture of what happened.
- Election reforms. McDonald focuses on early voting which has not, to this point, shown clear evidence of increasing overall turnout. Here is one study by Paul Gronke and colleagues. But maybe that is changing.
These are all plausible, not-mutually-exclusive explanations. I’ll conclude with a few thoughts about testing them.
1) It seems like a good time to extend Rosenstone and Hansen’s analysis by including these recent elections.
2) Wouldn’t it be great to get the Obama campaign’s internal data on voter contact? (And now my dorky side reveals itself.) If you ever hear high-pitched cries after an election, that is the sound of reams and reams of campaign-related data dying a slow death in filing cabinets and on hard drives. Simply put, the parties and candidates put enough resources into data collection to fund a kajillion NES surveys, but almost all of it never sees the light of day, much less a political science journal. (Daron Shaw’s work, much of which draws on data from the RNC, is an exception.) In my graduate school days, I had no luck prying data out of the hands of the 1998 Gray Davis gubernatorial campaign, despite good contacts inside the campaign and despite portraying myself as an innocent and hapless doctoral student. (I was told to blame this dude.) I am certainly not well-positioned to get Obama’s data, or McCain’s for that matter. But I hope someone is.
3) Related to #2: We need to build on the useful field experimental research on turnout and analyze some good observational data. As Andy pointed out the other day, both kinds of research designs are useful, and the latter is especially useful for showing not just what can happen but what actually happened. Joel Middleton and Don Green’s paper (gated) on the effects of MoveOn.org’s 2004 mobilization campaign is a good example.2
Perhaps Michael or others have additional thoughts?
1 So I guess I disagree with Paul Gronke in his comment on Henry’s post. I just don’t see much evidence of a decline between 1976-96. Sure, 1996 was a low point, but 1992 and 1984 were not. I think these upticks deserve to be taken seriously as more than just “exceptions.”
2 At the same time, I am told of another scholar who negotiated access to the GOP’s mobilization data from 2002 (i.e., the “72 Hour Plan”), wrote up an manuscript on its (significant) effects, and then had that manuscript rejected by a top-tier journal without its even being sent out for review. “Not enough theory,” was the explanation by the editors. God forbid political science actually tell an empirically rigorous and relevant story of a contemporary election without some perfunctory theory tacked up front.
Comments
This Presidential campaign will be known for what was the worst ever by a candidate. John McCain showed the absolute worst judgment, worst sense of what he was doing, total arrogance towards his supporters, and how in the world that his campaign manager will ever work in politics again will be the wonders of wonders. Never has such a display of old fashioned stupidity and blundering been out there for all to view in a campaign. Couple that with the execution that Obama showed in his campaign and it is a no brainer why he won over McCain, irrespective of the other variables in play for 2008.
Posted by: Joseph Bailey | January 13, 2009 10:27 PM
It’s irksome when campaign consultants, who dismiss political scientists as being out of touch, won’t share their data with those of us who could actually do interesting and important analysis with it. It’s particularly irksome when we do get the data, obtain interesting findings, but get rejected by journals for being insufficiently theoretical. This disconnect between consultants and academics is ultimately bad for both groups.
Posted by: Seth | January 13, 2009 11:56 PM
Seth: It’s also irksome to provide data (and ideas) to academics and then not hear back as to what they are using our work for, nor get credit for the contribution. I certainly disagree that academics are the only ones that can “actually do interesting and important analysis.” Ugh. Publication is often largely relevant only for building the careers of the academic authors, not for conclusive or even helpful policy or fieldwork. There has long been (if not plenty of) communication among activists, philanthropists and campaign staff (researchers, managers, etc.) on some of these issues (messaging, turnout, etc.). Granted the fugative nature of this work is frustrating, and the quality varies, but it can be timely and relevant. Physicists didn’t invent engineering; biologists didn’t invent agricultural science; political scientists didn’t invent political analysis.
Posted by: Doug | January 14, 2009 07:02 AM
Doug, it’s more than irksome if a political scientist uses data and doesn’t credit the source — it’s unethical. Also, I don’t think that political scientists invented political analysis or are the only ones who can do it. At least in theory, however, we should be able to do it well and in an unbiased manner. We all have our axes to grind. Consultants want to prove that their strategic decisions, advertisements, etc., made a difference. Academics want to get published. Fine. But for an academic, it’s just as interesting to show that, say, the Daisy ad didn’t affect the 1964 election as to show that it did.
Posted by: Seth | January 14, 2009 09:44 AM
Hi, Seth: Yes, I agree that the for-profit firms have motives, but they also have to respond to questions and hopefully the questions are properly asked and generate decent research. (There was a pscyh prof at Yale who had an excellent intro book on stats (which included almost no stats) that was called “Statistic as Principled Argument”.)
However, besides the profit seeking groups (and non-profits trying to polish their work to funders) there are non-profits (and again clients) that also really do want (and sometimes do) analysis that lets them know “what really works.”
Unfortunately, there are strategic (for the social movment types) and bottom-line reasons (for the for-profits and some non-profits) for not always sharing these data (or only selling it).
I think the key for academia is that more reward be given for applied work on social problems, not just journal articles. I don’t know if that is too much to hope for or how it could be “regularized” enough for tenure evaluations, etc.
As somebody once said, “society has problems, universities have departments.” Hopefully, with the recent increase in public policy programs and professorships around the world more credit will be given for “extension work” regardless of the department where the faculty reside.
-Doug
Posted by: Doug | January 14, 2009 01:07 PM
Another problem here, that John alludes to, is that members of the political science profession are not particually supportive of voting studies, either. I have twice had NSF proposals rejected to do a large-scale collection and dissemination of voter registration databases. The reason: reviewers did not find the proposals theoretically interesting. My sadly “favorite” review dismissed my proposal for a lack of graduate student training, at a time when my school did not have a graduate program. I am not alone in experiencing this indifference towards voting studies from NSF, but fortunately I have found greater funding interest from charitable foundations (but not on collecting voter registration databases).
Two points about why the current elections are not entirely comparable to the “high” 1950s and 1960s elections in the graph. First, post-1972 elections include lower participatory 18-20 year olds. Second, there are a large number of mail ballots (somewhere between half and three quarters of a million) that I estimate were rejected in the 2008 election. As mail balloting has increased in some states, turnout rates are being artificially depressed for persons failing to follow proper procedures. Ironically, the procedures for opening a mail ballot (e.g., matching signature, proper envelope, etc.) are much more rigorous than actually counting the ballot if is deemed properly cast (i.e., voter intent).
Posted by: Michael McDonald | January 14, 2009 02:14 PM
Interesting stuff, Michael. Do we know what number of provisional ballots were cast and counted yet, and by where?
-Doug
Posted by: Doug Hess | January 14, 2009 02:56 PM