Many of the articles in this week’s Forum draw on historical data or make historical comparisons (see, e.g., David’s post on the Caesar and DiSalvo piece). So for example, they pick a starting point (some earlier election year), take the range of data from that starting point to the present, and then put 2008 in the context of that range.
A frustrating thing about these articles, as well as other political science and much political commentary, is that they don’t discuss why that starting point and the resulting historical epoch is the best one, or correct one, or relevant one. Indeed, in political commentary, you often see authors picking the starting point that confirms their argument. Want to portray the Democrats as down-and-out? Well, they’ve lost 7 of 11 presidential elections since 1968. Want to portray the Democrats as more dominant? Start in 1992. To quote from the Klinkner and Schaller article, “For the fourth time in the past five cycles, the Democratic presidential nominee won the popular vote.” Want to portray things as evenly divided? Start in 1928. You get my point.[1]
Let me give you some examples of various starting points employed in these articles:
- Michael McDonald examines voter turnout from 1948-2008. Thus, the recent increase in turnout looks significant because it returns us to the high point of this period in the early 1960s. But if we go back beyond 1948, when turnout was much higher, the recent increase in turnout might look different.
- David Walker’s forecasting model encompasses 1920-2004.
- James Campbell’s analysis of election history begins in 1868.
- Caesar and DiSalvo’s analysis of election history begins in 1896.
- David Kimball compares interest group activity in 2004 and 2008. (Shafer and Wichowsky also do some comparisons of 2004 and 2008, as do Klinkner and Schaller.)
In some cases, data limitations impose constraints. But in other cases, the choice of comparison is clearly at the author’s discretion.
In all cases, there is little if any attempt by the authors to justify the particular period they are considering, whether data considerations are a factor or not. It would be better if political scientists provided such justifications and, in so doing, set an example for pundits.
The authors are free to defend their choices in comments or guest posts.
fn1. I recently had a similar debate with a colleague. He was teaching students about the “dealignment” of people from parties. But, I said, party identification has arguably become more relevant—i.e., a more potent predictor of vote choice (see here). But, he said, is it as potent as it was in the era of strong party organizations and machines? I said that it was hard to know. In any case, the underlying question here was, “Compared to what?”
{ 5 comments }
I had an econ professor who, when talking about program evaluation studies, repeatedly told the same joke, which he attributed to Henny Youngman.
First man: “How’s your wife?”
Secon man: “I dunno. Compared to what?”
I’ve often complained that poltical scientists use the ANES (conducted 1948-present) to justify creating theories to explain when turnout rates are higher or lower. John is mistaken with regards to my article, however. In the third sentence, I note “…participation has reverted to the high end of the trading range in the last century…” That is literally a century dating back to 1908, not 1948 as John mistakingly claims.
Obviously, authors have to make choices as to what to discuss given limited space. I have chosen to discuss modern elections following the dissolution of the 19th century political machines, which is the electoral environment we now operate in. For readers who wish to place turnout in the 2008 election into a broader context, I plot and discuss turnout rates over the entire history of the country here (which was also published as a Politico op-ed, a deeper commentary is forthcoming in an Oxford handbook chapter):
http://elections.gmu.edu/Election_of_a_Century.html
Michael, thanks for correcting me. That’s what I get for trying to surf across the articles for this post. I think starting with the dissolution of the party machines is very much a reasonable choice for historical comparison. The link to the entire historical period is very handy.
I’ve often found that the post-WWII cutoff is justifiable on the grounds of both data availability (which simply explodes) and at least a smidgen of theoretical justification (namely, the US government is a very different animal, at least in terms of size & scope).
However, one little thing I’ve discovered in submitting articles is that when I simply ignore the choice of cutoff dates, the reviewers generally ignore them as well, whereas when I devote a page to explaining why a particular cutoff (or choice of period) was made, I get static from reviewers on that. “Pay no attention to the man behind the curtain” works a lot better when you never mention either man or curtain in the first place.
John: I see what you’re saying but I want to make two points.
1. Talking about losing 7 out of 11 elections or whatever is silly. The ties don’t really count for the purpose of understanding public opinion. Things will always look more stable if you use vote proportions rather than win/loss.
2. In general, it’s better to go back more years. Sure, going back to 1948 isn’t as complete as going back to 1928, but both are much much much much better than the common journalistic practice of going back to 2004 and no further or of comparing to single past elections.
Comments on this entry are closed.