Gerrymandering again

On Wednesday I put together a slide show on American politics in the Vietnam era, including both Presidential and Congressional elections, for a class. In so doing (with the help of Wikpedia, which is outstanding for this sort of thing), I noticed that Wikpedia's results included something that I had long been looking for--the total party vote, nationwide, in Congressional elections. Since then I have spent a good deal of time playing with the data since 1946 to try to uncover some patterns.

My goal was primarily to evaluate the combined influence of gerrymandering and luck upon various elections. If the population of the United States were evenly divided into districts (which it is not) and political preferences were evenly distributed, an overall Democratic vote of 53% would produce a real landslide in the House of Representatives, just as a 53% vote for one Presidential candidate usually produces an electoral landslide. Such, however, is not the case, and the distribution of House seats is littered with anomalies. By entering all the data from 1946 to the present, however, I managed to pick out some clear patterns, even though all had exceptions. My goal was to correlate the total percentage vote for the winning party with the number of seats it won, although for simplicity's sake--and because minor-party totals in the last decade have actually amounted to a couple of percentage points in some cases--I settled on the winning party's percentage of the votes cast for Republicans and Democrats as the key variable. The range of that figure is from slightly below 50% (stay tuned for that one) to three Democratic landslides of 1958, 1964, and 1974, in which the Democrats won 56%, 58%, and 59% of the major party vote. Thanks largely to their long-time strength in the South, the Democrats have had of course had a built-in advantage for most of the 1946-2006 period, and the Republican popular vote record is 55%, all the way back in 1946.

Now as many of you know, I was extremely concerned by possible effects of gerrymandering on the 2006 elections and speculated late last summer that they would make it impossible for the Democrats to win control of the House. I turned out to be wrong. The Democrats gained 5.4% of the raw vote from 2004 to 2006, worth 33 new seats, and now hold a total of 233, almost exactly reversing the result in the last Congress. But I was not altogether wrong. In 2004, the Republicans won 232 seats with 51.4% of the major party vote; in 2006 the Democrats won 53.3% of the major party vote--almost two percentage points more--but emerged with just one more seat. The electoral math clearly favors the Republicans now.

The same result emerges from a different kind of comparison--how much is 53% of the major party vote usually worth? The answer, on average, is 246 seats, thirteen more than the Democrats collected (although to be fair, most of the data points come from the era of the Democratic solid south, which does seem, at first glance anyway, to have given the Democrats more seats with less votes, perhaps because southern turnouts were so low.) And as a matter of fact, when the Republicans took over the Congress in 1994, they won only 230 seats with the same 53% of the major party vote, suggesting that they overcame a considerable Democratic districting advantage. That advantage did not last, however--in 1996, the Republicans kept the control of Congress with 228 seats to 217, even though the Democrats actually outpolled them in the popular vote for Congress. (I do not recall seeing any mention of that rather anomalous result at the time, and this was the first I had heard of it.)

It occurred to me midway through this post, however, that I might be asking the wrong question. Perhaps it would make more sense to compare the net gain in the national Congressional vote from one election to the next, and the number of seats into which that gain translated. And that is what I have done, using the Democratic vote as a template (this would produce some minor anomalies when minor parties made a big jump, but they should be minor, and I'm running out of time.) This suggests that the Democrats should indeed have done much, much better than they did the last time out.

The Democratic vote actually increased by 5.4% of the total from 2004 to 2006, which is one of its three biggest gains since 1946. In 1948, an 8% gain (after a 7.5% loss in 1946) won 74 seats for the Democrats; in 1974, when they were starting from a much higher base, it won 49; but last time out it won only 31. The Republicans' three biggest corresponding gains were in 1946 (7.4%, 54 seats); 1966 (64.%, 48 seats); and, mirabile dictu, 1994, when a 5.9% gain translated into 44 new seats.

In any case, the return of the Democrats to majority party status was huge news and has enormous implications for 2008, especially as the Republican candidates' field looks weaker and weaker. But our democracy still faces a substantial road block in the apportionment of districts that left several states with razor-thin Senate races (such as Missouri and Tennessee) without a single contested House race. Let us hope we can move back in the right direction.