Gerrymandering Paper Problems
A paper has been written that
casts a shadow on our blog assumption that the decrease in House
centrism is because of computerized gerrymandering. So I read the paper.
Though I'm glad the author's looking at the problems, I think this
needs more work. I thought the paper had serious problems with facts,
Maybe the worst is that the author is charmingly ignorant about when
computers started to have serious horsepower.
The author believes that
computers were only powerful enough to gerrymander effectively
starting in the 90s; in reality, powerful minicomputers and
workstations were available well beforehand, and likely to have been
cheap enough for consultant purchase. I seem to remember Tully having
mentioned using workstations at least for his consulting work in the
past.
The error has fatal consequences, because he uses data from 1990 as a
pre-computer-gerrymander baseline for comparison, when in reality it
started well before then. Garbage In, Garbage Out. Indeed, one point
the paper makes is that alot more partisan stiffening happened in '80.
Hat tip,
Tyler Cowen.
Hey, Tully, if you see this, any thoughts on when the first serious
computer-based gerrymanders did begin?
Posted by Jon Kay at October 30, 2007 01:37 AM
Since redistricting happens every decade after a census, we'd be looking at 1981 and 1971 as the key years. The PC was well established in 1991, so the pre-computer era would have to be at least 1981.
The PC was starting to collect steam in 1981 and was likely used by then. You're talking mainframes and punch cards in 1971, which would be the baseline I'd use if I was designing a study.
It would probably be worth looking at a paper on party-competitiveness and polarization by David King (previously blogged here by Tully). Simply put, the only direct effect gerrymandering can have is to reduce party-competion. Therefore, any effect of gerrymandering on polarization would have to result from reduced competition being a cause for increased polarization. As King demonstrates, competitive districts are often just as polarized as non-competitive ones. The only difference is that whereas non-competitive districts are consistently represented by outliers in one direction, competitive districts seesaw between outliers in one direction and then the other. King accounts for this with the two-stage electoral process with primary voters giving the general electorate a choice between the extremes of either end.
If I were to speculate as to why this change arose between the post-war period and now, I would say the sorting of the parties. Whereas the post-war period had plenty of conservative Democrats and liberal Republicans, most of those types switched parties between that period and now. As a result, the primary electorate of each party has moved from the center as the Democratic party lost conservatives and gained liberals and vice-versa for the Republican party.
Some neat stuff in that paper. I'm going to read it slowly and carefully when I have more time.
As best I know the first truly serious attempts at computer-driven gerrymandering came out of the 1980 census. Earlier tries after the 1970 census wouldn't surprise me in the least, but I was a kinda young at the time. :-)
But the pols before then were hardly amateurs at the practice (the term itself dates to 1812) and it doesn't require computers to be effective. Of course computers lend new and deeper dimensions and greater precision to it, especially thanks to the accumulation of consumer marketing databases covering so many details and the processor power to crunch all those added dimensions.
And amusingly (to me, anyway) in Ireland it's called "Tullymandering." :-) My work is more concetrated towards working with the voters already in a district.
Just to add a bit--it's not that it took computers until the 1990's to be ABLE to crunch that data, though I can understand the confusion. It was that until the 1990's the many consumer databases with all those additional dimensions (a.k.a. personal habits and purchases) were not available. Once they WERE available, the additional information could be integrated to "hone" programs already in use.
Frankly, I think there's a lot of over-analyzing going on there. Party data people tend to start with the universe of people available. I start on the other end--people already known to actually vote. Therefore right out of the gate I have greater effectiveness.
Yes, that's a naked plug. For filthy lucre I am willing to share my expertise. Kids gotta eat! :-)
Sei pronto per una straordinaria avventura nel sito di casino online? Gioca online al tuo gioco preferito oppure, se non lo hai ancora scelto, consulta i nostri manuali per imparare a giocare qualsiasi gioco del casinò ed a allenarti con i giochi gratis. Potrai poi avventurarti nei tornei con soldi veri e sbancare il casinò...
casino online - http://www.quellefromage.com/
