Partisanship & Ideology in Southern CDs

I thought I would post a small tidbit from the data I have been looking at recently for my dissertation on congressional redistricting. My data set here consists of all congressional races from 1972 through 2008, a bit more than 8000 races.

Although somewhat tangential to my research, one of the major controversies in redistricting is whether 1982 amendments to the Voting Rights Act, thought to mandate the creation of “majority-minority” districts in states with significant racial minority populations (e.g. a districts where the majority of the population was African American), actually hurt the substantive political interests of those (presumable liberal) minorities by increasing the likelihood of electing Republicans in other districts.  This was particularly relevant in the South, where since 1992, many white moderate Democrats in Congress have been replaced by either African-American liberals or very conservative Republicans.  So do these additional Republicans more than offset the gains made by the African-American delegation in the South through majority-minority districting?  The chart below, while very crude in its measurements, offers some insight.

Effect of Southern region on partisanship and ideology of congressional delegation, pre- and post- majority-minority districting constraints

		1972-1990			1992-2008
Probability of Republican Seat (Probit)
State GOP Pres. Vote		2.573	**		2.560	**
		(0.266)			(0.203)
Nat'l GOP Cong. Vote		2.200	**		2.006	**
		(0.486)			(0.372)
South Indicator		-0.432	**		-0.154	**
		(0.047)			(0.055)
Constant		-0.003			0.036
		(0.044)			(0.027)
Estimated Ideology of Representative (Linear Regr.)
State GOP Pres. Vote		0.976	**		1.225	**
		(0.064)			(0.067)
Nat'l GOP Cong. Vote		0.568	**		0.343	**
		(0.119)			(0.128)
South Indicator		0.041	**		-0.023
		(0.011)			(0.019)
Constant		-0.046			0.076
		(0.011)			(0.009)
n		4342			3914
**p<.01

This chart measures the likelihood that a congressional district will elect a Republican, and how conservative that member of congress will be (measured by Poole & Rosenthal DW-NOMINATE scores), controlling for two electoral trends.  The first is the average Republican vote in presidential elections over the course of a decade at the statewide level (so that this value does not depend on the actual gerrymander).   The second is the Republican vote at the national level in a given congressional election year. 

All these variables are scaled such that positive values indicate a Republican/conservative leaning and negative values indicate a Democratic/liberal leaning.  In all cases, the coefficients on the two electoral trend controls are large, positive, and highly statistically significant, but this doesn’t really tell us anything that isn’t already obvious.  It simply indicates that (a) states that vote Republican in presidential elections are also more likely to elect Republicans and conservatives to Congress, and (b) more Republicans and conservatives are elected to Congress in years when Republicans win a greater share of the national congressional vote.

What we are interested in here is the “South indicator”, which takes the value 1 for all congressional district elections in twelve Southern states, and a 0 otherwise.  What we are essentially asking with this variable is “Do Southern states elect a more conservative congressional delegation than we would expect given their presidential voting trends?”  And as we see from the different coefficient values in the two columns, the answer heavily depends on what time period you are talking about.

In districts drawn prior to the enactment of the VRA amendments creating majority-minority districts (the 1970s and 1980s), we see a significant negative value on the probability of a Republican win, and a significant and large positive value for ideology.  This indicates that while the South was much more likely to elect Democrats than the rest of the country, these delegation were nevertheless substantially more conservative than we would expect even from presidential voting patterns. 

Yet in the 1990s and 2000s, when many more districts were drawn to assure African-American representation in the South, both of these coefficient values are slightly negative.  This indicates that while the Republican party achieved more success in the South during this period in terms of number of electoral victories, the overall delegations in these states were actually slightly more liberal than we would expect from the presidential voting data (albeit by a statistically insignificant margin).   This doesn’t mean more liberal overall, since the South was also becoming dramatically more Republican at the presidential level during this period.  But it would appear that to the extent that these delegations became more conservative, it is because voters in the region began voting more Republican at all levels, and if anything, the new redistricting regimes in the South impeded these trends in Congress rather than exacerbating them.

As racial gerrymandering is not the focus of my dissertation, I discovered this through informal exploration of my data and am aware of the many methodological criticisms that could be brought against it conclusion.  But I thought it was an interesting little finger on the scale in the debate over substantive and descriptive representation.