Everyone knows that there's a 'native son' effect. That is, the state where the candidate is from is more likely to vote for that candidate.
But how big is it?
Is it the same for Republicans and Democrats?
Answers below the fold
OK, there would be a lot of ways of measuring this, and this is a first shot, based on data I already have entered.
Here's what I did: I modeled Cook PVI (which is basically a measure of how the state voted in 2000 and 2004 presidential elections, discounting third party runs) and modeled it on a bunch of demographic variables and on whether it was a state where one of the presidential or vice-presidential candidates came from. This isn't ideal. Ideal would be to look at a LOT of elections, one at a time. Maybe all post WW2 elections. But I don't have all that data. So, this is what you get.
Specifically, I did a linear regression with Cook PVI as the dependent variable, and the following independent variables (by congressional district):
%Black
%Latino
%Other race (not White, Black or Latino
%veterans living in the district
Median income
%Urban (this is really more like %non-rural
and
whether it was NC, TX, WY, CT, TN, or MA (those being the home states of the presidential and vice-presidential candidates.
Here's what I found
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) - 15.25737 2.64072 - 5.778 1.47e-08 ***
congress$Black 0.39709 0.03755 10.574 < 2e-16 ***
congress$Latino 0.13480 0.03892 3.464 0.000587 ***
otherrac 0.37251 0.07013 5.311 1.76e-07 ***
congress$Veterans - 0.23493 0.07483 - 3.140 0.001810 **
congress$Med.Inc - 0.21351 0.05750 - 3.713 0.000232 ***
congress$Urban 0.26582 0.03483 7.632 1.56e-13 ***
EdwardsTRUE - 4.56849 2.64273 - 1.729 0.084594 .
BushTRUE - 13.63252 1.80704 - 7.544 2.82e-13 ***
CheneyTRUE - 11.90111 9.11256 - 1.306 0.192260
LiebermanTRUE 8.71844 4.12773 2.112 0.035259 *
GoreTRUE - 0.76623 3.09946 - 0.247 0.804863
KerryTRUE 15.03278 2.96854 5.064 6.15e-07 ***
Well, what does all that mean?
OK. If you wanted to predict the Cook PVI (with negative numbers being Republican and positive numbers Democratic) in a particular congressional district, you would use this formula
-15.26 + .40*%Black + .13*%Latino + .37*%OtherRace
- .23*%veterans - .22*MedianIncome + .27*%nonRural
- 4.57*NC - 13.63*TX - 11.9*WY + 8.72*CT - 0.77*TN + 15.0*MA
that is, e.g. congressional districts in TX were about 14 points more Republican than the demographics would suggest, and those in MA were about 15 points more Democratic. TN's districts were just about where the demographics sugget, etc. It is best, here, to ignore the question of statistical significance.....because, e.g., while WY has only one congressional district, it still has hundreds of thousands of people.
Now, we can't imply causation from correlation. But the data are suggestive. If someone knows of a place where I can get this sort of demographic information over time (say, for each census back to 1940 or 1950) by congressional district, that would be great