Measuring the Effectiveness of Political Parties at the State Level in the U.S. (Part 2)

I wanted to revisit the topic of comparing the dominance of a political party in charge of the state house and senate and looking at the pattern of internal migration. I will do two regressions. The first looks at the average exact percentage of the number of Democrats in both chambers and then takes the average of those percentages. So if the Democrats control 60% of the house and 70% of the senate in State A, then State A will have the number 65 assigned to it for the PolAvg variable. I look at two time periods. I take the average political dominance for the 1990s and the 2000s. I then run a pooled regression, using Is2000s as a dummy variable. The data shows coefficient, then p-value. The results are:

PopChange %	Intercept	IsYear2000s	PolAvg
	7.127 (.002)	-1.18 (.31)	-10.04 (0.013)

Adjusted R-squared: 0.068
N=98

 The results show that on average the population of each state is increasing by 7.127% when the state legislature is controlled by all Republicans.  If the state legislature turns from entirely Republican to entirely Democrat, the state's population increase will fall by 10.04 percentage points so that the state will over the decade be losing population at a rate of -2.913.  The result is statistically significant with a two sided T-Test with a P-Value of .0129.  The R-Squared is low at .068 which suggests that there are many other reasons for leaving a state besides the ruling political party.  The low R-Squared is a weakness in this regression.




I tried doing another regression when I take the change in political control from 1990 to 2000  and from 2000 to 2010 and run this change against the population change.  This difference regression should cancel out fixed effects.  The results are:

PopChange %	Intercept	IsYear2000s	PolChange
	1.1 (.279)	.1091 (.939)	-7.6256 (0.234)

Adjusted R-squared: 0.0198
N=98

The results show that while the PolChange variable is large at -7.6256, the results are not statistically significant with a large P-Value of .234.  The adjusted R-squared is also very low.  I believe the reasons for this are that states like California which lost a considerable portion of their population had the Republicans make slight gains in the state legislature during the 2000s, but the Democrats still controlled over 60% of the seats.  Thus a difference regression is not going to be useful unless it can account for actual changes in control of the majority in the state.