Every state gets money from the USA and every state gives; but the amounts given and received don't balance. The state that gets the most per amount given is New Mexico (which gets $2.03 for each dollar spent), the state that gets the least is New Jersey (which gets $0.61
More below the fold
At this site the Tax Foundation provides data on tax burden by state (note: I do not support the Tax Foundation - but this data seems sound) (also note, this is from 2005).
What might determine which states give vs. get the most? Well lots of things, but I chose these:
Poverty - states with low median incomes might get more
Republicanism - given that this data is from 2005, maybe states that supported Bush got more
Senatorial clout - hard to measure, but I used total years seniority of the two senators, as of 2004
First, a table of all the states, and what they spend for what they get
NJ 0.61
NV 0.65
CT 0.69
NH 0.71
MN 0.72
IL 0.75
DE 0.77
CA 0.78
NY 0.79
CO 0.81
MA 0.82
WI 0.86
WA 0.88
MI 0.92
OR 0.93
TX 0.94
FL 0.97
RI 1.00
GA 1.01
IN 1.05
OH 1.05
PA 1.07
UT 1.07
NC 1.08
VT 1.08
IA 1.10
NE 1.10
WY 1.11
KS 1.12
AZ 1.19
ID 1.21
TN 1.27
MD 1.30
MO 1.32
SC 1.35
OK 1.36
AR 1.41
ME 1.41
HI 1.44
MT 1.47
KY 1.51
VA 1.51
SD 1.53
AL 1.66
ND 1.68
WV 1.76
LA 1.78
AK 1.84
MS 2.02
NM 2.03
Next, some graphs:
First is a scatterplot of Median income (on the x-axis) and amount gotten per amount given to the feds, on the y-axis
the straight line is a linear regression line, the bendy line is a loess fit. There are boxplots of each variable, below and to the left of the graph, and states that are unusual are labeled.
There is clearly a relationship between these variables, it seems to be stronger at the extreme low income (left side of graph). Alaska, Hawaii, Virginia and Maryland get more than expected for their income; Nevada gets less. Mississippi and New Mexico get more than a linear model would predict.
Next, a similar graph, but with senatorial seniority
There is a relationship, and it's very nearly linear, but not nearly as strong as the one for income. Arkansas, Louisiana, Alaska, Mississippi and New Mexico get more than predicted; Nevada, Delaware, Connecticut, and Massachusetts get less than expected.
Finally, one with Bush vote:
Even less of a relationship, and it appears to be nonmonotonic .. that is, predicted money goes up as Republican vote goes up, but at the high end, it starts declining.
Now, let's put all three variables into one model.
First, let's try a linear regression model. That gives us:
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 1.722582 0.455551 3.781 0.000448 ***
Senate.seniority..2004 0.010258 0.002314 4.433 5.73e-05 ***
Med..Income - 0.030378 0.006270 - 4.845 1.48e-05 ***
X.Bush.2004 0.007983 0.004716 1.693 0.097293 .
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 0.2413 on 46 degrees of freedom
Multiple R-squared: 0.59, Adjusted R-squared: 0.5633
F-statistic: 22.07 on 3 and 46 DF, p-value: 5.315e-09
What does all this mean?
Well, if we want to predict the "whogets" variable, the we would use this equation:
0.01026*seniority - 0.03038*median income (in thousands) + .007983*%voting for Bush in 2004. The R^2 means that 59% of the variation in whogets is determined by this linear model.
[Update [2009-4-29 7:5:25 by plf515]:] in the comments, NCrissieB brings up an interesting point that maybe swing states get more than they should. I will investigate this.
UPDATE I computed a 'swinginess' variable (it was 30 - the absolute value of the difference between the Bush % and 50) and there is very little relationship between this and the DV. Oh well. Good idea to investigate.