I want to show that population density in the US correlates to voting habits (see attached csv).

I have data by county from the '08 election and '08 population density. Using R its easy to look at them, but what test should I use / algorithm should I apply to find:

what the "critical" population density is,
how accurately it predicts the counties vote,
and how might I visualise the data best?

currently, just playing around i have:

Code: 
library(ggplot2)
ddply(out,.(winner),function(x) {
	mean=mean(x$pop_density_09)
	median=median(x$pop_density_09)
	min=min(x$pop_density_09)
	max=max(x$pop_density_09)
	iqr=IQR(x$pop_density_09)
	lower=median-iqr/2
	upper=median+iqr/2
	return(data.frame(mean,median,min,max,iqr,lower,upper))
})

ggplot(out,aes(x=winner,y=pop_density_08))+geom_boxplot()+scale_y_log10()
ggplot(out,aes(x=pop_density_08))+geom_freqpoly(aes(y=..density..,colour=winner),binwidth=0.2)+scale_x_log10()
Thanks!

Justin

election_results_08.txt.zip