No correlation between college and income

This morning I was reading some articles from the Utah Department of Workforce Services' Trendlines publication. One article discussed the increase in median household income for the six most populous counties in the state. Another discussed the educational attainment of the same counties.

Naturally, I wondered about the correlation between the two. After all, if the people making the most noise about college are correct, there should be a distinct positive correlation. Right?

I did a quick regression analysis on the data. The first article listed median income for 2005 and 2007. The second included the percent of the population older than 25 that had at least a bachelor's degree. Since I don't know how current the data from the second article is, I did two analyses: one for 2007 and another for 2005.

Here's the data that Gnumeric (my favorite spreadsheet application) produced for me:

2007

County	% B.A.	2007 M.I.
Cache	34.2	$45,695
Davis	31.6	$65,686
Salt Lake	30	$56,350
Utah	34.3	$57,408
Washington	19.6	$46,822
Weber	22.5	$52,155

2005

County	% B.A.	2005 M.I.
Cache	34.2	$41,097
Davis	31.6	$56,809
Salt Lake	30	$48,068
Utah	34.3	$47,428
Washington	19.6	$43,980
Weber	22.5	$49,107

Regression (2007)

SUMMARY OUTPUT
Regression Statistics
Multiple R	0.3763818345189
R Square	0.14166328535581
Adjusted R Square	−0.07292089330524
Standard Error	7718.04054428782
Observations	6

Regression (2005)

SUMMARY OUTPUT
Regression Statistics
Multiple R	0.08550799026314
R Square	0.00731161639884
Adjusted R Square	−0.24086047950145
Standard Error	5948.6234792109
Observations	6

If you don't know and/or care much for statistics, the key value to look for is "R Square". The closer that number is to 1, the more perfectly correlated the data. The closer to 0, the less the numbers have to do with each other. For 2007 that number is about 0.14. For 2005 it is about 0.01. Even the 2007 number does not indicate any correlation.

I, a devout doubter of the return on investment of college, was surprised by this. No correlation at all? Amazing. I'll grant that such a small sample (n=6) doesn't mean much, but I found it interesting nonetheless.