Sources of rankings used as input for the correlation and principal component analyses:

Business costs and labor force rankings: Forbes Magazine, "The Best States for Business", Aug. 16, 2006.  States with lowest business costs ("cost of labor, energy, and taxes") have rankings closest to 1.  The labor-force ranking ("educational attainment, net migration, and population growth") was inverted so that the "best" state for business had a rank of 50.

Cancer risk from air pollution:  Number of additional cancer cases per million residents expected to be caused by air pollutants, from  Scorecard.com

Green capacity index:  A score computed from 24 variables (encompassing information such as development of climate-change action plans, support for public transport, and protection of wild landscapes) and expressed on a 1-to-100 scale in the Resource Renewal Institute's (RRI's) 2001 report (pdf) entitled "Assessing the Capacity of the States to Achieve Sustainable Development Through Green Planning"

Population:  2005 estimates by the US Census Bureau

Median income:   2002 estimates by the US Census Bureau

Income inequality:  The Gini coefficient, by state, from a fascinating 2006 working paper by James Galbraith and Travis Hale at the University of Texas Inequality Project (pdf).

Number of Wal-Mart Supercenters:  From (where else?) Wal-Mart

These eight variables for each of the 50 states were used to compute the rank correlation coefficients for all pairs of variables.   The correlation coefficient varies between -1.0, when a group of 'objects', in this case states, rank exactly opposite for the two variables, and 1.0, when they rank exactly the same.  The correlations discussed in the article were in the range 0.40 to 0.75 or -0.40 to -0.75 and were statistically significant (that is, they had a very low probability of having simply occurred by chance).

The 28 correlations among all pairs of the eight variables were used in a principal component analysis, which computed eight uncorrelated principal components, each one of them represented by an axis running through the eight-dimensional space defined by the variables.  The first graph in the article shows the first two of those components, which condensed almost 3/4 of the total variation in the original eight-dimensional space into the two dimensions shown. 

The eigenvectors for those two components, which show the degree of relationship between each component and each of the original variables, are as follows:

	Principal component 1 (Horizontal axis on graph)	Principal component 2 (Vertical axis)
Business costs	-.39	-.04
Labor force	-.31	+.38
Cancer risk	-.40	-.28
Green capacity	-.39	-.02
Population	-.26	-.54
Income	-.41	+.26
Inequality	+.12	+.59
Supercenters	-.42	+.25

A state's value for a principal component is computed by multiplying each coefficient above by the state's corresponding standardized value and adding up those eight products.  Values of the first principal component varied from -3.8 for New Jersey to +3.5 for Wyoming, and the second varied between -2.4 for Louisiana and +2.7 for Vermont and New Hampshire.  States were plotted using their US Postal Service abbreviations.  To form the four divergent groups of states shown in the second graph, I first drew a circle of radius 1.5 centered on the graph's central point (0,0) and deleted all states within the circle.  I also deleted Alaska, which lay exactly at the 0 value for principal component 1. 

The groups were assigned as follows:

	Principal component 1	Principal component 2
Deep Blue	negative	positive
Metropolis	negative	negative
Big Sky	positive	positive
Greater Dixie	positive	negative

Means of each of the four groups were then computed for the eight variables used in the principal component analysis and for eight additional variables.  All differences between means that were mentioned in the article were statistically significant at the 5% probability level.

Sources of additional data:

Composition of state Congressional delegations:  http://clerk.house.gov/members/olm109.html  and
http://www.senate.gov/general/contact_information/senators_cfm.cfm

Minimum wage:  http://en.wikipedia.org/wiki/List_of_U.S._state_minimum_wages

Unemployment rate:  http://www.bls.gov/news.release/laus.t03.htm

Percentage of population in poverty:  http://www.infoplease.com/ipa/A0104529.html

White, non-Latino population:  http://www.eeoc.gov/stats/census/majorgroups/us/us_state_rank.html
(excluding Hawaii)

Voter turnout:  http://www.census.gov/population/www/socdemo/voting/cps2004.html

Federal civilian employment:  http://ftp2.census.gov/govs/apes/02federalstate.txt
(omitting Maryland; that state and Virginia have huge numbers of federal employees, a spillover from Washington, DC, which was not included in any of these analyses).

Military deaths in Iraq:  http://icasualties.org/oif/ByState.aspx

Percentage of evangelicals: http://www.beliefnet.com/politics/religiousaffiliation.html

Except for the population figure, each of the values shown in the article is a mean of the state values, with each state given equal weight.  Therefore, the mean shown is a mean of the states, not the mean for the group of state's population as a whole.  The "Republican share of states' Congressional delegations" was calculated by computing the percentage of each state's two Senators who are Republican (0, 50, or 100%) and the percentage of its House members who are Republican and averaging those two percentages.  That number will be different, perhaps very different, in 2007.

Footnote:  One comment that might be made is this: "Well, blue states may have higher wages, but maybe that's because they're more expensive to live in."  While estimates of cost-of-living in different cities are not hard to come by, statewide estimates are.  Big cities have a slight tendency to be bluer, and to be more expensive than other places.  Of the 18 most expensive US cities, 7 are in red states and 11 in blue.  But that still leaves out most of the country.  I did include what I think was the more relevant characteristic for this analysis: the cost of doing business, which is definitely higher in blue states.  But remember, a big part of what a business considers a cost -- wages -- represents your means of survival if you don't own a business.  The fact that Manhattan has an astronomical cost of living is of little relevance to working people living in blue Portland or red Birmingham.  No state is easy to live in these days for households below the median income, and it's awful for those below the poverty line.

Table for the eight variables that went into making the map:

	Deep Blue	Metropolis	Big Sky	Greater Dixie
Wages, taxes, energy costs (from Forbes, 50 = highest)	36	38	25	9
Labor force size/education (from Forbes, 50=highest)	42	27	18	7
Air pollution cancer risk (added cases per million)	624	940	243	424
Green capacity (from RRI, 0 to 100 scale)	48	47	23	19
Total group population	32 million	138 million	9 million	32 million
Median income	$49,600	$46,600	$38,000	$38,700
Income inequality index (higher = greater rich-poor gap)	42	46	39	45
Wal-Mart Supercenters per million residents	0.3	5.4	12.9	16.7

In addition, here are some other measures that didn't go into making the map but vary strongly among the groups:

	Deep Blue	Metropolis	Big Sky	Greater Dixie
Voter turnout, 2004	68%	63%	68%	61%
Republican share of states' Congressional delegations	44%	43%	63%	63%
State minimum wage	$6.71	$6.45	$5.44	$5.22
Unemployment rate	4.1%	4.7%	3.4%	4.8%
Percentage in poverty	9%	12%	11%	16%
White population, non-Latino	85%	64%	91%	76%
Federal jobs per million residents	7910	7800	9140	9410
Iraq war deaths per million residents	9	8	12	15
Percentage Evangelical	21%	19%	25%	38%