## Cost-of-living confusion

A comment on Paul Krugman’s blog led me to this chart compiled by Steve Sailer of median income and cost of living by state. Sailer worked with median income for families of 4 and the Accra cost-of-living index and found the highest cost-adjusted incomes to be in Minnesota, Illinois, and Wisconsin, and the lowest in Hawaii, California, and New Mexico.

Awhile ago I was looking at state incomes over time and found this:

The lower-right graph plots cost-of-living-adjusted income vs. unadjusted income by state. My results are completely different from Sailer: I have the richest states as Connecticut, followed by Maryland and then Virginia. In general, my adjustments were much smaller than his. The two differences: I was using individual income, not family income, and I was using the Berry, Fording, and Hanson cost-of-living index, which makes much smaller adjustments. For example, the Accra index used by Sailer gives Minnesota a cost-of-living index of 100 and California gets a 151. In contrast, the Berry et al. index gives Minnesota a 105 and California a 109. Actually, if housing is included, this doesn’t seem right at all! I’d just made that graph without thinking too carefully about where these numbers came from. Can anyone offer some guidance here?

P.S. I recognized that cost-of-living adjustments do not and cannot cover all. In particular, if it costs xx more to live in New York than in Iowa, but people are still living in New York, there’s probably a reason for it.

If the y-axis is a log scale, does it still make sense to say trends in absolute income?

2. c. says:

Is there a reason for using average income here, as opposed to median (which Sailer's income statistics are)?

3. Andrew says:

When the actual axis numbers are on the original scale I think it's clearer just to label as income.

C,

I used what was easily available. But the key differences are in the cost of living measures, not the incomes.

4. Aleks says:

Andrew, interesting post! You've controlled for the global income trend. But one could also control for profession: clearly some states have more of a certain type of job than others. And this would be an interesting case for the controlling-for problem.