Casey Mulligan draws on Census data to suggest that credit-crunch and demand-based theories of the recession are off the mark.

Another theory of the recession is that it was caused by a lack of demand — fewer employees were needed because employers were selling less to their customers. The low-demand theory is a good description of a couple of industries, like manufacturing and home construction, but if it described the economy as a whole we would have seen all types of employment cut, not just employment of people making less than $2,000 a week.

Mulligan seems more sympathetic to replacement theories:

Another set of theories say that high-paid employees are replacing low- and middle-income employees. This replacement might come from employers’ attempts to cut personnel during the recession, rather than payroll spending. For example, employers might have worried about health insurance and other employment regulation whose costs are proportional to the number of employees they have. In this view, the bank bailout did little to prevent layoffs.
Some of the replacement could come from families, as they react to some new incentives presented during the recession. Unemployment insurance reduces incentives for unemployed people to accept a new job, because, for up to 99 weeks, the insurance pays them for being unemployed.
Unemployment insurance benefits are capped, so the disincentive is much less for people making $2,000 or more a week than for people making less. Some homeowners with underwater mortgages may react by earning more, while others earn less as they recognize that all their extra earning ultimately ends up with their mortgage lender.

He then offers a chart tracking the percentage change in the fraction of adults in the state who make $300-1,200 a week vs. those who make over $2,000 a week:

The low-demand and payroll-cutting theories say that the two should be positively correlated, because the states with the largest demand reductions (or the employers most motivated to cut payroll spending) would have cut employment across the board.
But the Census data suggest that they are negatively correlated (the correlation is also negative if the small or “outlier” states shown in the chart are excluded or downweighted according to population).

Very interesting. One gets the sense that shifting away from an employer-based system of insurance coverage and relaxing various labor market regulations, including minimum wage increases, might have dampened employment losses.