Among the more interesting reactions to today's decent jobs report (146,000 new jobs in November, unemployment down to 7.7 percent) was Matthew Yglesias's post at Slate, which notes a startling divide between housing starts and construction sector employment:

You should never get too worked up about one months' worth of anomalous economic statistics since there are always sample errors and other statistical noise to worry about, but the divergence between new housing starts and residential construction employment looks to have been going on for months now. In the chart above, the blue line is new starts of housing projects—an index that's clearly ticked well up over the past six months. The red line is employment in residential construction, which has been flat or even falling.

So what's going on? Is something being mis-measured? Is the construction industry experiencing a productivity boom? Are hours per worker surging but firms aren't adding staff for some reason?

Here is his graph:

This is an intriguing question, resolved by expanding the graph to include an additional decade's worth of data. Dividing the number of housing starts by the number of construction workers, a clearer picture emerges:

The housing market provided enough jobs for an average 2.1 residential construction workers per housing start per month from 2000 to 2006. When the housing market collapsed in 2006, taking the rest of the economy down with it, monthly housing starts plummeted. The number of construction jobs quickly followed, making construction one of the highest unemployment sectors during the recession.

Now, housing starts are rising again as real estate prices begin to recover. But while the trends are positive, we're still a long ways from the two construction workers-per-housing start-per-month equilibrium of the early 2000s. A basic polynomial regression suggests the labor market for residential construction workers may not begin to improve appreciably until early 2014.