by Calculated Risk on 7/11/2007 05:47:00 PM

Many of us have been trying to understand why BLS reported residential construction employment has only fallen 4% from the peak (March 2006), even though housing completions have fallen close to 30%.

Click on graph for larger image.

This graph shows starts, completions and residential construction employment. (starts are shifted 6 months into the future). Completions and residential construction employment were highly correlated, and Completions used to lag Starts by about 6 months.

Both of these relationships have broken down somewhat (although completions have fallen to the level of starts). The time between start and completion has increased recently. But the puzzle is why residential construction employment hasn't fallen further.

The main explanations have been:
1) The BLS has not counted illegal immigrants working in construction.
2) The BLS Birth/Death model has missed the turning point in employment, and therefore the BLS has overstated the current number of residential construction employees.
3) Some construction employees have moved from residential to commercial work, but they are still being reported as residential construction employees to the BLS.
4) Some companies are "hoarding" workers for the expected recovery.
5) Many workers are still employed, but they are working far fewer hours.

Yesterday, Greg Ip at the WSJ reviewed an analysis from Deutsche Bank economists suggesting that the illegal immigrant explanation accounts for most of the misssing job losses.

In a new report, economists at Deutsche Bank estimate construction employment should have fallen about 900,000 since early 2006 when in fact it’s only down 150,000. They conclude 500,000 of the unexplained gap is attributable to layoffs of illegal Hispanic workers.

The economists addressed most of the other explanations listed above:

Deutsche Bank dispute many competing explanations for the disconnect between home building and construction employment. Errors in the Bureau of Labor Statistics’ “birth/death” model, which estimates job creation and destruction at new firms, can only account for 20,000 construction jobs, they say. It’s unlikely many workers now classified as employed in residential construction have moved to commercial work because the type of work is so different, they say. Third, the notion that because of lags, the layoffs have yet to come, is “starting to wear thin ... We are now 1.5 years into the slowdown .... We find ourselves increasingly skeptical with the notion that construction companies have not yet recognized the severity of the situation.”

Although uncounted illegal immigrant workers might account for some of the puzzle, this analysis is not very satisfying.

I'm not sure why the Deutsche Bank economists think errors in the birth/death model can only account for 20,000 construction jobs. We know from the Business Employment Dynamics Summary that 77,000 construction jobs (NSA) were lost in Q3 2006, and yet the BLS reported a gain of 5,000 construction job (NSA) for the quarter. That is a difference of 82,000 jobs (1) (NSA) and we don't know the errors for Q4 2006, and Q1 2007 yet.

This brings us to a second potential flaw in their analysis. The economists wrote:

the notion that because of lags, the layoffs have yet to come, is “starting to wear thin ... We are now 1.5 years into the slowdown ..."

Actually employment tracks completions, not starts - and completions only fell off the cliff starting at the beginning of 2007. It is very possible that many workers are still employed - for now - but are working reduced hours. This would be combination of explanations 4 and 5 above.

And if the BLS missed by 82,000 workers (NSA) in Q3 2006 before completions started falling off a cliff, perhaps they missed by many more in Q1 and Q2 2007 when completions were dropping rapidly.

And finally, I believe there is some merit to the argument that workers have moved from residential to commercial work, and are still being reported as residential workers to the BLS. Commercial construction spending has increased 25% from January 2006 to May 2007, but the number of workers in non-residential construction has only increased 4% over the same time period. This raises the opposite question: why hasn't the boom in commercial construction created more jobs?

I think the answer will be combination of these explanations.

(1) Note: In a previous post, I used SA instead of NSA numbers and overestimated the BLS error.