Believe those who are seeking the truth. Doubt those who find it. Andre Gide

Wednesday, October 19, 2011

The great sectoral shock of 2006

There's a lot to disagree about when it comes to understanding the great recession and subsequent slow recovery. But at least one thing seems clear: the U.S. real estate sector has played, and continues to play, a significant role in the ongoing saga.

In this post I want to talk about an old theme: the idea that a shock in one sector of the economy can reverberate throughout the entire economy. This theme was highlighted by John Long and Charles Plosser in their classic paper Real Business Cycles (JPE 1983). But it was a view that really never gained much traction. The one-sector neoclassical growth model was, and continues to be, the preferred basic framework of modern macroeconomic analysis.

The fact that most economic sectors tend to expand and contract together has something to do with the willingness to approximate the economy as consisting of one sector. But in doing so, one is then led to search for explanations in terms of hypothetical "aggregate demand" and "aggregate supply" shocks. And so we hear today that the slow recovery is attributable to "deficient demand," and that this explanation is borne out by the survey responses of businessmen who report that their main problem is a lack of product demand. (I have spoken about the pitfalls of this interpretation here.)

In any case, what I want to talk about here is the work of my colleague Juan Sanchez (with Constanza Liborio) about the role of the construction sector in the great recession. Let's start off with a look at U.S. construction sector data over the period 2005-2010.

Construction sector GDP (value-added), gross output, and employment are all normalized to 100 in the year 2006. All variables decline by about 30% over the next four years. Note that the U.S. economy officially went into recession in December 2007. The decline in the construction sector occurred over a year before that.

The Direct Effect

Construction sector employment at it's peak in 2006:Q1 was 7,651,000 workers. By 2010:Q4, employment in this sector shrank to 5,505,000 workers; a decline of 28.1% (2,146,000 workers).

Total employment in 2006:Q1 was 135,401,000 workers. By 2010:Q4, total employment fell to 130,128,000 workers; a decline of 3.9% (5,273,000 workers). 

Consequently, 40.7% of the decline in total employment over this period of time is directly attributable to the employment losses experienced in the construction sector. 

This sounds like a big number--and it is. But there is reason to believe that it is, in fact, just a lower bound. That's because this direct effect is likely to have implications for product demand in other sectors that supply the construction industry. 

Indirect Effects

The following data shows the share of a sector's output used in the construction sector (for the year 2006). 

The way to read this graph is as follows. In 2006, roughly 3% of what was produced by the U.S. mining sector was used as an intermediate input in the U.S. construction sector; and so on. According to this data, Both manufacturing and retail depend heavily on product demand generated by construction.

Quiz: A business manager at the local Home Depot (retail sector) reports that her regular customers in the construction sector have scaled back their purchases. She would like to hire more workers but the problem, she explains, is a "lack of demand" for her store's product. It follows immediately and conclusively that the problem with the U.S. economy is "deficient demand" and that government stimulus is needed. Answer true/false/uncertain; and explain.

Input-Output Analysis

To get a rough sense of how the collapse of the construction sector may have spilled over into the rest of the economy via the intersectoral linkages described above, Juan uses the BEA input-output tables to construct a simple model.

To be conservative, he applies the actual decline of output in the construction sector from 2006-2007 (average of these two years) to 2009. Evidently, the effect will be larger if one considers the change from 2006 to 2010.

The result of this simple simulation exercise suggests that the collapse of the construction sector accounts for 46.4% of the decline in U.S. GDP and 51.9% of the decline in total employment (roughly 3.4 million jobs).

The following bar graph summarizes his results.

Juan uses the same model to ask how much of the expansion during 2002-2006 was attributable to construction. His analysis suggests that construction played a much smaller role in the expansion, accounting for only 7.4% of the increase in GDP over this period.


The rosy expectations that drove residential investment prior to the recession turned out to be overly optimistic. The "overbuild" in residential capital needs time to work off, through depreciation and population growth. The adjustments taking place in this sector will take time. To the extent that other sectors are tied to the fortunes of the construction sector, economic activity throughout the economy is likely to remain relatively depressed. What can or should be done about this remains an open question.

PS. If you would like to contact Juan to learn what he did in greater detail, send him a message here:


  1. Have you seen the work by Acemoglu et al that looks like it is related to this ?

  2. This is fine as far as it goes. However, it misses the balance sheet aspect of what happened. Leverage is great on the way up, horrible on the way down as it compounds the move. What is more, it ripples into other areas. It would be nice if economics started factoring that into the models...

  3. David,

    The broad use of the data from 2006-2010 overlooks some interesting variations that occur within that time frame. In particular, construction employment starts falling around April, 2006 but the rest of the economy's employment continues to grow through about April, 2008. In other words, the rest of the economy seemed to be doing fine (and maybe were absorbing some of the unemployed construction workers) for two years after collapse in housing. Only in mid-2008 do things start looking ugly, and they look ugly everywhere. It is hard for me to see the importance of housing and construction by itself in this crisis.

    I did a post awhile back where I show this graphically. On the link below, eee the third figure down:

  4. "only in 2008 do things look ugly."

    Calling James Hamilton
    Oil! $147 barrel will do that to you! See 1973,1979-80

  5. David (Beckworth):

    I do not believe the analysis here is necessarily inconsistent with your own hypothesis:

    Thus, the housing bust at best would have led to a mild recession. It was the failure of monetary policy that created the Great Recession.

    Juan's study here is suggesting that construction might account for up to 1/2 of the decline in GDP and employment. There is still another 1/2 to explain. I do not agree with you that it was a failure of monetary policy; but there is obviously room for other factors.

    One thing that Juan's study abstracts from is the effect that the collapse in the real estate sector had on the financial collapse (as per the comment by j.c. above). Once one factors in the role of MBS in the shadow-banking sector, the run in the repo market, etc., even more "blame" might be attached to the events relating to this one sector.

  6. When do you think this will be in working paper form? Very interesting stuff.

  7. Arguably, the asset price effects of the housing downturn were more important than the direct output effects. Falling house prices impaired collateral and restricted credit intermediation. The effects on credit reverberated throughout the economy, and cannot be captured through "sectoral" analysis.

  8. "The effects on credit reverberated throughout the economy, and cannot be captured through "sectoral" analysis."

    Partially agree, but sectoral analysis is still vary valuable. Some sectors have much more impact on credit than others, and reverberate much more strongly.

    What the sectoral analysis can show is that there is a repetitive character to recent financial bubbles, but that bubbles in different sectors have different magnitudes, depending on their effects on household leverage. There have been a series of bubbles since at least 1987, but most of them were more narrowly restricted in the U.S. to the investment community: LTCM, Latin America, Asia, Russia.

    In the 19th century, the most widely dispersed asset was land, and land speculation influence most economic downturns; during the 20th century, it was housing.

    The cumulative effect of the bubble depends on the number of those holding the asset, and the cycle of a depression depends on how deeply indebted households are, and how long it takes them to get from peak borrowing to deleveraging from the trough. This one may take a very long time.

  9. Hi David,

    How would you respond to Alan Blinder's claim in today's WSJ?

    "By now, massive underbuilding during the slump far exceeds the overbuilding during the boom. So, by rights, a shortage of houses should be pushing up house prices, incentivizing home builders, and boosting growth in gross domestic product. Instead, actual and prospective foreclosures hang over the housing market like a wet blanket."

  10. Casey Mulligan has argued that the mortgage modification plans have hampered the labor market. Ohanian documents through business cycle accounting that the gdp reduction is driven by a decrease in labor input, not a TFP shock. So this would be consistent with the sectoral shock thesis too.

  11. interesting post. thanks for highlighting juan's work.

  12. Greg Hill: It's hard to know what to say Blinder's comment. It would be nice to see some evidence supporting the claim, but of course, maybe that's too much to ask for an op-ed piece. As for the rest of the article, I think he has some interesting ideas.

  13. David, I think the new format needs a bit of work. At least for me it takes much longer to load this page. And I am not seeing the old link to most popular posts.
    Do you really mean to have all the posts on one page ?

  14. Does anybody else want to chime in here? I'm not sure what to make of this format either. Anon...there are several viewing options at the top of the page. To see past posts, click on the "Flipcard" view, for example.

  15. Yes, the viewing options are helpful. Thanks.

  16. I like the 'classic' format the best. It's easy to read for old people like me.

    See the link for some skepticism on the role of sectoral reallocation in our current troubles. There have been plenty of construction busts before in our history, which were not accompanied by a rise in the natural rate. For the construction bust to have moved the natural rate, either the transmission mechanism has to have changed relative to normal, or the shock causing the most recent construction bust has to have come from an exotic source.

  17. I find this stuff very interesting. The Long/Plosser ideas were picked up by Horvath and by me in the 1990s (most of my work never got pubblished) and then more recently by Vasco Carvalho (in some really outstanding work). Kling's notion of the macroeconomy is related too.

  18. I want to thank you for posting such information. The topic above helped me understand recession and its effect.

    Part Time Jobs in Brisbane

  19. It's now 2012 and things haven't improved in real estate much. i think it will go lower over the next 2 years.