International comparisons of construction employment in the recession

Many participants in the ongoing discussion ([1], [2], [3], [4], [5], etc) on international comparisons of the evolution of output and employment have raised the point that construction employment has played a more prominent role than usual in the most recent recession.

I've put together a few graphs that may be useful. I'd also like to point out that the original problem that Nick put forward was to come up with a story for why the US was so different that didn't amount to saying 'The US is different.'

Construction employment in the G7:

Construction employment in the G7 plus Spain and Ireland and minus France, Italy and Japan:

Ex-construction employment in the G7:

Ex-construction employment in the G7 plus Spain and Ireland and minus France, Italy and Japan:

The ratio of construction and ex-construction employment in the G7 plus Spain and Ireland and minus France, Italy and Japan:

 

I don't have a fully-formed story to tell here, but maybe these will help move the discussion along.

Comments

[Nick Rowe]

Thanks Stephen. I think this does move things along.

For starters, these graphs, while not totally inconsistent with my own story, are not fully consistent with it either. My story doesn't fit the facts nearly as well as I hoped it would.

Spain, the US, and Ireland (in that order) all broke Okun's Law. Spain had the highest rate of productivity (GDP/employment) growth during its recession, followed by the US, followed by Ireland. All three countries stand apart in your last graph above, in having the biggest declines in the ratio of construction to non-construction unemployment. So far so good. But the order isn't right for my hypothesis. Ireland had the biggest decline, closely followed by Spain, followed at a great distance by the US.

But look at Ireland and Spain. Wow! The whole countries must have been one big construction site. Which then closed down

[Patrick]

"The whole countries must have been one big construction site"

Alternatively, it was the only game in town? I suppose it comes to the same thing.

FWIW, it seems to me that Kling's recalculation story seems to make some sense for the US. Clearly people who made leveraged commitments based on the outlook of 2005 where reconsidering (i.e. running for the exits) in 2008/2009.

[Patrick]

Ugh.

*were* reconsidering.

[Determinant]

Reports from Spain made its seem like one big construction site. There was a decent trade in second homes for British expats, though given the EU I don't know if expat is the right word. But news reports do give the impression that there was a great deal of speculative development

[Nick Rowe]

Thinking again about my hypothesis (that it's the big decline in construction employment that caused some countries to have increasing rather than declining GDP/employment during the recession):

Spain and the US fit my hypothesis.

Spain had a very big decline in construction employment, and also had a very big increase in productivity. Check. I got that right.

The US had a big decline in construction employment, though not as big as Spain, and also had an increase in productivity, though not as big as Spain. Check. I got that right.

But Ireland doesn't. Ireland had an even bigger decline in construction employment than Spain, but only a small increase in productivity. I got that wrong. It's Ireland that doesn't fit my story.

[Highlander]

Would it be valuable (or possible) to tease-out FIRE employment and analyze its relative change over the same period? Maybe the Ireland productivity quandary is a function of the relative change in construction-related employment PLUS construction employment. (I.E. Did US and Spain FIRE employment fall by a greater degree than what was experienced in Ireland?)

[Stephen Gordon]

Unfortunately, no - at least, not with the OECD source I've been using. They only break it down by 2 or 3 sectors, and construction happens to be one of them.

Too bad, because that would be interesting as well.

[Phil Koop]

Nick: there is another problem with your hypothesis, which is that ex-construction employment in Ireland, Spain and the US shows the same exceptional qualitative pattern as construction employment. So even if you are right, that can explain only a part of what is funny about the these countries. And anyway, can construction really account for the anomalous productivity of the US? Comparing back to Stephen's original graphs, it doesn't seem to comprise a large enough fraction of employment/GDP to do this.

[Nick Rowe]

Phil: Yep. That's the trouble with empirical economics. It's too damned easy to see your theory doesn't work. Construction might still be part of the story, but it's nowhere near the whole story I once thought it was.

[Phil Koop]

THe WSJ has posted some detailed statistics about US unemployment by profession: http://online.wsj.com/article/SB10001424052748703791904576075652301620440.html. For what it's worth, it does seem that US construction workers had fairly high unemployment rates to begin with.

[Kosta]

Nick wrote: But Ireland doesn't. Ireland had an even bigger decline in construction employment than Spain, but only a small increase in productivity. I got that wrong. It's Ireland that doesn't fit my story.

But Ireland differs from Spain and the U.S. in the fact that after the bubble burst Ireland almost immediately went into austerity, while Spain and the U.S. have had more accommodating stances (the U.S. employed a fairly large fiscal stimulus). This likely accounts for much of the differences in the ex-construction employment, particularly in the second half of 2009.

[Nick Rowe]

Kosta: that might explain why Ireland had a deeper recession. But that's not what we are trying to explain here. We are trying to explain the ratio of GDP/employment, and whether it can be explained by the ratio of construction/non-construction employment.

[Kosta]

Hi Nick, I made a resolution to be less verbose on these blogs, but in this case, I think brevity was a mistake as I didn't clearly explain my assumptions. Essentially what I am arguing is that the depth of recession directly impacts the GDP/employment ratio (although I have been focusing on construction/non-construction employment).

Here's the model I'm thinking of. For countries with a large construction sector, the recession initially leads to a disproportionate number of low productivity construction workers losing their jobs and the GDP/employment ratio outperforms (this is your model). As the recession deepens (i.e., Ireland's case) high productivity workers start to lose their jobs (they could be construction or non-construction employees) and the GDP/employment ratio underperforms. Alternatively, as the recession deepens, the high productivity employess could start to lose their productivity as companies hoard labour. I guess I'm arguing that for countries with large construction sectors, Okun's Law is violated initially, but it starts to take effect during very deep recessions.

However, going back to the charts in your previous post on Spain and Ireland (http://worthwhile.typepad.com/worthwhile_canadian_initi/2011/01/spain-is-even-more-exceptional-than-the-us.html), it is interesting to compare the trajectories of the Employment:GDP ratios from 08:4 forward. Going forward from this date, Employment:GDP growth of the U.S., Spain and Ireland is largely the same. The difference between the three nations comes from their performance from 07:4 to 08:4, a period when Ireland's economy was already contracting.

As an aside, German growth in this ratio matches the above three nations from 09:1 going forward.

[need better stats]

It would be useful to have the graphs for different sectors. What are the three biggest construction sectors? Housing, infrastructure and commercial? In the USA, rapidly expanding upper-middle class housing expansion is a bubble unless they go back to a pre-Reagan economy. While in Chindia and some of Canada/Australia, some construction might be inlfationary (maybe bad if promotes future bubble whereas recessioned countries aren't worried about this; need different graphs).