Employment in residential and nonresidential construction during the crash

There’s been a lot of response to my argument that the loss of residential construction jobs did not play a major role in the recession, rather falling NGDP was to blame.  Bob Murphy countered with some graphs showing the level of employment in construction was quite different from building starts (or building completions for that matter.)  But what matters is employment in residential construction, so lets break out the data by category.  I will simplify this BLS data by adding both construction workers and trade workers, since most of the employment on both residential and non-residential structures is trade workers.  (These two categories include the vast majority of all construction workers, the remainder are building infrastructure.)  Here are the totals in thousands:

Category              January 2006            April 2008        October 2009

Residential                3422                        2931                 2172

Non-residential         3204                        3436                 2771

Total                         6626                        6367                 4943

Un-rate                     4.7%                       4.9%                 10.1%

If you look at the total level of construction employment, you see that data that caused Bob Murphy to reject my hypothesis.  Although housing construction was down sharply by April 2008, total construction employment had fallen only slightly.  But that data has no bearing on my claim that the decline in residential employment contributed heavily to the recession.  We need to focus on residential construction jobs.

Even in the residential sector, it is true that jobs fell more slowly than construction activity.  But I’d still argue the data strongly supports my hypothesis:

1.  Almost 40% of the job loss had occurred by April 2008, yet the national unemployment rate remained relatively low.  Those workers mostly found jobs in non-residential construction, or other fields, or in a few cases returned to Mexico.

2.  In mid-2008, economic forecasters were predicting fairly low unemployment for the year 2009, even though they already knew that housing starts had fallen much faster than housing employment.

3.  In mid-2008 commercial real estate prices were still quite strong, despite the fact that residential housing had been declining for more than 2 years.  No spillover was expected.

4.  Then NGDP fell sharply after June 2008.  Even if there had been no pre-existing subprime crisis, one would expect a sharp break in NGDP to severely depress the housing industry.  Not surprisingly, it was after mid-2008 that prices began falling in non-subprime markets like Texas.  Surely a big portion of the post-April 2008 housing downturn was caused by the fall in NGDP.  Australia did not see a decline in NGDP, and did not see a housing crash in 2008-09, despite even more inflated prices.

5.  So nearly 40% of the job losses had little effect on the national unemployment rate, and a big portion of the remaining 60 % were almost certainly caused by the drop in NGDP.  How much of the rise in the national unemployment rate can be plausibly attributed to job losses in housing not attributable to a fall in NGDP?  I’d say well under one percentage point of the more than five percentage point increase in the unemployment rate.  What do you think?

HT:  Marcus

PS.  I can’t get the BLS link to work, they are series CES2023610001, CES2023620001, CES2023800101, CES2023800201.

My flawless post showing a strong job market in 2007

Arnold Kling recently criticized my post arguing that the US job market was relatively strong in 2007 and early 2008.  I cited a low unemployment rate, and a reasonably good labor force participation rate (LFPR.)  Here’s what Kling had to say in response:

I am going to react to three things: Nick Rowe talks about the fact that housing transaction volume is higher when prices are rising; Scott Sumner’s latest attempted swindle; . . .

Scott tries to twist the labor market data to raise doubts that labor demand was weak prior to 2008. I am sorry, but I am using the same labor market indicators that I used years before the crisis, and I think that the decline in labor demand prior to 2008 is pretty evident. See this post.

I eagerly read the rest of Kling’s post looking for all the examples of how I attempted to “swindle” the reader and how I had “twisted” the data, and didn’t find a single problem with my post.  I’ll take that as implying there are no flaws.  Then I checked the post Kling linked to:

So here is, if you will, daughter of LUCY. You take total nonfarm payroll employment and divided it by the population over 16. (This is basically just an employment to population ratio.) Divided each monthly number by the peak value, reached in December of 1999, then multiply by 100. The resulting index behaves as follows:

This differs from my LFPR in several respects.  The LFPR uses population aged 16 to 65 in the denominator, whereas modified LUCY uses all population over 16.  Thus an increase in the number of women living to be 93 does not affect my indicator at all, whereas using Kling’s ratio it indicates (assuming those women don’t work) a loss of jobs.  I also believe the LFPR includes the total number working in the numerator, whereas LUCY uses nonfarm payroll employment.  I don’t object to Kling using this indicator, and never criticized his post.  I’m still a little puzzled as to why he criticized my post.  Perhaps in his next post he’ll tell us exactly how I twisted the data and swindled my readers.

Update:  Not so flawless after all.  There are several versions of the LFPR, some stop at age 65 and some go all the way up to infinity.  The BLS has no age limit.  Commenter Brent explains where the difference between Kling and I probably lies:

Actually, the labor force participation rate published by BLS uses the civilian noninstitutional population aged 16 and over as the denominator. So a 93 year-old woman will be included in the denominator unless she resides in an institution, such as a prison, mental hospital, or nursing home.

The differences between your measure and Kling’s are in the numerators. Your numerator is the number of persons classified as either employed (including the self-employed) or unemployed (that is, actively searching for work). Kling’s numerator is the number of wage and salary jobs“”that is, people with second jobs are counted twice and the self-employed are not counted at all.

The statistics suggest that second jobs grew pretty rapidly during the 1990s and declined during 2000s. It isn’t clear to me whether these changes were driven more by shifts in labor demand or in labor supply. At any rate, for the purpose of discussing missing jobs, I think people are usually thinking of primary jobs rather than second jobs, so I prefer your measure.

PS.  Just to be clear, I was not arguing that the job market in early 2008 was as strong as in 2000, just that it didn’t seem nearly as weak as the Jim Tankersley article I linked to had suggested.  I think it is quite possible I am wrong, but I’m still waiting for a commenter to successfully refute my post.

A missing explanation for why the National Journal says we have missing jobs

Tyler Cowen and Arnold Kling both linked to a National Journal article that implied America had lost (or failed to create) 15 million jobs in the last decade, even apart from the effects of recession.  Go back to April 2008, when unemployment was a quite low 4.9%.  The claim is that even then we were suffering from 10 to 15 million missing jobs.  The data is a bit vague, so it’s not clear exactly how many:

The Great Recession wiped out what amounts to every U.S. job created in the 21st century. But even if the recession had never happened, if the economy had simply treaded water, the United States would have entered 2010 with 15 million fewer jobs than economists say it should have. . . .

The forecasters said [in 2000] that the economy would create 22 million jobs over the next 10 years. At the decade’s economic peak, though, that number stood at only 7 million. Job growth in the 2000s was the lowest of any decade ever recorded by the federal government, stretching back to the 1940s. As a result, workers were extremely vulnerable to the tidal-wave recession that washed away all of the decade’s meager gains.

How could that be?  As a matter of pure arithmetic, missing jobs imply one of two things, a huge rise in unemployment, or a huge fall in the labor force participation rate.  Can you think of a third factor?

Since unemployment was only 4.9% in April 2008 (i.e. normal) there should have been a huge drop in the LFPR between 2000 (which was a red-hot boom year), and 2008.  Yet according to this graph, it looks like the LFPR merely edged down from about 67% to 66% between 2000 and 2008.  I’m surprised the decline was that small, given how the dot-com boom sucked anyone who could tie their shoelaces into the labor force.  Remember what fast food help was like back in 2000?

I notice that the LFPR rose sharply between the 1960s and 1990, I’d guess due to more women working.  Then it leveled off at about 66.5% during 1990-96.  Then it rose to 67% during the dot-com boom of 1998-2001.  Then it dropped to 66%, and leveled off until 2008.  Aren’t there lots of possible explanations for this tiny drop in the LFPR?  Recently I’ve noticed more adult women who could be working, but choose to stay home.  I don’t know if that’s a trend, but Jim Tankersley doesn’t provide us with any of the data we’d need to make sense of the claim about missing jobs.  He may be completely correct; I just can’t see where the numbers come from.

Now if you go up to 2010, then yes, I do see a worrisome loss of jobs.  This shows up as both much higher unemployment (perhaps 8 million lost jobs), and a bigger drop in LFPR (another 2 million lost jobs), both obviously related to the late 2008 plunge in GDP.  I just don’t see evidence that we had a massive jobs problem before the recession.  Does anyone know what data can support Tankersley’s claim, and why it doesn’t show up in either the LFPR, or the unemployment rate in April 2008?

I still think our unemployment problem is about 80% AD and 20% structural problems (99 week UI extension, 40% minimum wage jump, etc) but I have an open mind on the proportions.

PS.  In my earlier housing post, I should have cited this Nick Rowe post on a housing “Phillips Curve.”

Does unemployment actually lag output?

Everyone seems to think it does, so naturally I’ll argue the other side.  What surprised me is how easy it is to make the argument.  As you look at the following data, ask yourself what you’d expect to happen to unemployment if there were no lags.  Keep in mind that the trend rate of RGDP growth is around 3%:

2008:1 and 2008:2 — RGDP falls at about 0.1%

2008:3 through 2009:2 — RGDP plunges 4.1%

2009:3 — RGDP rises at a rate of only 1.6%

2009:4 and 2010:1 — RGDP rises at a 4.35% rate

2010:2 and 2010:3 — RGDP rises at only a 2.15% rate

I’d expect unemployment to rise modestly in early 2008, soar in late 2008 and early 2009, rise a bit more in the 3rd quarter of 09, fall in late 2009 and early 2010, and then rise a bit in the summer of 2010.  Here are the actual unemployment rates:

December 2007 (cyclical peak) — 5.0%

July 2008 — 5.8%

July 2009 — 9.5%

October 2009 (unemployment peak) — 10.1%

May 2010 (euro crisis begins) — 9.6%

November 2010 — 9.8%

Too early to know where it goes next, but I expect RGDP growth to pick up over the next few quarters, and unemployment to trend downward (although the recent 9.4% figure may have been a blip.  But in general isn’t that exactly the unemployment pattern you’d expect if there was no time lag at all between output and the unemployment rate?

I think the problem here is that during the last three recessions unemployment has not fallen significantly during the early stages of recovery.  One explanation for that raises no problems; growth has been slow.  The last rapid recovery we saw was in 1983, and unemployment fell almost immediately from the moment the economy started recovering.  The other issue is more complicated, productivity growth seems unusually high during recent recoveries.  Still I think it is possible to overdo this difference.  This table shows that while productivity during recent recessions has been much stronger than 1974 and 1982, there were also some fairly strong productivity numbers in garden variety recessions like 1957-58 and 1948-49.

Perhaps productivity growth was a bit higher in the early stages of recent recoveries for reasons unrelated to AD.  In that case the baseline job growth might be a bit lower, but even so any extra AD would show up as more jobs.  That could explain the close correlation in timing for the high frequency fluctuations in output and jobs discussed above, and the disappointing overall job growth.  And I still insist that this recovery is fairly weak in terms of both real and nominal GDP

Tyler Cowen has a slightly different view:

The AD-only theories, taken alone, encounter major and indeed worsening problems with the data.  Year-to-year, industrial output is up almost six percent, sales up more than six percent, but the labor market has barely improved.  How does that square with the AD-only hypothesis?  Has it been seriously addressed?

I think he is misreading the recent data on output.  Elsewhere in his post he cites productivity data showing very strong gains in 2009.  But the past 4 quarters only show 2.65% productivity growth.  If productivity growth is not high, and jobs aren’t being created, how can I explain the rapid output growth observed by Tyler?  I’d like to stick my head in the sand and deny it, but I guess that won’t do.  Seriously, I think the problem is that the industrial production data is not representative.  RGDP growth over the past 4 quarters in 3.25%.  That’s not horrible, but on the other hand it’s not that much above trend.  And unemployment has fallen a bit since the 10.1% peak of October 2009.  He’s got a point about productivity being somewhat unusual in this recession, especially 2009; but the 6% industrial production figure may overstate things.   Productivity gains in manufacturing tend to be much higher than in services, so even 6% manufacturing growth could coexist with both 2.65% overall productivity gains and also a relatively small gain in total jobs.

My previous critique of Tyler Cowen’s ZMP post was focused on one issue; I didn’t think the data supported his argument.  After reading his recent post, I don’t want to argue against the general idea that there may be some workers who (in the short run) have MPs much lower than the wage rate.  Some of the quarterly observations he discusses are very suggestive.  And I don’t have a good feel for this issue, indeed I may have read too much into his earlier post.  Tyler Cowen points out that Krugman once offered a similar hypothesis, and now rejects it out of hand, so I think Krugman may have also misread Tyler.

In earlier posts I argued that the sticky wage theory is often misunderstood.  If the Fed suddenly imposes 10% fall in NGDP, it is not true that factory workers can keep their jobs by accepting 10% wage cuts.  Why not?  Because other workers may not.  Suppose half of workers take 10% pay cuts (factory workers, etc) and half do not (teachers, health care workers, public employees, etc.)  The aggregate wage will fall 5% and we’ll have a severe recession.  During recessions people cut back on car purchases much more than health care (often paid for by insurance or Medicare), so it will be the factory workers losing their jobs, despite their willingness to take pay cuts.  Now see if that argument reminds you of Tyler Cowen’s point 7:

What does the zero MP hypothesis add?  First, the zero MP hypothesis explains why wage adjustments can’t do the trick for a lot of the unemployed, as wages won’t fall below zero.  Second, the zero MP hypothesis explains why you need steady real growth, boosting the entire chain of demand, to reemploy lots of workers and reflation alone won’t do the trick.  (I still, by the way, favor reflation because I think it will do some good.)  Those predictions are not looking terrible these days.

I agree the “entire chain” may need a boost, but I think reflation can do the trick.  On the other hand if we can’t get more people to buy cars, wage cuts for windshield makers in Toledo are not going to restore their jobs.

I completely agree with the following by Arnold Kling:

I want to reiterate that I would like to see the Fed behave as if this were an AD-caused recession. However, we should be prepared for the possibility that it is not, in which case expansionary policies will cause price bubbles in some sectors without doing much for employment and output.

What would get me to give up my AD-only explanation (actually 80% AD and 20% AS)?  Evidence that more NGDP would not result almost one for one in more RGDP.  How could we discover who’s right?  It would be simple—just create NGDP and inflation (or NGDP and RGDP) futures markets, and watch how they react to monetary shocks.  How do we identify monetary shocks?  Look for major Fed announcements, and see if the press interpretation is confirmed by market responses.  Example: Bernanke gives a speech strongly hinting at QE3.  The dollar plunges and stocks soar.  That’s a good indication the speech increased the expected future monetary stimulus.  To see how much of the recession is real and how much is demand-side we’d merely have to watch the reactions in the NGDP and inflation markets.  I say NGDP expectations would rise much more than inflation expectations.

This would be incredibly useful information to policymakers, and it would cost peanuts for the US government to set up and subsidize trading in such a market.  Why don’t they?  My wholesome and naive personality says they are well-intentioned, and just don’t know about my ideas.  Many would argue that Robin Hanson has a more clear-eyed and realistic take on what makes people tick.

What do you guys think?  Do elite macroeconomists and Fed officials enjoy presiding like high priests over a mysterious macroeconomy, or do they actually want to discover the truth?

PS.  One reason I hold my views so strongly is that even though we don’t have a NGDP futures markets, we do have enough reasonable proxies that I am pretty sure monetary stimulus “works.”

How plausible is the zero marginal product of workers hypothesis?

In a not so recent post Tyler Cowen made the following argument:

Matt Yglesias suggests the notion is implausible, but I am surprised to read those words.  Keep in mind, we have had a recovery in output, but not in employment.  That means a smaller number of laborers are working, but we are producing as much as before.  As a simple first cut, how should we measure the marginal product of those now laid-off workers?  I would start with the number zero.  If a restored level of output wouldn’t count as evidence for the zero marginal product hypothesis, what would?  If I ran a business, fired ten people, and output didn’t go down, might I start by asking whether those people produced anything useful?

It is true that the ceteris are not paribus,  But the observed changes if anything favor the hypothesis of zero marginal product. There has been no major technological breakthrough in the meantime.  If anything, there has been bad monetary policy and a dose of regulatory uncertainty.  And yet again we can produce just as much without those workers.  Think of “labor hoarding” yet without…the hoarding.

I don’t normally comment on old posts, but I was asked what I thought of this idea, which still seems to be attracting attention.  My initial reaction is skepticism. Why wouldn’t companies just lay off more workers?  But Tyler Cowen refers to the labor hoarding hypothesis, which might be able to explain that seemingly irrational behavior.

So I think we need see how the theory matches the data.  This post by Stephen Gordon shows US employment in 2010:3 falling about 5% below its 2008:1 peak, while output seems to have declined only about 0.7%.  This is what Cowen finds puzzling.

But I don’t see any puzzle at all.  If employment didn’t change, I’d expect US output to grow at about 2% a year, which is the trend rate of productivity growth.  Because we are looking at a two and a half year period, you’d expect output to grow roughly 5% with stable employment.  Now assume that employment actually fell 5%.  If the workers who lost jobs were similar to those who remained employed, I’d expect output to be flat over that 2.5 year period.  Because output fell slightly, it seems like the workers who lost jobs were slightly more productive than those who remain employed.

Do I believe this?  No, for several reasons I think they were less skilled than those who remained employed.  Labor productivity growth (assuming we were at full employment) probably slowed in the most recent 2.5 years, as investment in new capital declined.  Measured productivity continued to rise briskly, partly because technological progress continues in good times and bad, and partly because those workers still employed are somewhat higher skilled, or perhaps are trying harder in fear of losing their own jobs.  So Tyler is probably right that those workers who lost their jobs have a lower than average marginal product.  I just don’t see why zero is the natural starting point for consideration of the issue, as you only get that number by making some fairly extreme assumptions about technological progress coming to a screeching halt after 2008:1.

I’m certainly open to alternative views here.  My baseline assumption is not consistent with Okun’s Law, for instance.  And the three most recent recessions have seen slower recovery in employment than earlier recessions, although I think people often underestimate how much of that difference is because monetary stimulus has been much weaker during those recoveries (compared to a recession like 1981-82.)