Instead of a Peek, Scott Sumner Should Have Taken a Ponder
I am so naive. In response to what I think was a magnificent defense of Arnold Kling against Scott Sumner (and Brad DeLong), I was expecting Scott to either blow me up, or to admit that I had a good point and the Austrian theory went up one notch in his book.
But alas, all Scott did was elaborate on one of my own points, while thinking (a) this somehow helped his case and (b) that he was telling me something I didn’t realize.
I am truly strapped for time, so I can’t spell this out completely. In the interest of brevity, I’m going to simply repost the relevant sections from my article, and then give Scott’s response (which to repeat, in my mind is a simple elaboration of my own point). Then I’ll end with my wise-aleck comment that I left on Scott’s post.
So here goes. Remember the context, Scott thought he had dealt a crushing blow to the Austrian/Recalculation explanation of the recession, when he noted that housing starts fell by more than half from January 2006 to April 2008, whereas the national unemployment rate barely blipped upward. Scott said, “So housing starts fall by 1.3 million over 27 months, and unemployment hardly changes. Looks like those construction workers found other jobs, which is what is supposed to happen if the Fed keeps NGDP growing at a slow but steady rate.”
(1) I rushed to the rescue by arguing the following:
There’s just one problem with Sumner’s argument: housing starts are not synonymous with construction jobs. In fact, from January 2006 to April 2008 — the period when Sumner thinks construction workers must have been laid off in droves because housing starts fell by more than half — US construction employment only fell from 7.6 million to 7.3 million.
Then I went on to say:
In the years leading up to 2006, housing starts steadily expanded. Now if they had simply leveled off in January 2006 — so that (annualized) housing starts every month thereafter remained at the permanently high plateau of 2.3 million — construction employment would have continued rising after that date.
The reason is that there’s more to the construction sector than simply starting new houses. That fact alone buys us another few months, as a look at housing completions shows.
But more important, construction workers are needed to maintain an existing stock of housing. In other words, if the builders had kept constructing new homes at an annualized rate of 2.3 million from January 2006 onward, and if families moved in to them as in normal times, total construction employment would have needed to rise above its January 2006 level. Those workers were needed to keep cranking out the brand new houses, and it would have taken new workers (siphoned from elsewhere in the economy) to, say, add a new deck in the backyard or build a shopping mall down the street from a new housing development.
I am an economist, not an expert in housing or construction. I do not pretend to know exactly what construction workers were doing in the two years after housing starts peaked. But what I do know is that Sumner is wrong in his assessment of the labor markets. Contrary to Sumner, there is no huge reallocation of construction workers (from January 2006 to April 2008) that Kling or the Austrians must explain.
Note that the part I’ve put in bold above is new emphasis; this is my point that Scott himself will elaborate in his own response.
(2) Now here’s Scott’s response to my article, in a post titled “Bob Murphy wrongly assumes I won’t peek”:
Bob Murphy responded [to my point about housing starts like this]: “The reason is that there’s more to the construction sector than simply starting new houses. That fact alone buys us another few months, as a look at housing completions shows. [Bob has a link here] But more important…”
Waaaait a minute. Note how Bob tells me housing starts are the wrong data, because construction workers keep working for some time after the starts, and then tells us to look at housing completions. But then he merely provides a link, moving right along to something “more important.” I wonder why? Perhaps because housing completion data also supports my view? Here are the numbers. I’ve also averaged the two, as the average of starts and completions might be a good indicator of ongoing activity.
January 2006: starts 2,303,000 completions 2,058,000 average 2,180,000
April 2008: starts 1,008,000 completions 1,014,000 average 1,011,000
October 2009 starts 527,000 completions 745,000 average 636,000
Using the housing activity average, an even greater share of the total slowdown occurred between 2006 and 2008, when unemployment was stable, and an even smaller share occurred after April 2008. I want to thank Bob Murphy for further strengthening my argument.
Bob also makes another argument, citing data showing that construction employment declined much less than housing construction between 2006-08. But that’s easy to explain, as commercial real estate prices didn’t peak until late 2008. So the commercial RE sector may have picked up some of the workers laid off from building houses. (I don’t know about infrastructure and government building.) And even if commercial RE didn’t add housing workers, if housing is half of all construction then a 20% decline in housing construction jobs would translate into only a 10% decline in all construction jobs. All this of course supports my point. The big drop in housing construction between January 2006 and April 2008 did not cause a significant impact on the US unemployment rate. Doesn’t that suggest that those housing construction workers weren’t [sic] able to find jobs in other forms of construction, or other activities?
As I said, I am not going to connect the dots. If people don’t see that Scott is literally filling in the gaps of my own argument, I don’t know what to say. Those allegedly embarrassingly reallocated construction workers (who stopped cranking out new houses in 2006) didn’t go into fruit picking or software design, Scott hypothesizes that they apparently went into commercial real estate–one of the things I guessed in my quote above (“shopping mall down the street from a new housing development”).
(3) And as far as me hoping against hope that Scott wouldn’t click the link and verify my wild assertions, all I can do is repost my comment on Scott’s blog:
Scott,
What did you think I meant when I said this:
That fact alone buys us another few months,
I know Austrians aren’t good at math, but did you think I believed that from January 2006 to April 2008, only a few months had passed?
In closing, I should note that I’m not expecting Scott to capitulate. I just want him to admit that the Klingian/Austrian story isn’t as farfetched as he and DeLong originally thought. For example, I admitted (not going to dig up the link) that I was “shocked” by the behavior of TIPS yields in the fall of 2008, when Scott told me the Fed had a really tight stance and I thought he was nuts. So I concede that Scott’s theory can much more easily accommodate that fact than mine.
Bob,
You wrote:
“That fact alone buys us another few months,
I know Austrians aren’t good at math, but did you think I believed that from January 2006 to April 2008, only a few months had passed?”
The problem is that those few months go the wrong way.
Let’s review the numbers:
January 2006: starts 2,303,000 completions 2,058,000 average 2,180,000
April 2008: starts 1,008,000 completions 1,014,000 average 1,011,000
October 2009 starts 527,000 completions 745,000 average 636,000
Scott’s actually doing you a favor by averaging. The completion data is more damning than the start data.
The decline in starts between January 2006 and October 2009 is 1776 thousand. The decline in starts from January 2006 to April 2008 is 1295 thousand or about 73% of the overall decline.
The decline in completions between January 2006 and October 2009 is 1313 thousand. The decline from January 2006 to April 2008 is 1044 thousand or about 80% of the overall decline.
So the issue is not that completions data moves it only a few months, it’s that it moves it in the wrong direction.
My advice is give it up before you dig yourself in deeper.
The big problem that Scott allegedly unearthed was that housing starts peaked in January 2006, and then started falling precipitously, and yet the national unemployment rate didn’t start shooting up until much much later.
I looked at the numbers and realized that construction employment (as opposed to housing starts) held steady after the January 2006 peak. So how to explain that?
My first observation was that housing completions peaked (and then started declining) a few months after January 2006. This makes sense, since it presumably takes time to build a house.
So “that fact alone” buys us a few months.
Averaging numbers from January 2006 to April 2008 and October 2009 is completely irrelevant to the point I was making about housing completions.
Now that you understand the minor little point I was making about starts versus completions, you might say, “Why did you make such a big deal out of it then? That’s hardly crucial to your case; your big point was to try to wiggle out of Scott’s hammergrip by switching from house construction altogether, to other areas that employ construction workers.”
Exactly. That’s why I didn’t cite any figures, but just did a link to the housing completions data. And then I said, “More important,” to lead into the part of my argument that I considered more important.
Scott somehow took my failure to cite the stats, as evidence that I was bluffing.
Sumner has conceded now, right? He’s probably on his way to pick up Rothbard’s America’s Great Depression.
Bob,
The peak in completion buys you exactly two months as the peak occurred in March 2006. But Scott’s point is not when the peak occurred but how far things had fallen without unemployment going up. (And in fact now that I notice it, peak completions totalled 2203 thousand. Thus repeating the above calculations yields an even higher figure of 82%.)
More importantly, in order for Kling’s “Recalculation Story” to make any sense you have to see some sectors declining and other sectors growing (or at least not declining). We don’t see anything like that. in fact what we see is large scale declines in sectors accounting for over 85% of national employment, and all taking place at approximately the same time starting in late 2008. This sounds like a large negative AD shock much more than recalculation.
Other data also doesn’t support the recalculation theory either. If there were shortages in some sectors and surpluses in others we should see increases in job vacancy data and a divergence in compensation growth between sectors. There’s no hint of either. The recalculation theory holds about a thimble’s worth of water.
As for Scott implying that you were bluffing I wouldn’t take that too hard (or too seriously).
Hold on a second. ABCT would say that too many people moved into construction because of the fed manipulated boom and when the bubble burst these people needed to go into different fields. Sumner says that nonsense, look at the fall in construction starts and unemployment did go up in construction during that initial fall. So Dr. Murphy explains why that it took time for the jobs to start being shed. For example, he mentioned housing completions, commercial real estate, building decks, etc. Then he showed that once construction jobs did start getting shed that the national unemployment rate shot up too. Once their layoffs stopped the climb in the national numbers also started to rise. All this fits ABCT perfectly.
How does Sumner explain the unemployment charts moving with construction if this had nothing to do with a housing bubble?
Two errors above.
Should have read, didn’t go up…
And stopped rising (instead of started to rise)
Bob,
You aren’t paying attention.
Scott ALWAYS does this shuck & run move.
It’s constant.
He does a smoke and mirrors magic act — distracts attention by going off on irrelevant tangents.
As I say. He does this constantly.
And I don’t think its accidental.
He needs to distract from the topic, otherwise his world view come crashing down, and he’d have to work extremely hard to catch up.
The preference for leisure explains avoidance of Hayekian macro.
No one likes to think hard or work hard. It’s easier to crank out more math puzzles.
I know a couple of, now former, home construction workers. One an electrician, another that laid foundations. They both told me that from 2005 until 2009 when both got other jobs to fill their time, their incomes dropped by over half.
So, your version is that there was unemployment, but it didn’t show up in the official figures?
I can entirely that it was just that, but is there a way that we can convince others and ourselves of that?
There’s a difference between wage rates and unemployment. If you feel you’re underpaid then go make your labor worth more. If you’re unemployed then God Bless You Gentlemen.
Being a macroeconomists means never having to think economically or deal with the real world.
DeLong is so breathtakingly dishonest, I am stunned you even deign to mention his name.
Bob you wrote:
“But since Scott himself agreed that the former housebuilders moved into other construction projects–which may very well have been corollaries of the previous housing construction–I think Kling still has a leg to stand on. (Not surprisingly, so does Kling.)”
Let me quote again for emphasis:
“WHICH MAY VERY WELL HAVE BEEN COROLLARIES OF THE PREVIOUS HOUSING CONSTRUCTION”
Are you sure you want to make this the crux of your incredibly weak argument?
If so prove it!
Yes I’m sure.
Me too.
I have explained here what both Sumner and Murphy missed in this debate.
I suppose since you and Scott both misunderstood my original article, I must have been unclear. But I don’t see how my explicitly mentioning “shopping malls” made you guys think I was unaware of the possibility that commercial real estate could be the solution. How is building a shopping mall “work on old houses” which you said was my theory?
Actually, I did notice it But the point that you’re overlooking is that non-residential construction often (including the business cycle discussed) develop very differently from housing construction.
>Well, this is clearly related to the fact that real hourly wages was reduced by 1.5% during the first 6 months (after having previously dropped by 1% during the 6 months before that), only to increase as much as 5% during the following 6 months.
Real weekly wages rose somewhat less, 4%, because of a decline in the average work week.
>The big drop in employment was thus mainly the result not of a contraction in real output, but from a big real wage shock. The big decline in NGDP thus didn’t primarily lower real output, it instead because of a sudden change from high inflation to deflation and nominal wage rigidity redistributed from people who lost their jobs to people who held on to their jobs.
Bingo.
Murphy should agree with this. In his book “Politically Incorrect Guide to the Great Depression”, he favorably quotes Lowell Gallaway:
“While the initial increase in unemployment can be largely explained by the productivity shock, the very sharp rise in unemployment in 1931 was not related to further declines in output per worker. Productivity per worker changed little, actually rising somewhat.… Money wages fell, but rather anemically. Whereas in the 1920–1922 depression a roughly 20 percent fall in money wages was observed in one year, the 1931 de-cline was less than 3 percent. By contrast, prices fell more substantially, 8.8 percent, so real wages actually rose significantly in 1931, and were higher in that year than in 1929, despite lower output per worker. The 1931 price [declines], accompanied by a failure of money wages to adjust … seemed to be the root cause of the rise in unemployment to over 15 percent in 1931.”
Granted, there was no unemployment insurance back in 1920, nor minimum wage, nor was there a substantial Fed induced cultural psychology of constantly rising prices (by 2008 many worker’s great-grandparents experienced gradually rising prices). Both of these facts most likely facilitated the rapid fall in nominal wage rates. Today, wage rates are more “sticky” because of these government interventions.
Furthermore, it’s one thing to ask people to take a pay cut of 20% because they have expectations of 2% inflation, it’s another to take a pay cut of 25% with the same inflation expectations but in reality there will be price deflation of, say, 5%, and the option to support oneself on UI.
All this talk about sector specific employment and construction and you are the first person to introduce SUPPLY AND DEMAND to an economics discussion!
Come on, that’s funny.