We’re well into a severe housing downturn by every measure except for the number of people working in residential construction.
 |
The figure above is similar to one that has often been highlighted by one of our favorite bloggers, Calculated Risk. The graph shows the relation between the number of new housing units completed each month and the number of people employed in residential construction. The latter is calculated as the sum of the BLS series for those working in residential building (currently a little under a million people) and residential specialty trade contractors (over two million). It is puzzling that the decline in the number of people working in residential construction has so far been quite modest relative to the decline in the number of units completed.
One plausible explanation that is sometimes offered is that many of those laid off in the current episode may be undocumented immigrants. Although there must be some truth to that claim, it is not correct to assume that undocumented workers do not show up in the BLS series. According to the Wall Street Journal, the IRS has issued 11 million individual taxpayer identification numbers to people without a social security number. In practice, the penalties for hiring an undocumented worker appear to be lighter than the penalties for failing to withhold taxes on your workers. As a result, government counts of the number of people working may be more accurate than one might have supposed.
Another possibility is that some of those firms that had previously been classified by the BLS as involved primarily in residential construction are perhaps now being sustained at least in part by commercial construction. Nominal spending on nonresidential construction was up 15% in 2007:Q1 relative to 2006:Q1 (BEA NIPA Table 1.1.5).
Or, some may conclude that residential construction firms are simply postponing the inevitable layoffs, although it is unclear what these firms’ calculations and motivations would be.
I wanted to do some statistical analysis of this relation, but unfortunately the BLS does not provide the values for specialty trade residential construction employment prior to 2001, and 7 years is not enough data for the questions I’d like to ask. However, the series for residential building employment, though only 1/3 of total residential construction employment, goes back to 1985 and seems to have some similar properties. Certainly one is again struck by the failure of this series to fall as sharply as completions over the last six months:
 |
Let yt denote the residential building employment (excluding specialty trade contractors) in month t and let ct denote the number of completions in month t. An OLS regression over 1985:1 to 2007:3 yields the following estimates (t-statistics in parentheses):
This appears, both to the eye and to the regression, to be a very significant relation. But one feature that can fool both the eye and the regression is the fact that employment yt has a tremendous amount of persistence– its level last month is an extremely important predictor of where it is likely to be today. The same is true of completions ct. When one regresses one such variable on another (or just plots them together in a graph, inviting your eye to do the regression for you) it is very easy to mistake the common swings in the two series for a reliable predictive relationship.
One way to investigate the magnitude of these concerns is to examine the serial correlation of the residuals of the above regression, which we can do with a second regression of these estimated residuals on their own lagged values:
These coefficients sum to 0.96, meaning that when the level of employment deviates from the value predicted by the relation, it takes a very long time to return. In fact, one accepts the null hypothesis that these coefficients sum to 1.0, in which case a discrepancy between employment and completions in the original regression would never be completely eliminated. To get a sense of what this could mean, the blue line on the graph below plots the actual residuals through March 2007. At that time there were several hundred thousand more people working in residential building than one would have predicted on the basis of the number of homes completed. The green dashed line indicates what those residuals would be if from here they follow the dynamic pattern of the above equation. Even two years from now, we might still be seeing 100,000 more workers than predicted by the historical regression, and must accept the hypothesis that the true discrepancy could end up much bigger than that.
 |
How could there be a drop in the number of homes completed with residential construction employment never following it down? In addition to the issues about how we measure the number of people working in residential construction employment noted in the beginning of this post, there are changes each month in the types of units that are built and the means by which they are constructed. All of these changes would cause discrepancies between the numbers for residential construction employment and houses built in a given month. The statistical evidence is that these discrepancies can take a long time to be eliminated, and indeed may never be.
So is residential construction employment about to follow completions off the cliff? Maybe. Or maybe not.
Technorati Tags: macroeconomics,
employment,
housing,
construction employment
So is residential construction employment about to follow completions off the cliff? Maybe. Or maybe not.
YES! That was a great economist statement.
What do one handed economists think about construction employment?
Perhaps it depends on whether their remaining hand is the left or the right, Buzzcut.
I’d question the quality of the completions data. Public homebuilders’ deliveries in the 3rd and 4th quarters were only down 4% and 6% y/y, respectively. They’ve been shortening build times and accelerating their backlog conversion rates since 2Q06 to maintain cash flow generation despite slowing orders. Only problem is they’re running out of backlog, which is down ~30% y/y. So on the question of whether or not employment is about to fall off a cliff, I’m going to say maybe.
Perhaps it depends on whether their remaining hand is the left or the right
Now, now, your just trolling, like you were on the post about 35% of Democrats being certifiably insane!
It is really too bad that gas prices are so high. If they weren’t, a good proxy for construction employment would be sales of certain pickup truck models. Sales are WAAAY down, but it could be because gas prices are so insane.
I would put my money on a collapse too. Look at the minor downturn in 2000-01. Although much smaller, it produced a similar lag effect. It took jobs about four qtrs to catch up with completions at the bottom of the hill.
I am looking at the number of completions in the 1985-1988 years, which were much higher than the corresponding employment numbers, in the opposite direction from where they are now! Was this an indication of doing more with less? And if so, maybe we are inefficiently doing less with more workers now. I would be real interested to see these charts going back to around 1950.
JDH,
I have been asking myself the same question. I have assumed that because construction is a long term industry that building contracts will be winding down, but this does seem to be a longer process than I anticipated. I was hoping you would have a stong reason this is true, but it appears that you are as mystified as I. I would welcome reasonable specualtions on why construction employment seems to be holding in the fact of the slowdown.
I’m all for two-handed economists. “I don’t know” is a respectable answer.
I didn’t see any mention or analysis of the period around 1985 in the second figure. Completions high, employment low–based upon eyeballing the rest of the chart; just the opposite of what we’re seeing now.
I think this analysis requires an ecnomist with three hands.
Professor, folks probably take you for a right handed economist, after your mean question a few days ago about Democrat politicians.
Things sure seem slower — traffic on 5, traffic coming down into La Jolla — in San Diego based on eyeball. Can’t wait to see that retail sales data on Friday.
Having worked in construction 35 years, during good times and bad, there a number of things employers and employees can and will do. Number one is rotation. Employees take turns at working, on weeks then off weeks, often for months. Sometimes employees will file for unemployment insurance while temporarily off work. Some states classify temporary unemployment (a few weeks) differently than permanent unemployment. Not sure if or how this is reflected in labor statistics. Rotation plus unemployment insurance beats no job at all.
There are a lot of illegal immigrants in the home construction industry. Their numbers are not in the employment and they are not unemployed. In my area roofing crews speak Spanish.
Many legit workers in residential construction are now independent contractors. When they have less than full time work they are still “employed”. We use three residential construction firms. All three are now 2 or 3 employees with contracted work crews. One runs his staff through a company called Command Staffing mostly because they certify the immigration status and deal with all the employment forms and taxes.
But the net is this: All these workers are day workers. It is not clear to me that any are eligible for ordinary unemployment compensation.
The contractors are business owners and they are always employed until they close their business.
I’m enlightened somewhat by this claim:
Although there must be some truth to that claim, it is not correct to assume that undocumented workers do not show up in the BLS series. According to the Wall Street Journal, the IRS has issued 11 million individual taxpayer identification numbers to people without a social security number. [My estimate of 12M illegal aliens just got a terrific boost, your’s?] In practice, the penalties for hiring an undocumented worker appear to be lighter than the penalties for failing to withhold taxes on your workers. As a result, government counts of the number of people working may be more accurate than one might have supposed.
It seems to me that there are many residential contracts that are short term, esp short term when you have hordes rather than a skeletal crew. The payroll nature of the business is not going to capture the number of workers in the same way that they don’t catch a lot of the agricultural workers. Part of it is that it fails to meet the requirements of payroll employment and increasingly, much harder to monitor on the basis of a receipt that documents you paid foreman Hose $X for work performed. Was that Hose and a few chums or Hose and that gang lined up at the post office?
So we pay for this lack of regulation at the border not just in terms of displacing over-paid union carpenters for under-paid material movers and even more over-paid new truck driving contractors, but also in terms of econometrics that might inform policy…about whether to expect a flood of whistlin truck drivers looking for some one to boss or not.
It would be nice to see here a chart that CR has done comparing starts and completions. He aligns the series by shifting starts 6 monsths ahead. On those charts, it is apparent that in any slowdown, completions fall somehow differently compared with starts. It may well be that during periods of residential construction slowdown, builders somehow delay completions and layoffs. That would reduce de apparent inventory of completed homes for sale. Aj jg says, they can do it just by delaying rotations.
Howsabout you overlay remittances to Mexico on your charts? Or, the remittances delta (I believe the growth rate has slowed dramatically, but is still positive).
Like IM, I was wondering about starts as well. It seems that starts are more a measure of expected future prospects, and would correlate much more tightly with labor indicators for that reasons – i.e. employers would both hire workers and start construction based on anticipated demand.
For sure the chart shown seems to be backward to my intuition, which would be that employment should lead completions, not lag it. (i.e. if there is a “pipeline” of houses at various stages of construction, then as builders anticipate a slowdown they will stop starting new houses, but perhaps continue to finish the existing houses, resulting in a gradual slowdown of employment as less workers are redeployed at the beginning of the pipeline).
I was also wondering whether or not some sort of cointegration analysis might be appropriate here and what it could tell us, since presumably the overall level of both housing completions and housing employment could be driven by an ARIMA process more or less tracking GDP.
Robert Bell, what I was providing was precisely a cointegration analysis, without the jargon or further details reported. The sum of coefficients I referred to is the familiar rho statistic; Dickey-Fuller t-stat is -1.5. The conclusion is that one accepts the null hypothesis of no cointegration between c and y, meaning that it is problematic to interpret the results from that regression as a useful forecasting relation.
The notion that the illegal labor market has suffered a disproportionate number of constructon layoffs is not consistent with the graph our host has provided. Whatever that blue line represents, it tracked completions well on the way up. Some of the workers counted in that blue line are illegal. Maybe there are others who are not counted in the blue line. If there is a strong causal link, the blue line should track completions on the way down. Only if contractors added workers in one ratio between workers counted in that blue line and workers not counted, and are now laying them off in a very different ratio, would “hidden” layoffs be a reasonable explanation. The argument that layoffs of illegal workers explain the divergence between the two lines seems to rely on some contractors being awfully precise about who they are laying off.
I too think the employment of undocumented works in the home construction industry has been under-reported.
Within the last week I read that INS has noticed that there are fewer Mexicans attempting to come north without benefit of papers. Mexicans come here for work, when the word gets out that there are few opportunities coming to the US isn’t worth the cost and risk.
I love stats even though I am terrible at it. Could you reduce the serial correlation by looking at the year on year changes in construction employment v year on year changes in completions?
In most of the cosntruction data I follow housing starts are nearly always a more useful indicator than comletions. After all when is a house complete? According to my wife, never.
In Eastern Europe tax is payable on new houses when complete so guess what, there are always a few bits left to do.
When they built the condos next door, the crew arrived by tens in red vans from Sockton (70 miles each way. Payed in cash, don’t think there were any tax forms involved on their part.
How come no one has mentioned overtime. An in process building is eating into capital without providing any gain. Getting it done warrants overtime pay. Even when it is still warranted, overtime will be cut before layoffs occur. This factor will mess up your stats.
David, yes, you can look at the relation between the change in employment and change in completions and that would solve all the problems I mention here. However, when you do that, the relation disappears!
Correction– I should say mostly disappears. I just rechecked the regression– t stat falls from 26.5 to 2.3 when you go from levels to changes.
Current survey instruments also fail to pick up surplus labor uptake in booming occupations such as race-war militiaman and abortion-clinic bomber.
I think it’s a safe bet that a large portion of the 11 million are construction *foremen* who have relatively permanent ties to their employer, short of being deported. From what I’ve seen, they are trusted by the owner/operator, and they in turn shield the owner and hire/fire their crew, who seem to be paid in cash daily/weekly or job by job basis. I wouldn’t be suprised to find that the foreman has a trusted handful that they keep for all jobs who also monitor the rest of the crew.
That handful may account for some of the 11 million also.
And before you ask how I’d possibly know this, I’ll just add that in high school, many years ago, I was trained in my job by an illegal alien who got deported more than once. Pretty normal stuff out here on the left coast.
The point of my comment above was simply to offer some advice on how to interpret the data. Figure for every foreman + trusted handful — say 6 core folks — there are another 6-18 depending on the side/labor
involved.
So if the graph “looked” right, how many unaccounted employees would that translate into?
Maybe we shouldn’t trust govt stats.
Professor Hamilton,
Easter Island’s civilisation is thought to have collapsed because the statues were considered to be so valuable they kept building them until they ran out of resources to do anything else – a collective mania.
If homes aren’t selling but construction employment isn’t falling then something’s definitely amiss in the works.
JDH: much obliged – and still learning! 🙂
Jim,
Interesting post. Another possibility to examine is that hours worked per week may have fallen, thereby reducing labor input even with employment little changed.
Also,your graph is striking; one series is very smooth, the other is highly volatile. Is it possible that in constructing the data, the BLS is somehow imparting spurious high frequency smoothness to construction? Or, conversely, could the Census folks be adding spurious volatility to completions?
Buzzcut,
Gas is not insane yet. A cup of gasoline costs thiry cents while latte’s at Starbucks are how much? I hardly think the price is insane and I would be curious to see your reaction next summer when it’s $5/gal.
I guess it depends on your definition of “insane”!
I look at it this way. When the price of gas gets high enough that Chuckie Schumer crawls out from underneath his rock to conduct Congressional hearings into “price gouging”, prices are insane.
Okay, the prices aren’t insane. Democrats are.
JDH, that ~150-200k discrepancy shows up at both ends of the period…subduing my concerns somewhat over the reluctance of the blue line to follow the red line dive on the RH side. [ie, shoot what a lucky period to achieve such great correlation.]
As much as I like Rp’s anecdotal (and calmo confirming post of course) remarks, kharris’s note is a bother, no?
I’m no conspiracy nut, but if the IRS can deduct taxes on 11M undocumenteds, is it that much of a stretch to think that the numbers of undocumenteds were determined by pencilling them in, not according to the non-existant empirical data, but to the requirements of the correlation?
Maybe you need to check and see what the birth-death adjustment does to you employment data.
Over the course of an expansion the birth-death model works fine. but it is almost certain it will cause the employment data to miss the next cyclical downturn in employment.
many construction workers are illegal immigrants. Many of them from Mexico. Many of them send remittance back to Mexico regularly. If you look at the Mexico remittance number (yoy growth), you will be surprised how high correlation with the the US housing starts yoy growth too.
Thanks calvin, see my post above requesting an overlay of remittances.
T J said: “Having worked in construction 35 years, during good times and bad, there a number of things employers and employees can and will do. Number one is rotation. Employees take turns at working, on weeks then off weeks, often for months…”
This post gets my vote for the answer to the “why construction employment hasn’t plunged” conundrum.
My naive interpretation: If housing starts double from 1 million to 2 million, it doesn’t require double the construction workers to get the job done, but more of the existing workers working more often with *some* new additional workers.
When housing starts drop, the last-hired are let go completely and the others “throttle-back” to more-frequent “rotations” of being underemployed but not entirely out of work for extended periods.
My own additional wrinkle: There may be a housing “recession”, but it’s not part of a general economic recession. Construction employment would be more likely to *really* plunge if the entire economy were entering a recession, but since it isn’t…
Sebastian
In many cases, small, home contractors also are doing remodelling work so that some framing crews are also remodelling crews, working on everything from piers and bulkheads to cabinetry. After all, if it is made out of wood, that is what a carpenter works with, and that is what they do.
I realize that Michigan represents the extreme of the housing market fall off, but one might also see it as a bellcow for the rest of the nation.
Tonight I talked with an associate who works for one of the area’s largest home improvement companies and he indicated that he has only written $80K of orders since January whereas in a “normal” year he would have had $750K by now.
Some new home builders are buying homes from their customers at 10% below appraised values if the home owners buy a new home from them. They are exchanging higher cost assets for lower cost ones just to keep going. Then they move higher risk customers into the older homes they bought up to try to keep their cashflow up.
There are many stalled developments and many more where land has been acquired, but no building commenced. The situation is about the same as 1980-82 when interest rates were 16%. The difference is that the higher paying jobs to afford the homes typically being developed are no longer as available. So, people are circling the wagons.
Relatives in Florida are saying that they are seeing signs of the same thing. Could it be that there is a core element of our economy that is being eroded and the housing malaise is just a symptom? We know what that is in Michigan and it looks as if the rest of the nation may find out as well.
I am getting confused by what i am seeing in this picture.
Some things i appear to know:
1. A friend of mine in Real Estate in
Cincinatti says they are very very busy since early February. They have had some of their best ever months for sales. Prices did fall 10% however.
2. A builder adds value to land and creates a product that is still massively over priced by historical levels of affordability. If he were profitable in 2003 then why not now? If this view is correct he can carry on building even as prices fall since he was already making a massive profit, and now his profit is just less massive. His builds will drive the rest of the market down but he still profits.
3. Builders are more or less saying they are in survival mode at the moment.
Put it together and i am guessing that the builders are not doing as bad as they say they are, because they want to be seen as the good guys in this scenario.
On the other hand if builders are *really* doing as bad as they say they are then if prices are down 10% in some markets and nobody seems aware of that i find that very frightening given the debt levels involved here.
Undocumented workers, however, would shift the curve, not move us along it. The interesting thing here is that the two curves correlate strongly, then suddenly diverge. It’s not like there were no illegals in 1988-1992.
Could it be that there is a core element of our economy that is being eroded and the housing malaise is just a symptom?
Yes, of course. That core element is homebuilding.
Look, homebuilding was dead in the ’80’s and much of the ’90’s. In terms of new housing, the post war boom ended in the ’70’s.
Something changed in the very late ’90’s. Maybe it was higher birth rates from boomer echoes, maybe it was massive immigration, lower interest rates, pent up demand, whatever, but the home building industry revived. There was a rebuilding of the housing stock. Home building employment increased as a result.
Now, something has changed. Maybe we were in a bubble that was popped by higher interest rates. Maybe prices have risen too fast for incomes. Who knows. But homebuilders are hurting financially, and that part of the economy is in the tank.
It’s circular reasoning, I know. But the post war real estate boom was itself a reaction to the dearth of building in the Depression/ WWII era. Maybe the boom we just went through was a reaction to the ’80s/ ’90s drought. We’re back to even, and moving forward homebuilding won’t be the industry that it was in the past decade.
Could it be a shift in the labor-intensity of the remaining homes under construction?
If factory-built houses are down and can be presumed to be less labor-intensive, while custom, on-site construction continues, the ratio of completions to man-hours changes upward.
Probably a second order effect at about the same level as a shift to remodeling from new construction.
Alex – it’s not like we were building an insane amount of housing in 1988-92, right?
I remember a supervisor in Glendale who was challanaged to build a trac home in less than 30 days. He got his subs all working together and finished in 28. About 2400 sqft., one level on a finished lot in a subdivision. Now think about the realtionship between construction and empoloyment. To build an “average house”, talk to an estimator about how many hours of labor are needed. Concrete, electric, plumbing, framing, masonary, finish ect. You can easily come up with a number that say to build for example a 1800 sqft home for $80 a sq ft. will cost 144,000 (hard cost of labor and materials – not including site work).
Then you calculate how much is labor. It is very easy to correlate Construction costs to labor costs. What are all of you people thinking? Of course if construction is down, labor is also down. Dont be so stupid.
No doubt. Aggregate hours worked must be down in proportion to the decline in building activity. If the number of people working has not declined as much, then the hours worked per employee is down.