Several people have asked me about a new NBER working paper that takes a look at construction productivity (a popular topic around here), “The Strange and Awful Path of Productivity in the US Construction Sector,” by Austan Goolsbee and Chad Syverson.
The treatment of specialized-trade inputs, which seems both not-well-defined and non-intuitive, is a big red flag for me -- it seems like the resulting measures depend heavily on the in-house vs contracted practices of GCs over time, which is only loosely related to 'productivity.' That being said, the figure for cost of highway construction seems much easier to measure and unambiguously bad, even starting from 2003, so that might be a good place to start further exploration.
I also wonder if there has been a change in the distribution of the relative 'quality' of new construction -- in the sense of new housing that costs more or less than the area median. Affordable housing advocates claim that construction in cities is dominated by high-end developments (chasing higher margins), but I don't know 1) if that's true or 2) how that compares historically.
I agree, the way subcontractors are accounted for really complicates understanding a lot of this data. That said, this paper tried to estimate subcontractor labor for several areas of construction and in most cases the measured productivity got worse (though its once again hard to extract signal from the data) https://www.bls.gov/opub/mlr/2018/article/measuring-productivity-growth-in-construction.htm
I doubt the problem is monocausal, but one cause not identified is the difference in regulatory compliance. It takes a lot more to meet codes and standards than it might have 50 years ago. The cause could actually be a net social good if we value more efficient buildings and better worker safety.
More economists would benefit from becoming regular readers of Brian's Substack, but I won't hold out much hope for improved understanding of the unique complexities of construction processes. That Goolsbee and Syverson lump together lateral construction (i.e. infrastructure) with buildings makes any conclusions they draw nearly worthless. Brian, and many of the commenters on this post have noted the confusion of the contractor/sub-contractor delivery method as well as the increasing challenges of our regulatory environment. Also noteworthy is the great sham associated with productivity gains we have been promised by the software/IT products that we use in the industry. Mobile devices on the jobsite are used most frequently by trades calling to apologizing that they can't show up when they're supposed to.
The big picture is that construction inputs reach a scale limit faster than any other manufactured good humans have ever produced. One yard of ready mixed concrete delivered to a jobsite is less efficient than 10 yards, but from that point on any additional 10 yards requires another truck---which is traveling through traffic, waiting in line, placing, cleaning, and then heading back to a plant that's typically located in a marginal area at least 15 miles from an urban center. And, if the wait is too long, the batch turns to junk and the driver has wasted an entire day. Efforts like these define every step of the building process, and the linear sequencing is tyrannical.
It could be that the flood of cheaper foreign born workers into construction trades over the past few decades has deterred labor saving capital investments, so productivity falls even as construction firms claim a larger surplus.
No need to spend heavily to save labor if labor is already pretty cheap.
The muddy confusion here reminds me of some job sites I’ve been on.
And also of the notion of the “fog of war”—ever changing cultural values and standards, rapid churn in technology and materials, sudden economic lurches from boom to bust and so on.
At the ground level where workers wield nail guns and hang sheetrock, I’ve noticed something that seems obvious and important. How our society deploys talent and skill seems to have changed significantly over time—to the detriment of the building trades. Talented motivated people tend to get funneled and sorted into other economic niches which pay better and have higher status. And there is a high churn rate in and out of trades that might take a decade or more to learn.
Simultaneously educational programs focused on the trades seem to do a poor job preparing those who go down the path. European trades programs which combine theory and practice in a rigorous and effective way contrast markedly with careless half-assery here in the US.
The result of this is a “fog of construction”—an expensive creaky inefficient machine with a thousand hiccups instead of a well-oiled one. The constant accumulation of errors, defects, rework, and waste mire projects in overruns, and the root cause is hard to see or measure.
Reading historical accounts of building processes makes me think that this was not always the case. Paul Starrett’s autobiographical account of the construction of the Empire State Building is interesting in this context. For a lot of reasons it’s impossible to imagine a project of that scale coming in under budget and ahead of schedule in today’s world.
A lot of things have changed in the way houses are built since the 1960s. For example, windows were often built on site by carpenters and then glaziers called in the install the glass. Modern windows are purchased modular units with much better insulation and mechanisms. Walls and attics are filled with insulation, often by specialized contractors. Kitchens tend to have stone counters and fancier fittings. Those stone counters require specialized subcontractors and a higher material cost.
You could still build much more modestly priced houses, but they'd use more energy and appear rather bare bones and pokey to the modern eye.
One big driver is land price which has been rising for a number of reasons. If you need financing, as most builders and buyers do, the land price sets a floor on how much has to be spent on the house itself. The need to spend more on land and hence construction to satisfy the bankers could be one of the reasons overall costs are rising.
Imagine if we measured agricultural productivity only by measuring the productivity of the most marginal land itself added every year.
The biggest productivity gains of housing people in high productivity areas flows to the already existing stock. Without breakthroughs in transportation speeds or large productivity gains in lower cost areas or easier building in high productivity areas, it will be hard to see construction productivity gains.
Subcontractors may be intermediate inputs from the perspective of a single firm, but they are not intermediate inputs (rather, they are part of value added) for the construction industry as a whole. Intermediate inputs for the construction industry would only include items produced in other industries, such as lumber and fuel (manufacturing) or architectural services (professional services).
Interesting, seems to pretty thoroughly destroy some my priors, most especially that American firms underinvest in capital goods as a result of strong cyclical swings.
So, with that out...
I think there's an open question around compliance/permitting/design, etc. costs; these are things which *should* have benefitted from the IT revolution but I think we all can see that they did *not.*
But that's 10% or so of project costs, maybe 20% in sensitive cases. The major thing that they screw up badly is urban housing, and that less because of direct costs and more because the massive uncertainties they introduce prevent projects from ever happening at all.
Actual construction productivity just seems to be extraordinarily sensitive to Baumol's Cost Disease. We have no clue how to automate most site processes, prefabrication and off-site manufacturing are of limited use in replacing site processes, and those site processes which can be automated are either very expensive to do with machinery or are directly related to management, compliance, and design.
I used to think there were answers, now my working theory is "suck it up and build what we need, land use is most of the problem for urban housing markets anyway."
I will note that a flat metric of productivity per worker should directly translate to a falling metric of productivity per dollar paid given the increase in prevailing wages, again courtesy of Baumol.
The treatment of specialized-trade inputs, which seems both not-well-defined and non-intuitive, is a big red flag for me -- it seems like the resulting measures depend heavily on the in-house vs contracted practices of GCs over time, which is only loosely related to 'productivity.' That being said, the figure for cost of highway construction seems much easier to measure and unambiguously bad, even starting from 2003, so that might be a good place to start further exploration.
I also wonder if there has been a change in the distribution of the relative 'quality' of new construction -- in the sense of new housing that costs more or less than the area median. Affordable housing advocates claim that construction in cities is dominated by high-end developments (chasing higher margins), but I don't know 1) if that's true or 2) how that compares historically.
I agree, the way subcontractors are accounted for really complicates understanding a lot of this data. That said, this paper tried to estimate subcontractor labor for several areas of construction and in most cases the measured productivity got worse (though its once again hard to extract signal from the data) https://www.bls.gov/opub/mlr/2018/article/measuring-productivity-growth-in-construction.htm
I doubt the problem is monocausal, but one cause not identified is the difference in regulatory compliance. It takes a lot more to meet codes and standards than it might have 50 years ago. The cause could actually be a net social good if we value more efficient buildings and better worker safety.
More economists would benefit from becoming regular readers of Brian's Substack, but I won't hold out much hope for improved understanding of the unique complexities of construction processes. That Goolsbee and Syverson lump together lateral construction (i.e. infrastructure) with buildings makes any conclusions they draw nearly worthless. Brian, and many of the commenters on this post have noted the confusion of the contractor/sub-contractor delivery method as well as the increasing challenges of our regulatory environment. Also noteworthy is the great sham associated with productivity gains we have been promised by the software/IT products that we use in the industry. Mobile devices on the jobsite are used most frequently by trades calling to apologizing that they can't show up when they're supposed to.
The big picture is that construction inputs reach a scale limit faster than any other manufactured good humans have ever produced. One yard of ready mixed concrete delivered to a jobsite is less efficient than 10 yards, but from that point on any additional 10 yards requires another truck---which is traveling through traffic, waiting in line, placing, cleaning, and then heading back to a plant that's typically located in a marginal area at least 15 miles from an urban center. And, if the wait is too long, the batch turns to junk and the driver has wasted an entire day. Efforts like these define every step of the building process, and the linear sequencing is tyrannical.
It could be that the flood of cheaper foreign born workers into construction trades over the past few decades has deterred labor saving capital investments, so productivity falls even as construction firms claim a larger surplus.
No need to spend heavily to save labor if labor is already pretty cheap.
Just a thought.....
The muddy confusion here reminds me of some job sites I’ve been on.
And also of the notion of the “fog of war”—ever changing cultural values and standards, rapid churn in technology and materials, sudden economic lurches from boom to bust and so on.
At the ground level where workers wield nail guns and hang sheetrock, I’ve noticed something that seems obvious and important. How our society deploys talent and skill seems to have changed significantly over time—to the detriment of the building trades. Talented motivated people tend to get funneled and sorted into other economic niches which pay better and have higher status. And there is a high churn rate in and out of trades that might take a decade or more to learn.
Simultaneously educational programs focused on the trades seem to do a poor job preparing those who go down the path. European trades programs which combine theory and practice in a rigorous and effective way contrast markedly with careless half-assery here in the US.
The result of this is a “fog of construction”—an expensive creaky inefficient machine with a thousand hiccups instead of a well-oiled one. The constant accumulation of errors, defects, rework, and waste mire projects in overruns, and the root cause is hard to see or measure.
Reading historical accounts of building processes makes me think that this was not always the case. Paul Starrett’s autobiographical account of the construction of the Empire State Building is interesting in this context. For a lot of reasons it’s impossible to imagine a project of that scale coming in under budget and ahead of schedule in today’s world.
A lot of things have changed in the way houses are built since the 1960s. For example, windows were often built on site by carpenters and then glaziers called in the install the glass. Modern windows are purchased modular units with much better insulation and mechanisms. Walls and attics are filled with insulation, often by specialized contractors. Kitchens tend to have stone counters and fancier fittings. Those stone counters require specialized subcontractors and a higher material cost.
You could still build much more modestly priced houses, but they'd use more energy and appear rather bare bones and pokey to the modern eye.
One big driver is land price which has been rising for a number of reasons. If you need financing, as most builders and buyers do, the land price sets a floor on how much has to be spent on the house itself. The need to spend more on land and hence construction to satisfy the bankers could be one of the reasons overall costs are rising.
Some of the gap is a compositional fallacy.
Imagine if we measured agricultural productivity only by measuring the productivity of the most marginal land itself added every year.
The biggest productivity gains of housing people in high productivity areas flows to the already existing stock. Without breakthroughs in transportation speeds or large productivity gains in lower cost areas or easier building in high productivity areas, it will be hard to see construction productivity gains.
Subcontractors may be intermediate inputs from the perspective of a single firm, but they are not intermediate inputs (rather, they are part of value added) for the construction industry as a whole. Intermediate inputs for the construction industry would only include items produced in other industries, such as lumber and fuel (manufacturing) or architectural services (professional services).
Interesting, seems to pretty thoroughly destroy some my priors, most especially that American firms underinvest in capital goods as a result of strong cyclical swings.
So, with that out...
I think there's an open question around compliance/permitting/design, etc. costs; these are things which *should* have benefitted from the IT revolution but I think we all can see that they did *not.*
But that's 10% or so of project costs, maybe 20% in sensitive cases. The major thing that they screw up badly is urban housing, and that less because of direct costs and more because the massive uncertainties they introduce prevent projects from ever happening at all.
Actual construction productivity just seems to be extraordinarily sensitive to Baumol's Cost Disease. We have no clue how to automate most site processes, prefabrication and off-site manufacturing are of limited use in replacing site processes, and those site processes which can be automated are either very expensive to do with machinery or are directly related to management, compliance, and design.
I used to think there were answers, now my working theory is "suck it up and build what we need, land use is most of the problem for urban housing markets anyway."
I will note that a flat metric of productivity per worker should directly translate to a falling metric of productivity per dollar paid given the increase in prevailing wages, again courtesy of Baumol.
I may have missed it, but I didn't see a comparison with other first-world countries in this regard.
If it is Baumol's, I would assume that they'd be suffering similarly.
If they have largely avoided the same fate as us here, it's probably something else.
Have you shared this comment with Goolsbee and Syverson?