A low ratio does not necessarily mean that there is little room for efficiency gains. There could still be huge wins by making the *inputs* more efficient. Instead of seeing the cost of the battery in the BMW as "Well we can never get our input costs down lower than this", you could instead see it as "wow, if we can get this cost down then huge efficiency/profit gains are possible". Then you figure out why that battery is so expensive and optimize there.
It is the same for homebuilding. It is not that there are no opportunities for efficiency gains, it is that you are looking at problem only from the fixed level of the current large homebuilders. These companies outsource almost all the tasks that could be made more efficient - they are basically just project managers and financiers rather than home builders.
I think the more interesting question is: Why do current large US on site home builders all seem to fall into this model rather than following the typical producer model of looking for efficiencies and capturing the value add in these processes? Why are there no vertical (actual) home building companies that are training their own crews, buying up their suppliers, and developing new technologies to replace their costly inputs? It is a big enough industry to support this model, so there must be reasons why it does not happen.
Drawing on some experience in multifamily, the obvious answer is that there is no such thing as a standardized product in housing. Sure, homebuilders can definitely modularize some things. But every site is going to be just different enough with just enough need for customization to lead to some amount of friction in the building process. And that's not even accounting for local building codes and zoning rules - even if I were to make a house factory that produces some optimal product for the City of Houston, the moment I end up in another jurisdiction that changes. And in our highest demand areas, nobody wants a tract home. You can probably get away with a row of the same townhome copy+pasted ten times, but you're not going to be successful dropping the same townhome 5,000 times (assuming you could even get enough land in a neighborhood that fits the income level/demographics your design is targeting!)
Seeing economies in housing basically requires nationalized zoning and building rules, or at least statewide mandates in the larger states. And I don't think most people would find that loss of local control acceptable
I suspect it has something to do with risk mitigation on the part of large homebuilders. If they vertically integrate in a cyclical industry such as home building, they are subject to lots of pain when demand decreases. When they are simply project managers and financiers, they can adjust their overhead costs quickly so as to weather the economic down turn. This allows them hold their company together and wait for the next market up swing. They accept the inefficiencies of their material and labor suppliers as a tradeoff for flexibility.
Correct. Brian notes this in the essay's close: "You could vertically integrate backwards into the production of raw materials and components, in the hopes of driving down *those costs."
...but applies to *all* inputs, not just materials and components. Specifically costs that are currently widely distributed across small scale and low capital subcontractors could be greatly optimized.
Removes personality room for Greenspace, impermeable surfaces with parking issues. Living next to a green hill does not exist in most areas. We have to allow for how people would really like to live. Do you attend your neighborhood meetings or Council meetings? People are not happy living in rectangle edifices. We can and should do better
That is too simplistic. Once green infrastructure is removed neighborhoods reconfigured and rectangle boxes installed people have no choice as that's all that is available. It's urban planning at its worst. Even urban developments need peaceful oasis and structures that are architecturally interesting as well as green built and cost effective. Therein lies the problem.
Even if it's not cheaper, the fact that giving many thousand more people homes in just 6 years is pretty amazing.
Sharing the building structure helps saving costs (foundation, heating/cooling, plumbing, etc), makes the urban area more compact (makes it possible to go to places without a car, which makes it possible for a large population to have relatively few cars), and so on.
Also, simply more people can live next to a nice green hill.
Good stuff as usual, thanks Brian. For people who have been following your work and have actual experience in construction this all makes sense. Inevitably, someone will post a link to a video or an article of a pre-fab building that was erected in five hours and say: "See, it can be done!"
I've been following the work of Bensonwood and Unity Homes with a certain amount of cautious optimism. The goal there, if I grasp it correctly, is to deliver better quality than site built/stick built housing. Every project is unique and every site imposes challenges that require adjustments to design and delivery.
I wonder if there is a concept of a "no-cut" home assembly - where all lumber, laminate, tiles, drywalls parts, etc are pre-sized to the layouts, or conversely the layouts and walls are designed around standard drywall and material sizes.
Say everything is designed on a 2ft grid unit with like 10 approved lengths total ( mostly multiples ), that lead to standard wall lengths, ceiling heights, window placements, cabinet widths, floor layout, stair geometry etc where these all derive from the same grid.
If taken to extreme, this would require some vertical integration or optimizations upstream that would be done before materials or the lumber are delivered, like 8" or 10" high holes on all studs for running wire and cables
I assume this practice ( reducing cuts ) is already done to some extent today. But if taken to an extreme, with all facets, materials and aspects, I think it could change home building.
I am looking forward to your essay on reducing the overall cost to material inputs cost ratio.
Isn’t this essentially what Sears, Roebuck, Montgomery Ward, and the other kit house makers attempted in the first decades of the 20th century?
Not sure if they ever produced a “no-cut” kit, but my understanding is that the components were substantially pre-cut and sent to the builder (often the individual homeowner) on 1-2 rail cars, which significantly reduced the need for skilled labor at the building site. The same manufacturer could also supply the plumbing fixtures, furnace, tile, etc., although I expect the builder would still need to engage local contractors for plumbing, foundation work, etc.
My grandfather told me that his kit home which was built in the late 1950s or early 1960s was a no-cut set up, needed 6 people, and they put in together in a day. Was probably only 1000 sq feet, and sourced from around Red Deer in Alberta.
There aren't economies of scale in construction for the same reason the law recognizes a hard distinction between real property and personal property: the latter is more-or-less fungible (or at least it can be, in most of its forms), but the former most definitely is not.
The reason the law looks at it this way is that each plot of land is necessarily unique. Two different plots may be similar, but they aren't numerically identical, and location is important.
But that uniqueness of location spills over into the physical properties of real property, not just its abstract, philosophical uniqueness. Every car that rolls off a Toyota production line is pretty much identical, even allowing for customization of particular units. Supposed to be, anyway. If one swapped out one unit for another with identical specs, there would be no meaningful difference between the two that could not be attributed to manufacturing defects.
The same isn't true of houses. No two are the same. Even when manufactured, the integration process (i.e., turning products off a line into an actual dwelling on actual land), will necessarily render all of them unique in irreducible ways. Constructing an actual building, in the real world, involves simply too many variables, with too many margins for error, too many of which are almost entirely outside human control, for construction projects to ever reach sufficient uniformity for economies of scale to ever emerge.
"Driving nails," for instance, is a repeatable, scalable process only if one eliminates most of the potential variables involved in the task, i.e., relative position, angle, material properties, etc.
If nothing else, production lines gain efficiency because one can do the same task an arbitrary number of times without ever moving from one location. It is, in effect, abstracting a physical task from real space and eliminating most of the variables that go along with changing locations. Construction involves doing similar tasks many times, yes, but even on the same site, the location of each of those tasks in space is different. There may be some gains to be had in pre-fab/manufactured housing, but there still has to be work done to take that pre-fab/manufactured unit and turn it into a real building that someone can live in.
One thing I kept thinking about while reading this is that the constraint might not be purely economic or manufacturing related.
Even if factory production became cheaper, homebuilding still has to interact with a landscape that was divided long before modern industrial construction existed. Every parcel comes with its own geometry, utilities, zoning rules, and local conditions. In that sense, houses aren't products. They are deployed into thousands of unique environments.
Manufacturing works best when you're shipping the same program to millions of users. Homebuilding is closer to custom implementation, where every installation requires adjustments no matter how standardized the core product becomes.
That doesn't explain all of the industry's productivity challenges, but it does seem like an additional reason why scaling housing has proven so much harder than scaling automobiles or consumer electronics.
I have bee building for a number of decades, not professionally. Over time I have noticed some trends in this direction. I now buy "stud length" 2x4's for wall construction, rather than cutting 8' lengths for every stud. Roofs are now built from pre assembled trusses, delivered to the site, rather than building piece by piece. On large jobs the trusses are built into roof sections on the ground, and craned into place on the roof. Cabinets are now all pre assembled. There was a time when they were built from scratch. So there have been some changes. But the price of labour has gone way up, so....
Genuinely enjoying your articles Brian. I’d like to push on an idea though: gross margin seems like the wrong lens for testing economies of scale. Scale effects show up below the gross line, the whole mechanism is spreading fixed costs (overhead, SG&A, plant depreciation) over more units, and gross margin sits above most of that. It also nets out price, so a large builder pricing low for share (as Lennar says it does) would show a flat gross margin even if its per-unit production cost genuinely fell with volume. Wouldn’t COGS per square foot, or operating margin, be the more revealing comparison?
Separately, I’m still not sold that factory production isn’t inherently cheaper than site-building. I take the Phoenix point, multiple plants inside one shipping radius is real evidence against scale. But I’d want to see per-unit production cost decomposed before concluding the curve is flat, rather than inferring it from margins that may be hiding the effect.
Maybe Potter already covered this in another post, but are there economies of scale in building RVs? They are essentially efficiency apartments on wheels, and they don’t seem like they are that expensive to build.
I realize that even asking such a question of Americans is sort of a fool’s errand, but what if at least part of the problem is leaving so much of the process to private enterprise and “the markets” in the first place? I am thinking here for example of Singapore.
Some of the largest inputs in homebuilding are actually the lots themselves. The Levitts built on fairly uniform, inexpensive lots. Maybe design software that better utilizes the lots would bring substantial benefits. The other is 'owner fit.' People will pay multiples for custom, at least on the upper end. Once again, this typically involves many rounds of design Jenga with architects. Savings in this area might unlock substantial value. Both of these shift the cost from design labor to design computation. None of them are scale in the construction process; but there could be spill over so that design was compatible with efficient construction methods, supply chains, instead of going into the artisanal and bespoke.
LNG plants have achieved huge productivity gains by using modularization designs. Modules are built in yards with high levels of relatively cheap, but skilled construction crews, such as in Thailand, Philippines, and Malaysia, and then transported thousands of miles to the site of the LNG plants in Australia, Africa and Middle East.
"the ratio of the costs of the output to the costs of the physical inputs (drywall, lumber, concrete, etc.) is around 2. This is a very low ratio: it’s only slightly higher than the ratio we saw with various BMW i3 subassemblies." This sentence looks incorrect, or am i misunderstanding this section of the article? Aren't the i3 ratios labor/materials not output/materials? The optimized system is a materials/ labor ratio of 1 and construction has achieved it! There is no spread like Elon saw in rockets. This dramatically supports the whole point of the article.
Thanks for this post, very informative. Looking at the overall cost of finished homes in place, there can be substantial costs (land, approvals, permitting, utility hook-ups, roads, etc.) that aren’t captured in the labor and materials that go into the physical home, but coming from a career in construction, I think there are two main factors at play here - 1) The low ratio indicates we already build homes relatively efficiently. There are no major economies of scale, because we are already at scale in most areas (lumber is milled to size, drywall is used instead of lath and plaster, pre-hung windows and doors, etc.) Factory built homes can’t compete because you can’t ship a 1600SF 3br 2ba home with a physical footprint of 40’X40’ on any road, it has the be shipped (mostly empty space BTW) in large subassemblies and assembled on site, on the field prepared foundation with sub-foundation utilities and sewer connections pre-installed. 2) Optimizing factory processes are Complicated - many steps, sequences and processes, but ultimately learnable and controllable. Site construction is Complex - weather, terrain, large equipment, interferences, labor shortages, skill limitations, supply chains, local regulators and inspectors - a huge number of balls to juggle simultaneously, many not under the control of the builder, and all have to be learned and optimized for each site. It’s really amazing many homes are as affordable as they are (California, where I live excluded…).
A low ratio does not necessarily mean that there is little room for efficiency gains. There could still be huge wins by making the *inputs* more efficient. Instead of seeing the cost of the battery in the BMW as "Well we can never get our input costs down lower than this", you could instead see it as "wow, if we can get this cost down then huge efficiency/profit gains are possible". Then you figure out why that battery is so expensive and optimize there.
It is the same for homebuilding. It is not that there are no opportunities for efficiency gains, it is that you are looking at problem only from the fixed level of the current large homebuilders. These companies outsource almost all the tasks that could be made more efficient - they are basically just project managers and financiers rather than home builders.
I think the more interesting question is: Why do current large US on site home builders all seem to fall into this model rather than following the typical producer model of looking for efficiencies and capturing the value add in these processes? Why are there no vertical (actual) home building companies that are training their own crews, buying up their suppliers, and developing new technologies to replace their costly inputs? It is a big enough industry to support this model, so there must be reasons why it does not happen.
Drawing on some experience in multifamily, the obvious answer is that there is no such thing as a standardized product in housing. Sure, homebuilders can definitely modularize some things. But every site is going to be just different enough with just enough need for customization to lead to some amount of friction in the building process. And that's not even accounting for local building codes and zoning rules - even if I were to make a house factory that produces some optimal product for the City of Houston, the moment I end up in another jurisdiction that changes. And in our highest demand areas, nobody wants a tract home. You can probably get away with a row of the same townhome copy+pasted ten times, but you're not going to be successful dropping the same townhome 5,000 times (assuming you could even get enough land in a neighborhood that fits the income level/demographics your design is targeting!)
Seeing economies in housing basically requires nationalized zoning and building rules, or at least statewide mandates in the larger states. And I don't think most people would find that loss of local control acceptable
The villages in Florida has built on that scale.
I suspect it has something to do with risk mitigation on the part of large homebuilders. If they vertically integrate in a cyclical industry such as home building, they are subject to lots of pain when demand decreases. When they are simply project managers and financiers, they can adjust their overhead costs quickly so as to weather the economic down turn. This allows them hold their company together and wait for the next market up swing. They accept the inefficiencies of their material and labor suppliers as a tradeoff for flexibility.
Correct. Brian notes this in the essay's close: "You could vertically integrate backwards into the production of raw materials and components, in the hopes of driving down *those costs."
...but applies to *all* inputs, not just materials and components. Specifically costs that are currently widely distributed across small scale and low capital subcontractors could be greatly optimized.
Hong Kong tends to build residential high rises in identical batches. I'm curious if they successfully get any cost savings off that
Removes personality room for Greenspace, impermeable surfaces with parking issues. Living next to a green hill does not exist in most areas. We have to allow for how people would really like to live. Do you attend your neighborhood meetings or Council meetings? People are not happy living in rectangle edifices. We can and should do better
If people don't like them, they won't buy them.
That is too simplistic. Once green infrastructure is removed neighborhoods reconfigured and rectangle boxes installed people have no choice as that's all that is available. It's urban planning at its worst. Even urban developments need peaceful oasis and structures that are architecturally interesting as well as green built and cost effective. Therein lies the problem.
Even if it's not cheaper, the fact that giving many thousand more people homes in just 6 years is pretty amazing.
Sharing the building structure helps saving costs (foundation, heating/cooling, plumbing, etc), makes the urban area more compact (makes it possible to go to places without a car, which makes it possible for a large population to have relatively few cars), and so on.
Also, simply more people can live next to a nice green hill.
https://en.wikipedia.org/wiki/Kin_Ming_Estate completed in 2003, started in 1997 (by demonlishing the old village, and flattening a hill)
https://en.wikipedia.org/wiki/Tiu_Keng_Leng#/media/File:Tiu_Keng_Leng_1955.jpg
Good stuff as usual, thanks Brian. For people who have been following your work and have actual experience in construction this all makes sense. Inevitably, someone will post a link to a video or an article of a pre-fab building that was erected in five hours and say: "See, it can be done!"
I've been following the work of Bensonwood and Unity Homes with a certain amount of cautious optimism. The goal there, if I grasp it correctly, is to deliver better quality than site built/stick built housing. Every project is unique and every site imposes challenges that require adjustments to design and delivery.
I wonder if there is a concept of a "no-cut" home assembly - where all lumber, laminate, tiles, drywalls parts, etc are pre-sized to the layouts, or conversely the layouts and walls are designed around standard drywall and material sizes.
Say everything is designed on a 2ft grid unit with like 10 approved lengths total ( mostly multiples ), that lead to standard wall lengths, ceiling heights, window placements, cabinet widths, floor layout, stair geometry etc where these all derive from the same grid.
If taken to extreme, this would require some vertical integration or optimizations upstream that would be done before materials or the lumber are delivered, like 8" or 10" high holes on all studs for running wire and cables
I assume this practice ( reducing cuts ) is already done to some extent today. But if taken to an extreme, with all facets, materials and aspects, I think it could change home building.
I am looking forward to your essay on reducing the overall cost to material inputs cost ratio.
Isn’t this essentially what Sears, Roebuck, Montgomery Ward, and the other kit house makers attempted in the first decades of the 20th century?
Not sure if they ever produced a “no-cut” kit, but my understanding is that the components were substantially pre-cut and sent to the builder (often the individual homeowner) on 1-2 rail cars, which significantly reduced the need for skilled labor at the building site. The same manufacturer could also supply the plumbing fixtures, furnace, tile, etc., although I expect the builder would still need to engage local contractors for plumbing, foundation work, etc.
My grandfather told me that his kit home which was built in the late 1950s or early 1960s was a no-cut set up, needed 6 people, and they put in together in a day. Was probably only 1000 sq feet, and sourced from around Red Deer in Alberta.
Advanced Wood Framing is pretty much what you're describing, based on a 2ft grid: https://buildingscience.com/documents/insights/bsi-030-advanced-framing
There aren't economies of scale in construction for the same reason the law recognizes a hard distinction between real property and personal property: the latter is more-or-less fungible (or at least it can be, in most of its forms), but the former most definitely is not.
The reason the law looks at it this way is that each plot of land is necessarily unique. Two different plots may be similar, but they aren't numerically identical, and location is important.
But that uniqueness of location spills over into the physical properties of real property, not just its abstract, philosophical uniqueness. Every car that rolls off a Toyota production line is pretty much identical, even allowing for customization of particular units. Supposed to be, anyway. If one swapped out one unit for another with identical specs, there would be no meaningful difference between the two that could not be attributed to manufacturing defects.
The same isn't true of houses. No two are the same. Even when manufactured, the integration process (i.e., turning products off a line into an actual dwelling on actual land), will necessarily render all of them unique in irreducible ways. Constructing an actual building, in the real world, involves simply too many variables, with too many margins for error, too many of which are almost entirely outside human control, for construction projects to ever reach sufficient uniformity for economies of scale to ever emerge.
"Driving nails," for instance, is a repeatable, scalable process only if one eliminates most of the potential variables involved in the task, i.e., relative position, angle, material properties, etc.
If nothing else, production lines gain efficiency because one can do the same task an arbitrary number of times without ever moving from one location. It is, in effect, abstracting a physical task from real space and eliminating most of the variables that go along with changing locations. Construction involves doing similar tasks many times, yes, but even on the same site, the location of each of those tasks in space is different. There may be some gains to be had in pre-fab/manufactured housing, but there still has to be work done to take that pre-fab/manufactured unit and turn it into a real building that someone can live in.
One thing I kept thinking about while reading this is that the constraint might not be purely economic or manufacturing related.
Even if factory production became cheaper, homebuilding still has to interact with a landscape that was divided long before modern industrial construction existed. Every parcel comes with its own geometry, utilities, zoning rules, and local conditions. In that sense, houses aren't products. They are deployed into thousands of unique environments.
Manufacturing works best when you're shipping the same program to millions of users. Homebuilding is closer to custom implementation, where every installation requires adjustments no matter how standardized the core product becomes.
That doesn't explain all of the industry's productivity challenges, but it does seem like an additional reason why scaling housing has proven so much harder than scaling automobiles or consumer electronics.
I have bee building for a number of decades, not professionally. Over time I have noticed some trends in this direction. I now buy "stud length" 2x4's for wall construction, rather than cutting 8' lengths for every stud. Roofs are now built from pre assembled trusses, delivered to the site, rather than building piece by piece. On large jobs the trusses are built into roof sections on the ground, and craned into place on the roof. Cabinets are now all pre assembled. There was a time when they were built from scratch. So there have been some changes. But the price of labour has gone way up, so....
Genuinely enjoying your articles Brian. I’d like to push on an idea though: gross margin seems like the wrong lens for testing economies of scale. Scale effects show up below the gross line, the whole mechanism is spreading fixed costs (overhead, SG&A, plant depreciation) over more units, and gross margin sits above most of that. It also nets out price, so a large builder pricing low for share (as Lennar says it does) would show a flat gross margin even if its per-unit production cost genuinely fell with volume. Wouldn’t COGS per square foot, or operating margin, be the more revealing comparison?
Separately, I’m still not sold that factory production isn’t inherently cheaper than site-building. I take the Phoenix point, multiple plants inside one shipping radius is real evidence against scale. But I’d want to see per-unit production cost decomposed before concluding the curve is flat, rather than inferring it from margins that may be hiding the effect.
Maybe Potter already covered this in another post, but are there economies of scale in building RVs? They are essentially efficiency apartments on wheels, and they don’t seem like they are that expensive to build.
I realize that even asking such a question of Americans is sort of a fool’s errand, but what if at least part of the problem is leaving so much of the process to private enterprise and “the markets” in the first place? I am thinking here for example of Singapore.
Some of the largest inputs in homebuilding are actually the lots themselves. The Levitts built on fairly uniform, inexpensive lots. Maybe design software that better utilizes the lots would bring substantial benefits. The other is 'owner fit.' People will pay multiples for custom, at least on the upper end. Once again, this typically involves many rounds of design Jenga with architects. Savings in this area might unlock substantial value. Both of these shift the cost from design labor to design computation. None of them are scale in the construction process; but there could be spill over so that design was compatible with efficient construction methods, supply chains, instead of going into the artisanal and bespoke.
My first post on substack, so please excuse any breaches of protocol.
Are you asking for the mathematical, political or human response to that question.
They will undoubtedly be different, depending on your point of view.
LNG plants have achieved huge productivity gains by using modularization designs. Modules are built in yards with high levels of relatively cheap, but skilled construction crews, such as in Thailand, Philippines, and Malaysia, and then transported thousands of miles to the site of the LNG plants in Australia, Africa and Middle East.
A completely different approach to the problem, even if not suitable for everyone: https://www.amazon.com/dp/B00U0C9HKW
Think of it as the third and final stage in the suburbanization of the metropolitan complex.
"the ratio of the costs of the output to the costs of the physical inputs (drywall, lumber, concrete, etc.) is around 2. This is a very low ratio: it’s only slightly higher than the ratio we saw with various BMW i3 subassemblies." This sentence looks incorrect, or am i misunderstanding this section of the article? Aren't the i3 ratios labor/materials not output/materials? The optimized system is a materials/ labor ratio of 1 and construction has achieved it! There is no spread like Elon saw in rockets. This dramatically supports the whole point of the article.
Thanks for this post, very informative. Looking at the overall cost of finished homes in place, there can be substantial costs (land, approvals, permitting, utility hook-ups, roads, etc.) that aren’t captured in the labor and materials that go into the physical home, but coming from a career in construction, I think there are two main factors at play here - 1) The low ratio indicates we already build homes relatively efficiently. There are no major economies of scale, because we are already at scale in most areas (lumber is milled to size, drywall is used instead of lath and plaster, pre-hung windows and doors, etc.) Factory built homes can’t compete because you can’t ship a 1600SF 3br 2ba home with a physical footprint of 40’X40’ on any road, it has the be shipped (mostly empty space BTW) in large subassemblies and assembled on site, on the field prepared foundation with sub-foundation utilities and sewer connections pre-installed. 2) Optimizing factory processes are Complicated - many steps, sequences and processes, but ultimately learnable and controllable. Site construction is Complex - weather, terrain, large equipment, interferences, labor shortages, skill limitations, supply chains, local regulators and inspectors - a huge number of balls to juggle simultaneously, many not under the control of the builder, and all have to be learned and optimized for each site. It’s really amazing many homes are as affordable as they are (California, where I live excluded…).