Why Aren't There Economies of Scale in Building Size?

(Third in a series on economies of scale in construction, for previous entries see Part I and Part II)

In general, large buildings aren’t much cheaper to build on a per-square foot basis than small buildings are. Why is this the case?

We might expect a few different economies of scale to be at work with regards to building size: geometric effects, fixed effects (spreading the fixed costs of construction over a building larger area), learning effects (workers getting better at building as they progress), and volume discounts. Let's take a look at each.

Geometric effects

We do see a few examples of geometric scaling effects for larger buildings. For instance, an expensive and difficult building assembly is the exterior wall. A building with a square floor plate that doubled its area would reduce the amount of exterior wall per square foot of occupiable space by 30%. Minimizing exterior wall use would also reduce operational costs, since the costs of climate control would (theoretically drop). Following the development of fluorescent lighting, which allowed buildings to be effectively lit without using daylight, skyscrapers in some cases were able to use larger, more efficient floor plates with less exterior wall. Distribution centers and other industrial buildings that have large volumes of interior space can also take advantage of this.

But the ability to minimize exterior wall use is limited. For one, people like having windows, so buildings designed for human occupancy have limited ability to separate people from them (consider the backlash from the low-window dorm designed by Charlie Munger). In some cases, bedrooms are required to have windows for egress reasons.

The vast majority of geometric effects on building size are in fact diseconomies.

Many of these diseconomies are the result of building taller. Each additional floor adds load to the floor below, requiring heavier structural framing and more extensive foundations. Lateral forces rise non-linearly with height (bending moments rise with the square of height; sway and deflection rise with height to the 4th power), requiring stronger lateral systems and a stiffer structure.

The taller your building gets, the more complex your mechanical systems become, and the more elevators you need to add, which encroach on the rentable space below (shorter buildings, on the other hand, might not need elevators at all). A taller building has more stringent fire safety requirements, requiring the use of expensive, non-combustible materials or fire protection systems. Every additional floor in a building requires proportionally more infrastructure, and requires proportionally more construction time (since it takes longer to move people and materials to the top.)

Cost of building with regards to height for US construction, via Eriksen 2021. The jumps at 8 floors are from additional fire safety requirements once your building is a “high rise”. Adding even more floors results in more costs per floor.

(For more on the economics of building height, see my Why are Skyscrapers So Short? piece on Works in Progress.)

A building that just got larger horizontally would also face diseconomies (though fewer, and they might not apply to certain types of buildings such as industrial buildings). Larger buildings require larger parcels of land that will be more difficult to assemble. They face more opposition from the community. They may have increased market risk since they’ll be adding a large volume of available space to the market all at once. They take longer to construct, making the investment returns less attractive. And larger occupancies have stricter fire code requirements even outside of building height.

Fixed effects

There are fewer opportunities for fixed cost scaling in building size than you might expect. Most costs tend to increase as the building gets larger.

For instance, design costs tend to be proportional to building size. Partly this is due to buildings getting more complex as they get larger (offsetting any savings from repetition), partly this is due to design fees also covering things like “inherent risk of taking responsibility for the project” and “helping to resolve issues during construction”, which scale somewhat linearly with project size. There are some cases where this isn’t true (a 400,000 square foot industrial building might not cost much more to design than a 200,000 square foot one), but they’re somewhat limited. You also start to see fixed cost effects once buildings get below a certain size (a 500 square foot building is probably as expensive to design as a 1000 square foot building), but these taper off somewhat quickly. And some small buildings can have extremely low design fees - single family homes, for instance, can be extremely cheap to design because they typically don’t require architect or engineer stamped drawings.

Building permit fees are also often a function of building size.

Land cost seems like an obvious candidate for fixed cost scaling, since you could theoretically build a 1 story or a 100 story building on the same parcel of land, but in practice land cost also tends to scale with building size. Land price tends to be a function of how intensely the plot of land can be used - multifamily developers, for instance, tend to price land in terms of dollars per unit that can be built on the property, or in terms of allowable square footage. Doubling the number of units that can be built on a plot might double the price of it, regardless of its physical size. Similarly, zoning rules such as maximum floor-area-ratios couple allowable building size to the size of the land parcel.

We also don’t see terribly many production methods that have high fixed or upfront costs but are economical for larger buildings. Production methods (pieces of steel craned into place one by one, timber frames built stud by stud, concrete poured form-by-form) also tend to have costs that increase in proportion to building size, though there are some exceptions (a tower crane might require a pad to be poured first, which is a fixed cost independent-ish of building size). Similarly, crane size/expense scales somewhat linearly with building size - larger buildings tend to need either larger, more expensive cranes, or more of them (or both). 

Production methods that do become efficient for large buildings (such as SAM, the masonry robot that worked best on large, uninterrupted sections of wall) do not seem to have taken off.

(One potential counterexample is tilt-up concrete, which is a popular method of building large, low-rise buildings, and is very inexpensive once you’re above a certain building size. Once again, this is a system largely used by industrial buildings).

I think we should find this somewhat surprising - it seems to me that there is a lot of potential for production methods that would have high upfront costs but lower variable costs. There are enough examples of building fabrication equipment that is either portable or could be made portable (things like metal roof machines, CFS roll formers, the Joyn machine, other CNC machines) that it seems like there is potential for much more “on-site building factories” that could be used to produce large buildings cheaply. (Of course, as usual, the attempts at trying this haven’t worked.)

Part of the difficulty is that it's perhaps hard to have a fabrication system like this where costs don’t rise in proportion to building size. Gantry-based building 3D printers seems like an example here - the technology seems like a potential candidate for fixed cost scaling, since there are presumably substantial setup costs for the 3D printer itself that might best be recouped on a large building. But the larger your building gets, the larger and more expensive your gantry system becomes, and the more involved the setup/teardown gets.

The nature of construction work (with much of the value being added to the interior of a partially-built building) also means that the equipment for new production methods will need to be small and portable enough that it can be moved through a doorway, which means it would work just as well on a small building as a large building. Systems such as Dusty Robotics wall layout robot, Canvas’ drywall robot, and Apis-Cor’s center pivot 3D printer for instance, don’t have any obviously high upfront costs, and could effectively be used on small buildings as well as large ones.

Learning curve effects

Learning curve effects do seem to take place in some instances, especially if the building is very repetitive. For multistory buildings, upper floors typically go faster than lower floors, and in at least some cases contractors seem to have had disputes with owners because delays disrupted the learning curves that they had expected (and based their bids on).

Volume discounts

Volume discounts also do seem to exist at the level of the firm, though not necessarily the individual building. Firms who buy in larger quantities are sometimes able to receive lower per-unit costs. However, these discounts seem to be inconsistent - some developers mentioned to me they did exist, while others were unsure. And the origin story of Katerra centers on a developer who failed to receive volume discounts.

Conclusion

Overall it seems like learning curve effects and volume discounts are doing most of the work for scale effects that do occur, but these are likely less than the negative cost impacts of increased size.

(Next week, we’ll look at why builders who build large numbers of buildings don’t seem to do it much cheaper than small volume builders.)