Construction cost is a big deal

A hard problem that pro-housing advocates need to think about more

I closed the last post with a note on how housing costs are driven by two factors - the cost of the land and the cost of construction. Our housing crisis is in part this conundrum:

• Solving land cost is technically easy but politically hard.

• Solving construction cost is politically easy but technically hard.

I think this observation warrants a little more exploration, so let’s do a little math. I think many pro-housing folks are walking around with a mental equation something like this in their heads:

Given:

l = cost of the land
a = floor area of the building(s) (in square feet)
c = cost of construction (per square foot)
n = number of housing units on the lot
T = the total transaction cost, as a percentage of the project¹

Then, the minimum cost of a housing unit, H, is:

\(H = \frac{(l + ac)(1 + T)}{n}\)

This equation is correct and useful, as it makes the YIMBY argument very clear:

• That n in the denominator can do a lot of work for us, so let’s permit more units on each lot.

• Multiplying the project cost by T could make things a lot more expensive, and regulatory compliance is a major factor of T, so let’s streamline the approval process to drive that down.

That argument is correct and we should do both of these things.

But I think as we’re all mentally focused on driving up n and driving down T it’s easy to miss that sneaky little c in the numerator. So just to illustrate how powerful the cost of construction is, let’s do something completely unreasonable. Let’s imagine we have free land, no transaction cost, and can build any number of units we want. Our equation is now very simple:

\(H = \frac{ac}{n}\)

So how affordable can we make housing? To get some numbers for this equation we need to pick a location, so I’ll pick on Denver.

Suppose I wanted to build a 2,000 sqft home. Residential construction in Denver ranges from $230-465/sqft. Even anchoring on the bottom end $230/sqft, that 2000 sqft house will cost $460,000 — again, that’s without and land or transaction cost. If we’re a little more pessimistic and say that we’ll get a construction cost in the exact middle of this range ($347.5/sqft) that house will cost $695,000.

Is that affordable? The median household income in Denver is $85,883. The rule of thumb² is you can afford 4-5x your income, if we go with 5x we get $429,415. So our cheapest construction is stretching the limit of affordability, even without any land or transaction costs, and mid-market construction is well out of reach.

Well, if we can’t make construction cheaper, we can make homes smaller. At $347.5/sqft our median household can afford to build about 1,200 square feet (for $417k). That’s a little tight if you have multiple kids, but workable for most families and plenty for singles and couples. So building smaller is a good idea, and we should do that.

But wait, we assumed that we magically paid nothing for land and transaction cost, so what happens if we add those back in?

Let’s assume that the total transaction cost is 40% (see footnote 1), that we can buy a good lot for $1 million, and that we have permission to build as many units as we want — more units will spread that cost around.

Let’s start with 5 units on that lot. Sticking with 1,200 sqft homes, that’s a 6,000 sqft building:

\(\frac{($1M + (6,000 × $347.50)) × 1.4}{5units} = $863,800\)

That’s $863k per unit, nowhere near what we need.

What if we built 20 units (24,000 sqft building)?

\(\frac{($1M + (24,000 × $347.50)) × 1.4}{20units} = $653,800\)

Still not close. Forget about parking garages, elevators, and all the rest, what if we could fit 100 units (120,000 sqft building)?

\(\frac{($1M + (120,000 * $347.50)) × 1.4}{100units} = $597,800\)

You may notice we’re approaching a limit. No matter how cheap the land, no matter how many units we build, and no matter how efficient we can make the rest of the transaction, we can’t drive the unit cost below the cost of construction. So that exercise of magically zeroing those out to see where we stand is actually pretty useful.

Let’s wrap this up with a bit more napkin math.

In an unregulated environment we could get the land to be whatever percentage of a project we wanted by adding more units.³ We could combine that with our total transaction cost in to a single cost multiplier, and say something like “the land and total transaction cost will add about 50% to any project.”⁴

Using that 50% added cost, at $347.50 per square foot our Denver median household can only afford to buy an 825 square foot home for their max budget of $430k. That’s pretty tight.

If we could deliver construction at the lowest price ($230/sqft) we’d get them back up to 1,200 sqft for their $430k budget. To get to Denver’s median home size around 2,000 sqft we’d need to reduce construction costs to around $145/sqft.

Obviously if we look at households below median income we’d need some mix of smaller units or even cheaper construction, but I think its more historically realistic for us to aspire to build sufficient quantities of desirable new housing for median income or higher households, which would eventually reduce demand for older housing such that lower income households could afford “used” housing.

In conclusion, I’d encourage pro-housing advocates to think this way: the cost of construction puts a floor under our collective ability to deliver abundant, affordable housing, and right now that floor is high enough that regulatory reform alone is not likely to achieve our desired outcome. So in addition to pushing hard on the political levers, we need to be interested in what makes housing construction expensive, and searching for opportunities to reduce those costs.

1. Let’s talk about total transaction costs! I hate glossing over this in the article, but it doesn’t change the story and I made an editorial decision that it was distracting from the point. So, where did I get this number from?

In my mental model, the total transaction cost (T) includes these three components:

• All the professional and service fees of a project, including the land purchase and unit sale transaction costs, financing cost, design, consulting, etc.

• The direct regulatory fees associated with a project, but more importantly all the professional and service fees that cities inflict on projects by making developers redo their project plan over and over again for an unknowable amount of time as they beg for permission to be allowed to build anything.

• The profit margin necessary for a developer and investors to bother attempting the project in the first place.

These costs are all proportional to the size of the project, so we can approximate them as a percentage added to the hard cost of the project.

(Aside: The first two components are usually called the “soft costs” of a project, but when people talk about soft cost they aren’t usually including the return on investment for the developer and investors. Since ROI does contribute to the final cost of a delivered unit I’m rolling all of that together into T.)

Now, T varies widely depending on the details of a particular project, especially the location of a project, which determines the regulatory regime it must comply with. And since we’re multiplying all our other costs by 1+T, T is a really big deal, and hand-waving it away seems unreasonable. But we can do this because we know those costs are actually bounded between some minimum and maximum.

I suppose the minimum is theoretically 0, but in real life soft costs are never less than 10-15% of a project.

As for the upper bound, there really isn’t a limit — as

Kevin Erdmann

would say, for many projects the approval cost would be infinity. But those projects will never be built. A lot of the value added to a project is the developer’s ability to navigate the hostile vetocracy and emerge with a building permit. The longer and harder this is, the higher T rises, until the project is no longer feasible and/or the developer and investors run out of money. This is how cities prevent projects.

So, all of that acknowledged, it’s safe to say that for the majority of projects that will actually be built, T ∈ [0.2, 1], meaning this adds 20-100% to the cost of a project. 20-30% is often given as a rule of thumb for soft costs excluding return on investment, so I’m going to arbitrarily add another 10% padding on top for ROI and claim that T = 0.4 is a reasonable enough assumption for napkin math.

If you got this deep into this epic footnote and think I’ve made a mistake, please let me know!

2. Just to sniff test this, I also tried using the Zillow Affordability Calculator, which tells me that for a household earning $85,883 per year and a healthy $100k down payment, they can afford $413,306 (very close to 5x the income).

3. In fact developers usually look at projects this way, assuming land should be 20-25% of a project cost.

4. This is pretty arbitrary but it’s easy to do mental math with and plausible given the assumptions about T in footnote 1 and the idea that we’re intentionally trying to drive the cost of land down as a share of the project down by building more units.