Mini Post #4: We Unbuilt This City

Could "bad vibes" explain the housing crisis?

I’ve decided to try something new: once a week, to write a shorter, narrower post focusing on one specific paper. Last week, I wrote about how having a baby affect mothers and other family members. This is the tab with all the previous (and current) posts.

Onto the actual post: do people’s attitudes towards new housing affect housing policy?

---

This Kind Of Smart, Walkable, Mixed-use Urbanism Is Illegal To Build In Most SimNationese Cities

Housing is, as you might note, something of a pet interest of mine. In particular, I am of the (niche, unpopular, heterodox belief) that the quantity of housing a given place builds has a close relationship to the cost of said housing. In this sense, a city having a higher population with no more housing units to accommodate them should expect to face skyrocketing housing costs. The reason attributed to this housing shortage is, usually, the overregulation of housing construction.

This week’s paper is “Built out cities? A new approach to measuring land use regulation” (2024) by Paavo Monkkonen, Michael Manville, and Michael Lens. The study aims to utilize quirks in California state law to identify how cities’ attitudes towards new construction impact their regulatory output and outlook.

How can we measure housing regulation is a central question, if one is empirically minded and therefore wants to find evidence for our claims.A simple way would be to take a direct indicator, like % of land zoned for single families, or the floor-to-area ratio - but this is, obviously, very limited. One traditional way to do so is with regulatory indeces, such as the Wharton Residential Land Use Regulatory Index, which utilizes data such as requirements of community/legal involvement, the presence of zoning boards, the number and type of permits required for different types of construction, etc. These indeces capture the general gist of the regulatory stance, but chokes up comparing a city like Houston, which has loose zoning regulations but tightly controls parking, density, and floor-to-area ratios, to cities with loose de jure regulations but tight de facto zoning by demanding extensive environmental impact reviews, zoning board presentations, design review, and community input. Additionally, if you simply ordered a city to loosen certain regulations, it could just tighten others - requiring more parking, reduce height limits, density controls.

It is important to find a separation between housing regulation reducing the supply of housing (and thus increasing prices “badly”) and housing regulation increasing the quality of housing (a “good” reason) because they would both have the exact same observable effect on housing costs, but would carry entirely different policy prescriptions. This would create a positive bias for whatever variable measures more regulation and higher prices. Likewise, housing prices could have a negative bias vis-a-vis regulation because, firstly, it could be that more expensive areas end up building more even with regulation constant, driving down prices, but additionally, because regulations might emerge as a reaction to increased supply. This means that capturing the impact on housing production, rather than prices - because prices.

A different way to do it would be to develop some sort of measure to the number of people the city has decided it can acommodate, and compare it to the number of people it currently zones for. This figure would capture the city’s own “anti housing” bias, since it relies on the city’s own estimates and judgments - if Bridgeport residents declare that they can welcome another million people, while Pleasantview decides that the city is full and won’t be turned into Manhattan, that should trickle out into their zoning regulations and thus into their housing output. Put another way, if you could estimate a series of cities’ own estimates for how many people they want to live there, you could capture their relative hostility to new construction, which would inform the political process and thus their regulatory output.

Of course, doing this would require telepathy, or insane amounts of data and estimators thereof - wouldn’t it? No, it turns out, in California - because of the Golden State’s Housing Element law, which requires all jurisdictions to estimate their own unbuilt capacity - i.e. the maximum number of people they chould house. This item is fundamentally subjective, even if it often relies on objective estimators: vacant land available, terrain, population growth, etc. But it also includes zoning regulations - meaning that not only it includes a city’s ability to build more, but also its willingness to do so. Los Angeles, for instance, was zoned to hold up to 10 million people in 1960, but by 1990 it had reduced its capacity to 4.5 million - without changes to terrain, land area, or seismic risk. The city simply made the political decision to rule out 5.5 million inhabitants more than it had in 1990.

Starting with the choice of variables, the paper focuses on housing production and not housign costs, simply because housing prices have a mostly positive relationship with regulation (at least in a naively straightforward way), while production has a negative one. This means that more housing regulation, as stated above, could improve the quality of the housing stock, or be a response by privileged neighbors wanting to keep their ritzy homes valuable, and would generate a positive bias. Meanwhile, housing production is negatively related to regulations, because the possibility that neighbors would want lower development reduces the estimate. For a careful academic study, this is better than a bias for larger effects, because it means you’re drawing the most limited conclusion possible.

The reason why the paper selects unbuilt capacity is that focusing on single regulations results in counterintuitive or flat out wrong results (such as, in one study, San Francisco being less stringently zoned than Chicago) because the thicket of regulations as a whole is the variable of interest, not the specific channels involved. Likewise, use of indeces is hindered by the fact that they rely on questionnaires and surveys of planning and zoning authorities, who usually give inconsistent or contradictory responses between and within studies. Generally, these survey-based indeces are considered useful because they tend to systematically underestimate the severity of regulations, and because planners generally have an accurate assesment of the broad outlines of regulation - meaning that, even if the subindeces might be wrong, the overall picture they paint would be more correct than the sum of its parts.

One final consideration is that, overall, there are two types of supply-restricting policies: those who focus on prohibiting certain types of builds, versus those who demand long, strenuous processes to do so. Their joint impact could be proven clearly negative, but the incidence of process-based consideration could be nonbinding unless prohibitions are very lax. Put another way, if a city has very tight regulations around what developers can build, then they’ll only propose things that can pass all those hurdles and the endless hearings will not be very important, but if a city has broader guidelines, community and expert input might actually bog down a development. Process constraints are also hard to measure, because they change frequently, can be so complex that not even planners understand them, and only bind where prohibition doesn’t, which may mean they’re seldom approved. Lastly, process is more endogenous to supply than prohibition: if developers rarely need to request approval, since they’re already getting prohibited out of that project, then process restrictions will be lax; in the contrary, less stringent prohibition-type regulations might be replaced with more severe process-based obstructions. Lastly, demand plays a role too, since projects with high demand but little chance of being approved won’t need complicated community hearings, and projects with little demand regardless of prohibitive regulations will not be bound by restrictions through process. But the combined effect of these factors on measurements will be uniformly negative, not positive, meaning that once again it can be considered less harmful than the converse.

To measure unbuilt capacity, and thus evade complicated considerations about housing regulations, the researchers utilize California’s mandatory Regional Housing Needs Assessment (RHNA). The state forces municipalities to zone for a given number of housing units, broken down by income segments, and based on projections of future population growth in that area. Local authorities, thus, have to comply by providing a government program that can feasibly add those units, identifying specific parcels and presenting an analysis of their unbuilt capacity - which must exceed the state’s housing target by some amount. Being compulsory, this means the entire state is covered; because it’s consequential, it means it actually compels the municipalities to realistically zone for more population. And because it’s embedded in the political process, it measures the city’s general stance towards building.

This means that cities will consciously choose to include places that will never actually be redeveloped as future cites of mass social housing projects - to quote a pastor that allowed her church to be listed as a future redevelopment site: “what's the harm in letting our property be listed as one of the imaginary sites where it could be built, for the sake of submission to the state’? So we did the city a favor.” Additionally, most cities who don’t want to allow more construction simply report back that they are at the maximum possible level of development - even when they could increase their housing targets by, for example, replacing all their existing housing with massive apartment buildings.

The paper uses as its central measure the unbuilt capacity of 414 out of California’s 482 cities between 2014 and 2021, with 346 cities reporting multifamily housing specifically. The unbuilt capacity, therefore, is the city’s total estimate of the number of new housing units its existing zoning allows. The authors also supplement their measurements with indeces measuring restrictions through both Process and Prohibitions, and with the Wharton Index as a general index on constraints.

The core assumption is, therefore, that higher housing costs will in turn attract higher housing construction, unless there are regulatory obstacles for new development. Thus, they draw regressions between units permitted, the regulatory measures, and demand for new housing. They also include controls for city characteristics, like size, job market, density; a variable for demographics; and fixed effects (i.e. controls) for each metropolitan area - so Palo Alto, San Francisco, Oakland, Monterrey, etc; will all get the same effect, being roughly part of one big city despite being individual municipalities with different housing regulations. There are also a second set of regressions that interact the measures of regulation with rent, to test for the idea that regulation is only binding when demand exceeds supply.

Now, to results: first, the determinants of unbuilt capacity. To start, cities with more density have less unbuilt capacity, same as cities with more valuable housing, older housing, and higher homeownership - “vindicating” the idea that rich dream hoarders squatting on their mansions are the ones driving the decline in housing. Interestingly, there is no association between unbuilt capacity and either the age of the residents or the share of White residents, meaning that the stereotype of white boomers being the biggest NIMBYs is not necessarily true in the aggregate. Surprisingly, neither of the three regulatory indexes have associations with unbuilt capacity, with or without controls, which points to the important distinction between what cities claim they allow (their regulations) versus what they actually allow (the existing construction); simply put, unbuilt capacity measures the housing that could actually get built in numbers, and not the general idea of more or less housing - especially important since many cities claimed to not have any physical space available for new homes.

Looking at the number of permits issued, with separate regressions for unbuilt capacity and for each index, the authors find that cities with less unbuilt capacity permitted less housing, with an elasticity of 0.4 - that is, a 10% decrease in unbuilt capacity is associated with 4% fewer permits. The joint index for regulation, as well as the separate coefficients for Process and Prohibition, are not significant in either case, meaning that an onerous restrictions could or could not be binding, and could or could not have complex effects on actual supply. For mutlifamily permits only, the coefficients are all significant and have the expected coefficients: a 10% decrease in unbuilt capacity results in a 2.4% drop in multifamily permits, while a one unit climb in Prohibition results in 35% fewer permits. The Process Index is positive and significant, largely due to the aforementioned endogeneity.

Including rents as an interaction term does not significantly affect the results: the term is not important for the separate Prohibition and Process indeces, but is quite significant for unbuilt capacity or the Wharton Index, lending credence to the idea that regulation is more binding given higher demand. However, the idea that the P&P indeces are non-binding separately but binding together is interesting, and might be explained by the idea that zoning choice in vacant parcels is strategic. Regressing permitting by both rents and unbuilt capacity results in a positive and significant interaction term: unbuilt capacity matters more at higher rents, meaning that housing will only be built in large numbers by cities where it would be profitable and where it would be legal. This outcome is similar for multifamily permits only.

In conclusion, there is new evidence to predict that land use regulations decrease housing production, which in turn reduces housing affordability. This doesn’t mean that the proximate means of restricting housing supply don’t matter, just that the specific forms taken by supply restrictions might distort overall estimates, and research should be forewarned to use estimates of regulation that don’t include it.

---

To finish up, some links:

• The paper in question

• The 2006 paper setting up the Wharton Land Use Index (fun fact: Larry Summers’ mom, Anita Arrow Summers, is one of the coauthors)

• An empirical paper estimating the effect of being forced to move to a new city after a volcanic eruption (spoiler alert: positive)

• A previous post about the overall importance of housing

• An article I wrote for Liberal Currents about housing costs and wealth inequality