Cracking The Code Part 2: Is Bozeman’s Zoning Preventing Housing Creation?
In our previous post, we examined key provisions of Bozeman’s zoning code in a comparative context. The code did not appear stringent relative to national averages or to a select sample of college towns. We promised that we would subsequently take a different approach to the attempt to detect whether Bozeman suffers from a restrictive and distortive zoning code. Namely, we would look for the indicia of restriction not in the code itself, but rather in data on housing stock and housing production in the city. Specifically, we find the following indicia worth looking for:
Low per-capita housing unit production;
A composition of housing production tilted to one particular form (usually SFH-detached);
A high level of "missing production" in the rankings published by "supply-ist" housing activists like Up For Growth;
Low "zoning margin"; that is, most units zoned for high intensity already host high-intensity uses;
Low elasticity of housing supply, as demonstrated by econometric analysis;
Absence of other factors that would also constrain elasticity, because production hits "zoning limits" before it hits limits on other constraining factors.
Bozeman’s Housing Production: Feast or Famine?
Permit data shows that Bozeman has been producing housing at a per-capita rate with few rivals in the USA. Production has leaned heavily toward multi-family projects.
We will take these indicia one by one and see how Bozeman fares. First, we look at per-capita housing unit production. If zoning interferes with housing production, housing production should be low. The chart below compares Bozeman's housing production (using Census Bureau collation of data on permitting, which is the standard metric in the literature), smoothed over 5-year windows, to several other jurisdictions: Austin, TX (often talked about as a high-production city), Minneapolis and Portland (two cities cited as aspirational examples of aggressive zoning reform), and Palo Alto, CA (a jurisdiction that almost everyone would agree has highly restrictive and distortive zoning).
Bozeman blows away the competition. The city has produced housing at one of the highest rates in the country for at least the last decade. It is worth noting that the closest competitor in this small sample is Austin, a city that notably failed to pass its ambitious and highly contested zoning reform, the so-called CodeNext initiative. Also, Minneapolis and Portland's production peaks, low as they are relative to Bozeman, preceded their zoning reforms.
To get a broader picture of Bozeman's housing production in the national context, we need to look at housing production at the county level. A comparative data set at the county level is easier to construct than at the city/town level, as timely permits and population estimates are not always available at a more granular level. Our data set captures 2013-2021, which if anything may understate Gallatin county's housing production, since 2022 saw a large spike in production at the level of Bozeman. Gallatin County is at the 99th percentile of housing permits per capita among all counties in the data set, the 18th highest out of 2,745 counties. If we look at counties of 25,000 population or more, Gallatin is the 12th out of 1,615 counties. Of the counties with higher production rates, seven are from Texas or Florida, two states renowned for permissive land use rules. To get a sense of how great an outlier Gallatin County is, consider the chart below, which shows the rate of housing permits for the number one county in the national sample (Walton FL), for Gallatin County, for the county at the 90th percentile (Buncombe, the county of Asheville, NC), for the USA as a whole and for a few states, including Montana.
The comparison to Walton County perhaps does an injustice to Gallatin. For whereas Gallatin County is the economic center of the entire region and thus cannot draw on the construction labor and other resources of larger neighbors, Walton County is a small jurisdiction (population around 70,000) bookended by Pensacola and Panama City, which are part of counties with a combined population over over 500,000. Gallatin County has managed to generate housing at a prodigious clip without the ability to draw on the physical resources of larger nearby economic centers. The relevance of this observation will become apparent when we turn to physical limits to production.
But regardless of whether Gallatin should be assessed as in the 99th percentile or the 99.5th percentile, the fact remains that if a city can produce housing at a rate over four times the US average (which has hovered around 4 units per 1,000 residents for the last few years) and over double the rate of the county at the 90th percentile, it would tend to suggest that city does not impose overly onerous restrictions on housing production.
It is worth taking a look at the composition of housing production by form. Advocates for up-zoning will often describe restrictive zoning codes as “outlawing multifamily housing.” In jurisdictions under those kinds of onerous codes, what production occurs will tilt away from multifamily units. Let’s look at the composition of Bozeman’s production.
In the last half-decade, not only has the majority of housing production come in the form of multifamily units, but we see production both of high-unit-count buildings (5+ units) and smaller “missing middle” forms. The new production leans more heavily also toward multifamily than does the existing housing stock, in which single-family (attached or detached) represents a little over half of all housing units in Bozeman. That non-SFH housing represents a large share of the overall stock of housing, regardless of when built, also suggests that zoning has not proven a serious barrier to multifamily production in the more distant past.
Mind the Gap!
Looking beyond raw housing production, we must consider how production aligns with the formation of households. According to anti-zoning group Up For Growth’s national analysis of housing production gaps by market, Bozeman appears to have generated enough housing to keep up with the growth in households.
A slightly more sophisticated approach to the same question involves looking not merely at outright housing unit production per capita, but rather at the gap between housing production and the production required to keep pace with desired household formation. The rub is that one cannot observe this "production gap" directly, because one cannot know precisely the counterfactual amount of household formation that would prevail absent the suppression of household formation resulting from the allegedly inadequate housing production. Analysts use demographic data to try to determine the number of "missing households" and use that as a proxy for underproduction. This is not a perfect methodology, because there is always some degree of voluntary household combination, for example households consisting of roommates by choice or voluntarily multi-generational households. The prevalence of such households varies in an unobservable way (unobservable because we have no easy way to distinguish between voluntary and involuntary “household combination”) between jurisdictions and over time. UFG's methodology tries to control for the first form of variation by comparing prime-age headship rates today to prime-age headship rates in a base period (2012, which was a time when UFG believes housing in the US was closer to having been adequately supplied). It has no way to control for intertemporal variation and simply assumes any negative changes to headship rates reflect "missing households." Missing households in turn reflect underproduction of housing. Additionally, UFG has to build up its calculations from the Census Public Use Microdata Areas (PUMAs). Unlike larger metros, Bozeman itself is part of a larger PUMA, and as such the number for Bozeman covers an area that stretches beyond the city (although the city represents the lion's share). We think that this is a weak methodology, but we understand why UFG pursues it (1). The 2023 report, their latest, uses 2021 data.
So how does Bozeman fare in UFG's study? Bozeman PUMA had no underproduction from 2012 to 2020. In 2021, it slipped into underproduction... of a whopping 0.1% of total units! Basically, statistically indistinguishable from adequate production. The same cannot be said for several jurisdictions that have undertaken zoning reform: Los Angeles (the state’s SB9 initiative), Portland, and Minneapolis. Whether zoning reform has failed or simply has not had enough time to work, one must admit that these jurisdictions do have at least one indicator compatible with the diagnosis of zoning-mediated restriction in supply. Bozeman, not so much.
Zoned Capacity: Signs of Slack?
When parcels zoned for higher density are kept artificially scarce, we should expect more pressure for those lots to be used to capacity. Yet, in Bozeman we see most lots zoned for more than single-family detached housing wind up used for… single-family detached housing. Meanwhile, the multifamily and attached housing that does exist is well distributed over the whole city.
If zoning represents a binding constraint on development in a jurisdiction, a corollary is that land zoned for higher-intensity will be scarce relative to demand for it. Under those conditions, developers should face economic pressure to use up “zoned capacity”. One would not expect to see a lot of parcels zoned for high-intensity use ending up used for only low-intensity use like single-family housing. We can visualize how fully zoned capacity has been used by mapping housing forms and zoning. The map below shows different combinations of housing form and zoning. We consider single-family detached housing in districts zoned only for SFH, single-family detached housing in districts zoned for more intense use, and multifamily housing wherever it may be. For good measure, we also map single-family attached housing.
The amount of “red” – parcels zoned for multifamily that only host single-family detached housing – is startling. Red shows up as more prevalent than yellow (MFH/Mixed Use). It suggests very low pressure to fully utilize zoned capacity, and undercuts the claim of zoning-mediated supply restriction. In terms of numbers, there are over 5,200 parcels zoned for greater than single-family housing that contain nothing more than a single-family detached home. That exceeds the 3,800 parcels actually zoned R-1 or R-S (the traditional "single family zoning") that contain SFH-D! Hardly a picture of full utilization of existing zoning margin.
A similar mapping exercise also shows that multifamily housing options exist all over the city. We do not see a situation where the choice of housing form and of geographic location within the jurisdiction cannot be made independently of each other.
The map below divides the city into a grid of approximately 0.5x0.5 mile blocks. For each grid, we calculated the total number of residential units, the total number of SFH (detached or attached) and the total number of units coded multifamily in the city’s database. We plot for each block what percentage of total units consist of multifamily and color the block accordingly. One can see that other than on a few fringes of the city, nearly every grid square has at least 20% of units in the form of MFH units. For many squares, distributed throughout the city, over 50% of units are MFH. The idea that Bozeman has created “enclaves of single-family-only housing” is so divorced from reality as to be nearly delusional.
Fantastic, Elastic, But Alas Expensive
A real estate market may be expensive for a number of distinct reasons. Zoning reform can make a claim to help if the richness of the market is a product of a lack of supply elasticity. Overly restrictive zoning prevents the market from generating a supply response to higher prices and profitability of development. Bozeman may be expensive, but it has a very high elasticity of supply of new housing.
Now, we will get a little more technical. We mentioned at the beginning of our first post that the economic consequence of a binding, restrictive land-use regime amounts to an artificially low elasticity of housing supply. That is, for a given level of profitability of development (ex-land) production, the amount of housing actually produced ends up lower than in a counterfactual low-restrictiveness regime. The supply curve ends up “flatter” than it would otherwise be, and as a result for a prevailing level of demand, quantity produced will end up lower and price higher than in counterfactual.
So, the heart of the matter is elasticity of supply. We can measure it by seeing how robust a supply response arises from a demand shock. We cannot observe demand shocks directly, but there are statistical methods available to "identify" demand shocks. Citizens for Responsible Planning wrote a whitepaper on supply elasticity (2), which measured elasticity for all US counties according to three different specifications. Here's how Bozeman compares to some other places (we use the percentile rank of elasticity among counties for each specification)
Bozeman is among the highest elasticity areas in the country, on all three specifications. You can see how the legendarily strict jurisdictions in California rank very low. Places with a "build what you want" reputation like Austin, Phoenix and Cape Coral (Lee County, FL) also show high elasticity, but not quite at Bozeman's level. And the YIMBY-cited "upzoning" stories of Portland and Minneapolis show lower elasticity than Bozeman. It suggests that either Bozeman's zoning is not interfering with production, or perhaps that zoning is not one of the most important factors determining elasticity (and in fact, many developers in California will tell you that zoning is a much less salient problem for them than other regulatory brambles like CEQA challenges and the California Coastal Commission -- environmental rules are important, but are often the enemy of housing production in a much more powerful way than zoning).
Now, let's look more directly at elasticity in Bozeman. We created a "profitability metric" using local price history and the Census Bureau's Construction Price Index . We then did a regression to see if housing permit activity was correlated with profitability. We found a statistically significant relationship (r2 = .20, p = .004). Perhaps the easiest way to see the relationship is graphically. Below is a chart of profitability metric (lagged one year) and Bozeman's permitting approvals.
You can see that housing units per capita permitted responds quite quickly and dramatically to changes in the profitability of development. This is elasticity visualized. Now to see inelasticity realized, here is the same chart for Santa Clara County, CA.
We extended CFRP’s analysis for a few selected counties, by including two more years of data and performing our slightly more sophisticated way of generating a development profitability metric. The result gives a statistical estimate of the additional housing unit production per capita (units per 1000 residents) associated with a positive profitability shock of 10 percentage points. We performed the analysis for our entire time series (1988-2023) and then for a shorter history (1999-2023) as a robustness check.
According to this analysis, too, the Bozeman area shows an extremely high elasticity of housing supply. We believe that this is probably the best single metric of whether a jurisdiction is experiencing artificial supply restrictions of any kind. Once again, evidence that Bozeman’s housing affordability symptoms do not arise from a zoning-related malady.
Is the Speed Limit the Problem When A Downed Tree Blocks the Road?
If zoning is the key constraint on production, we should not see other forms of constraints binding. But in Bozeman, there are signs that we are hitting physical limits to production, for example the availability of skilled labor in construction.
Finally, we want to consider whether other constraints on production might start to bind well before zoning. Past some level of quantity, the supply curve will flatten out for one reason or another. That’s the nature of real-world constraints. Bozeman potentially faces worse physical constraints than some other high-production jurisdictions because it represents the economic center of gravity of the region. Bozeman cannot easily call upon productive capacity of neighboring areas the way some other high-production areas can. Think, for example, of the counties north of Dallas, with their ability to call on the rest of the Dallas metroplex, or Hays County Texas, which can call upon the resources of the Austin metro.
Construction labor often becomes noticeably scarce when housing production activity picks up. Skilled tradespeople cannot be trained up quickly and may not fluidly move from one part of the country to another. And, of course, if a jurisdiction already suffers from high housing prices, it can prove even more difficult to attract skilled construction workers without boosting wages to a point that depresses development profitability such that many projects cease to “pencil out”. The chart below shows what percentage of the workforce in a given county works in the residential construction industry (3).
Gallatin County’s workforce already tilts very heavily toward residential construction. How much higher can this go? And what would that mean for construction wages, an input into the profitability equation that drives production at any given level of zoning constraint? We already see that the residential construction industry enjoys an unusual wage premium in Gallatin County.
Even if one believes in the theory that zoning is a major cause of housing affordability challenges, and that upzoning can cure affordability challenges in places with a restrictive zoning code, it is not sensible to take the risk of “side effects” of blanket upzoning in a specific jurisdiction unless it can be established that zoning is restrictive and having a distortive effect there. We have attempted to test that proposition for Bozeman. The data show, both at the level of the code itself and of the resulting patterns of development in the city, little evidence of a zoning-driven problem. As Bozeman re-engages in a discussion of updating the UDC, we urge the city leadership not to make the mistake of leaping right into discussions of treatment before taking the time to make an accurate diagnosis of what drives housing affordability challenges in our city.
Author Benjamin Heller, Bozeman Resident
The biggest problem with this methodology is an unspoken assumption that the reason housing production fails to keep up with household formation is a non-economic production constraint. It is entirely plausible – maybe even the most likely case – that many new households can’t afford their own dwelling unit because of labor market outcomes. This might be the case If population growth is driven, for example, by recent immigrants without assets or established jobs, or if wages in non-construction industries are not keeping up with the purely hardcosts (i.e. labor and materials, leaving aside land and red tape) of production. It is beyond heroic to assume that the only reason for underproduction is land-use regulation. It would perhaps be instructive for the city of Bozeman to compare moves in the Construction Price Index to, say, the evolution of clerical and technical worker wages at MSU. If we perform this comparison for Bozeman City schools, we find that the salary for a 5th year BA+60 teacher rose 6.1% from the 2018-19 school year to the 2023-24 school year, a period in which the Construction Price Index increased 38%.
Wage and employment numbers from 2023 annual reported statistics from BLS Quarterly Census of Employment and Wages.