The Toughest Places to Build: Behind the Scenes of a Wall Street Journal Analysis

The Wall Street Journal recently asked me to identify and rank the nation’s toughest places to build housing. This blog post provides readers with the full set of rankings, as well as commentary and a rundown of the data and methods used.

Download Rankings

• Metropolitan areas
• Zip code areas, ranked nationally
• Zip code areas, ranked within metropolitan areas

How to identify tough-to-build places?

One might think that places which are slower to permit construction or reject a greater share of plans are tougher to build in than others, but the opposite is often true. The toughest places to build tend to see few, if any, proposals for new construction because developers know to conserve their energy by pursuing more feasible alternatives. For example, if expensive Silicon Valley suburbs like Palo Alto were to sprout glass condo towers in the midst of single-family homes, they would sell out in no time. However, the fact that no developer wastes their time proposing such a project doesn’t mean it would be easy for a developer to swoop in to build it. On the contrary.

A better way to gauge the toughest places to build is to ask “where does an increasing willingness to pay for housing fail to result in more housing being built?” If people are willing to pay increasing amounts of money for housing, then a paucity of new homes indicates that construction is obstructed, i.e. that it is a “tough-to-build” area. The obstruction can stem from natural geography, from man-made constraints on what may be built, or – as is generally the case – from a mixture of both. Therefore, stunted or delayed construction plans do not indicate that a place is tough-to-build. In fact, it often indicates that plans are likely enough to succeed that they’re worth a shot for developers.

An exception to the rule is the rust-belt scenario in which existing homes – left over from better economic times – are sufficiently abundant relative to current needs to keep the price of housing below the cost of new construction. In this case, there is little impetus to build.

Metropolitan Areas

The extent to which American metro areas are tough-to-build is closely related to their ability and tendency to expand their developed footprint and to densify within that footprint.

Below is the metro area chart, published by the Wall Street Journal. Data-visualization master Edward Tufte would surely argue that their chart is more effective than my original chart at delivering the information visually (it is even entitled “When Less is More”). Aside from squaring the rectangle, switching the axes and clearing the visual clutter, there are three noteworthy differences between The Wall Street Journal’s chart and my original version, beneath it:

First, The Wall Street Journal omitted plotted metro areas that were not labeled in the original; they can be found in the rankings available for download, above.

Second, The Wall Street Journal also omitted metro areas in which the impetus to build was weak, which appear with triangular markers in the original chart (I have referred to these cities elsewhere as legacy cities). Whereas Detroit and Cleveland have seen declining home prices and little new construction, Pittsburgh and Buffalo have experienced price increases, but these price gains are not yet sufficient to trigger significant amounts of new construction. New Orleans is more complicated because of Hurricane Katrina.

Finally, the original chart emphasizes that it is each metro area’s direction with respect to the origin that represents how tough it is to build. Thus:

• It is tougher to build in Honolulu and Los Angeles than in San Francisco and New York.
• New York and Philadelphia are tougher-to-build than Boston and Miami, which are tougher than Washington, which is tougher than Seattle.
• Seattle and Riverside – whose metro area spans Los Angeles’ so-called “Inland Empire” – are roughly on par with each other.
• Denver and Houston are roughly on par with each other as well but are both tougher to build than Dallas, Phoenix, and Austin.
• Except for Las Vegas, the large metros in which it is easiest to build are located in the Southeast, and Atlanta stands out as the easiest among them.

Zip Code Areas

The best arbiters of the zip code area rankings are the long time residents of each metro, who know the places from personal experience. Nevertheless, a handful of broad observations are in order:

• The toughest places to build are not downtown. It is expected and accepted that U.S. downtowns be dense, and once density is accepted in an area it is easy to build more there.
• The toughest-to-build places tend to be in the inner suburbs. Local land use rules typically codify as taboo dense construction outside of downtowns and in the vicinity of transit hubs. Because the inner suburbs have been around longer than more distant suburbs, the inner suburbs are more likely to have depleted their supply of vacant lots, leaving no room for “acceptable” new construction. The three toughest-to-build neighborhoods featured in The Wall Street Journal – Venice Beach in Southern California (90291), Prospect-Lefferts Gardens in Brooklyn (11225) and the Fishtown section of Philadelphia (19125) – all fall into this category.
• The toughest-to-build places are often in gentrifying neighborhoods. While the process of gentrification is in progress, neighborhoods experience sharp housing price appreciation. However, because gentrification is often closely tied to the neighborhood’s physical charm, housing price appreciation is rarely met by equally large increases in the rate of new construction. As a result, gentrifying neighborhoods often elicit an increasing willingness to pay for housing while failing to get more housing built, i.e. they are tough-to-build. The three toughest-to-build neighborhoods featured in The Wall Street Journal fall into the gentrifying neighborhood category as well.
• Another variety of tough-to-build places consists of exclusive and wealthy low-density enclaves. When such enclaves are sufficiently mature that they no longer harbor vacant land eligible for construction, they often fail to produce any new housing. In metro areas that continue to sprawl, gentrification is less common and the sharpest housing price appreciation often occurs in exclusive, wealthy enclaves such as these. The Villages in Houston (77024) and Brookhaven in Atlanta (30319) are good examples.

---

Data and Methodology for the Metro-level Analysis

Data:

Data on the change in the number of housing units are drawn from comparing the 2000 Census and the 2015 1-year ACS. Data on housing price changes are drawn from the work of Bogin, Doerner and Larson at the FHFA. Population data used to determine the 50 largest metros are drawn from the 2010 Census. Data on housing price levels, used in identifying metros in which housing supply exceeds current demand, are drawn from Zillow.

Method:

The percent change in housing units (dQ) is plotted against the percent change in housing prices (dP) for the top 50 US metros by population, plus Honolulu. Metros are ranked as toughest-to-build based on the angle at the origin between the metro and the dP axis, (i.e. metros with a smaller change in quantity relative to the change in price are taken to be tougher-to-build). Metros with less than a 10% change in quantity and median home values below 75% of the sample median (of the 51 metros) are labeled as having legacy housing supply that exceeds current demand.

Rationale:

Places that are “tough-to-build” are interpreted as having a low price elasticity of housing supply. Inasmuch as supply curves are fixed, and changes in price and quantity are driven only by increasing demand, the dQ/dP ratio reflects the (arc) elasticity of supply. For metros whose demand has never fallen in the past, the “inasmuch” assumption clause is a rough but reasonable approximation, especially with respect to preserving the ranking of elasticities as opposed to their levels. Intuitively, when the supply curve is fixed and changes in price and quantity are driven only by increasing demand, one can think of rising prices as a measure of “unmet demand” or “insufficient housing supply.”

In metros that have seen demand fall in the past, including before the year 2000, the dQ/dP ratio does not reflect their elasticity of supply. Because housing is an extremely durable good, legacy housing in metros whose demand has fallen can exceed current demand and, as a result, they may have little impetus to build new housing. Inasmuch as demand is rebounding in such metros, it drives up prices well before an impetus to build new housing emerges, as in Pittsburgh and Buffalo. Metros where legacy housing supply exceeds current demand are identified as having little change in housing supply (<10%), and a housing price level that – even if it has risen over the period – remains low relative to other metros.

Note that the method does not capture ease-to-build in the negative dP-positive dQ quadrant, because in this quadrant it is not reasonable to assume that demand for housing is increasing. No metro areas fall into this category (except some with legacy housing in excess of current demand), so this is a moot point with respect to metros, but it is important with respect to the ZIP Code-level analysis.

---

Data and Methodology for the ZIP Code-level Analysis:

Data:

Data sources are similar to those used for the metro-level analysis, with one exception: data on the change in the number of housing units are drawn from comparing the 2000 Census and the 2010-2015 5-year ACS, not the 2015 1-year ACS, because the latter does not provide information at the ZIP Code level.

Method:

The toughest-to-build ZIP Code areas are ranked similarly to the toughest-to-build metros, except that the angle between each ZIP Code and the dP axis is not taken at the origin, but at the point (dP,dQ) = (0,-20%). ZIP Code areas with less than a 10% change in quantity and median home values below 75% of the within-metro median (across all ZIP Code areas within the metro) are labeled as having legacy housing supply that exceeds current demand. The ZIP Code-level analysis was not conducted in metro areas whose legacy housing supply was deemed in excess of their current demand.

ZIP Code areas in which the change in housing units is more negative than -10% are omitted because they are likely to reflect changes in ZIP Code area definitions. Such ZIP Codes amount to about 3.84% of the total. ZIP Code areas in which the change in housing units is between -10% and 0% are modified as having a dQ value of zero. Such ZIP Code areas may have experienced events such as tear downs, non-residential re-purposing, or the combination of multiple smaller units into fewer large ones. The modification prevents the extent of diminished housing from dominating the toughness-to-build ranking. However, non-negative net changes in housing unit counts may reflect some degree of housing unit reduction.

Rationale:

The rationale is similar to that of the metro-level analysis, but the modification whereby the toughest-to-build ZIP Code areas are ranked using the angle between the metro and the dP axis at the point (dP,dQ) = (0,-20%) instead of the origin is important. Without the modification, the ZIP Code area with the greatest price increase among those with non-positive dQ always ranks as the toughest to build. Measuring the angle at the lowered point allows ZIP Code areas with slightly positive dQ but much higher dP to emerge as tougher to build than ZIP Code areas with non-positive dQ and much smaller dP.

The choice of (dP,dQ) = (0,-20%) is a matter of discretion. It reflects an implicit weighting of the importance of dP versus dQ in qualifying toughness to build. Choosing a point with more negative dQ would amount to placing more weight on dP instead of dQ. Furthermore, fixing the point (dP,dQ) = (0,-X%) for any X results in more weight being given to dQ in metro areas that have seen greater overall price increases, like San Francisco, and more weight to dP in metros where prices have remained more stable.

As noted above (see metro-level analysis rationale), the method does not capture ease-to-build in the negative dP – positive dQ quadrant, because in this quadrant it is not reasonable to assume that demand for housing is increasing. For this reason, only the top 10 ZIP Code areas in terms of the toughness-to-build measure in each metro are ranked, and only the top 100 are ranked nationally. In metro areas with fewer than 20 ZIP Code areas, only those with toughness-to-build measures above the metro median were included.