This content was first published on BuildZoom.com on December 17, 2018.

• After falling or remaining flat for three decades, real construction costs have increased sharply since the mid-2000s. The rise was driven by both material and labor costs.

• Construction costs are lowest in the South, and are highest in the expensive coastal cities and in remote locations like Honolulu and Anchorage. Material costs vary relatively little across cities, but labor costs differ substantially and account for the bulk of geographic variation in overall cost.

• Over the last decade, the mix of construction work in some of the expensive coastal cities has shifted away from new construction and towards renovation, which tends to be more labor-intensive. Alongside faster overall wage growth, the shift has resulted in labor cost appreciation in these cities outpacing that of material costs. In contrast, labor and material cost growth in the South has been more balanced.

• Rising labor costs can fuel a vicious cycle whereby costly construction dampens the supply of housing and exerts even greater upward pressure on labor costs (in several distinct ways).

The housing market appears to be reaching a cyclical peak, with falling home sales suggesting that already-slowed price appreciation may soon turn negative. Yet even now, residential construction still hasn’t recovered to its average historical levels. Rob Dietz, Chief Economist at the National Association of Home Builders, summarizes the headwinds confronting homebuilders as “the five L’s” of labor, lots, lending, lumber, and local regulations. Although lots and local regulations pose the most fundamental challenges to new residential construction, and as lending is finally sorting itself out (the way it tends to ironically at just the wrong time), this study is focused on direct construction costs, which comprise the two remaining L’s of labor and lumber.

To what extent have construction costs risen in recent years, and how do they compare to historical trends? How have material costs fared relative to labor costs, and how has this played out in different parts of the country? This study uses publicly available data as well as construction cost data published by R.S. Means and data from BuildZoom’s national building permit database to sketch out some answers.

The Long View

Construction costs have risen substantially since 2004, to the tune of 23.6 percent. Historically, real construction costs increased by 44.7 percent from 1950 to 1975, as wages in the industry improved, and then fell by 13.4 percent from 1975 to 2004 as inflation rose faster than both labor and material costs.[1] The rise since 2004 was primarily driven by material costs in the first few years, coinciding with the run-up to the 2000s housing boom, but more recently it has also been driven by labor costs.

Where is construction most and least expensive today?

The following chart lists the ten most expensive cities in the U.S. for construction.[2] Five of the top ten, New York, San Francisco, Boston, Philadelphia, and Los Angeles—or six, if we include Sacramento—are expensive coastal cities. Chicago, also on the list, tends to have construction labor that is more unionized than elsewhere. The high cost of construction in Honolulu and Anchorage, on the other hand, probably owes to their remote locations, and in the case of the latter also its climate.[3] Finally, Rockford, Illinois, made the list because it is essentially an offshoot of the Chicago metro area, despite being just beyond its official bounds.

In contrast, the ten least expensive cities in the U.S. for construction are all located in the South. The list includes cities in large metros such as Austin, Jacksonville, and Miami, but it consists mostly of places that are much smaller.

Both labor and material costs are higher in the cities with the most expensive construction costs, but as we will see in a moment, differences in construction costs across the U.S. are driven primarily by the cost of labor, not materials.

Labor costs vary greatly across cities; material costs do not

The cost of construction labor differs substantially across U.S. cities. At the same time, the relative ranking of cities in terms of their construction costs appears to be fairly stable over time. In New York, labor has ranged between 64.6 and 78.2 percent more expensive than the 30-city average during the 10 year period observed.[4] At the other extreme, San Antonio’s construction labor costs have hovered between 36.3 and 34.2 percent below the 30-city average.

In contrast, material costs hardly differ across U.S. cities. The next chart plots material costs against the same scale used in the labor cost chart in order to highlight the smaller magnitude of the differences across cities.

The differences in material costs are so much smaller than the differences in labor costs because materials are a tradable good, whereas labor is not. Goods that can be traded across locations tend to converge towards a common price everywhere. Of course, discrepancies can persist inasmuch as prices are driven by the cost of transporting the goods, but otherwise, they tend to be similar everywhere. If they weren’t, whoever was selling the goods at a low-price location would be compelled to sell some of their inventory in a higher-priced location, raising the cost in the former and reducing it in the latter until the prices net of transport costs equalized between the places. Construction materials, despite their bulk, are no match for today’s advanced logistics, and they fall squarely in the category of tradable goods. The wages of construction workers, on the other hand, vary by location and are influenced by the cost of living and the different labor market conditions in each city. Despite the relative ease of geographic mobility in the US, most people remain in their state of birth–especially those without a college degree. Moreover, when workers migrate their compensation tends to shift as well.

Labor cost appreciation has exceeded that of materials in the expensive coastal cities

More expensive metro areas have seen labor costs appreciate substantially more than material costs. The following chart plots the ratio of (central cities’) labor and material cost appreciation over the last decade against metropolitan housing price levels. When the ratio is greater than one it means labor appreciated more than materials and vice versa. The San Francisco Bay Area leads the pack with labor costs that appreciated 32.8 percent faster than its material costs, followed by the Los Angeles and New York metro areas. At the other extreme, metro areas in the South and in the Rust Belt saw more balanced labor and material cost appreciation. The greater appreciation of labor costs in the expensive coastal cities is most likely related to generally tighter labor markets and faster wage growth in these cities.

Renovation is more labor-intensive than new construction, resulting in greater pressure on labor costs in growth-restricted metros

Renovation work–defined broadly to include maintenance–tends to be more labor-intensive than new construction. This is because working in existing conditions tends to be slower and less efficient than building from scratch, and it often also uncovers surprises that need to be remedied to pass inspection. Thus, if we could run a (purely theoretical) experiment in which two identical markets shifted so that one had a greater share of renovation work than the other, we would expect the renovation-heavy market to experience greater demand for labor and reduced demand for materials relative to the other market. The result would be a higher ratio of labor to material cost appreciation in the renovation-heavy market.

The pace of new construction and that of existing property renovation have both risen in recent years.[5] BuildZoom’s national building permit database allows us to observe the mix of new construction and renovation work in different places and see how it has changed over time.[6] An interesting finding emerges when we plot the ratio of labor to material cost appreciation against the change in the share of building permits corresponding to new construction. Metro areas which shifted towards renovation work have seen labor costs increase more quickly than material costs, whereas metros which shifted toward new construction have seen more balanced labor and material cost growth.[7]

A Vicious Cycle?

Inasmuch as it means that those working in the construction trades are better off, construction wage growth in the expensive coastal cities is a welcome development, which ought to be celebrated.

However, construction wage growth could also be part of a dangerous, self-reinforcing cycle. Rising construction costs tend to reduce the supply of new housing because they shift marginally-worthwhile residential development into the red.[8] This helps raise housing costs, which in turn feeds back into construction wage growth for at least three reasons:

• When housing prices rise, workers require greater compensation to subsist.

• When housing prices rise and the supply of homes is smaller than it would be otherwise, it is harder for people to find suitable homes for sale (and to afford them). This creates a greater incentive for renovation, which exerts upward pressure on labor costs as suggested earlier.

• When housing prices rise, the pool of potential home buyers becomes more financially select, which likely corresponds to higher expectations with respect to homes’ level of finish.[9] Once again, this means that more renovation is bound to take place (including home-flipping), exerting even more upward pressure on labor costs.

And, higher labor costs reduce the supply of new housing. Repeat ad infinitum.

It seems likely that this vicious cycle is real in America’s expensive coastal cities. Although rising construction costs may not be the single most important factor responsible for curbing the supply of housing in these cities—restrictive local land use policy almost certainly holds that title—it is still worth paying attention to them. If local land use restrictions on the provision of new housing were ever relaxed, the cost of construction and particularly the cost of labor would quickly come into play as binding constraints on housing supply. From a YIMBY’s perspective, it is tempting to treat rising construction costs as a secondary issue as long as restrictive land use policies continue to plague American cities, but that could end up being a rather costly mistake.

Notes

1. The data associated with each year typically reflect conditions in the prior year, e.g. R.S. Means’ estimates of 2019 construction costs were published in late 2018. This means that the 2004 figure actually reflects conditions in 2003.

2. The list is limited to metro areas’ central cities. Given the mobility of construction services within a metro area, the central cities’ figures ought to be roughly representative of the entire metro. The underlying list of cities for which R.S. Means provides historical city cost data consists of 200 U.S. cities.

3. The cost of construction in Anchorage used to be in a league of its own. However this distinction has leveled off, as construction costs in Anchorage have fallen from 63.2 percent above the 30-city average in 1955 to just 24.7 percent above it in 2000, and 14.2 percent above it in 2018.

4. R.S. Means reports installation (labor) and material costs in relation to the average cost among a selection of 30 cities nationwide dating back to 2008. See Data and Methodology section for the list of 30 cities.

5. Residential renovation is intimately related to the volume of home sales.

6. The spatial and temporal coverage of BuildZoom’s building permit data is at the industry forefront, however, it is not comprehensive. The data do not span every jurisdiction within each metro area, however for 29 of the 30 metros presented there is substantial coverage within each metro, including the central city as well as some peripheral suburbs, the majority of which are observed consistently over the 2008-2017 period. The 30th metro area, Memphis, TN, has been omitted from the chart.

7. Regression results indicate that this result persists and is statistically significant at the 99 percent confidence level even when controlling for differences in the levels and changes of metropolitan housing prices (as well as the share of building permits corresponding to new construction, as opposed to the change thereof), i.e. this result is distinct from the previous one.

8. Increasing construction costs can prevent projects from being worthwhile even if they are small relative to the value of the finished real estate. See penultimate section here.

9. When house prices rise, affordability tends to worsen, which renders the pool of viable home buyers more financially select. Whether lending standards are tightening or being relaxed as housing prices rise can have a separate effect on the affluence of the viable home buyer pool.

Data and Methodology

1. R.S. Means:

   • R.S. Means is a construction cost estimating database whose operation dates back to 1940 (it is currently a brand of Gordian). It is routinely used by urban and housing economists as the standard for observing construction costs in North America, e.g. by Harvard’s Ed Glaeser and Wharton’s Joe Gyourko, here.

   • The historical cost indices used in this study measure input prices paid by builders for a fixed market basket of labor and materials used in the construction of a number of building types which the producers of R.S. Means deem representative. The quantities of labor and materials associated with each building type have been “…combined to form a composite building representing as closely as possible the actual usage of materials, labor, and equipment in the North American building construction industry.” Additional information, including the list of representative building types, is available here. Whether the selection of buildings and their specifications has remained constant over time, or has been updated in some way that reflects the changing nature of North American construction is not clear, however the reputation of R.S. Means suggests that these issues have been appropriately addressed, and this study takes the historical cost indices at face value. The building types span different types of commercial construction, include apartment buildings which fall under “multifamily”–a commercial construction category. They do not include single-family homes, but this study makes such homes’ cost of construction is sufficiently correlated over time and geography with that of commercial construction so as to make the historical cost indices useful in the residential context as well. For a variety of alternative sources of construction cost data, see Ed Zarenski’s blog post entitled “Construction Inflation Index Tables”.

   • The 30-city average reported by R.S. Means includes the following cities: Atlanta, GA Baltimore, MD Boston, MA Buffalo, NY Chicago, IL Cincinnati, OH Cleveland, OH Columbus, OH Dallas, TX Denver, CO Detroit, MI Houston, TX Indianapolis, IN Kansas City, MO Los Angeles, CA, Memphis, TN Milwaukee, WI Minneapolis, MN Nashville, TN New Orleans, LA New York, NY Philadelphia, PA Phoenix, AZ Pittsburgh, PA St. Louis, MO San Antonio, TX San Diego, CA San Francisco, CA Seattle, WA Washington, DC.

   • R.S. Means reports three types of data that were used in this study:

     • City-level historical construction cost indices, available in 5-year intervals from 1950 to 1995 and annually from 1998 to 2018.

     • City-level breakdowns of total construction costs into materials and labor, with the latter referred to as “installation”, as well as a weighted average of the two, from 2008 to 2018 (the second data set also breaks down construction cost into several Masterformat categories, such as concrete, masonry, openings, finishes and so forth, which were not used in this study). The broken down costs of material and labor in each city are reported in relation to the contemporaneous cost of that element for the 30-city average, e.g. the material and labor costs in New York City as of 2018 are reported as 100.6 and 178.2, meaning 0.6 and 78.2 percent above the average material and labor costs in the list of 30 cities, respectively.

     • National labor cost rates for the period 2007 to 2018 for open shop labor, repair and remodeling labor, residential labor, and standard union labor.

   • The following method was used to obtain the levels of material and labor cost in each city each year:

     • The equation of interest is: (i) {{t_c = a_c l_c + (1-a_c)m_c}}, where {{t_c}} is the total construction cost index for city {{c}} (and is stated in terms of some base level, e.g. such that the 30-city average as of 2008 equaling 100), {{l_c}} and {{m_c}} are the labor and material cost indices (again, stated in terms of some base levels, e.g. such that the 30-city average labor and material indices as of 2008 equal 100), and {{a_c}} is the weight of the labor cost index in determining the total cost index. In this equation, only {{t_c}} is observed, using R.S. Means historical city-level construction cost indices.

• The equation observed is: (ii) {{ t_c / \bar{t} = b_c(l_c / \bar{l}) + (1-b_c)(m_c/ \bar{m}) }}, where variables with a bar denote the 30-city average, and {{b_c}} is a weight. The ratios {{ l_c / \bar{l} }}, {{ m_c / \bar{m} }}, and {{ t_c / \bar{t} }} are the relative labor, materials and weighted (total) cost indices reported by R.S. Means in the broken down, city-specific dataset. Because all ratios are observed, the weight {{b_c}} can be obtained algebraically.

• Equation (i) can be expressed in terms of the observed ratios, as follows: (i’) {{ t_c = a_c \bar{l}(l_c / \bar{l}) + (1-a_c) \bar{m} (m_c / \bar{m}) }}, which implies that (iii) {{ \bar{t}b_c = a_c\bar{l} }}, and (iv) {{ \bar{t}(1-b_c) = (1-a_c)\bar{m} }}.

• Using the simple average R.S. Means national labor cost rates in lieu of {{ \bar{l} }}, the value of {{ a_c }} can be obtained from (iii), and then the value of {{ \bar{m} }} can be obtained from (iv), after which the value of {{ m_c }} can be obtained from the observed ratio {{ m_c / \bar{m} }}. Note that equation (iv) is slightly different for each city, leading to 30 slightly different derived values of {{ \bar{m} }} (denoted {{ \hat{\bar{m} } }}). In practice, the different values of {{ \bar{m} }} are remarkably similar across the 30 cities each year. None of the charts included in the study report values of {{ \bar{m} }}, but if they were, a simple average of {{ \hat{\bar{m} } }} over {{ c }} would suffice.

• Over the time period observed, 2007-2018, the 30-city average labor share of construction cost, {{ (1/30)\sum_c a_c }}, remained stable, ranging from a minimum of 42.2 percent to a maximum of 43.8 percent. The city-specific labor shares, {{ a_c }}, ranged from a minimum of 40.1 percent (Pittsburgh, PA in 2009) to a maximum of 48.7 percent (Cleveland, OH in 2016).

• Federal Reserve: The linked charts in the section entitled “The Long View”, which help break down the trend of real construction costs into its labor, materials and inflation components relies on data from the Federal Reserve Economic Data (FRED) system. Specifically, construction wages are drawn from the “Average Hourly Earnings of Production and Nonsupervisory Employees: Construction” series (AHECONS), material costs are drawn from the “Producer Price Index by Commodity for Special Indexes: Construction Materials” series (WPUSI012011), and inflation is drawn from the “Consumer Price Index for All Urban Consumers: All Items” series (CPIAUCSL).

• FHFA, Zillow & Census: A crude measure of metropolitan housing price levels was determined as follows: 2017 median home values for each county within metro areas (CSAs where applicable; CBSAs elsewhere) were obtained from Zillow, as the simple average of monthly values. Earlier county-level median home values were obtained by adjusting the 2017 Zillow level using the county-level, FHFA housing price indices in Bogin et al, dating back to 1975. Each metro areas housing price level measure (no longer strictly a median) was obtained using a contemporaneous-population-weighted average across the counties within each metro areas current boundaries. Historical county populations were drawn from the U.S. Census.

• BuildZoom: BuildZoom’s national building permit database was used to estimate the share of building permits in the metro areas (CSA where applicable; CBSA elsewhere) corresponding to the 30 cities each year, which were classified as new construction. The new construction classification is produced by a proprietary BuildZoom algorithm which has been trained on cities’ own classifications thereof.