As if the Census Bureau had planned out their release schedule for maximum impact, the week David Card was awarded a the Nobel Prize in Economics, the Bureau published a new paper coauthored by Card, Jesse Rothstein, and Molses Yi, concerning migration and wage premiums. The paper uses two high quality US administrative data sets, the American Community Survey (ACS), and the Longitudinal Employer-Household Dynamics (LEHD). The authors use a variety of statistical methods to estimate increase or decrease in wages associated with a move, and specifically, what explains the “wage premium” associated with a move.
This paper exhibits some of the best practices of academic writing. Sentences are crisp and concise, the analysis and results are clearly explained, and the methods section is clear and approachable for a wide variety of audiences. I will let econometricians with greater knowledge discuss the methods employed, however, I will highlight the findings and discuss some of the implications and questions they raise from a practitioners perspective.
The first major takeaway is a probably the most plain, that industry premiums and location premiums are to a large extent separate. So for example, a banker’s wage in Waco TX, is likely to have a similar industry premium to a banker in New York, NY; but both will have different geographic premiums based on their location. This aligns with anecdotal evidence and even policies put in place by tech firms whose workers are choosing to move to lower cost cities.
The second takeaway is a surprising one. Card, Rothstein, and Yi find that although college educated individuals “sort” into high cost cities at a higher rate than those without a college education, this sorting doesn’t result in increased wages as a result of agglomeration and skills matching.
I think there are several reasons that this matching doesn’t result in increased wages. First, these “college education” industries likely have ‘national’ labor markets in which hiring can capture potential laborers from outside the Commuting Zone (CZ). This aligns with the first takeaway that shows that the industry premium and the geographic premium are separate. In essence, the additional pay an individual would receive as a result of their education and skills has already been “priced in” and being in a thicker labor market does not result in greater wages (or vice versa).
For local policy makers, this means that in some cases, trying to grow establish a clustered sector may not lead to greater returns, and that tax breaks or subsidies for companies could be a dicey proposition. Instead, this paper points to “economic gardening”, and building a diverse group of industries, as a more sustainable growth model. Furthermore, this might indicate the “placemaking” strategies which focus on building desirable communities might have a greater effect on migration than previously thought. As the authors note: “One potential explanation [of workers moving] is that large cities have even greater consumption amenity advantages than they do productivity advantages (Albouy 2011; Albouy, Cho, and Shappo 2021).”
The final takeaway is fairly well known “truth” but does bear repeating; that significantly higher housing costs in the high wage and large metro regions eats up the location premium to workers. The authors note that any increases in wages gained by workers moving from Waco to New York will be lost as a result of the increased cost from housing. In fact, it was found that moving to a high cost location resulted in a loss in the real income of a worker.
There are a number of ways to interpret this. One interpretation is that employers have increased bargaining power in setting wages which means that they are able to negotiate a discount in exchange for allowing workers access to thicker labor markets. However, that seems counter indicated by the fact that the paper finds that the effect of a move is relatively symmetrical. A possible caveat to this is that national labor markets allows employers (particularly national employers) to limit gains in wages that might ordinarily result in high wage areas. There is some evidence that gains in productivity have not translated into gains in wages, but it is likely more evidence is needed.
Another interpretation is that this loss in real income is driven by land use policy. In essence, migration between areas will happen regardless of housing production, but the result will be that local economies will grow slower as disposable income is shifted towards housing costs. Instead, the land use policy drives gentrification and housing price gains in high wage areas. This would mean that high wage areas need to change their land use policies to increase housing to drive economic growth. A future research question might examine if areas that with lower housing costs relative to similar metro areas have higher rates of job growth and economic growth.
This paper also raises some interesting future questions. First, I think it is important to highlight the LEHD data is from employers. So the data does not capture self employed contractors (i.e. Uber drivers, Door Dash workers, and other app workers). This would be an interesting experiment to see if there is a similar wage premium for app employees across localities or if app companies are able to “flatten” that premium.
Another questions I have as a result of this paper is related to resource rich areas. The authors observe, ” There are some resource-intensive CZ’s (on the Texas Gulf Coast and in the Permian Basin, for example) that have relatively high CZ effects but only average earnings levels.” This to me signals that these areas actually are functioning as some sort of ‘transitional hub’ with income not ‘sticking’ in an area. However, the authors also note that resource areas in South Dakota are high wage areas. I would like to see further research on what processes might be causing these differences.
The final questions is related to the size of an area (in population) and its relation to wages. The authors find, “The ACS analysis indicates that skill is only weakly related to CZ size, explaining only 17% of the size effect on earnings, while the LEHD indicates a much stronger skill-size gradient, explaining nearly two-thirds of the total.” The authors seem to think this relates to the fact that a larger CZ has more educated workers scattered among different industries, but I am unsure why this would not show up similar in the ACS Data. Perhaps the LEHD data is using the 4-digit NAICS code? I would be interested in understanding this more.
Card, Rothstein, and Yi have put forward a really interesting paper that takes a clear step forward in applying statistical methods to pull apart increases in wages by industry and geography. for practitioners, it provides some significant evidence that placemaking and economic gardening strong economic development strategies. A diverse economy that provides high quality amenities will attract and retain workers rather than luring specific clusters through governmental incentives.
Additionally, in high wage regions, land use policy is possibly providing a drag on economic growth, as movers to the region are willing to spend significantly more on housing (that is to say, high housing costs are not affecting their decision to move). These regions should see building housing (of all types) as a priority.