Concentration of High-Tech Workers Has Benefits — and Costs
Scientists and engineers tend to work more productively in high-tech-intensive cities. With more opportunities to network with their peers, they appear to benefit from knowledge spillovers, and their firms benefit from a larger labor pool.
Productivity gains — defined as the number of patents produced in a year — increase steadily when firms in cities with smaller concentrations of research and development activity are compared to those in localities with larger concentrations, but production costs also increase. Gains outpace costs at low- to mid-level R&D concentrations, but the picture changes in urban areas with the highest agglomerations, such as Boston and San Francisco, where the difference between productivity gains and cost increases is close to zero.
Productivity increases at a steady rate as clusters of scientists and engineers grow larger, but associated R&D costs rise more rapidly at higher levels of agglomeration.
In Place-Based Productivity and Costs in Science (NBER Working Paper 30416), Jonathan Gruber, Simon Johnson, and Enrico Moretti estimate R&D costs for 133 metro areas and assess how productivity varies with the density of scientists. They build on Moretti’s earlier work to show that researchers located in an area with 10 percent more scientists in their field produce 0.7 percent more patents per year.
Productivity increases at a steady rate as clusters of R&D workers become more concentrated; however, R&D costs rise more rapidly at higher levels of agglomeration. This means that increased productivity delivers diminishing net returns to the firms carrying out this research.
The researchers find that, overall, R&D spending in an area with 10 percent more scientists is 0.1 percent more costly — well below the productivity gains delivered by agglomeration. But while the productivity gains from adding more scientists do not vary much by area, the costs of R&D do vary. In particular, in the cities with the highest concentration of scientists, each additional 10 percent increase in the number of scientists increases costs by 0.53 percent. This estimate is somewhat less than the estimated increase in productivity in these cities, but the researchers cannot reject the possibility that the two effects are of the same magnitude, that is, that cost increases are equal to the productivity gains.
The most expensive US metro areas in which to conduct R&D are San Jose-San Francisco-Oakland; Honolulu; and Boston-Worcester-Manchester. The least expensive are Dayton-Springfield-Greenville; Milwaukee-Racine-Waukesha; and Grand Rapids-Muskegon-Holland. The cost of doing business is 38 percent higher at the top than at the bottom, and the difference in the cost of housing is even greater, ranging between 49 and 99 percent depending on the location.
R&D has become increasingly concentrated. As of 2009, the top 10 metro areas for computer science accounted for 70 percent of all US inventors, for semiconductors 79 percent, and for biology and chemistry 59 percent.
Congress is considering legislation that could provide incentives for R&D activity in new tech hubs instead of what the study calls “existing coastal superstar cities.” The researchers identify, but do not quantify, three potential benefits associated with greater dispersion of technology-intensive R&D jobs: bringing new jobs to depressed regions, reducing the national economy’s vulnerability to geographic shocks such as hurricanes and earthquakes, and generating broader political support for science spending by increasing the voter base it would benefit.