Knowledge Shocks Diffusion and the Resilience of Regional Inequality

This blog post was written by Alexandra Lopez-Cermeno, PhD, Universidad Carlos III de Madrid

In the last decades, economic historians have shown that high-value added sectors locate in particular regions fostering growth at the expense of the de-industrialization of the rest of the economy. Part of the explanation is that comparative advantage comes from large population concentrations and all the perks from the traditional Marshallian hypothesis. This process enhanced long-term regional divergence and motivate the search for policies to reduce income inequality.

The case of Idaho, a notorious potato exporter, provides a good motivation to study the impact of new institutions to attract skilled workers to relatively poorer regions. The creation of a national nuclear research facility during the 1950s in this deserted area led to a presumably unexpected upswing in terms of population and income: in less than two decades Idaho`s population growth rate multiplied by five and made it to the top-100 biggest metropolitan areas in the US. The case of the United States is interesting on its own, but particularly because it is the paradigm of skills and human capital growth. Its academic institutions have not only turned the Human Capital Century into the American Century (Goldin and Katz [2009]), but have also driven the divergence of regional economic performance.

To continue the brainstorm raised by Fujita and Krugman [2003], a recent EHES working paper challenges the conventional wisdom that knowledge spillovers act locally by exploring the regional effect of knowledge shocks on growth through the relationship between the creation of new universities in the United States during the twentieth century and market size.

Figure 1: Distance-weighed total sum of GDPs by county, 1980. Author’s estimations. Click to enlarge

The effects of education on growth are directly visible on labor productivity through an increase in the quality of the workforce (see Caselli and Coleman [2002]), but there are also indirect effects: higher income generated by labor productivity raises physical capital investment and the capital to labor ratio; also, the quality of the workforce facilitates the diffusion of innovations and ideas. Thus, regions with higher levels of education are expected to grow faster.

At the same time, regional specialization determines the average level of human capital: while mining regions are associated with relatively lower effects of knowledge spillovers and have remained small, cities that grew around the textile industry were crowded with unskilled labor and only grew at the beginning of the century. In contrast, commercial towns that specialized in skill-intensive activities like accounting, advertising and law (Chicago, Boston, New York, …) were bigger and continued increasing over the same period. This way, in the 1930s the population in Idaho Falls was specialized in the production of agricultural products and their low wages corresponded to their skills level. After the establishment of the Nuclear Research Center, their production bundle diversified by including valuable knowledge intensive services, attracting scientific employees earning higher salaries and fostering the creation of new businesses that, eventually, raised overall living standards.

Figure 2. Foreign layer of Distance-weighted sum of GDPs by country, 1980. Author’s estimations. Click to enlarge

This paper provides a simplified method of exploring the geographical limits of a knowledge shock over the long run using a geographically decomposable distance-weighed sum of GDPs by county analogue to the concept of Market Potential developed by Harris (1954). This measure accounts not only for local GDP, but also for all potential trade with neighboring counties within a state, with the nation as a whole, and even with other nations; this is useful because it provides a detailed view of the extension of the market based on bilateral transport costs and the size of other markets for each county, and allows to assess the impact of the shock in each of the layers (local, neighbor, domestic, foreign and total).

Figure 3. Local component of Distance-weighted sum of GDPs by county, 1980. Author’s estimations. Click to enlarge

The experiment consists of finding a causal effect between the establishment of new academic institutions and the change on the distance-weighed sum of GDP by comparing a set of counties with and without new universities. I account for the fact that counties where new universities are established might have been already expected to grow by finding a control group that is synthetically similar to the treatment group in terms of market potential, urbanization and primary sector share.

Results show that the impact of the shock is relevant not only locally, but also in the neighbor counties’ layer, the domestic and even the foreign layer (ceteris paribus). In other words, after a new university is established in Fresno, not only local GDP increases, but also other counties in California eventually experience an upswing in their GDP significant at 1 per cent on average. Moreover, these results show that these shocks eventually impact distant counties, although the effect is statistically smaller than for state-neighboring counties. Additionally, the shock seems to make the domestic economy more competitive, creating a negative and significant relative decrease on the foreign layer by reducing the distance-weighted GDP sum of trade partners. However, it is essential to consider the assumption of ceteris paribus to be increasingly unrealistic for larger geographical layers.

Figure 4: Before and after knowledge shock effect on neighboring counties. Click to enlarge

The long run effect is equivalent: the closer in time the shock is, the higher its impact. The local impact seems immediate but lasts less than the domestic impact, in other words, a knowledge shock implies the acceleration of local GDP growth that expands to nearby regions, creating a ‘shock-wave’ effect that eventually gets to faraway counties. The local shock of a new university institution might generate immediate local growth; however, spillover effects may take some time to reach neighboring counties by increasing the demand of other services (Moretti [2004]), and reach distant counties even later. In practical terms, new researchers rapidly arrived to the Idaho Falls city as soon as the job positions opened, but the expansion of the city took several years. Similarly, the short span of the effect on the Foreign component might is explained by the many other variables that affect foreign GDP including its own economic policy and the response to the opening of new academic institutions, which might have an analogue negative effect for the US.

Figure 5. Long term knowledge shock effect on neighboring counties. Click to enlarge