Urban occupational structures as information networks: The effect on network density of increasing number of occupations

Abstract

Urban economies are composed of diverse activities, embodied in labor occupations, which depend on one another to produce goods and services. Yet little is known about how the nature and intensity of these interdependences change as cities increase in population size and economic complexity. Understanding the relationship between occupational interdependencies and the number of occupations defining an urban economy is relevant because interdependence within a networked system has implications for system resilience and for how easily can the structure of the network be modified. Here, we represent the interdependencies among occupations in a city as a non-spatial information network, where the strengths of interdependence between pairs of occupations determine the strengths of the links in the network. Using those quantified link strengths we calculate a single metric of interdependence-or connectedness-which is equivalent to the density of a city's weighted occupational network. We then examine urban systems in six industrialized countries, analyzing how the density of urban occupational networks changes with network size, measured as the number of unique occupations present in an urban workforce. We find that in all six countries, density, or economic interdependence, increases superlinearly with the number of distinct occupations. Because connections among occupations represent flows of information, we provide evidence that connectivity scales superlinearly with network size in information networks.

Fig 1. Network density versus network size.

Among cities in the six countries studied, the density of a city’s occupational network increases superlinearly with the network’s size, measured as the number of unique occupations within the city. The exponent of a power law function for each country is given as β. Note that, for comparability, network size has been normalized by maximum possible size.

Fig 2. Network density vs. standard deviation of zeta for U.S. metropolitan areas.

Increasing density (mean zeta) is correlated with increasing standard deviation of zeta driven by the appearance of rare and highly interdependent pairs of occupations.

Fig 3. Occupational network size vs. density for U.S. cities at different employment aggregation levels.

When occupational network size is compared to its density, the resulting scaling exponents differed little when 2013 U.S. employment is aggregated at the 6-digit, 5-digit, or 4-digit occupation code. The number of distinct occupations in each case are 812 (6-digit), 455 (5-digit), and 107 (4-digit). Note that, for comparability, network size has been normalized by maximum possible size.

Fig 4. Occupational network size vs. density for German cities at different spatial aggregation levels.

When the size of an occupational network is compared to its density, the resulting scaling exponents differed little when German employment is geographically aggregated into 258 LLMRs, 141 LMRs, or 96 SPRs. Note that, for comparability, network size has been normalized by maximum possible size.