IT and Urban Polarization

Jan Eeckhout, Christoph Hedtrich and Roberto Pinheiro, "IT and Urban Polarization," Federal Reserve Bank Of Cleveland, September 2021, https://www.clevelandfed.org/en/newsroom-and-events/publications/working-papers/2021-working-papers/wp-2118-it-and-urban-polarization.aspx

"...Fact 2. Routine Cognitive Occupations Decline Faster in Expensive Cities. We now turn to the second result: high cost locations feature a decline in the share of workers in routine occupations, whose tasks can presumably be automated after the introduction of new technology. We use 1980 as the pre-technology period in order to construct the control variables and compare it to the occupational composition in the period 1990-2015. Our focus on such a long span of time is motivated by the fact that in the model, we compare steady-state predictions and ignore short-term dynamics. Furthermore, the national trend shows a decline in the share of routine cognitive jobs starting in the late 1980s. Table 3 presents the results of linear regressions of the change in the MSA’s share of routine cognitive occupations between 1990 and 2015. Specification (1) indicates that a one standard deviation increase in the local price index (an increase of 21.4 percent in the local price index) is associated with a 0.8 percentage point larger drop in the routine cognitive share over 1990-2015. Thus, the most expensive places have about a 4.5 percentage point larger drop in the routine cognitive share relative to the cheapest locations. This is one quarter lower than the average routine-cognitive share of 23 percent in 2015….”

"...A counterfactual exercise where we simulate a fall in IT prices by 65 percent - corresponding to a similar change in the data between 1990 and 2015 - explains both the fall in employment in routine cognitive jobs and the rise in non-routine cognitive jobs. Quantitatively, the exercise explains about 28 percent of the change in employment shares in cognitive occupations...."

"...the employment share of routine cognitive occupations falls substantially more in expensive locations, about 40 percent more relative to the average. Similarly, the wage gap between routine and non-routine cognitive jobs widens even more in expensive locations, about 20 percent more compared to the average. Overall the results indicate a strong role for IT in the displacement of routine cognitive employment and a rise in non-routine cognitive employment and the accompanying polarization of earnings across jobs and cities..."

Evidence"...Fact 1. Stronger IT Adoption in Expensive Cities. Figure 2 visualizes the positive correlation between local rental prices and the average IT budget per worker. Mere inspection shows that the magnitude of the change in IT spending as the rent index changes is sizable. Furthermore,Table 1 shows the results for MSA-level linear regression models of the log of the average IT budget per worker, adjusted for plant employment interacted with three-digit SIC industries, following Beaudry et al. (2010) and Doms and Lewis (2006). The regression results provide support for the hypothesis that IT expenditure per worker is increasing in the cost of housing. The elasticity is highly significant and its value barely changes under different regression specifications. The MSA’s rental price index in 1980 helps to explain the variation in IT budget per worker across MSAs, even after controlling for the presence of natural amenities, housing supply elasticity, and industry composition.9 In specification (1), a one standard deviation increase in the local price index (an increase of 21.4 percent in the 1980 local price index) is associated with an increase of $107.43 in the MSA’s average IT budget per worker. This magnitude corresponds to an increase of 3.67 percent in the average IT budget per worker. Specification (2) finds no statistically significant correlation between the MSA’s share of routine cognitive jobs in 1980 and the average IT budget per worker in 2015. Specification (3) finds a statistically significant correlation between the area’s ratio of college equivalents to non-college equivalents and the average IT budget per worker in 2015, constructed as suggested by Beaudry et al. (2010). However, as we include all controls presented in specifications (1)-(3) together in specification (4), the area’s ratio of college equivalents to non-college equivalents loses statistical significance. Differently, the impact of local rent prices shows just minor change in statistical significance between specifications (1) and (4). Finally, specification (5) controls for the MSA’s average degree of offshorability of local jobs in 1980 - using the task offshorability index presented by Autor and Dorn (2013). We find again that the impact of local housing prices is robust to the addition of the controls...."

Bottomline, "We show that differential IT investment across cities has been a key driver of job and wage polarization since the 1980s. Using a novel data set, we establish two stylized facts:IT investment is highest in firms in large and expensive cities, and the decline in routine cognitive occupations is most prevalent in large and expensive cities. To explain these facts, we propose a model mechanism where the substitution of routine workers by IT leads to higher IT adoption in large cities due to a higher cost of living and higher wages. We estimate the spatial equilibrium model to trace out the effects of IT on the labor market between 1990 and 2015. We find that the fall in IT prices explains 50 percent of the rising wage gap between routine and non-routine cognitive jobs. The decline in IT prices also accounts for 28 percent of the shift in employment away from routine cognitive towards non-routine cognitive jobs. Moreover, our estimates show that the impact of IT is uneven across space. Expensive locations have seen a stronger displacement of routine cognitive jobs and a larger widening of the wage gap between routine and non-routine cognitive jobs….Figure 2 visualizes the positive correlation between local rental prices and the average IT budget per worker. Mere inspection shows that the magnitude of the change in IT spending as the rent index changes is sizable….”