What local factors best predict where new work emerges?
Core argument: The initial local share of college-educated residents and log population density are the two most significant predictors of employment growth in "new work" across US commuting zones and decades, in an analysis spanning 1980 to 2010.
Our “new work” measure exploits the evolution of lists of job titles [from] US government agencies. Text analysis transforms job titles into multidimensional vectors: for each job title, we calculate the vector-distance from all titles existing in the previous period and consider the most similar title (minimum distance). If this value is above a threshold parameter, we classify the added job title as “new.” We document the importance of the initial local share of college-educated and log population density as the two most significant predictors of [employment growth in] “new work” across [Commuting Zones] CZs and decades. We analyze how exposure to global shocks and the presence of local factors correlate with the emergence of “new work” across commuting zones between 1980 and 2010. To investigate the effect of the local supply of human capital on the creation of “new work,” we use the historical presence and density of 4-year college institutions in the CZ. The importance of the initial local share of college-educated and log population density [are] the two most significant predictors of “new work” across CZs and decades. High manufacturing share in 1980 also strongly (but with a negative sign) predicts “new work.”




