The Urban Wage Premium in Imperfect Labor Markets

Notes: IEB and BHP, 1985–2010. Estimates show the second-step regression (Equation 5). The dependent variable is the estimated wage elasticity of the labor supply to the firm obtained from the first-step separation equation (Equation 4), which we model as a Cox model with a worker–region-specific baseline hazard. Further region controls are the shares of low-skilled and high-skilled workers, the log employment share of the largest two-digit industry, and the log Herfindahl index of employment at the industry level where all second-step regressors are centered around their means. In the Cox model, worker controls consist of real experience (linearly and squared), as well as groups of dummies for education, one-digit occupation, and non-German nationality. Employer controls are the shares of part-time, high-skilled, low-skilled, female, and non-German workers among the plant’s workforce, as well as groups of dummies for plant size and two-digit industry. We also add time dummies. In Model III, we further include the plant wage effect from Card, Heining, and Kline (2015) interacted with its reference period. In Model IV, the wage regressor is the deviation of the log wage from the plant wage effect from Card, Heining, and Kline (2015). Robust standard errors are given in parentheses.

Notes: IEB and BHP, 1985–2010. Coefficients from second-step regressions as in Equations 7 and 8. The dependent variable is the local wage level obtained from the first-step wage regression (Equation 6). The labor supply elasticity is estimated using Model II from Table 2. The F-statistic of the first-stage regression for Panel C is 23.38. Further region controls are the shares of low-skilled and high-skilled workers, the log employment share of the largest two-digit industry, the log Herfindahl index of employment at industry level, and the unemployment rate, where all second-step regressors are centered around their means. In the first-step wage equation, worker controls consist of real experience (linearly and squared), as well as groups of dummies for education, tenure, one-digit occupation, and non-German nationality. Employer controls are the shares of part-time, high-skilled, low-skilled, female, and non-German workers among the plant’s workforce, as well as groups of dummies for plant size and two-digit industry. We also add year dummies. Robust standard errors are given in parentheses.

Notes: IEB and BHP, 1985–2010. Coefficients from second-step regressions as in Equations 7 and 8. The dependent variable is the local wage level obtained from the first-step wage regression (Equation 6) in first differences. The labor supply elasticity is estimated using Model II from Table 2. The F-statistic of the first-stage regression for Panel C is 23.38. Further region controls are the shares of low-skilled and high-skilled workers, the log employment share of the largest two-digit industry, the log Herfindahl index of employment at industry level, and the unemployment rate, where all second-step regressors are centered around their means. In the first-step wage equation, worker controls consist of real experience (linearly and squared), as well as groups of dummies for education, tenure, one-digit occupation, and non-German nationality. Employer controls are the shares of part-time, high-skilled, low-skilled, female, and non-German workers among the plant’s workforce, as well as groups of dummies for plant size and two-digit industry. We also add year dummies. Robust standard errors are given in parentheses.

Notes: IEB and BHP, 1985–2010. Coefficients from second-step regressions as in Equations 7 and 8. The dependent variables are the region-specific coefficients of real experience and its square (times 100), respectively, obtained from a first-step wage regression akin to Equation 6 including worker–region fixed effects. The labor supply elasticity is estimated using Model II from Table 2. The F-statistic of the first-stage regression for Panel C is 23.38. Further region controls are the shares of low-skilled and high-skilled workers, the log employment share of the largest two-digit industry, the log Herfindahl index of employment at industry level, and the unemployment rate, where all second-step regressors are centered around their means. In the first-step wage equation, worker controls consist of real experience (linearly and squared), as well as groups of dummies for education, tenure, one-digit occupation, and non-German nationality. Employer controls are the shares of part-time, high-skilled, low-skilled, female, and non-German workers among the plant’s workforce, as well as groups of dummies for plant size and two-digit industry. We also add year dummies. Robust standard errors are given in parentheses.

Notes: IEB and BHP, 1985–2010. The first column shows the coefficient of the agglomeration measure in the second-step regression (Equation 5) where the first-step separation equation includes worker controls, employer controls, and a worker–region-specific baseline hazard—akin to Model II in Table 2. The second column gives the estimated urban wage premium when not conditioning on the labor supply elasticity as in Equation 7, where the first-step wage equation includes worker controls and fixed effects, as well as employer controls—akin to Model IV in Table 3. The last two columns present the corresponding estimates when conditioning on the labor supply elasticity as in Equation 8, where we instrument the elasticity as in Panel C of Table 3. In the final row, we use the share of job exits into nonemployment as instrumental variable; the F-statistic of the first-stage regression underlying this row is 22.77. Robust standard errors are given in parentheses.