Full Citation
Title: Discrimination in Hiring: Evidence from Retail Sales Online Appendix
Citation Type: Journal Article
Publication Year:
ISBN:
ISSN:
DOI:
NSFID:
PMCID:
PMID:
Abstract: In this section we discuss the sensitivity of our results in several dimensions. Productivity. In Section 4.2 we estimate our baseline productivity regression (11) via weighted-least squares. Tables 3 and 4 shows the resulting supermodularity statistics and confidence intervals from 10,000 bootstraps. As stated in the main text, mean productivity is supermodular, especially for white-Hispanic and black-Hispanic pairs, and productivity variance is log-submodular, especially for black-Hispanic pairs. Thus, the black-Hispanic pairs indicate screening discrimination (97%) while white-Hispanic indicate screening or complementarity (99.1%). Table B1 shows the regression coefficients underlying these results. As discussed in Section 4.2, Table B2 conducts several robustness checks and Figure B2 recalculates Figure 4 for (a) the first six months of workers' tenure, and (b) later months. Turnover. Section 4.3 considered turnover after 6 months, showing that x ⇥⇥ 0 < 0 for all three pairs, with white-Hispanic being significant. The underlying regression is shown in Table B3. Table B4(a) shows that the results are robust to measuring turnover at 3, 9 and 12 months. All the coefficients are negative except black-Hispanic at 6 months; only the white-Hispanic coefficients are significantly negative. Table B4(b) then includes the worker's average productivity, ¯ e i , as a control. As expected, this has a negative sign; e.g. with the "6 month" regression, the coefficient is 0.0609, meaning a one standard deviation increase in ¯ e i of 0.439 lowers turnover by 2.7% from a base of 36.1%. In the model, a worker is fired when their realized productivity falls below a threshold. Under screening discrimination and complementary production, same-race workers have higher productivity than cross-race workers, so controlling for SPH should increasê x ⇥⇥ towards zero. Under taste-based discrimination, same-race workers have lower productivity than cross-race workers, so controlling for SPH should lowerˆxlowerˆ lowerˆx ⇥⇥. Table B4(b) shows thatˆxthatˆ thatˆx ⇥⇥ tends to increase for white-Hispanic and black-Hispanic pairs.
User Submitted?: No
Authors: Benson, Alan; Board, Simon; Meyer-Ter-Vehn, Moritz
Periodical (Full):
Issue:
Volume:
Pages:
Countries: