Separate and Unequal in the Labor Market: Human Capital and the Jim Crow Wage Gap Appendix: For Online Publication

A key contribution of this paper is a county-level dataset to study how differences in school quality across locations and races in the South ultimately affected labor market outcomes. Prior to the construction of this data, investigation of this question would have been limited to the use of state-level averages of school quality metrics. These state-level averages are suboptimal for two reasons. First, as discussed in Section 2 of the main text, the level of aggregation for school quality data matters in the modern literature (Betts, 2010), and the use of aggregate data flattens important variation in school quality within states. Second, because multiple measures of school quality are reported (term length, teacher salaries, class size, and so forth), and because states did not unanimously report any particular statistic, relying on state-level data would limit the scope of our study to uniformly available metrics or a smaller set of states that reported most of the same metrics. With county-level annual data on several domains of school quality we have the breadth and depth of information necessary to produce a standardized index of school quality for all ten states, as described in Section 3 of the main text. Table 1 quantifies the advances from utilizing county-level versus state-level data. First, we highlight the aggregation impact coming from state-level data. In Column 1, we regress the outcomes of interest (here limited to weekly and annual wages) on a cubic of countylevel measures of two statistics with close to universal coverage in the data: average teacher salary and teachers per pupil. We continue to include a cubic function of educational attain-