Statistical Discrimination Statistical Discrimination Discrimination in absence of

Statistical Discrimination • Statistical Discrimination: – Discrimination in absence of prejudice. – Employers use actual average labor market attachment differences by sex as a signal of what to expect from individual workers. – Causes gender gap even for women who never leave LF to raise kids.

Regression • Regression model to test for discrimination: – Multivariate regression: wage as dependent variable (on left hand side) with FEMALE as an exogenous (right hand side) variable. – With actual hourly wage as dependent variable, coefficient on FEMALE is average $ wage difference from being female, holding constant other relevant factors. • In table of results – See FEMALE coefficient as # other controls. – Statistical significance: effect we estimate with data is a true difference, not one arising just from our particular sample.

MBA Study by Montgomery and Powell • Unique data for study: – GMAT Registrant Survey – Longitudinal survey of 4285 GMAT test-completers. – Surveyed 3 times from 1991 to 1994. • Focus on test-completers helps to statistical problems results more reliable. • Authors improve even more by separating sample into two groups: – Those who completed MBA; – Those who did not complete MBA; – Use statistical correction for this selection.

Focus of Study • Focus on statistical discrimination: – Look at coefficient on FEMALE. • Note: model has lnwage as dependent variable so coefficient on FEMALE is %wage difference by sex. • See Table 11. 3 – – Very good list of control variables See two sets of results. See t-statistics (big is good). See difference in FEMALE coefficient: • In MBA sample, females do not earn less than males. • In test-taking sample (but no MBA), women do earn less than men

Results cont. – Conclusion: • Employers use MBA degree as a positive signal that helps to lessen the negative signal of being female. • Supports idea of statistical discrimination.