180 likes | 273 Views
Sexually Integrated Workplaces and Divorce: Another Form of On-the-Job Search. By Terra McKinnish Journal of Human Resources, 2007. Central Question. Are men and women who work in more sexually-integrated workplaces more likely to divorce? Theoretically:
E N D
Sexually Integrated Workplaces and Divorce: Another Form of On-the-Job Search By Terra McKinnish Journal of Human Resources, 2007
Central Question Are men and women who work in more sexually-integrated workplaces more likely to divorce? Theoretically: • Integrated workplaces might lower search costs for married individuals, increasing their likelihood of divorce. • But also lowers search costs for singles, which could lead to higher-quality matches and reduce the likelihood of divorce.
Previous Literature (what do we already know?) • Large increase in divorce after WWII • Rising labor market opportunities of women partially responsible • Divorce more common when sex ratios unbalanced (South and Lloyd 1995) • Two other studies of sex-mix at work and divorce: Aberg (2003) and South, Trent, and Shen (2001). Find a relationship but do not account for endogenous selection into industry/occupation.
How might increased contact with opposite sex affect divorce? Individual finds a more appealing spouse and divorces to marry that person. Contact leads to an affair that leads to divorce, even if no subsequent marriage. Changes individual’s perception of the outside alternatives. Also: “the random search process creates a meeting externality whereby one divorce (marginally) increases the remarriage probability of other divorcees.” –Chiappori and Weiss (2001)
Empirical Analysis: 1990 Census Data First, how does McKinnish measure sex-segregation? • Each working person is identified as working in one of 235 industries and 501 occupations. • For each industry-occupation cell, calculate the fraction of workers 18-55 who are female.
Empirical Analysis: 1990 Census Data Sample: * All ever-married, non-widowed individuals 18- 55 who report an industry & occupation. * Drop if too few people in their ind/occ cell. * 1,907,701 women and 1,853,243 men.
Empirical Analysis: 1990 Census Data OLS Regression Model: Dependent variable = 1 if currently divorced. Note subscripts, vector notation.
Empirical Analysis: 1990 Census Data Interpret OLS results:
Empirical Analysis: 1990 Census Data Statistical significance: A coefficient is statistically significant at α% if we can reject the null hypothesis at the α% level. Null hypothesis in this case: What does it mean to say that the coefficient is statistically significant at 5%? How do we determine if the coefficient is stat. sig.?
Empirical Analysis: 1990 Census Data Fixed Effects Analysis “Choice of occupation and industry is potentially endogenous.” What exactly is the problem? Solution: include industry and occupation-specific fixed-effects. How is this done? How does it solve the problem? This is the paper’s identification strategy (along with IV). The identification strategy is the method by which the causal effect is identified and estimated.
Empirical Analysis: 1990 Census Data Interpret Fixed Effects results:
Empirical Analysis: NLSY79 Data The NLSY79 is a panel data set. What does this mean? Advantages? Disadvantages?
Empirical Analysis: NLSY79 Data OLS regression model: Differences from Census model?
Empirical Analysis: NLSY79 Data Interpretation:
Conclusion People who work with more members of the opposite sex are more likely to be divorced. Effects seem to be larger for women. Discussion: * Do you believe the results? * What are the weaknesses of this study? * What are the implications? * How might you improve or extend the paper?