« Why Merit Pay for Teachers Doesn't Work... And Why We Shouldn't Stop Trying | Main | The "Commonsense" Bill and Its Nonsensical Implications »



Feed You can follow this conversation by subscribing to the comment feed for this post.

Really enjoyed your guest post! I took a closer look at the GAO report you referenced. The report describes the comparison of male and female managers' earnings; that comparison adjusted for age, hours worked beyond full time, race and ethnicity, veteran status, education, industry sector, citizenship, marital status and the presence of children in the household.

The inclusion of age in the model may introduce gender bias. While age is commonly used as a proxy for prior work experience, it’s flawed and does not accurately reflect prior experience. It fails to consider that women typically experience more absences from the labor force than men, due to childbirth and child rearing.

The inclusion of age in a compensation model can introduce an artificial gender bias, which gives the appearance of gender discrimination.

Combined with the issues you raised, one is led to question the relevance of the report. Much of the discussion on gender pay equity is based on flawed analysis or misinterpretation of the analysis results. The GAO report is just one more example of "bad math" leading to "bad policy".

Stephanie: Would not the factor measuring the "presence of children in the household" be a control or permit cross-correlation with your "age" argument? Knowing if there were children, one could test to see if that factor were either a positive or negative predictor of pay or if it had any effect on the age measure as a proxy for experience. The absence of comment about "children" is interesting, suggesting that it did not correlate with age or pay in any significant degree.

Perhaps the hypothesis that child-bearing reduces the work time sufficiently to justify a pay penalty (the "mommy track") could be easily proven or disproven here. Whether any such resulting negative differential (if it exists due to that factor) is legal, substantial, or desirable from either an economic or sociological standpoint is another question.

Regardless, Census data is poor justification for public policy decisions, but it's used for that all the time.

Stephanie & Jim -- Thank you for sharing your reactions to the blog. From reading your own writings on compensation issues, I greatly value your expertise and insights.

Stephanie -- Thanks for the generous feedback. Its an honor to hear from someone who has such a commanding knowledge of economic and litigation issues.

Jim -- I could see the value of such a special analysis of "cross-correlation". But such extra effort for the special analysis would only be valid if the foundation of the compensation analysis were valid, which it sounds like we agree -- it is not.


The presence (or absence) of children in the household certainly could be used to examine the "mommy track" issue you raised. But it doesn't really get at the artificial gender bias introduced by age. To see how age as a proxy for experience is problematic, consider the following example.

A 35-year old male employee and a 35-year old female employee have identical educational backgrounds. Both entered the labor force at age 22, right after college. The male employee has thirteen years of prior experience, while the female employee has eight years of prior experience because she left the labor force after giving birth and did not return until her child began elementary school.

Because of the five year difference in experience, the female employee earns $2,500 per year less than the male employee. And for sake of argument, let’s assume that we know with 100% certainty that the difference in earnings is attributable to a difference in experience and nothing else. Using age as a proxy for experience does not, and in fact cannot, account for the situation described above. It assumes that both individuals have thirteen years of experience. If we compare the compensation of the male and female in the above example and control for gender and age, the model will tell us that the $2,500 difference is attributable to gender. More specifically, we might infer that the $2,500 difference is attributable to gender discrimination. That’s all the model can tell us – we’ve made an assumption that age captures all the relevant information about prior experience, but it doesn’t.

My next Cafe post will look more closely at labor market intermittency and the gender wage differential. Thanks for the inspiration!

The comments to this entry are closed.