Does the Navy discriminate against Hispanics?

The Village People made good music and distinctive music videos. I'm having trouble getting their sound out of my head as I read this new paper by Golan, Greene and Perloff . I'm reading that paper because I'll be at the NBER for the next 3 days attending this conference and I wanted to see what recent research the conference organizer (Jeffrey Perloff) has been up to. I apologize (this is what nerds do with their finite time).

Now, please recall that in 2009 that I published a prominent book on the U.S military . So, my opinions count!

Here is the abstract for this new Perloff paper:

"The Navy’s promotion-retention process involves two successive decisions: The Navy decides whether an individual is selected for promotion, and then, conditional on the Navy’s decision, the sailor decides whether to reenlist or leave the Navy. Rates of promotion and retention depend on individuals’ demographic and other characteristics, wars and economic conditions and factors that the Navy policy makers can control. Using estimates of these decision-making processes, we examine two important public policy questions: Do Navy promotion and retention rates differ across race and sex? Can the Navy alter its promotion and other policies to better retain sailors, or do war and civilian labor market conditions determine retention?"

Here is the cool part:
"To estimate the model, we use data on virtually all Navy enlisted personnel from January 1997 through May 2008."

The interesting piece to the econometrics here is that a Navy person's probability of remaining in the Navy is a function of whether the Navy promotes him (not kicked out) and whether he chooses to stay. A labor economist would say that this probability will be higher if the officer is happy and productive in the Navy and if his next best "outside opportunities" (i.e teaching at UCLA) are not so hot.

Controlling for a large number of observable characteristics such as aptitude scores on standardized tests, age, education, sex, Macro trends such as whether the attacks of 9/11/2001 had recently taken place, Perloff and co-authors show that blacks are less likely to be promoted and less likely to be retained than whites.

How large are these effects? To judge this they conduct the following thought experiment, "If the Average black male soldier had the average white male's observable characteristics, his probability of being promoted would be 35% while the average white male soldier with the average white male's observable characteristics is predicted to have a 37.6% chance of being promoted. Now, the average Hispanic given the average white's characteristics only has a 26.6% chance of being promoted. That's a huge difference in this "apples to apples" comparison.

Why are the authors "standardizing" by observable attributes. The ideal experiment here is to compare an identical soldier who happens to be black, Hispanic, white , male, female and test for differential outcomes. Statistical techniques allow you to do this based on "observables" such as education, age, AFQT test scores.

Is this evidence of discrimination?

"The annual probabilities of promotion and retention could differ across racial and sex for three reasons. First, demographic groups could be treated differently by the Navy in the sense that people with the same characteristics but who differ in terms of race or sex have different probabilities (that is, the coefficients on individuals’ characteristics are the same across demographic groups). Second, these groups could have different mixes of observed characteristics such as education and experience. Third, there could be differences in unobserved characteristics across the demographic groups.

Differences in coefficients (the first hypothesis) play roughly twice as large a role as the difference in observed characteristics (the second hypothesis) in explaining the overall difference in promotion probabilities between Whites and other races. The difference in coefficients is most pronounced for Hispanics. We cannot explicitly examine the third hypothesis. However, because our bivariate probit analysis includes an objective ability measure, the AFQT score, as well as a large number of other observed characteristics, it is relatively unlikely that racial and sex differences in promotion rates reflect unmeasured ability differences across demographic groups."

Add/view comments on this post.

------------------------------

The Christian Science Monitor has assembled a diverse group of the best economy-related bloggers out there. Our guest bloggers are not employed or directed by the Monitor and the views expressed are the bloggers' own, as is responsibility for the content of their blogs. To contact us about a blogger, click here. To add or view a comment on a guest blog, please go to the blogger's own site by clicking on the link above.