Friday, July 23, 2010

Matching Marine officers to MOS

A reader writes:
"Dear Al,

I was a grad student a few years ago in your econ 2010a segment.

I came across a market design problem recently that I thought you might find interesting.  My brother is a Marine officer and is approaching the point when he gets matched to his "Military Occupational Specialty"-his future job (infantry, engineering, logistics etc).  As in matching medical residents, upcoming officers provide a ranking of their preferences.  Then the Marine administration ranks all the upcoming candidates on the basis of their performance in basic training (their overall average grade in "military skills, leadership, and academics").

The matching process is as follows: in a class of 100 officers, the officer ranked #1 gets his first choice, the 33rd ranked officer gets his first choice, then the 67th ranked officer gets his first choice.  Then the #2 ranked guy gets his top choice of the remaining jobs, then 34th, 68th, 3rd, 35th, 69th etc.

The intended purpose is clearly to ensure that there is an even distribution of "quality" in all specialties (you don't want all the best officers doing infantry (highest prestige) and all the worst officers managing logistics or supply).  (Actually this seems like a good idea for doctors too.)  However, the mechanism is clearly not incentive compatible-the 32nd ranked guy tries to fail his last tests to drop in rank to 33...

More details are here:  I also believe that the Army and other branches use similar systems.

Anyway, perhaps you've already seen this. And perhaps the Marines are totally uninterested in changing a "tried and true" system. But it sounded like an interesting question on a socially important topic.



John said...

At West Point, they do it purely on class rank, w/ Cadets first picking branch, then post. However, they recently starting allowing cadets to jump the queue by agreeing to accept an additional service obligation (e.g., you serve another 2 years in exchange for getting your top pick).

dWj said...

Perhaps they would be willing to entertain the apparently minor tweak of ordering them 1, 63, 25, 86, using the golden ratio; not only does it make the outcome more homogeneous (if preferences are highly correlated and reported honestly), but in practice it would likely improve incentive compatibility a great deal, insofar as one is more likely to know that one is near the bottom of the top third and wishes to drop than that one is exactly 24th and wants to engineer a drop by exactly one position in class rank.