Monday, October 24, 2011

Tayfun Sonmez on matching officers to military branches

On Friday Tayfun Sonmez gave a stimulating presentation of his analysis of West Point's system for assigning graduating officers to military branches: Tayfun Sonmez and Tobias Switzer, Matching with (Branch-of-Choice) Contracts at United States Military Academy




In a second paper, he was more critical of the ROTC assignment system.

9 comments:

Tayfun Sonmez said...

I can certainly tell from the picture I was more critical of the ROTC mechanism :)

Al Roth said...

Now that you mention it, I can sort of see that too...:

John said...

I was quite glad to go through this process with the old system (2001).

When they introduced the new system, I remember hearing that it was "strictly voluntary" but cadets are/were fully aware that the number of slots is fixed and that someone getting getting their branch or post by agreeing to an ADSO was imposing an externality on others (and if that's not the word they would have used).

Also, from a practical standpoint, the new system removes one of the prime incentives for doing well at the Academy. I also don't like how it forces cadets to make a decision about long to serve before they have spent even one day in the real Army as an officer.

Tobias said...

Tayfun you do look pretty worked up...I like the passion though.

John, I understand how you feel but I analyzed multiple years of matching data for cadets-branches. Because the BrADSO contracts are only paid for the last 25% of the branch allocation virtually everyone in the top quartile received their #1 choice without paying an extra three years. Really, the top half of the distribution was practically immune from competition with lower-ranked BrADSO contracted cadets. There is still a lot of incentive to do well and get a good OLM score.

John said...

What about post? Getting your branch was pretty easy unless you wanted to do something like medical service corps (given that I think 85% of males had to go to a combat branch). I think everyone in my class that wanted armor, field artillery, infantry etc. got it. Post outcomes were very different. Your chance of getting Hawaii, Germany, Washington State (highly desired) vs. Korea, Fort Polk, Ft. Bliss (almost universally less desired) etc. were very OML-dependent.

John said...

One comment - I definitely over-stated my case when I said "removes" incentives for doing well - I should have said "weakens"---it's clear that there is still a incentive to do well and its monotonic until you get near the very bottom of the list, when you start getting close to the goat money :).

Tobias said...

So for your cohort and up until the first two years(ish) of OCSP the Army A1 gave roughly (1.1-1.2)*N number of branch allocations per year group. There were far more branch allocations than cadets. Catching on to the need to create market scarcity the number of branch allocations were reduced to N. So it was a lot easier to get one's desired branch back in the day.

To the post issue....I don't remember how that was handled but I don't believe it was treated as a zero-sum game. However the post-of-choice contracts were not offered until after the grad school and then the branch-of-choice incentives were offered. At first cadets were allowed to contract for only one incentive, then two. The number of post-of-choice contracts was always a much smaller number than the grad school and branch contracts. I would be curious to know if cadets were displaced during the post-of-choice matching.

John said...

Interesting - thanks for following up. I didn't know that the types of offers were sequenced, but ordering it branch, grad school and then post seems the most aligned with cadets taking an ADSO for the "right" reasons & promoting what's good for the Army.

I'm still not crazy about the ADSO-for-X plan, mostly because I'm skeptical that cadets that stage know their long-term preferences and my worry about incentives & selection effects, but I feel a bit better about it.

Tayfun Sonmez said...

Hi John,

Thanks for your perspective. I am not in a position to comment on whether it is adequate for the Army to "auction" the last 25 percent of slots at each branch but it is quite an innovative approach to mitigate Army's retention issues to say the least. Our job as market designer is making sure that, the mechanism used for this process is well-behaved and does not have any adverse impact. USMA mechanism in that sense is "close" to an optimal design although I cannot say the same for the ROTC mechanism. There is a serious risk that the ROTC mechanism might undermine Army's efforts in human capital accumulation (since cadets in dead zones are encourage top perform just below the median.)