An Experimental Evaluation of Deferred Acceptance : Evidence from Over 100 Army Officer Labor Markets

We present evidence from a randomized trial of the impact of matching workers to jobs using the deferred acceptance (DA) algorithm. Our setting is the U.S. Army's annual many-to-one marketplace that matches 10,000 officers to units. Officers and jobs are partitioned into over 100 distinct markets, our unit of randomization. Matching with DA reduced officers' attrition in their first year in their new match by 16.7 percent, but we can rule out more than a 10 percent reduction in attrition by the end of their second year. Matching with DA had precise zero effects on performance evaluations and promotions. Although matching with DA increased truthful preference reporting by a statistically significant 10 percent, many officers matched by DA misreport their true preferences. We present new evidence suggesting that communication and coordination of preferences may limit the benefits of DA in matching markets where each side actively ranks the other