An Experimental Evaluation of Deferred Acceptance: Evidence From Over 100 Army Officer Labor Markets
通过随机试验比较延迟接受算法与传统管理者主导匹配在美国陆军内部劳动力市场中的效果，发现延迟接受算法能减少行政负担、提高匹配质量，但对军官留任和绩效的总体影响有限，仅在管理者经验不足的市场中有显著改善。
Internal labor markets are increasingly important for matching workers to jobs within organizations. We present evidence from a randomized trial that compares matching workers to jobs using the deferred acceptance (DA) algorithm to the traditional manager‐directed matching process. Our setting is the U.S. Army's internal labor market, which matches over 14,000 officers to units annually. We find that DA reduces administrative burden and increases match quality as measured by reduced justified envy, increased truthful preference reporting, and officers' and units' preferences over their matches. The overall impact of DA on officer retention and performance in the two years after officers started their new jobs is limited by strategic preference coordination between officers and units. However, DA leads to significant improvements in officer retention and promotions in markets with inexperienced managers. Our findings suggest that cross‐market communication between agents in internal labor markets can attenuate the benefits of strategyproof matching algorithms.