Research for Estimation Bias

Syndicate content
The Effect of Enlistment Bonuses on First-Term Tenure Among Navy Enlistees
April 1, 2003
While enlistment bonuses (EBs) have traditionally been used to affect accession decisions, it seems likely that offering a recruit a bonus that is payable at the end of training could also reduce attrition. This study attempts to assess the relationship between the size of the enlistment bonus offered a recruit and the likelihood that the recruit attrites—holding all else constant. Following recent studies of enlistment incentives, we employed non-experimental data (data generated from the administration of the EB program) to explore this relationship. Using this type of data creates an empirical challenge because both the size of enlistment bonuses and attrition behavior are likely to be substantially affected by unobserved variables. While there are various empirical techniques to control for the effects of unobserved variables under specific circumstances, we find that these methods have only limited applicability to the current analysis—that is, they can only control for some of the effects of omitted variables. Among our empirical findings, we find limited evidence that enlistment bonuses reduce attrition. While we believe that these statistical results are not substantial enough to guide policy, we suggest that they are sufficiently compelling to justify the Navy pursuing experiments on the issue.
Read More | Download Report
The Time is Now: Proposals for Two Enlistment Incentive Experiments
April 1, 2003
When Congress initiated Enlistment Bonuses (EBs) in the early 1980s, it required the Department of Defense to conduct experiments on how these incentives affect recruiting. More recently, the only policy guidance for those who set enlistment incentives has come from analyses based on nonexperimental data. These analyses have employed widely recognized statistical estimators and, in many instances, have produced findings that seem both plausible and precise. Unfortunately, the statistical estimators used in these works have significant potential for bias when they are applied to nonexperimental data. This analysis describes several of the estimation problems that arise from using nonexperimental data to analyze the effects of enlistment incentives, discusses some of the econometric tools that researchers have used in their attempts to overcome these problems, and indicates the shortcomings of these tools. It also argues that the potential for estimation bias in these works is so severe that the Services should view the results from the recent analyses as inconclusive. Finally, the study proposes two experiments the Navy could undertake that would reveal important information about enlistment incentives, that would be low cost, and that would not hamper the Service’s ability to meet its accession goals.
Read More