Valuing Non-Wage Amenities: Survey and Experimental Evidence from Army Enlistments

• In Event: Beyond Wages: Worker Choice, Autonomy, and Wellbeing in Flexible and Nonstandard Work

Abstract

Survey-based discrete choice experiments are widely used to estimate willingness to pay (WTP) for non-wage amenities. Recent innovations in elicitation methods, such as Bayesian Adaptive Choice Experiments (BACE), can generate precise, individualized WTP estimates, making possible a more stringent field test of survey-based discrete choice methods against person-level realized choices. In this paper, we estimate WTP across two key amenities in the U.S. Army: (i) the compensation recruits require to accept longer initial service obligations, and (ii) recruits’ willingness to pay for an assignment location of their choosing. We implement two complementary approaches. First, we conducted a randomized controlled trial (RCT) from March 2024 to June 2025 in which we randomized enlistment bonuses across these amenities, allowing us to estimate WTP for location of choice and for extending recruits’ contractually obligated service from 3 years to 4, 5, and 6 years. Using information from 12,201 enlistment contract choices, we estimate an average WTP for assignment location of choice of $4,900 and a required bonus of $1,872, $14,740, and $16,223 to select a four, five, or six year contract, respectively, relative to a three year initial contract. Second, we administered a BACE survey from January through June 2025 to estimate the same parameters. Of the 1,731 applicants who completed the survey, 729 (42 percent) went on to enlist, allowing us to compare individual WTP estimates against observed choices. In the survey we estimate an average WTP for assignment location of choice of $5,909 and a required bonus of $889, $3,739, and $7,469 to select a four, five, or six year contract, respectively, relative to a three year initial contract. While the magnitude of the WTP estimates for location and a four-year contract length are similar to the RCT, the survey estimates for five and six-year term lengths are substantially smaller than the RCT estimates. Furthermore, we find that though our parameter estimates are strongly correlated with observed choices, we correctly predict the term length for those who sign contracts in only 39 percent of cases. To determine if prediction errors are inherent in the survey design and estimation or better explained by the environmental and behavioral factors influencing the observed choice, we implemented a set of survey questions that more closely mirror the decision made by applicants when signing an enlistment contract. In on-going work, we find that our ability to accurately predict contract length (59 percent) and station of choice selection (64 percent) is substantially better inside the survey, suggesting that the environment in which applicants make their choice may explain some of the deviations of observed choices from our survey-based predictions. We have recently implemented various information treatments in the survey to shed further light on the information environment and how applicant decisions vary with choice complexity, though we do not yet have enough responses to provide estimates along these dimensions. Together, these analyses provide a field test of survey-based valuation methods and suggest that targeted changes to hiring incentive structures could yield substantial cost savings in our setting.

Author

• Michael Thomas Baker, United States Military Academy
  Presenting Author