Craig Copeland

A Behavioral Model for Predicting Employee Contributions to 401(k) Plans

Abstract Previous research on employee contribution behavior to 401(k) plans has often been limited by lack of adequate data. This is primarily because of the types of matching formulas utilized by sponsors. While these formulas are often complicated because of the desire of sponsors to provide sufficient incentives to non-highly compensated employees to contribute in order to comply with technical nondiscrimination testing, this complexity makes it virtually impossible to appropriately analyze the employee’s behavior if we are forced either to observe aggregate plan data or to use information on the plan contribution formulas provided by the participant. The purpose of this paper is to provide preliminary findings introducing new methodology to expand the usefulness of modeling these data as well as a better understanding of contribution behavior by 401(k) plan participants. We utilize a sequential response regression model to allow for the differing incentives faced by the employees at various levels of contributions. Based on findings from 137 distinct matching formulas, we have estimated a behavioral model that is able to control for the tendency of employers to substitute between the amount they match per dollar of employee contribution and the maximum percentage of compensation they are willing to match.