Modeling Trajectories in Social Program Outcomes for Performance Accountability

Abstract

Government and public focus on accountability for program outcomes, combined with practical and ethical constraints on experimental designs, make nonexperimental studies of social programs an increasingly common approach to producing information on program performance. In this paper, we compare the effectiveness of alternative nonexperimental evaluation methods used with longitudinal data to produce information about social program outcomes. The analysis is applied in an evaluation of a demonstration program designed to help young parents and their partners make the transition from receiving public transfer income to earned income. Empirical findings indicate that modeling outcome trajectories using multilevel methods generates more complete information about the nature of the program effects, relative to standard econometric alternatives as commonly applied in evaluations. With increased access to sources of administrative and management information system data covering multiple time points, we urge researchers to help social policy and program decision-makers appropriately model and produce more accurate information on program outcomes.