The Difficulty of Worker Monitoring in the Service Sector: Simulation and Empirical Evidence of the Value-Added of Middle School Teachers

Abstract: Accurately monitoring workers in some service sector jobs presents distinctive challenges because: (a) workers interact with, and receive utility directly from, heterogeneous clients; and (b) client contributions to output vary dramatically and may be correlated with the same client characteristics that affect worker utility. The selection bias problem this causes is particularly severe with public sector services where clients are not voluntary participants (e.g., students are required to attend school) and therefore market indicators of productivity are less informative. This may explain, for example, why it appears that elementary school students are non-randomly assigned to teachers, which has led to concern about the validity of “value-added” teacher productivity measures. We investigate these problems by studying how students and teachers sort into tracks in Florida public middle schools. Using theory and evidence, we generate a data set in which teacher productivity is known. We will use these simulated data to identify value-added specifications that minimize bias relative to actual teacher productivity.