A Data Envelopment Analysis Approach to Evaluation of the Program Follow Through Experiment in U.S. Public School Education

A method called Data Envelopment Analysis (DEA) is used to decompose the efficiency of Decision Making Units (DMU's) into two parts: (1) a component resulting from managerial decisions and (2) a component resulting from constraints (called programs) under which management operates. The DEA approach accomplishes this by enveloping the input-output observations with extremal relations developed in terms of a specified nonlinear programming model (and/or its linear programming equivalent). Differences between the observations and the progam specific envelopes -- called alpha-envelopes--are imputed to managerial inefficiencies. An inter-program envelope is then constructed from 2 or more such alpha-envelopes and used to identify 'program' inefficiencies, which are the inefficiencies that remain after the previously determined managerial inefficiencies have been eliminated. Numerical illustrations accompanied by suggested tests of a probabilistic/information theoretic character are provided by means of recenty released data from 'Program Follow Through.' Designed as a study of possible ways of reenforcing or extending Program Head Start - an ongoing pre-school program for disadvantaged children -- the Program Follow Through experiment provides data on agreed upon inputs and outputs for both PFT (Program Follow Through) and matched NFT (Not Follow Through) participants in various parts of the U.S.