Regression-based adjustment of student outcomes for the assessment of the merit of schools is considered. First, the basics of causal modeling and multiple regression are briefly reviewed. Then, two common regression-based adjustment procedures are described, pointing out that the validity of the final assessments depends on: (1) the degree to which the assumed adjustment model accurately reflects the actual causal processes of schooling in the target school district; and (2) the validity and reliability of the measurement of all necessary variables. In the final section, it is argued that assessment of school merit should not be based solely on regression-based adjustment of student outcomes because state-of-the-art knowledge of causal processes and measurement skills is not adequate to ensure reasonable accuracy. Based on the contention that each approach to merit assessment currently has serious flaws, it seems likely that the greatest confidence in the fairness of any merit assessment would result when the assessment system is based on several different approaches, including adjusted student outcome comparisons, and is tempered by informed judgment. A simple example is used to illustrate the possible severity of bias introduced by a single "specification error" in the model. However, it is suggested that regression-based procedures might play other, less central, roles in the assessment of school merit. Two tables and one figure are provided. (Author/SLD)