THE MODEL BEGAN WITH A SECOND ORDER APPROXIMATION TO AN ARBITRARY MULTI-OUTPUT, MULTI-INPUT PRODUCTION POSSIBILITIES FUNCTION. COMBINED CROSS-SECTIONAL AND TIME SERIES DATA FROM 35 CITIES WERE USED TO TEST FOR THE EXISTENCE OF AGGREGATE INDEXES OF POLICE OUTPUT, FOR THE DISJOINTEDNESS OF OUTPUTS, AND FOR THE CONSISTENCY OF THE ESTIMATED COST FUNCTION WITH OPTIMIZING BEHAVIOR AS PREDICTED BY CLASSICAL ECONOMIC THEORY. RESULTS INDICATE THAT POLICE ADMINISTRATORS' DECISIONS ARE INCONSISTENT WITH COST MINIMIZATION, THAT THE HYPOTHESIS OF NONJOINT PRODUCTION IS INCORRECT, AND THAT RETURNS TO SCALE VARY CONSIDERABLY WITH ACTIVITY LEVELS. IN ADDITION, NO CONSISTENT INDEX WAS FOUND FOR ANY OF THE SUBAGGREGATES OF OUTPUTS, WHICH INDICATED THAT AGGREGATION OF POLICE OUTPUTS INTO ONE MEASURE RESULTS IN LOSS OF INFORMATION. RETURNS TO SCALE, MARGINAL COSTS, AVERAGE COSTS, AND AVERAGE RATES OF TRANSFORMATION AT THE SAMPLE MEAN WERE ALSO CALCULATED. FURTHER RESEARCH IS RECOMMENDED, ESPECIALLY ON THE SUBJECT OF CRIMES AGAINST THE PERSON. TABLES, FOOTNOTES, AND A BIBLIOGRAPHY ARE INCLUDED. FOR THE COMPLETE VOLUME OF WHICH THIS STUDY IS A PART, SEE NCJ61013. (CFW)