Specification errors and empirical estimation of generalized crop - fertilizer production functions

THE DATA COMES FROM POTATO FERTILIZER EXPERIMENTS CARRIED OUT IN 73 DIFFERENT LOCATIONS IN THE PERUVIAN SIERRA REGION FROM 1967 TO 1971. MODELS AT TWO LEVELS OF AGGREGATION WERE CONSIDERED, GROUP MODELS WITH FIVE DATA GROUPS LEASED ON BUOL"S SOIL MANAGEMENT CLASSIFICATION SYSTEM AND AGGREGATE MODELS WITH ONE DATA GROUP. THE GROUP MODELS WERE ASSUMED TO BE TRUE AND THE AGGREGATE MODELS TO BE WRONG. FERTILIZER RECOMMENDATIONS AND PROFITS USING EACH OF THE ESTIMATED MODELS WERE MADE FOR A NUMBER OF SOIL CLASSES USING THESE PROCEDURES: 1) FOR SOILS NOT KNOWN TO BELONG TO A GIVEN GROUP, THE LA MOLINA SOILS DATA AND THE FIXED EFFICIENTS MODEL WERE USED, AND 2) FOR SOILS KNOWN TO BELONG TO A GIVEN GROUP, THE EXPERIMENTAL SOILS DATA AND THE GROUP MODELS WERE USED. RESULTS INDICATED THE IMPORTANCE OF CONSIDERING ALL THE FACTORS IN THE RESPONSE RELATION.