THE DEPENDENT VARIABLES RELATED TO FIRES OF AN INCENDIARY ORIGIN ARE ESTIMATED LOSS BY INCENDIARY AND SUSPICIOUS FIRES, NUMBER OF INCENDIARY AND SUSPICIOUS FIRES, ESTIMATED LOSS BY UNKNOWN CAUSES OF FIRES, AND NUMBER OF UNKNOWN CAUSES OF FIRES. THE NUMBER OF PREDICTOR VARIABLES IN THE DATA SET WAS REDUCED FIRST TO 18 BY DISCARDING VARIABLES HAVING A CORRELATION COEFFICIENT OF LESS THAN 0.6 WITH AT LEAST 1 DEPENDENT VARIABLE, THEN BY APPLYING PRINCIPAL COMPONENTS ANALYSIS TO ELIMINATE REDUNDANT VARIABLES. THESE VARIABLES WERE ANALYZED USING THE MAXIMUM R-SQUARE IMPROVEMENT TECHNIQUE FOUND IN STEPWISE PROCEDURES OF THE STATISTICAL ANALYSIS SYSTEM (SAS) AND CONSOLIDATED IN THESE SUBSETS: (1) FINANCIAL CONDITIONS OF THE INSURED, (2) LABOR COSTS OF FIRM INSURED, (3) FEDERAL GOVERNMENT SURPLUS AND DEFICIT, AND (4) OTHER VARIABLES ASSOCIATED WITH THE BALANCE OF PAYMENTS, WHOLESALE PRICE INDEX, AND MONEY SUPPLY. MODELS GENERATED BY THE TECHNIQUE WERE USED AS INPUT INTO THE GENERAL LINEAR MODEL PROCEDURE OF SAS TO OBTAIN THE DURBIN-WATSON STATISTIC FOR AUTOCORRELATION AND PLOTS OF RESIDUALS. THE STUDY FOUND A STATISTICAL RELATIONSHIP BETWEEN ARSON LOSSES AND SELECTED ECONOMIC INDICATORS. HOWEVER, IT IS IMPOSSIBLE TO CONCLUDE THAT ARSON LOSSES ARE DIRECTLY RELATED TO ECONOMIC RECESSIONS AND DEPRESSIONS, AS THESE ECONOMIC PHENOMENA NEVER HAVE BEEN DEFINED PRECISELY. SOME OF THE ECONOMIC INDICATORS TRADITIONALLY ASSOCIATED WITH ARSON LOSSES ARE LESS SIGNIFICANT THAN A SMALLER SET OF VARIABLES. 'BANKRUPTCY-OTHER' WAS THE BEST SINGLE INDICATOR. IMPLICATIONS FOR THE INSURANCE INDUSTRY ARE THAT THEY SHOULD BE AWARE OF THE INSURED'S FINANCIAL CONDITIONS; USE EXTENSIVE METHODS TO DETECT AND AVOID FRAUD DURING BOTH THE UNDERWRITING AND THE CLAIMS ADJUSTING PHASE OF THE OPERATION; AND EXPAND EFFORTS TO IMPROVE THE REPORTING, INVESTIGATING, AND PROSECUTING OF POSSIBLE ARSON-CAUSED FIRES. REFERENCES AND TABULAR DATA ARE PROVIDED. (DAG)