HISTORICAL DATA FROM 65 TIME POINTS COVERING 4-WEEK PERIODS BETWEEN 1970 AND 1974 WERE USED TO FORECAST MANPOWER NEEDS FOR 26 PERIODS OF 4 WEEKS EACH FOR THE CALENDAR YEARS 1975 AND 1976. FORECASTS WERE THEN COMPARED WITH ACTUAL CALLS FOR SERVICE. TWO FORECASTING METHODS WERE USED: THE TRADITIONAL MATHEMATICAL MODEL WHICH COMBINES TREND PROJECTION WITH SEASONAL ANALYSIS AND A NEWER TECHNIQUE DEVELOPED BY G.E.P. BOX AND G.M. JENKINS. THE BOX-JENKINS METHOD IS A MODEL-BUILDING RATHER THAN A MODEL-FITTING PROCESS BECAUSE THE MODEL IS SPECIFIED BY THE DATA RATHER THAN BEING HYPOTHESIZED FOR THE DATA. THE MATHEMATICAL CONCEPTS UNDERLYING THE BOX-JENKINS METHOD ARE BRIEFLY SUMMARIZED. THE DATA COLLECTED AND THE PROJECTIONS MADE WITH EACH METHOD ARE GIVEN IN TABULAR FORM. THE BOX-JENKINS METHOD TENDS TO UNDERFORECAST A RAPIDLY RISING SERIES. THIS BECAME APPARENT WHEN FORECASTING BURGLARY RATES IN SEATTLE. THEREFORE, WHEN THIS STUDY WAS MADE, UPPER LIMITS WERE ADJUSTED TO COMPENSATE FOR UNDERFORECASTING ERROR. HOWEVER, THE METHOD IS MORE SENSITIVE TO ITS OWN ERROR, AND ADJUSTED FORECASTS WITH THE BOX-JENKINS PROGRAM WERE MORE ACCURATE THAN THE ADJUSTED FORECASTS ACHIEVED WITH THE TIME-TREND METHOD. DURING 'GROWTH' PERIODS, THE BOX-JENKINS METHOD SHOULD BE SUSPECT. HOWEVER, EARLIER FORECASTS CAN BE QUICKLY CORRECTED BY THIS METHOD. FOR ACTUAL FORECASTING OF PATROL NEEDS, THE LESS SOPHISTICATED TREND METHOD WAS SELECTED BECAUSE CALLS WERE SHOWING AN INCREASING RATE AND IT WAS FELT THIS METHOD WOULD MORE ACCURATELY REFLECT HIGHER CALLS FOR POLICE SERVICE IN SEATTLE. TABULAR DATA AND GRAPHS ARE INCLUDED. (GLR)