The program was carried out in Denver, Colo.; Long Beach, Calif.; and St. Louis, Mo. It included security surveys of 430 commercial establishments located in 10 commercial areas throughout the 3 cities. In addition, five visits were made to each surveyed establishment. The final overall compliance level -- percent of total recommended changes that were complied with -- was 59.1 percent. For each establishment, the survey staff reviewed which survey recommendations had been made and which had been followed. The staff then judged whether the followed recommendations were substantial enough to reduce the establishment's risk-to-burglary, thus designating the establishment as treated or untreated. The nearly even split -- 194 treated and 236 untreated -- and the comparability between treated and untreated establishments made possible a split-area research design. Applying this design to burglary statistics showed that security surveys with compliance accounted for a 64.8-percent reduction in burglary in Denver. No such impact was observed in Long Beach or St. Louis. Development of a risk-based security survey instrument and the conduct of additional evaluations of security survey programs are recommended. In addition, development of a training manual on the conduct of security surveys is also advised. Statistical models -- including equations and narrative descriptions -- for assessing an establishment's risk-to-burglary and for the split-area design are appended. Ten exhibits and 25 references are included. (Author abstract modified)