A study estimated nine cumulative logistical regression models using data from a city-wide survey of urban conditions and services and compared the effects of 11 independent variables across the nine models. Findings indicate differences for seven of the nine models. The article discusses issues in assessing police performance and models of policing, including community policing, and citizen assessment of police duties. The most consistent differences across models are for race-ethnicity and marital status. In addition, the parameter estimates for income, education, age and gender are significant for some of the models. The study found no significant parameter estimates for the models for investigating drug crimes and investigating gang activity. Nearly all respondents, regardless of personal and social characteristics, gave high ratings to those two duties. Overall, findings indicate that most citizens rate both crime control and order maintenance duties as important. Notes, appendix, references