The model of the crime deterrence process is based on an assumption that the police objective in allocating resources among various programs is to minimize the social losses from crime. The independent deterrent effects of two such programs, involving patrol and investigation units, are analyzed with linear regression analysis and data from police departments in the Los Angeles region. The two subprograms, identified as visible patrol and investigation, are assumed to affect offender risk perceptions via two different channels--patrol through a threat of immediate arrest for all offenses, investigation through a long-run threat of arrest for more serious offenses. Enforcement levels in patrol and investigation are measured by the density of visible patrol units and the citywide arrest rate. A third indicator, the ratio of filings to arrest, is introduced to measure the joint production of both subprograms. The evidence tends to support the view of the deterrence process--that offenders form subjective risk estimates based on the visible presence of the police, and that increasing patrol visibility can therefore have a significant deterrent effect. The study refutes both the extreme positions frequently taken in the public debate over crime control policy--that crime is beyond the control of the police, that increasing police expenditures will be ineffective, and that only wholesale expansion of police manpower will solve the urban crime problem. The most cost-effective police policy appears to be one that requires the least additional new resources or new spending but merely redeploys existing police resources. Such changes by local police, combined with attacks by the Federal Government on poverty and racial discrimination in labor and housing markets, appear to offer the greatest hope for reducing urban crime. Tables, an appendix showing data sources, and 24 references are provided.