The study used pooled county-level time-series data for 57 Florida counties from 1980 to 1998. To be consistent with measures used by Marvell and Moody, the study used the total number of police employees as its police variable. "Crime" was defined as the number of offenses known to police that comprised the UCR Crime Index (excluding arson) per 100,000 population. The findings of this study were generally consistent with those reported by Levitt (1997) and Marvell and Moody (1996), i.e. that police personnel levels had significant and substantial impacts on robbery, burglary, and larceny. The elasticities in the current study, however, were somewhat smaller than those reported by Marvell and Moody as well as by Levitt. Regarding homicide, Levitt and Marvell and Moody found strong evidence that more police reduced homicide rates at the city and State levels of analysis. The current study found little impact on homicide at the county level. The most likely explanation for the different results was that during the late 1980's Florida released many offenders from its prisons due to overcrowding. If part of the impact of police levels on crime was through incapacitation, then increased police levels might have helped overload Florida's prisons. Early-release programs designed to alleviate prison overcrowding might have had the unintended consequence of releasing large numbers of criminally active offenders back into the community. 2 tables, 17 notes, and 71 references