Does Measurement Error Distort Our Perception of Crime Rates?

Accurate data on crime rates is essential for evaluating many social programs. However, in the United States, officials often impute crime rates to compensatefor the 20--40% of agencies that do not report crime statistics. I show that all commonly used methods underestimate crime. First, I show that passage of state laws requiring agencies to report crime leads to a 16--23% increase in local jurisdictions' crime rates, depending on the methodology. Thus, current methods underestimate local jurisdictions' crimes rates in states that do not require crime reporting. Second, I compare commonly used murder estimates to the National Center for Health Statistics (NCHS) homicide measure, which is widely believed to be accurate. I find that county murder estimates reflect only 78--87% of homicides, and that state reporting requirements increase the estimated number of murders by 8%--19%. Moreover, undercounting crime can lead one to underestimate the impact of crime-inducing policies. I illustrate this finding by showing that the introduction of casinos in rural counties increases the number of homicides by 1.17 per 100,000, but the number of murders rises by only 0.36--0.85 per 100,000. Finally, I provide empirical justification for studies that restrict their analysis to large jurisdictions, by showing that this sampling restriction significantly reduces measurement error