The methodological risk of relying on official statistics to construct crime and other deviancy rates

This study theoretically and empirically examines the risk of relying on official statistics to construct crime rates. It assumes that a critical methodology demands the continual reappraisal of one's research methods and especially of one's data. It examines the risk of using Census data in social research when constructing crime rates. More specifically, it examines whether adjusting for differential undercount in racial minority populations significantly alters the distribution of violent crime among the major American cities.;Because American racial minorities are subject to differential undercount, the question arises as to whether the use of such population data unduly biases social research involving their populations. In order to examine this issue, the violent crime rates of 76 U.S. cities with populations over 200,000 in 1990 were computed using three different population bases: (1) the Census based on "enumeration" (the official population); (2) a Post-Enumeration-Survey adjusted Census (PES); and (3) a synthetic, population adjustment strategy used in combination with the first two strategies that includes an additional proportional rate adjustment for the Mexican and Black populations of each city. The synthetic rate construction is informed by the Census Bureau's (CSMR) coverage evaluation studies and earlier Census Bureau research.;The analysis of these different rates of crime rates for each city produced mixed results leading to questions requiring further research. Generally, violent crime rates using the synthetic adjustment rather than the official population lowered the rate of crime an average of 11% with the largest decreases, 15%--20%, occurring in the largest metropolitan-area cities. Some significant shifts also occurred in the rankings in terms of city-level crime rates. Results from regression analyses indicated only minor shifts in the relationships of city-level characteristics to city-level crime rates based on using these three different measures of crime.