Establishing the "Statistical Accuracy" of Uniform Crime Reports (UCR) in West Virginia.

• NCJ Number: 214961
• Author(s): James Nolan Ph.D.; Stephen M. Haas Ph.D.; Theresa K. Lester M.A.; Jeri Kirby M.A.; Carly Jira B.A.
• Date Published: July 2006
• Length: 20 pages
• Grant Number(s): 2004-BJ-CX-K004
• Sponsoring Agency: Bureau of Justice Statistics (BJS)
• Corporate Author: West Virginia Criminal Justice Statistical Analysis Ctr
• Sale Source: West Virginia Criminal Justice Statistical Analysis Ctr
• Publication Format: Document
• Publication Type: Report (Study/Research)
• Language: English
• Country: United States of America
• Note: Downloaded July 24, 2006.

Annotation

The methodology used in this report illustrates a valid approach for assessing the statistical accuracy of Uniform Crime Reports (UCR) crime statistics in West Virginia and the Nation.

Abstract

The methodology involved presampling, selecting and reviewing sampled records, and analyzing the results. The presampling stage involved the partitioning of records and the calculation of the appropriate sample size. In the partitioning of records, the 12 largest municipal police departments in the State composed the population for the study. From the 12 agencies, 3 were randomly selected to participate. The records in each of these 3 agencies were divided into 15 categories, some of which were single UCR offense categories while other categories included several types of crimes, such as "all sex offenses other than rape." A random sample of 100 reports was selected from each category. A systematic procedure for the assessment of each record was established to ensure a high level of reliability between reviewers. Multiple reviewers examined those records with a suspected classification error. Upon reviewing the records, investigators found five larceny reports (5 percent) misclassified that should have been recorded as burglaries. Five burglary reports (5 percent) should have been recorded as larceny. No other errors were found in the review of the 1,500 records of the sample. Determining statistical accuracy for the entire year's records required a calculation that estimated the impact of the error found in the sample on the population of records for the year. This paper describes this calculation. Based on this calculation, of the 31,084 reported crimes in the population, 1,297 were estimated to contain a classification error. Both the violent crime and index totals for the State were significantly undercounted in reported UCR statistics. The discussion highlights the primary reasons for the presence of classification errors in UCR statistics. 7 tables, 8 notes, and 16 references