NCJ Number
134809
Journal
Journal of Research in Crime and Delinquency Volume: 28 Issue: 4 Dated: (November 1991) Pages: 400-417
Date Published
1991
Length
18 pages
Annotation
The methodological approach of Cantor and Land (1985) in determining the unemployment and crime relationship is critiqued regarding statistical validity.
Abstract
The equations Cantor and Land estimated are shown to be statistically misspecified within their terms of reference. In their model, Cantor and Land explored various specifications of linear and log-linear relationships between detrended annual U.S. time series (1946-1982) for each of seven FBI index crimes and similar annual series of the levels and first differences of the unemployment rate. The claim is based on the logic of recent work in linear time-series/econometric methodology, particularly the conintegration and error connection models. Based on the assumption that the time series consisting of levels of the variables are nonstationary and those consisting of the first differences are stationary, and that the levels series are I, the specification of an I series, the unemployment rate (as regressor in a model of the first-differences) or I, crime rate series would be a misspecification, and, consequently, the results become invalid. In addition, the attempt to overcome the difficulties of specifying and estimating a full structural model by concentrating on the reduced form is viewed as limited because other possible exogenous variables are ignored. 3 tables and 33 references