Content area
Abstract
Specht and Warren (1975) have presented a technique by which the same causal model may be tested across various populations, provided that the assumptions of ordinary least squares regression are satisfied. The purpose of this study is to demonstrate that OLS assumptions are not critical to the Specht-Warren technique. In the presence of auto-correlation, heteroscedasticity, or multicollinearity, generalized least squares (GLS) regression assumptions can still be satisfied and the technique remains valid.
After demonstrating this, the Specht-Warren technique is applied to time series riot and economic deprivation data in four western nations. Data of this sort are typically analyzed either across time in a single nation or across nations at a single point in time. Under its relaxed GLS assumptions, the Specht-Warren procedure allows the simultaneous comparison across time and across nations, thus providing more information concerning the operation of the social process within and between nations.
No significant differences are found in the operation of the causal model across nations. The Specht-Warren procedure, however, remains valid even when applied to data which violate the assumptions made by the original authors.
