WEIS vs. COPDAB: Correspondence Problems

The problem of ‘empirical validity’ in events data research is addressed by comparing WEIS and COPDAB measures of conflict, on a year by year basis, between 1966 and 1978. Correlations between the two sets are moderate to high for all years except 1971, 1972 and 1975. In order to determine for which countries large scoring differences exist, the two data sets are standardized (z-scores) and differences analyzed on a year by year basis. It appears there is about an 80% probability that if COPDAB under-reports relative to WEIS the country is a major power or is located in Europe or Asia while if WEIS under-reports relative to COPDAB the country is located in the Middle East, Africa or Latin America. These differences tend to persist over all years, even those years in which there is a strong correlation between the two data sets. In general, the differences seem to be a function of the multi-source basis of COPDAB and the single source basis of WEIS. Users are warned that there is a very low relationship between the data sets in 1972 and a weak relationship in 1971 and 1975 and that some empirical findings may be ‘set bound’, even in the years of high correlation. This latter point is established by comparing correlation results of the three best attribute predictors of conflict for the two sets. In addition, the sets show no agreement, other than chance results, when they are used to estimate the rise or decline of total conflict in the system when compared to previous years' totals. Thus, the case for ‘empirical validity’ in the years of high correlation is undercut by apparent regional biases in reporting, partially inconsistent correlation results and the disagreement between the sets in identifying increasing or decreasing conflict in the system.