Measuring Social Class with Changing Occupational Classifications: Reliability, Competing Measurement Strategies, and the 1970–1980 U.S. Classification Divide

Periodic changes in occupational classifications make it difficult to obtain consistent measures of social class over time, potentially jeopardizing research on class-based trends. The severity of this problem depends, in part, on the measurement strategies used to address those changes. The authors propose that when a sample has been coded partly with one occupational classification and partly with another, Krippendorff’s index α be used to identify the best strategy for measuring class consistently across the two classifications and to assess the reliability of the class measure employed in the final analyses. This index can be computed regardless of the metric of the class variable; it can be used to compare measures based on different class schemes or that use different metrics; and statistical inference is straightforward, even with a complex sampling design. The authors put the index to work in conducting a case study of the effects of the switch from the 1970 to the 1980 U.S. Census Bureau Classification of Occupations on the reliability of Erikson–Goldthorpe–Portocarero class measures. Their findings indicate that measurement strategies that seem a priori equally reasonable vary substantially in terms of their reliability, and that the bulk of this variation is accounted for by the extent to which the strategies rely on subjective judgments about the relationships between occupational and class classifications. Most importantly, as long as the best-performing measurement strategies are used, the switch in occupational classifications appears to be substantially less consequential than has been previously argued. A computer program made available as a companion to the paper makes estimation of Krippendorff’s α, and statistical inference, very simple endeavors for nominal class variables.