The Census Occupational Taxonomy: How Much Information Does It Contain?
Concludes that, judged on the basis of wage and job status information, most of the detail contained in the 297 occupational codes used in 1960 and expanded to 443 for the 1970 Census is irrelevant. To explore efficiencies achievable through reclassification, the authors adopt a quadratic function--a second-order approximation to more general functions--as a measure of the information lost with aggregation. In fact, the authors use two indexes, weekly wages and socioeconomic status, as scalar measures of information, taking the existing detailed codes as the point of departure. Fully 98 percent of the variance explained with 298 codes is retained after aggregation to only nine groups. As a first step in a search for efficient taxonomies, the authors present a simple quality index to be used for judging alternative aggregation schemes when criteria other than wages and socioeconomic status are added. (See also R-1666.) 36 pp. Bibliog.