Are estimates of earnings inequality sensitive to measurement choices? A case study of Canada in the 1980s
While it is necessary that researchers make choices in order to estimate inequality, the reasons for the measurement choices and their impact on inequality estimates have not been widely assessed. This paper uses Canadian data from the 1980s to analyse whether inequality estimates are sensitive to common measurement choices. Seemingly minor technical choices about the treatment of outlying observations, such as the use of top-income coded data, exclusion of very high and low observations, and differences among data sets in the capture of very high observations affect estimates of inequality. Further, the impact of the treatment of outlying observations on inequality estimates are at least as large as the impact of measurement choices of a conceptual nature, such as the income definition and population selection. The sensitivity of inequality estimates to measurement choices, which often remain invisible, affect inferences about the relative degree of inequality at a given point in time among countries and changes over time.