On October 22nd Barbara Jancewicz and Dr. Zbigniew Karpinski presented a paper: ‘Testing Measurement of Perception of Income Inequality’.
Abstract:
Measures of income inequality, such as the Gini coefficient or the quintile ratio, are becoming increasingly popular among non-specialists, such as journalists or bloggers. These measures satisfy a number of requirements that define how changes in the characteristics of an income distribution translate into changes in the degree of income inequality. Unfortunately, these requirements are rarely supported by non-specialists, as shown in early research by Amiel and Cowell (1992, 1999). In other words, the way ordinary people perceive inequality is largely inconsistent with the way social science defines it. However, a considerable proportion of responses provided by participants in Amiel’s and Cowell’s study was puzzling or even self-contradictory.
We suspected that a major reason for these results is an imperfect questionnaire construction. Therefore, we analyzed the questionnaire designed by Amiel and Cowell and modified it significantly. We changed: the order of questions; the way of presenting income distributions by replacing numerical vectors with bar charts and expressing income amounts in the Polish currency rather than abstract monetary units. We also standardized the set of possible answers and merged pairwise comparisons into short ranking questions. Also, we added an introduction to the questionnaire and included a “hard to say“ option in the individual questions.
This paper analyses results of an experimental test of the modified questionnaire against the original one. 134 undergraduate students took part in our study. Respondents were randomly divided into two groups, each filling a different version of the questionnaire, so that differences in the results between the two groups – in terms of the frequency of contradictory responses and support for the inequality axioms – can be attributed to the changes made in the questionnaire.
Modifications of the numerical questions (bar charts and ranking questions) reduced the number of inconsistent and puzzling answers significantly and increased support expressed for income inequality axioms. Simultaneously, changes introduced to the verbal questions had a more mixed impact that heavily depended on the exact formulation and length of the questions asked. In general it seems that our modifications go in the right direction, yet further changes and experiments are needed.