Material deprivation from 1999 to 2013 in Switzerland: How index construction impacts on measured patterns of evolution

This paper discusses the consequences of various index constructions and of using panel data on the measurement of the evolution of material deprivation in Switzerland at macro level over the last two decades. In the larger purpose of providing reliable conclusions about the evolution of economic and social inequalities, we compare the patterns of evolution computed with data from the Swiss Household Panel and various deprivation index constructions. We discuss the topic of weighting deprivation scores by the social importance of the items used in the index (both consensual weights based on the social opinion on how much each item is necessary for a decent living, and prevalence weights based on the part of the population having each item). We also discuss how to deal with partial non-response, the number of items in the index, and the consequences of considering the part of households with no deprivation rather than the mean score of deprivation index. In addition, we examine how attrition effects due to the panel structure of the data impacts the patterns of evolution of material deprivation. The conclusion is that the way in which the deprivation index is constructed has a limited influence on the level of deprivation measured each year, and no impact on the pattern of evolution (time series). This confirms, at least on the macro level, the robustness of deprivation measurement to methodological choices. The panel structure of the data poses a greater challenge, since attrition tends to reduce the measured deprivation score even when using sample weights. As a consequence, the introduction of refreshment samples produces clear ruptures in the time series. Despite these limitations, patterns of evolution drawn by the data are coherent with further results on the evolution of socioeconomic inequality in Switzerland.