Entropy-Based Consumption Diversity—The Case of India

Abstract

In recent years, there has been growing research in analysing the spending diversification of households in applied demand analysis using disaggregated household-level data. Taking cue from Engel’s (Die Lebenskosten Belgischer Arbeiter Familien frfther und jetzt, Bulletin de l’institut international de statistique, tome IX, premiere livraison, Rome, 1895) findings that large share of income is spent on basic goods such as food for lower-income decile, the applied demand analysts also observed that with increasing income, there is an increase in spending on other non-food commodities, implying a hierarchical structure of consumption pattern. Evidences also supported positive correlation between household income and the dispersion of household spending both at cross-country-level analysis and at household-level analysis. These findings justify the use of consumption-based measures such as food share (Anand and Harris in Am Econ Rev 84:226–231, 1994) and consumption diversity (Clements et al. in Empirical Econ 31:1–30, 2006; Chai et al. in J Econ Surv 29:423–440, 2014) as indicators of household welfare. In this paper, we attempt to examine the stylized facts of behavioural heterogeneity across disaggregated commodity groups by employing entropy-based Theil’s measure. Using National Sample Survey household expenditure data of urban sector of four major states of India for the year 2011–2012, we show the extent to which income (measured through monthly per capita expenditure and thereby controlling household size) and other demographic characteristics such as number of children explain the variation in consumption diversity. We also capture commodity group-wise variations for explaining consumption diversity within commodity group by considering not only inherent characteristics of commodity groups such as income elasticity as control variables but also as random coefficient models varying randomly across commodity groups. The incorporation of between-commodity heterogeneity via random coefficient model is our contribution in this literature on consumption heterogeneity. The random coefficient models establish significant heterogeneity across commodity groups, mainly through intercept change, but not so much in income and demographic factors’ effect.