ABSTRACT
Equivalence scales are used to adjust household income for household size and composition differences. The literature on poverty calculations have found sensitivity of poverty measurements to the choice of equivalence scales. In this study, we aim to estimate the sensitivity of the Turkish poverty rate and its change over time to the choice of equivalence scales. To that end, Household Budget Survey micro-level data are used and poverty rates are calculated for different equivalence scales. Results show that poverty rates differ sharply with the scale parameter. The change in poverty over time varies in absolute terms with most scale parameter values but only for high values of scale as a ratio of the initial period. Characteristics of poor households also change with scale alternatives.
JEL classification:
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 An earlier study by Buhman et al. (Citation1988) notes that 34 different scales were being used at that time in different countries.
2 The OECD-modified scale assigns value of 1 to the household head, 0.5 to other adult household members and 0.3 to each child.
3 They provide two alternatives. The one with a classification that is similar to the OECD-modified scale has adult equivalence parameter 0.65 and child equivalence parameter 0.35, rather than 0.5 and 0.3, respectively, in the OECD-modified scale.
4 The absolute poverty thresholds are under revision by Turkstat since 2009 and the figures have not been announced officially since then. Figures provided for the years after 2009 are calculated by the author using micro data sets obtained from Turkstat.
5 The relative poverty rate, calculated as those earning less than 60% of the median household income, decreased from 25.4% in 2006 to 22.4% in 2013.
8 The FAO scale parameter is 0.78 with no adjustment for children.
9 Gini coefficient was 0.42 in 2003 and improvement in income inequality may have played a role in poverty reduction. Yet, Seker and Jenkins (Citation2015) find that the decrease in poverty is largely due to economic growth than distributional changes.
10 We have also conducted a stochastic dominance test following Chen and Duclos (Citation2011) using all the pairs of scale and child parameters from 0 to 1 with 0.01 intervals (a total of 10,000 alternatives) and tested whether the change from 2003 to 2013 is significant. The change at all pairs were negative and the pair with the lowest t-value is found to be significant at 5% level.