Partial Deprivation – identification and aggregation for uni- and multidimensional poverty applications
Krause, Peter (2016). 'Partial Deprivation – identification and aggregation for uni- and multidimensional poverty applications' Paper presented at the annual conference of the HDCA, Tokyo 2016.
This paper offers a counting approach to consider partial deprivation within the established fgt-measurement framework for uni- and multidimesional ordinal and cardinal poverty applications. Following Sen we face two general steps in the measurement of poverty. At first we have to choose indicators and to set up thresholds to identify someone as poor or non-poor – the identification step. At second we apply measures to notify the incidence, intensity and inequality of poverty – the aggregation step.
This paper challenges the simple use of dummy variables as poverty indicators, when further informations on distributions are available. In case of relative income based poverty measures the deprivation thresholds refer usually to 60% [EU-standard] or 50% [sdg-standard] of median equivalent household-net-income. Everyone living in a household with lower income is counted as poor – with increasing intensity (and inequality) of poverty, when the income distance to the poverty line increases. Everyone with higher income is regarded as non-poor – even if his/her income is just above that line. However, in reality people may experience at least some degree of partial deprivation, even if their incomes are (just) above that line. This kind of partial deprivation is neglected in our standard measures on income poverty and on other dimensions of deprivation.
The paper includes five sections. The first section defines the identification function using a fuzzy parameter (epsilon), which triggers the shape of the functional decline in deprivation scores above the poverty line. Two ways are discussed to set up the final identification function – (a) the choose of the fuzzy parameter (epsilon), and/or (b) the use of a second poverty (fuzzy) threshold to define the area of partial deprivation and to separate it from the areas of full deprivation and no deprivation. The second section describes the integration of the derived partial deprivation scores into the measurement framework for ordinal and cardinal fgt-measures on poverty.
The third section offers empirical results for several indicators comparing deprivation rates with and without the consideration of partial deprivation. The empirical findings are based on the German Socio-economic Panel study (GSOEP) for long-term applications over the last two decades and on EU-Silc data for cross-national comparisons. The empirical findings refer to the ordinal (fgt-0) and cardinal (fgt-1 2) case. The fourth section illustrates the same for multidimensional poverty measures. Therefore two alternative approaches of multidimensional measurement are applied, (1) the well-establish Alkire-Foster Dual-Cutoff procedure [Alkire/Foster 2011a,b; Alkire/Foster et al. 2015] and (2) the CSPI-approach (correlation sensitive poverty index, [Rippin 2014,2015]). Both approaches indicate, that the consideration of partial deprivation has substantial impacts on the measurement and decomposition of multidimensional poverty profiles due to the high share of individuals affected by overlaps of marginal deprivation scores just above the full dimensional deprivation thresholds. Section five offers a brief summary and discussion.