Estimating poverty and inequality in the absence of consumption data: an application to the Middle East and North Africa

ABSTRACT

Measures of consumption and poverty are critical metrics of the wellbeing of individuals, their households, communities, and countries. Collecting data on consumption and poverty is challenging and costly, and therefore these measures are only infrequently available in survey data. In this paper, we demonstrate how information commonly available in household surveys can be used to impute consumption, even recovering the original variance, which is crucial for assessments of poverty and inequality. Our application adds consumption estimates to the publicly available Labor Market Panel Surveys for Egypt, Jordan, and Tunisia, which can act as a valuable resource for researchers interested in the intersection of inequality, poverty, and a host of labor market behaviors in the Middle East and North Africa.

KEYWORDS:

• Inequality
• consumption
• poverty
• labor market panel surveys

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Caroline Krafft http://orcid.org/0000-0001-6906-9418

Ragui Assaad http://orcid.org/0000-0002-2608-5335

Notes

1 Since our goal is prediction, not interpreting coefficients, included X_ch variables may be endogenous without yielding estimation problems.

2 Specifically, we bootstrap not only over 100 repetitions of the imputed consumption, but also we redraw the LMPS sample five times for each r (and different redraws as we move through the different imputed consumptions) in order to incorporate the variability from using a second survey rather than the census. This yields 500 repetitions of the bootstrap, which are redrawn accounting for the sampling structure (PSUs) of the various surveys.

3 See Krafft et al. (2017) for data on the poverty lines.

4 See Krafft et al. (2017) for the underlying equations for these inequality indices.

5 The degree to which the surveys find similar characteristics, given the harmonization, can be assessed with the summary tables in the supplemental Appendix. In general, the differences are small.

6 In all of the analyses of Egypt, because the ELMPSs exclude the Frontier governorates, we likewise exclude these areas from the samples of the HIECS in implementing the poverty mapping.

7 Replication files available at https://sites.google.com/site/carolinekrafft/publications

Additional information

Funding

This research was partially supported by Carnegie Corporation of New York [grant number B 8939].