Comparing Poverty Over Time

Quantitatively comparing poverty over time is a critical aspect of measuring the impacts of policies and other sources of change. However, directly comparing poverty indicators from separate time periods can be misleading since data collection and interpretation methods often change over time. Applying the treatments to make such data comparable can be analytically difficult and, lacking adequate funding, time and/or knowledge, analysts often default to direct comparison. Nonetheless, the consideration of these issues is critical; their neglect can potentially lead to large errors in the magnitude and sign of poverty changes.

This site provides a guide for some of the basic issues in comparing poverty levels over time. While the discussion focuses on income and consumption measures, some of the issues apply more broadly to all types of quantitative survey data. In addition to a guide on how to identify the problems, this site provides a basic discussion and some options for dealing with the five most common issues when comparing poverty over time:

1. constructing the consumption aggregate,
2. accounting for price changes,
3. selecting the poverty line,
4. adjusting for changes in sampling and,
5. dealing with changes in survey administration.

All five discussion sections provide links to a glossary of terms and to theoretical and regional documents, which give more details and example applications. In each section, we provide some guidance on thinking about robustness of poverty comparisons over time. However, all of the measures discussed below require some assumptions and/or alterations to the data. Hence, it is worthwhile checking how the measured change in poverty changes if these assumptions change or a different method is used to make the data comparable.

While this site is meant to be primarily used by those seeking to deal with existing data, it is also of use to those seeking to update and improve household surveys. As Deaton and Kozel (2004) observe "there is always a conflict between updating and improving a survey instrument on the one hand, and consistency of estimation on the other." Hopefully, the discussion that follows can provide some considerations on how to maintain consistency while improving survey instruments.

Back to Analyzing Poverty