Sponsored Projects for Undergraduate Research

The standard approach to measuring poverty in low income countries relies on measuring households' total expenditures on non-durable consumption and services and prices for these goods, an approach implemented at the World Bank through the design and implementation of comprehensive "expenditure modules" in the Living Standards Measurement Surveys (LSMS).

Though the Marshallian theory of demand provides sound theoretical foundations for the standard approach, the task of measuring /all/ expenditures is both expensive and challenging, particularly since it seems clear that some important kinds of expenditures are quite difficult to measure (e.g., expenditures on housing services for populations with varying tenurial arrangements).

Satisfying the data requirements for measuring welfare and poverty would be much less expensive if one needed to measure only a selected subset of goods and services.  But the same Marshallian theory that justifies using prices and total expenditures to measure welfare also tells us that using a subset of expenditures is invalid save under highly implausible assumptions regarding households' utility functions.

However, the Marshallian formulation of the theory of demand is not the only possible formulation.  In particular, instead of relying on prices and total expenditures, the Frischian theory relies on prices and the marginal utility of expenditures, and Ligon (2017) shows how this latter quantity can be recovered as a latent variable from an incomplete demand system.  This incomplete demand system needs to have data on only a small selection of consumption goods and services.

Taking the logarithm of this latent variable and adjusting for some household characteristics gives a measure of consumer welfare Ligon calls "neediness."  Thus, the Frischian theory and our approach to estimation offers a theoretically sound and empirically practical basis for measuring household welfare using data on prices and expenditures on only a small number of goods.

In principal the resulting welfare measures should be comparable across time, countries, and even differently designed expenditure modules.  Since we don't rely on total expenditures for the welfare calculation there's no need to calculate things like `real' expenditures or to use PPP indices.

The long-term vision for this research is a set of methods which allows one to estimate household welfare in a way which is internationally and intertemporally comparable, using only quite modest data on a selected handful of household expenditures on different items.  The theory which justifies this is mostly in place, but of course theory will fail to anticipate issues that arise in application.  Below are two separate proposals, each of which is designed around particular applications using LSMS data from Africa to extend the methods which seem sound in theory to real-world application.

The standard approach to measuring poverty in low income countries relies on measuring households' total expenditures on non-durable consumption and services and prices for these goods, an approach implemented at the World Bank through the design and implementation of comprehensive "expenditure modules" in the Living Standards Measurement Surveys (LSMS).

Though the Marshallian theory of demand provides sound theoretical foundations for the standard approach, the task of measuring /all/ expenditures is both expensive and challenging, particularly since it seems clear that some important kinds of expenditures are quite difficult to measure (e.g., expenditures on housing services for populations with varying tenurial arrangements).

Satisfying the data requirements for measuring welfare and poverty would be much less expensive if one needed to measure only a selected subset of goods and services.  But the same Marshallian theory that justifies using prices and total expenditures to measure welfare also tells us that using a subset of expenditures is invalid save under highly implausible assumptions regarding households' utility functions.

However, the Marshallian formulation of the theory of demand is not the only possible formulation.  In particular, instead of relying on prices and total expenditures, the Frischian theory relies on prices and the marginal utility of expenditures, and \cite{Ligon15} shows how this latter quantity can be recovered as a latent variable from an incomplete demand system.  This incomplete demand system needs to have data on only a small small selection of consumption goods and services.

Taking the logarithm of this latent variable and adjusting for some household characteristics gives a measure of consumer welfare Ligon calls "neediness."  Thus, the Frischian theory and Ligon's approach to estimation offers a theoretically sound and empirically practical basis for measuring household welfare using data on prices and expenditures on only a small number of goods.

In principal the resulting welfare measures should be comparable across time, countries, and even differently designed expenditure modules.  Since we don't rely on total expenditures for the welfare calculation there's no need to calculate things like `real' expenditures or to use PPP indices.

The long-term vision for this research is a set of methods which allows one to estimate household welfare in a way which is internationally and intertemporally comparable, using only quite modest data on a selected handful of household expenditures on different items.  The theory which justifies this is mostly in place, but of course theory will fail to anticipate issues that arise in application.  Our current research plan involves implementing these methods using a number of household surveys from various low-income countries to extend the methods which seem sound in theory to real-world application.