Skip to main content

Empirical Demand and Welfare Analysis

Periodic Reporting for period 3 - EDWEL (Empirical Demand and Welfare Analysis)

Reporting period: 2019-10-01 to 2021-03-31

Measurement of consumer welfare plays a central role in economic evaluations such as the calculation of price-indices, the formulation of tax/subsidy policies and regulation of mergers. The project aims to develop tools for demand analysis and welfare-evaluation corresponding to policy-interventions in the common real-life setting of discrete-choice. Examples include the impact of tuition subsidies for college entrants, fare-hikes for passengers and access to new channels for TV viewers. Previously existing methods in the literature have either made strong ad-hoc assumptions for such calculations, leading to potentially incorrect inference, or have not clarified the fundamental connection between the economic theory of choice among discrete alternatives and their empirical implications for demand and welfare predictions. The project EDWEL attempts to fill this gap. In particular, it aims to develop methods for four distinct but related problems, viz. (i) welfare analysis of price and quality change in choice settings involving multinomial, ordered or non-exclusive alternatives, (ii) predicting demand and welfare effects of price interventions in settings where an individual’s choice affects their peers’ utilities, such as health product adoption that can generate externalities and schooling decisions that produce peer-effects, (iii) understanding the connection between the economic theory of utility maximization by heterogeneous consumers and observable discrete choice demand data, and the implications thereof for predicting demand and welfare resulting from future policy changes, and (iv) application of rational choice theory and welfare results to understand two socially important discrete choice decisions in the educational sphere, viz. (i) how universities select applicants for admission, and (ii) the implied cash-value of publicly provided schooling in developing countries.
I have by now completed several of the sub-projects stated in my original proposal. These are (i) welfare analysis in multinomial choice models under general unobserved heterogeneity, (ii) welfare analysis in binary choice models with interval-reported income, and (iii) demand and welfare analysis in binary choice models with social interactions.
The first project developed methods for calculating consumer welfare effects of economic changes in multinomial choice settings, under general unobserved heterogeneity in consumer-taste. These results cover simultaneous price change of multiple alternatives (e.g. as after mergers), quality change and introduction of new goods or removal of existing alternatives.

The next subproject, in collaboration with Dr Ying ying Lee of the University of California Irvine, developed methods for calculation of consumer welfare as in the above problem, but where individual consumers' income data are reported in intervals, as is often the case with survey data. This practical problem that one faces in calculating empirical welfare estimates, nonetheless, lends itself to a bounds analysis. In our project, we show how to construct the best-approximation bounds on welfare effects, taking into account shape restrictions implied by economic theory.

I have largely completed the work on welfare analysis in discrete choice models with social interactions, in collaboration with Pascaline Dupas of Stanford and Shin Kanaya of Aarhus. In this project, we have shown how to estimate preference parameters in interactive decision models with large number of individuals interacting with each other. A specific contribution is to account for spatial dependence, i.e. common unobserved shocks, in estimating these interactive discrete choice models. We have also developed methods to calculate demand predictions and bounds on welfare effects resulting from a hypothetical price change, such as an income-contingent subsidy. The key results are that (i) spatial dependence can be ignored under certain "increasing domain asymptotics" in estimating preference parameters, and (ii) exact welfare analysis is not possible in presence of social interactions even when expected demand can be point-identified. These theoretical results have been illustrated with an empirical example of mosquito-net adoption in rural Kenya. The resulting paper is now under review at the Review of Economic Studies.
The key contribution of this project has been to develop empirical methods for demand and welfare analysis in discrete choice models, where very few, if any, assumptions are imposed on behaviour, except those implied by economic theory. In the traditional literature, there was no previous result on welfare analysis in nonparametric settings, none on rationalizability in discrete choice models with general heterogeneity or on individual welfare analysis under social interactions.

In the remaining period of the grant, I expect to develop (i) results on rationalizability in general discrete choice models beyond binary choice, (ii) methods for understanding admissions to elite universities (a binary decision for each applicant taken by admission tutors) via rational choice theory, (iii) methods to calculate the implied cash-equivalent of publicly provided schooling in developing countries with implications for poverty and inequality calculations, and (iv) methods for welfare analysis in situations involving uncertainty and/or nonlinear budget sets, such as choice of health insurance plans with varying marginal prices and food subsidies in developing countries with rationing. Parts (ii)-(iv) may be thought of as practically and socially important applications of the theoretical tools developed in the first part of the project.