Essays on the econometric analysis of welfare

This thesis explores several econometric approaches to making empirical welfare comparisons. It is divided into three separate, but closely related chapters. In the first chapter, we propose empirical likelihood-based inference for decomposable (additively separable) poverty measures which utilize relative poverty lines (i.e., poverty lines which are some fraction of either the mean or the median of the underlying income distribution). The primary advantage of using the empirical likelihood framework in this setting is that no estimates of scale parameters (i.e., standard errors) are required. Since estimates of poverty measures which utilize relative poverty lines involve a nuisance parameter (the poverty line), the asymptotic variance of these estimates are quite complex, as they depend on the underlying density function. Simulation evidence provided here suggests that the proposed methods may offer some improved performance over both asymptotic approximations and the bootstrap-' T' procedure. These methods are illustrated with an empirical example using Canadian household survey data on income. In the second chapter, we consider methods of statistical inference for vectors of inequality and poverty measures. The use of vector measures recognizes the fact that there is often no single measure of either inequality or poverty which is completely satisfactory to researchers. For example, there is sometimes no clear choice of the "best" (scalar) measure of inequality or poverty. Alternatively, when measuring poverty, there may be some disagreement over which poverty line should be used. The uses of vector measures also allows us to approach the issues of multidimensional inequality and poverty. That is, we may consider vectors of measures which represent inequality and/or poverty in different dimensions (e.g., income, wealth, health, educational attainment, etc.). Specifically, we propose hypothesis tests for such measures using a general framework for testing inequality constraints. Our proposed method has the advantage of allowing us to obtain unambiguous welfare orderings between two different populations. We present some simulation evidence which suggests that such tests have good size and power properties. Finally, this approach is illustrated with an empirical example using Canadian household survey data on income and educational attainment. In the final chapter of this thesis, we turn our attention from specific measures of welfare towards the more general approach of stochastic dominance. Specifically, we propose tests of bivariate stochastic dominance using a generalized framework for testing inequality constraints. Our proposed methods are illustrated with some Canadian household survey data on income and educational attainment.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.