| Abstract: |
This thesis addresses welfare measurement issues, with an emphasis on the
measurement of happiness and inequality. It contributes to the economic
literature in both methodological and empirical terms, with the empirical
analysis employing the PACO/CHER, ECHP and GSS datasets. Although human
welfare is a multidimensional concept, a classical approach is to simply
investigate the distribution of wealth and/or income. Our first chapter
analyses income distribution in Poland, using comprehensive data from the year
2000. We use the concept of stochastic dominance to investigate the extent to
which the income of certain subgroups (based largely on combinations of
gender, education, and region) unambiguously exceeds that of others, and
examine and formally assess hypotheses of stochastic dominance using recently
developed statistical tests. The results of this approach are contrasted with
simple scalar measures of inequality that are conventionally used. We find
that males, the higher educated and those living in the urban areas are better
off while the regional dominance relationship are difficult to establish.
However, to a large extent human welfare draws on subjective feelings of
happiness or similar subjective well-being concepts. While self-assessments of
well-being can be elicited, the relation of such expressions to the underlying
concept is intrinsically problematic. Consequently, in our second and third
chapters we present a semiparametric framework that allows for the modeling of
latent variables. This item response theory methodology is first applied to
assess the differences in \happiness"across selected European states. A more
detailed analysis suggests that the genesis of happiness is affected by
relative social status; income is more important to high status individuals
for example. The third chapter concerns further challenges in happiness
measurement in the presence of framing effects and/or differential item
functioning (\DIF"). The impact of the ordering of questions on subjective
well-being responses is studied under an extended item response theory model
incorporating the DIF feature of the survey. Contrary to previous studies, the
results indicate that individuals' happiness estimates are largely unbiased
when the framing experiment is ignored. The methodology we develop allows for
the assessment of framing and DIF effects and permits inter-subject comparison
and analysis even when such effects are large. |