Economic applications of nonparametric methods

This thesis deals with the subject of nonparametric methods, focusing on
application to economic issues.
Chapter 2 introduces the basic nonparametric methods underlying the
applications in the subsequent chapters.
In Chapter 3 we propose some basic standards to improve the use and
reporting of nonparametric methods in the statistics and economics literature
for the purpose of accuracy and reproducibility. We make recommendations
on four aspects of the application of nonparametric methods: computational
practice, published reporting, numerical accuracy, and visualization.
In Chapter 4 we investigate the effect of life-cycle factors and other demographic
characteristics on income inequality in the UK. Two conditional
inequality measures are derived from estimating the cumulative distribution
function of household income, conditional upon a broad set of explanatory
variables. Estimation of the distribution is carried out using a semiparametric
approach. The proposed inequality estimators are easily interpretable
and are shown to be consistent. Our results indicate the importance of interfamily
differences in the analysis of income distribution. In addition, our
estimation procedure uncovers higher-order properties of the income distribution
and non-linearities of its moments that cannot be captured by means
of a "standard" parametric approach. Several features of the conditional
distribution of income are highlighted.
Chapter 5 we reexamine the relationship between openness to trade and
the environment, controlling for economic development, in order to identify
the presence of multiple regimes in the cross-country pollution-economic
relationship. We first identify the presence of multiple regimes by using
specification tests which entertain a single regime model as the null hypothesis.
Then we develop an easily interpretable measure, based on an original
application of the Blinder-Oaxaca decomposition, of the impact on the environment due to differences in regimes. Finally we apply a nonparametric
recursive partitioning algorithm to endogenously identify various regimes.
Our conclusions are threefold. First, we reject the null hypothesis that all
countries obey a common linear model. Second, we find that quantitatively
regime differences can have a significant impact. Thirdly, by using regression
tree analysis we find subsets of countries which appear to possess very
different environmental/economic relationships.
In Chapter 6 investigate the existence of the so called environmental
kuznets curve (EKC), the inverted-U shaped relationship between income and
pollution, using nonparametric regression and a threshold regression methods.
We find support for threshold models that lead to different reduced-form
relationships between environmental quality and economic activity when
early stages of economic growth are contrasted with later stages, There
is no evidence of a common inverted U-shaped environment/economy relationship
that all country follow as they grow. We also find that changes that
might benefit the environment occur at much higher levels of income than
those implied by standard models. Our findings support models in which
improvements are a consequence of the deliberate introduction of policies
addressing environmental concerns. Moreover, we find evidence that countries
with low-income levels have a far greater variability in emissions per
capita than high-income countries. This has the implication that it may be
more difficult to predict emission levels for low-income countries approaching
the turning point.
A summary of the main findings and further research directions are presented
in Chapter 7 and in Chapter 8, respectively.