Research projects

Project CEMAPRE internal

TitleTail index regression model and Quantile Order Regression
ParticipantsJoão Nicolau (Principal Investigator)
Summary(1) In extreme value statistics, the tail index is an important measure to gauge the heavy-tailed
behavior of a distribution. Our goal is to link the tail index to the linear predictor induced by
covariates, which constitutes the tail index regression model. This model may have interesting
applications in finance and wages, income, etc. In particular, we want to use it in connection to
Current Population Survey (CPS). The CPS is, however, subject to an important limitation: the data
are right censored ('topcoded'). Topcoded data cause problems for inequality analysis because they
censor the range of incomes that are observed. Inequality is underestimated because very high
incomes appear as less high incomes. Also, due to the complex sampling design for the CPS, we need
to account for the "weights" to produce representative statistics.
(2) We also want to focus on a new models that we call Quantile Order Regression (QOR). Standard
linear regression models allows us to estimate the impact of an explanatory variable on the
dependent variables. In contrast, the QOR approach allows us to estimate the variation in the order
of the quantile of the dependent variable. The QOR model is therefore suitable for applications to
income, health and other types of inequalities, and more generally, the factors that explain the
relative position of individuals, households, firms, cities, states, countries, etc., with respect
to the dependent variable.