Efficient Minimum Distance Estimation of Pareto Exponent from Top Income Shares

In 2018, Yulong Wang, who is an econometrician with expertise in extreme value theory, gave a talk at the UCSD econometrics seminar. Later he invited me to give a talk at Syracuse University. When we were talking about research during dinner, I mentioned to him that it would be cool to estimate income Pareto exponents from top income share data (which I didn’t know how to do properly), and after briefly explaining the problem, he said that’s easy because top income shares are essentially sums of extreme order statistics, and the normalization technique in Müller & Wang (2017) can be applied. So we decided to write a paper together.

First Yulong developed the theory of weighted sums of order statistics in Section 2. I wanted to study econometrics, so I asked him for references and worked out the asymptotic theory of the minimum distance estimator in Section 3. It was a lot of fun.

Later, I learned that the top income share data that we used (which we took from Piketty & Saez) had some issues, so at some point I might write a follow up paper that uses the original IRS data.