The Computational and Financial Econometrics conference was just held in London. Here are three talks from the large menu.
Lars Helge Hass
The objective is to find a way to do an asset allocation optimization that includes private equity. A problem of course is that private equity is seriously opaque. To highlight that, using one private equity index in the optimization led to essentially no allocation to private equity while using a different index led to consistently hitting the upper bound of allocation to private equity.
Lars and his co-authors found a way of combining sources of information to get more plausible results.
The paper on SSRN is “Private Equity Benchmarks and Portfolio Optimization”.
The question is about bubbles in equity and property markets, and possible connections of the bubbles in the different markets.
The methodology is a regime switching model. Now, when I hear “regime switching”, my eyes roll with probability almost 1. I think in order to fit such a model, two conditions should hold:
- multiple regimes should be plausible
- estimating the regimes should be feasible
An example where I doubt the first condition: variance. There are indeed high volatility periods and low volatility periods, but I’m not convinced that two high volatility periods are necessarily much alike.
Even when regimes are plausible, it often seems thoroughly optimistic to think we have the ability to identify the regimes.
However, the present case seems to me — at first glance, at least — to be a solid use of a regime switching model. The bubbles are quite explicit (though I’m sure there are arguments about the fundamental value) so estimation should be reasonably easy.
The working paper is “Housing and equity bubbles: Are they contagious to REITs?”
Valeriy Zakamulin (a.k.a. Zakamouline)
A topic of several recent blog posts here has been that volatility from one day gives bigger estimates than from 20 days.
Valeriy is investigating the same thing, except he is looking at years rather than days. He finds what seems to be mean reversion on the scale of decades.
The SSRN version of the paper is “Very Long-Term Mean Reversion and Predictability of the U.S. Stock Market Returns”.