I am currently working on a variety of topics related to pension funds’ investments, corporate bond strategies, and ESG’s effect on asset markets. In one of the papers entitled ‘Choosing Investment Managers’ we study the determinants of pension funds’ hiring of external investment managers. We find that two factors play an influential role in selection: pre-hiring returns, and pre-existing personal connections between personnel at the plan (or consultant advising the plan), and the investment management firm. However, while relationships are conducive to asset gathering by investment managers, they do not appear to generate commensurate benefits for plan sponsors via higher gross returns or lower fees.
In another project, we investigate return predictability across stocks and bonds with Machine Learning and Big Data. We believe that because of the nonlinear payoffs of corporate bonds and the high correlation between many of the stock and bond characteristics, machine learning approaches are well suited for such challenging prediction problems by mitigating overfitting biases and uncovering complex patterns and hidden relationships. We find that machine learning models substantially improve the out-of-sample performance of stock and bond characteristics in predicting future stock and bond returns.
We are heartened by the fact that, since the publication of our 2008 paper, researchers pay more attention to the out-of-sample predictability of equity premium. Almost all published papers now attempt to cross the dual hurdles of in- and out-of-sample predictability. Despite these advances, I remain sceptical for at least two reasons. One, researchers still have a lot of latitude in (perhaps inadvertent) experimental choices that may influence the results. Two, statistical predictability does not always translate into economic predictability. In my mind, therefore, the jury on equity premium predictability is still out.
This class was designed to be the first class on empirical asset pricing. I covered the basics of asset pricing tests as well as the vast literature of anomalies (mostly in stocks but also in corporate bonds and options). We also took some time to discuss the latest in multiple hypothesis testing as well as machine learning, and big data applied to finance. The class was an econometrics class. Rather the goal was to see the interplay between theory and data; how well do theoretical models work in the data and/or how the empirical patterns in the data can lead to development of richer theoretical models.
There is an obvious difference between professionals and full-time students in a PhD programme. The advantage that professionals have is that their work experience has exposed them to real-life problems. Thus, they typically are not short of research ideas. On the downside, some (but not all) of the professionals need a bit more time to retool and refresh their academic knowledge which might have been acquired a while ago.
My only advice is to collect a list of topics which the students find interesting. Then discuss this list with your advisor to narrow it down to two maximum. In other words, I would refrain from choosing topics based on what is hot and/or what will be deemed to be hot in a few years. Based on my own research, market timing does not usually work!