Research
Research
Asset pricing is based on the paradigm of rational expectations, which assumes that investors observe fundamentals and use them to forecast returns. However, investor herding onto mispriced stocks could reflect that investor appetite is driven by imitation and not only by macro and market fundamentals. I propose a measure that makes investors comparable, and I use it to estimate the influence that U.S. institutional investors have on each other's trades by using SEC 13F data from 1980-2023. I build an empirical framework that allows for heterogeneity across investors who select portfolios based on the lagged information of their competitors' holdings. I sort a large set of investors into 'bins' according to their portfolio size and historical correlation with the market. Findings show that the overlapping of assets is low and that investors influence each other, yet not in the same magnitude. Passive investors exercise influence over active investors, regardless of size.
(Presented at 2024 Financial Management Association Doctoral Consortium, Europe Chapter)
Machine learning can address high dimensionality issues in asset pricing as the list of predictor variables continues to grow. While research has focused on predictive performance, the discussion on hyperparameter tuning methods usually needs to be addressed. Using 94 firm-level characteristics of publicly traded companies in the U.S., I estimate machine learning regression models with different hyperparameter tuning methods in performing two tasks. First, predict U.S. monthly returns with data from 1980 to 2022 for non-micro caps and all stocks; second, identify which firm-level characteristics provide high explanatory power. Different methods of penalty selection for lasso (adaptive lasso, plugin estimator, and BIC criterion) make a marginal contribution to predictive performance with fewer covariates.
[The effect of monetary policy on institutional investor demand]