Interactive parametric portfolio policies

: does complexity reward?

Abstract

I find that extending the breadth of a simple linear asset allocation model, by increasing the number of characteristics analyzed and/or including two-way interaction terms, leads to better performing portfolios in terms of risk-adjusted returns and active-portfolio management suitability. Of the characteristics analyzed, Size, Value, Momentum and Bid-Ask Spread are the main ones driving cross-sectional returns, with 5Y Jensen’s Alpha and Standard Deviation playing a minor role. Moreover, I explore the effect of using interaction terms built on standardized and unstandardized characteristics and find that these different formulations impact on the portfolios’ results with an In-Sample statistically significant difference in performance. Furthermore, both formulations result in statistically meaningful coefficients, with some exceptions. The results obtained are satisfactory enough Out-of-Sample when looking at the Sharpe Ratio of the optimized portfolios, with an average loss of 17.28percentage points and a maximum loss of 23.15 percentage points with respect to the models’In-Sample performances; most of them outperform the Baseline PPP portfolio in this setting, yet the difference is not statistically significant. Looking instead at the Information Ratio, none of the optimized portfolio yield a higher Information Ratio than the Baseline PPP portfolio, although some come close. Nevertheless, the difference in performance between models using standardized and unstandardized interaction terms is still significant Out-of Sample: I find that standardized models outperform unstandardized ones.

Date of Award	28 Jun 2023
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Pedro Barroso (Supervisor)