A Data-Driven Categorization of Investable Assets

Abstract

An important objective of asset allocation strategies is the construction of diversified portfolios. In order to approach their target, they strongly depend on the classification of assets. Conventionally, investment decisions are taken on the basis of a partition of the investable universe into asset classes, e.g. cash, equity, bonds, real assets, and alternative investments (hedge funds and private equity). Over the last years this framework has been under revision, and the industry has begun to think more in terms of risk factors or risk classes. This paper aims to contribute to this discussion by presenting a data-driven categorization of investable assets. The methodology used is hierarchical cluster analysis, a data mining technique, which allows grouping objects into categories by similarity. We show that the separation of equities and bonds is reflected in the data-driven categorization, while more complex investments and in particular alternative investments are exposed to a range of diverse risk factors.

TOPICS: Portfolio construction, analysis of individual factors/risk premia, big data/machine learning