Schur Complementary Allocation: A Unification of Hierarchical Risk Parity and Minimum Variance Portfolios

Peter Cotton

2024-10-30

Abstract:Despite many attempts to make optimization-based portfolio construction in the spirit of Markowitz robust and approachable, it is far from universally adopted. Meanwhile, the collection of more heuristic divide-and-conquer approaches was revitalized by Lopez de Prado where Hierarchical Risk Parity (HRP) was introduced. This paper reveals the hidden connection between these seemingly disparate approaches.

Portfolio Management,Mathematical Finance

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is how to bridge the gap between optimization - based methods (such as Modern Portfolio Theory, MPT) and divide - and - conquer heuristic methods (such as Hierarchical Risk Parity, HRP) in portfolio construction. Specifically:

1. **Theoretical Background**:
- **Optimization - based methods**: Modern Portfolio Theory (MPT), represented by Markowitz (1952), constructs portfolios through optimization techniques. However, this method faces many challenges in practical applications, such as being sensitive to covariance matrix estimation errors and having high computational complexity.
- **Divide - and - conquer heuristic methods**: The Hierarchical Risk Parity (HRP) method proposed by De Prado (2016) constructs portfolios by recursively partitioning and recombining assets, avoiding the problems brought by direct optimization, but in some cases, covariance information may be lost.

2. **Core Contributions of the Paper**:
- **Revealing Mathematical Connections**: The paper reveals the hidden mathematical connections between these two seemingly different methods. In particular, by introducing the Schur Complement to modify the sub - covariance matrix, the hierarchical method can be made closer to the Minimum Variance Portfolio (MVP).
- **Proposing a New Hierarchical Minimum Variance (HMV) Method**: By adjusting the sub - covariance matrices \(A\) and \(D\), the paper proposes a new Hierarchical Minimum Variance (HMV) method. This method can better utilize the information in the covariance matrix while maintaining the hierarchical structure.

3. **Specific Implementation**:
- **Formula Representation**: The new formula for allocation weights is
\[
w(\Sigma) \propto \begin{cases}
\left( \frac{1}{\nu(A')} w(A''), \frac{1}{\nu(D')} w(D'') \right)^T & \text{if } \dim(\Sigma) > m \\
w_{\text{term}}(\Sigma) & \text{otherwise}
\end{cases}
\]
where \(A'\) and \(A''\) are modifications of the original sub - covariance matrix \(A\), specifically:
\[
A'' = A - BD^{-1}C
\]
\[
b_A = \mathbf{1} - AC^{-1}\mathbf{1}
\]
\[
A' = (A_c) * b_A
\]
Here, \(A_c\) represents the Schur Complement, that is, \(A_c = A - BD^{-1}C\), and \(b_A\) is a vector.

4. **Experimental Verification**:
- The paper verifies the effectiveness of the HMV method through simulation experiments. Especially in a low - information environment, the HMV method can significantly reduce the variance of the portfolio and is superior to the traditional HRP method and simple equal - weight allocation.

In conclusion, this paper aims to enhance the hierarchical portfolio construction method by introducing the Schur Complement, enabling it to better utilize the information in the covariance matrix while remaining simple and easy to use, thus providing a practical intermediate solution between direct optimization and heuristic methods.