What dislocation of correlations? – invest.ch

“Simply Put”, the weekly column from the Multi Asset Group team at Lombard Odier Investment Managers

By Florian Ielpo, Head of Macro, Joshua Voelkel, Investment Analyst and Julien Royer, Quantitative Analyst

At certain points in every investment cycle, we as market observers begin to notice phrases like “the correlations are breaking down.” This observation is crucial because the structure of most of our portfolios is based on the complex concept of diversification. Ideally, correlations should be concentrated enough so as not to be disrupted by ambient market noise, but not too concentrated so as to reap the benefits of diversification across and within asset classes. There has been a lot of talk lately about this phenomenon of correlation breakdown, which prompted us to study the issue for ourselves, from an econometrics perspective. In this week’s “At a Glance” edition, we examine the current state of correlations based on eigenvalue and clustering analysis. These two solid methods allow the data to speak for itself: let’s listen to what it has to say to investors.

Equity values ​​and diversification

To analyze the fluctuation of correlations, a first approach is to examine the eigenvalues ​​of the correlation matrix of asset returns. Although the term “eigenvalue” may seem complex, the concept is closely related to the principle of “principal component analysis” (PCA), a technique used in data analysis. The eigenvalues ​​of a correlation matrix are used to measure information about each common factor that explains the cross-section of returns. For example, if the market factor predominantly explains variations in returns, the eigenvalue associated with this factor will significantly overshadow the others. These “information measures” indicate that the higher the eigenvalue, the more information the corresponding factor contains. There Figure 1 illustrates the explanatory power of the first factors for a wide range of assets, classified by asset class and then aggregated. In the long term, and from the perspective of the current situation, it is evident that the correlations are indeed breaking down. Typically, the first few crossed factors account for 95% of all correlations; However, this figure has recently fallen to 86%, with an uptick during the market’s brief correction. Where is this rupture most apparent? The chart on the right provides the answer: mainly in the area of ​​stocks. Different regions and sectors show that investors’ attention has recently shifted from the market factor to other, as yet unknown, areas. At the same time, the bond world continues to experience duration stress, while credit spreads have remained stable for 18 months.

From 2 to 12 clusters

Another way to understand the fluctuation of correlations is to use clustering methods. Econometrics offers a wide range of these methods, all designed with the common goal of grouping similar objects together based on a defined distance measure. This allows users to analyze how these similarities change over time. The connection with the previously discussed approach using eigenvalues ​​is clear: when the first few market factors explain a significant part of the dynamics of asset returns, large clusters can be formed. Conversely, when early eigenvalues ​​have lower explanatory power for these common market factors, the number of clusters increases because assets begin to look less similar. There Figure 2 presents a heatmap that displays the correlation level between different assets along with the corresponding clusters, such that assets with high correlation are placed close to each other, while assets with low correlation are placed far from each other. Over the long term, similar to Figure 1, the number of clusters is generally limited to two main types: risky assets and hedging assets. However, the more recent data in the bottom chart reveals a dramatic change: the number of clusters has increased significantly. In particular, distinct clusters are now forming around specific assets such as the Hang Seng Index, Japanese bonds or the energy sector, providing insight into the factors that explain past fluctuations observed in values. clean. The current number of clusters found in the correlation matrix analysis has increased from 2 to 12 – a significant increase.

Simply put, correlations are breaking down as idiosyncratic stories have recently gained prominence.

Macro/Nowcasting Corner

This section brings together the most recent developments in our nowcasting indicators for global growth, global inflation surprises and global monetary policy surprises. These indicators make it possible to monitor the most recent macroeconomic developments that move the markets.

Show article disclaimer