Gulamabbas Lakha,CFA, Chief Executive Officer, Providentia Analytics
April 2015
A variety of investors currently face a common set of challenges for building robust and well diversified portfolios, due to changes in uncertainty (volatility of volatility), stretched relationships between markets (volatility of correlations) and unusual behaviours between risks (correlation of volatilities). All of this against the backdrop of increasing regulatory requirements, including the AIFMD and UCITS. It appears that in the midst of this hurricane of activity and potential confusion, investment managers need to stand in the relative stillness of the eye of the storm and methodically distil vast amounts of data into useful specific information, resulting in implementable strategies for alpha generation within a risk framework that responds to the changing demands upon it over time. This central objective is seductively intuitive yet all too often illusive.
Modelling dynamic relationships
Over the last decade the team at Providentia Analytics have developed various tools to analyse the ever changing economic and market landscape, so as to empower portfolio managers. They can be considered various types of lenses to view the market. One of the tools, named ‘Correlation Elasticity’, can examine any relationship between economic or market variables, treating it like a piece of elastic; asking (i) how flexible is it and (ii) how stretched is it currently? For instance, the charts below relate to the quarterly correlation between US bonds and equities over time. The histogram on the right provides an indication of the distribution of that correlation, showing first that the relationship is quite flexible, ranging from highly correlated to highly uncorrelated; and second that at that point in time it was rather stretched (as indicated by the dark blue bar on the extreme left).
It is important to stress that mean reversion assumptions should not be made, since dislocations can persist or indeed extend. Rather, the role of fundamental judgement, empowered by such quantitative information, can be harnessed for structuring better informed trades for alpha generation. In addition, such tools can be used for dynamic stress tests regarding exposure to commodity, currency, interest rate or other macro risks (also useful for regulatory reporting).
Using ‘big data’ to identify dislocations
By harnessing powerful computational analytics, the exercise above can be undertaken for thousands of relationships, between equities, fixed income, currencies, commodities, funds, or any aspects of a given investment universe. Such an approach can identify alpha opportunities from dislocated relationships, where instruments may be behaving in a counter-intuitive way, resulting in much lower correlations relative to history. Furthermore, this lens can be focused on the high end of correlation distributions to provide an early warning system for concentration risks, style drift, or declining quality of diversification.
The following image of what we call a ‘Diversification Matrix’ shows the results of examining all the cross relationships of an institutional European equities portfolio with just under sixty holdings (1653 in total). The results enable the manager to focus attention on potential concentration risks from a few specific holdings.
Focusing attention on changing influential factors
From a strategy perspective, an ‘Influence Matrix’ can help monitor changes in the importance of economic, market or risk factors on a set of target investments or sectors. The following chart highlights which specific factors are influencing different equity sectors, by providing the historic percentile of current correlations, thereby giving a historic context to changing influences. From an accessibility perspective, this tool is potentially more intuitive than multi-factor regression models with static significance tests. Instead, with this approach managers have a wider picture to determine for themselves which factors are currently influencing the instruments they wish to invest in.
A ‘quality of diversification’ approach to portfolio construction
Utilising the above methods to systemically identify dislocated relationships, for both alpha generation and concentration risks, can empower a manager to distill ‘big data’ into specific information and distinguish between effective versus perceived diversification. The result is usually more robust portfolio construction.
