Regime-Switching Framework For Asset Allocation
When interest rates near zero, traditional approaches to asset allocation fall short as the risk-reward trade-off for bonds breaks down.
Fixed-weight allocation, such as 60/40, does not take into account the fact that bond returns fall faster than risk as interest rates approach zero. Typical risk-based portfolio construction techniques, such as minimum variance, risk parity, maximum diversification, and the like, tend to produce an overallocation to fixed income when rates are low, thereby harming performance in periods of low and rising rates. Classical mean-variance optimization (MVO), elegant as it is, requires investors to enter expected returns, or views if a Black-Litterman approach is used, for assets in the portfolio. This can help reduce fixed income exposure as rates fall, but it introduces the risk of estimation errors for the other assets in the portfolio.
Using the ALPIMA platform, we built and tested a generic, risk-based asset allocation framework that adapts to low interest rates thanks to regime-switching using well-known indicators, such as the US yield curve.
Such a framework offers several advantages:
A historical analysis performed on the ALPIMA platform reveals that such a framework can effectively capture regime changes and perform better than the strategies on which it is based, as well as other multi-asset indices and benchmarks based on data going back to 2001.
Regime Switching and Variable Geometry
Regime switching is a rich topic with many parallels in industries other than finance. It is closely related to the notion of variable geometry, which has been widely used in aeronautics for decades, and, more broadly, in engineering for centuries.
“Swing wing” jets are a popular example of variable geometry. The wing of a commercial ariliner before landing is another good example. Depending on where the aircraft is on the glide path, one, two or three notches of flaps are applied, thereby changing the geometry of the wing as the ground nears. Mechanical engineering is filled with brilliant examples of variable geometry, from ancient castle gates to modern spaceships.
The idea consists of assigning different strategies, or geometries, to different regimes. From a mathematical perspective, it can be described as a one-to-one correspondence, as shown below.
Regime 1 <=> Geometry 1
Regime 2 <=> Geometry 2
Regime n <=> Geometry n
The effectiveness of a regime-switching approach depends first, on the accurate identification of different regimes, and, second, on the correct selection of a suitable geometry for each regime. This can be expressed as follows:
Effectiveness ≈ Identification × Suitability
In other words, an effective regime-switching strategy must, firstly, identify regimes well and, secondly, propose a suitable geometry for each regime.
In investment management, the notion of geometry is replaced with that of a particular investment strategy.
The parallels with the idea of variable geometry are fitting. After all, zero is a very significant boundary condition for interest rates, and, as it approaches, it produces its own form of “ground effect” on asset allocation.
Identifying regimes can be based on more or less complex logic based on directly observable market measures (security prices, P/E ratios, interest rates, curve steepness, credit spreads, implied volatility, skew, etc.), derived indicators (price volatility, moving averages, momentum, etc.), or any combination thereof. Some investment firms have been using the concept of “investment clock” for decades to determine what type of investments to recommend to their clients based on where we are in the macro-economic cycle. In recent years, data science has helped practitioners introduce new ways of identifying regime changes. It has recently been suggested that Hidden Markov Models (HMM), for example, can be effective at identifying market regimes when applied to financial time series such as US large-cap indices.
The choice of which strategy to apply to which regime is essential to the effectiveness of a regime-switching strategy. It consists of selecting strategies that are well suited to the different regimes being considered. For example, a low-risk investment strategy may be designed for periods of high volatility by an investment manager looking to create an all-weather low-risk product. Strategy choice must be validated with data, i.e. tested historically, and, if possible, using Monte Carlo simulations.
Examples - Two- and Four-Regime Switchers
Our study examines two examples, one based on a simple two-regime switcher, and another based on a four-regime switcher. Results are interesting and reveal that, when carefully thought out, systematic regime-switching can add measurable value to asset allocation on a risk-adjusted basis, producing higher returns, lower volatility, lower drawdowns than one-regime approaches. This should come as no surprise given that regime-switching is meant to make strategies more in tune with the environment they operate in.
The ALPIMA platform makes the design, customisation and deployment of robust regime-switching strategies remarkably fast and intuitive. Today, such strategies can be deployed as model portfolios, into managed accounts, or into funds or notes. Expect them to become increasingly available as digital tokens in DeFi apps, which, hopefully, will make them even more accessible in the future.
The key is to ensure they are well designed, based on sound principles and solid foundations.
Contact us at firstname.lastname@example.org to find out more.