Macro Research
SUMMARY
We endeavored to develop a robust, data-driven approach for identifying and classifying inflation regimes to guide future determinations of when to position portfolios appropriately to hedge against the damaging effects of inflation.
The recent surge in inflationary pressures has become a leading concern of central banks, investment managers, and households alike. The U.S. economy experienced nearly three decades of sustained low inflation prior to 2020. The Covid-19 pandemic resulted in significant growth in the Federal Reserve’s balance sheet, massive fiscal deficit spending, and a breakdown in global supply chains. These events have led to a rapid and sustained increase in inflationary pressures which in turn have widespread effects on the real economy and financial markets.
Exploring solutions
A recent study by the Man Group (Neville, Draaisma, Funnell, Harvey, Van Hemert. “The Best Strategies for Inflationary Times.” (August 2021) provided a 95-year review of inflation’s impact on asset prices. The study sought to identify asset classes and factor exposures that performed well during historical periods of high inflation.
One key finding by the Man Group study that resonates in today’s market environment is the beneficial role commodities have played historically as an inflation hedge. The current inflationary regime has exhibited commodity price increases consistent with past periods, providing inflation protection to portfolios.
We endeavored to develop a robust, data-driven approach for identifying and classifying inflation regimes to guide future determinations of when to position portfolios appropriately to hedge against the damaging effects of inflation.
Our approach to measuring inflation regimes differs from others in that we focused on a purely data-driven approach. Financial and macroeconomic conditions often change behavior over extended periods of time creating regimes. Having a set of tools to model changes in behavior can help both improve asset allocation and risk management.
There are multiple approaches one could take to identify regimes. Expert judgment and historical knowledge, studying business-cycles, and time-series analysis can all help to determine states of the economy or market regimes. These approaches all require a priori assumptions.
"One accurate measurement is worth a thousand expert opinions." - Grace Hopper | computer science pioneer
In contrast, we implemented a machine learning approach that lets the historical data itself determine where regimes occur. We utilized a set of unsupervised learning methods which can discover hidden patterns in the data without asserting any initial assumptions such as the current state of financial markets or inflation’s relationship to other macroeconomic forces.
Specifically, we used a combination of K-means Clustering, Gaussian Mixed Methods, and Dynamic Time Warping. These models are especially robust at identifying clusters within a data series that share similar behavior, thus allowing us to identify regimes. The initial analysis was performed on the headline consumer price index reported by the Bureau of Labor Statistics.
The machine learning algorithm we implemented iterates to determine an optimal number of regimes it will use for classification. Four unique inflation regimes were identified in U.S. headline CPI. We classified those regimes as: low, medium, high, and extreme. Our research identified that the high and extreme regimes were periods where utilizing a commodity index as a portfolio hedge was most useful. All identified high/extreme regimes are listed below along with the cumulative return from holding an equal-weighted commodity basket over the identified periods.
What does all of this mean from an investment perspective? Sustained periods of high inflation are rare over the past 60 years, but when they have occurred, traditional 60-40 portfolios have underperformed as have many other equity and bond strategies. Commodities have tended to outperform during these periods and the recent Man Group study illustrated the effectiveness of adding a broad basket of commodities, including energy, metals, and agriculture to an existing portfolio as an inflation hedge. More importantly, knowing when to reallocate a portfolio or add a hedge is critical for risk management.
To examine the potential impact of commodities in a portfolio, we simulated a 60% equity/40% bond portfolio against a 60%/30%/10% portfolio where 10% is allocated to a diversified commodity index. The simulation ran from April 2020 through May 2022 which corresponds to the most recently identified inflationary regime by the machine learning algorithm. The inclusion of commodities improved the portfolio returns and Sharpe ratio, while decreasing volatility and drawdowns.
Stategy Capital’s approach to classifying inflation regimes is a move away from traditional modeling approaches and towards more data-driven, machine learning methods. We understand that periods of sustained high inflation in the U.S. have been rare and therefore we do not have many historical periods to study, but we understand the critical need for tactical portfolio allocation and risk management during these uncertain periods.
We believe in continual learning and, as always, strive to apply responsible data science and subject-matter expertise to tough problems.
Research Insights
