Analogous Market Moments
How to use data to contextualize our current macroeconomic climate
By: Chris Satterthwaite, Lionel Smoler Schatz, and Oleg Laskov
Perhaps the most fundamental assumption of quantitative investing is that we can learn from history, that studying historic market data will teach us something about the future.
In recent years, we have been focused on researching how macroeconomic signals can help us predict expected returns across asset classes.
We are, essentially, reasoning by analogy. The challenge is to find the right analogues.
To do this, we created a measure of macroeconomic similarity that incorporates the full range of predictive economic signals that we rely on like high-yield spreads, inflation, stock-bond correlation, and the yield curve. We converted the economic data at each point into a vector (a vector is a multi-dimensional mathematical object that represents a list of data points in a specific order. Each vector is defined by its components, which can be used to uniquely identify its position in a multi-dimensional space). To measure how extreme each observation is, we calculate the distance between that vector and all historical vectors. This distance measurement is known as the Euclidean distance (or L2 norm).
To build intuition, here’s a chart illustrating how we would measure the distance between two hypothetical months, using only at interest rates, inflation, and GDP growth:
Figure 1: Illustration of Measuring Distance Between Two Months
Source: Verdad analysis
The smaller the Euclidean distance between two months of macro data, the more similar those moments are, and vice versa. We can see in the above chart that there are numerous points closer to our current month than the one we’ve drawn a line to, for example.
This exercise can be scaled beyond just three illustrative signals. Below we show a comparison, using Euclidean distance, of June 2024 to every month prior dating back to 1960, using our most significant macroeconomic signals.
Figure 2: June 2024 Market Moment Similarity
Source: Verdad analysis
Today’s market conditions are most similar, in reverse chronological order, to the following 8 periods: 2019, 2007, 2000, 1995, 1989, 1979, 1973, and 1969. These periods were generally defined by conditions that encourage risk taking: tight high-yield spreads that lead to high-risk borrowing, subdued volatility that encourages investors to lever up, increasing stock-bond correlation which makes bonds less useful as a diversifier, and a relatively inverted yield curve which means long-duration government bonds are less attractive.
The worrisome thing about these dates is that, of the 8 closest historically analogous periods, 4 preceded major market crashes within 12 months.
Figure 3: June 2024 Analogous Moments
Source: Verdad analysis, Capital IQ
A 50% hit rate for negative forward 12-month S&P 500 returns and a negative average return over all 8 analogues are impressive, considering the S&P 500 has averaged a 9% annual return from 1969 to 2024.
This type of analysis is far from dispositive: we would not say that this data point alone means we are currently on the precipice of the next great financial crisis. But we do think this provides a more data-driven way to think about the macroeconomic climate than narratives drawn from cable news—and, in this case, the data seems to tell quite a different story than the popular narrative.
Equipped with this approach and related ideas, we can cluster similar market moments into “regimes” and examine potential implications for asset class returns in each regime. We will explore this concept in further detail next week.
Acknowledgment: Our intern Oleg Laskov has been working on this research. He is a rising junior at Yale University majoring in applied mathematics. He is interested in quantitative finance, trading, and machine learning.