Get Started with Scientific Portfolio
Self-register today or contact us to see how it strengthens your investment process.
Macroeconomic Regimes for Conditional Simulations of Equity PortfoliosWhitepaper | February 2025
Abstract
Changing macroeconomic conditions have the potential to strongly influence equity portfolio returns. This paper examines how key macroeconomic regimes affect the statistical characteristics of equity returns. We find that across a limited number of regimes, equities exhibit stable and well-defined properties, in- and out-of-sample. These findings are critical for investors wishing to incorporate their macroeconomic views in their investment decisions; they also facilitate reliable portfolio simulations and out-of-sample projections. Furthermore, we demonstrate that long-term factor models provide robust insights into portfolio behaviour within different macroeconomic contexts, even for portfolios with limited historical data.
Key takeaways:
- How macroeconomic shifts influence portfolio risk and return profiles
- Using factor models to estimate portfolio behavior across economic cycles
- The role of conditional simulations in improving investment decision-making
Introduction
Equity markets are sensitive to shifts in macroeconomic conditions, prompting investors to form views on forthcoming regimes and adjust their strategies accordingly. As changes in the economic environment influence firms’ cash flows and cost of capital, they fundamentally shape investors’ opportunity sets (Flannery and Protopapadakis 2002). However, given the short track records typical of equity portfolios, it can be difficult for investors to gauge how these portfolios might behave during rare, cyclical economic phases. To address this critical lack of information, it is often necessary to resort to simulations, which provide a spectrum of plausible alternative scenarios for historical returns to probe risks and opportunities in varied market environments beyond those revealed from the past. As such, simulations constitute a key input to the investment decision-making process. Unfortunately, producing reliable expectations about a portfolio’s behaviour in different macroeconomic regimes is challenging for a variety of reasons. For one, the extreme nature of events, such as financial crises, makes them rare, thereby complicating statistical estimation. Furthermore, relevant risk factors to which a portfolio is exposed are not independent, hence it is necessary to account for changes in their joint behaviour across different macroeconomic states. In this paper, we identify key differences in equity portfolios across macroeconomic environments and show how to efficiently estimate regime-dependent parameters with long-term risk factor models.
Simulations provide a way to access information that is not confined to the historical track record. To do so, they require two foundational elements: an accurate statistical distribution capturing a portfolio’s return properties, and a suitable sample to calibrate these distributions. While suitable distributions to simulate equity returns have already been proposed (e.g. Bouchaud and Potters 2003; Jondeau, Poon, and Rockinger 2007), selecting appropriate samples for calibration—particularly within distinct macroeconomic regimes poses a greater challenge. In particular, return distributions are not strictly stationary, i.e. their distribution is time dependent.1 In practice, however, simulations often rely on static distributional assumptions, which by construction reproduce long-term return properties, but fail to capture the full spectrum of short-term return variability. For instance, the average (annualised) 10-year performance of the US stock market between 1984 and 2024 is 11%. However, depending on the time of measurement, this performance varies between -5% (around 2009) to 20% (at the beginning of the 2000s).
One of the most common methods for modelling dynamic return behaviours is the use of Markovian processes, which allow the parameters of a distribution to change according to transition probabilities. However, while theoretically sound, this approach faces several practical issues that reduce its appeal for simulations aimed at supporting investment decisions: transition probabilities are difficult to link to a clear macro-economic context (Blitz and van Vliet 2011), and the number of states considered must be limited both to preserve interpretability and to avoid the so-called curse of dimensionality (Bellman 1966), which in this case denotes the issue of the state grid growing exponentially with the addition of further state variables, often resulting in computational infeasibility.
In this paper, we develop an approach that addresses these issues by focusing on the selection of economically meaningful sample periods that exhibit stable, but significantly different, statistical properties compared to long-term returns. Instead of modelling transitions between distinct states, we identify samples that reflect macroeconomic conditions with sufficient stability for effective parameter calibration. This enables us to retain economic relevance and enhance the reliability of model parameters without relying on traditional state-based simulations.
Our contribution to the literature is three-fold. First, we identify distinct macroeconomic scenarios that yield statistically different return distributions compared to long-term averages. Within-regime expected returns, and especially volatility, diverge significantly from ‘normal times’, highlighting the importance of adjusting expectations according to the specific macroeconomic regime.
Second, we find that certain regimes exhibit stable and robust statistical properties, with return distributions that remain consistent both in- and out-of-sample. This insight is valuable for investors wishing to align their portfolio’s performance expectations with their views on the forthcoming regime (Elkamhi, Lee, and Salerno 2023): simulating returns using regime-dependent parameters, i.e. performing conditional simulations, leads to more reliable outcomes than simulations based on long-term information only. This approach is close in spirit to (Hoevenaars et al. 2014), and (Bekkers, Doeswijk, and Lam 2009) who suggest deriving expected returns from a combination of long-term historical data, economic theory and current market circumstances for strategic asset allocation.
Third, we address the practical limitation of short historical records that hinder a sound statistical assessment of a portfolio’s behaviour within specific regimes. To sidestep the potential lack of data spanning multiple economic cycles, we show that for equity portfolios, linear factor models can efficiently extrapolate returns with reasonable accuracy, even with as few as five years of information.
This paper contributes to a rich body of literature exploring the impact of macroeconomic conditions on equity returns and risk premia, which supports the common practice of using macroeconomic variables to define market regimes. Since macroeconomic changes impact investment opportunities (Flannery and Protopapadakis 2002), they represent undiversifiable risk factors (Ross 1976) and thus should be priced in equilibrium (Merton 1973; Breeden 1979). While evidence on real-sector aggregates is more nuanced aggregates (Chen, Roll, and Ross 1986), macroeconomic conditions are widely recognized for their role in driving returns. For instance, inflation and money growth have shown a negative impact on market returns (Bodie 1976; Fama 1981), while industrial production, consumption, and labour income yield positive abnormal returns (Lamont 2001). There is also a vast literature that documents that returns and factor models behave very differently on days with announcements regarding e.g. inflation, unemployment and interest rates, such as in (Savor and Wilson 2014; 2013; Lucca and Moench 2015; Brusa, Savor, and Wilson 2020; Cujean and Jaeger 2023).These studies further confirm the ideas that macro-based regimes, regardless how they are defined, have a strong influence on market outcomes. Macroeconomic variables are deeply intertwined with equity markets and frequently serve as market predictors (Welch and Goyal 2008), Goyal et al. 2024). In the context of equity portfolios, studies by (Amenc et al. 2019) and (Esakia and Goltz 2023a) propose protocols to identify macroeconomic variables and market regimes significantly influencing equity risk factor returns. If macroeconomic aggregates affect risk premia, they likely influence other statistical properties, including volatility. Supporting this, research shows that macroeconomic conditions impact return volatility and distributions across regimes (Hamilton and Susmel 1994; Sinha 1996), with similar evidence documented for European markets (Errunza et al. 1994). Building on this literature, our study provides effective methods to reliably estimate equity returns’ behaviour across distinct macroeconomic regimes. Our approach, described in detail in the following sections, reduces dependence on historical data, captures multidimensional risk interactions and allows for a broader spectrum of outcomes that can be later employed for forward looking portfolio management through regime-dependent simulations.
Authors
Matteo Bagnara, PhD
Quant Researcher,
Scientific Portfolio ……………………………………….
…
….
Shahyar Safaee
Deputy CEO and Business Development Director,
Scientific Portfolio
Deputy CEO and Business Development Director,
Scientific Portfolio
Read the Whitepaper
"*" indicates required fields
