Introduction
Traditional financial models often struggle to explain the persistent cycles of outperformance and decline observed in investment styles like Value and Growth. While macroeconomic factors are frequently cited, this research argues that endogenous interactions—the way strategies influence market prices and each other—are equally critical. The authors propose Evology, a novel agent-based model (ABM) that conceptualizes the US mutual fund market as a biological ecosystem where trading strategies compete for capital like species in a niche.
The Evology Framework
Evology is designed as a heterogeneous ABM where strategies (Value Investors, Noise Traders, and Trend Followers) are treated as biological species. These agents exploit market inefficiencies and compete for "survival," defined by their ability to attract and retain capital.
Unlike many ABMs that rely on discrete orders, Evology utilizes a continuous excess demand function smoothed by a tanh function. This technical choice allows for realistic simulations of leverage, short-selling, and margin calls while ensuring that orders remain within budget constraints. The simulation operates in a closed loop, as illustrated in the model's operational flow:
Empirical Calibration and Datasets
To ensure the model reflects real-world dynamics, the researchers calibrated Evology using high-fidelity financial datasets:
- ICI (Investment Company Institute) 2022 Fact Book: Provided aggregate data on total net assets (TNA) and the number of funds across Growth, Value, and Blend styles.
- SEC Form N-SAR Filings: Offered fund-level data on assets, redemptions, and sales from 1994 to 2015.
- CRSP (Center for Research in Security Prices): Used to correlate fund returns with SEC data.
Through linear regression, the authors quantified "return-chasing" behavior. They found that a 1% excess return over a 10-year horizon increases a fund's TNA by approximately 0.53%, a key parameter for the model's investment flow mechanism.
Market Ecology and Strategy Competition
A central insight of the Evology framework is density dependence or "crowding." As a strategy's "biomass" (the wealth following it) grows, it reaches the carrying capacity of its niche, leading to diminishing returns. This creates predator-prey dynamics similar to those found in natural ecosystems.
The rotation between Value and Growth styles over the last 30 years resembles the oscillations seen in population dynamics (e.g., wolves and rabbits). The model's testing on a three-dimensional simplex revealed that final wealth distribution is highly sensitive to initial conditions. While Value investors often dominate, capturing up to 70% of wealth, high concentrations of Trend Followers can induce market instability and drive other "species" to extinction.
Key Results and Stylized Facts
Evology successfully reproduces several "stylized facts" of empirical markets:
- Heavy Tails: Extreme price movements occur more frequently than a normal distribution would predict.
- Volatility Clustering: High-volatility events tend to occur in bursts.
- Leverage Effect: A negative correlation exists between returns and volatility.
- Aggregational Gaussianity: While daily returns are non-Gaussian, yearly returns smooth out to resemble a normal distribution, matching historical observations.
Conclusion and Future Work
Evology provides a robust "gym" for training machine learning agents to navigate endogenous volatility and crowding. Future iterations aim to expand the model to include multiple assets, a broader diversity of strategies (including Blend and market-cap-specific styles), and participants from the bond and ETF markets. The ultimate goal is to provide an open-source toolkit for optimizing trading strategies and financial regulations to prevent systemic market failures.
Summary updated