The Fragmentation of Theory
The evolution of quantitative finance has long been marked by a fundamental dichotomy. On one side, macroscopic asset pricing models (rooted in Bachelier and Black-Scholes) rely on the assumption that price processes are semi-martingales. This reflects the absence of arbitrage and limits return predictability.
On the other side, market microstructure—the study of how latent demands translate into executed trades—uncovered robust empirical regularities that seemed to clash with simple diffusive models:
- Long MemoryPersistent signed order flow where the direction of trades correlates over time.
- Square-Root ScalingThe non-linear, concave market impact of large orders.
- Rough VolatilityExtreme roughness of volatility paths, far jaggeder than Brownian motion.
Historically, these were studied in isolation. The Muhle-Karbe framework unifies them. By identifying a single structural statistic, H₀, which quantifies the persistence of institutional trading, the authors prove these phenomena are mathematically bound together through no-arbitrage requirements.
The Two-Layer Hawkes Architecture
The primary innovation is describing order flow through a dual-layer architecture, distinguishing between Core Orders and Reaction Flow. Both are modeled using Hawkes processes—self-exciting point processes perfect for capturing the clustering and feedback mechanisms in financial data.
| Feature | Core Order Flow | Reaction Order Flow |
|---|---|---|
| Origin | Institutional metaorders, fundamental views | HFT, market making, liquidity provision |
| Primary Driver | Autonomous investment decisions | Response to observed market activity |
| Temporal Horizon | Low to medium frequency (hours/days) | High frequency (milliseconds to seconds) |
| Mathematical Role | Generates long-term persistence (H₀) | Generates martingale / high-freq noise |
The Anchor Statistic: H₀
The Scaling Limit & Fractional Dynamics
By analyzing the large-time asymptotics of this two-layer model, the theory establishes that rescaled signed order flow converges to a "mixed fractional Brownian motion".
This provides a brilliant theoretical resolution to a long-standing paradox: Why do Hurst exponent estimates depend on the sampling scale?
High Frequencies (Ticks)
The memoryless martingale component of the Reaction Flow dominates the signal. The flow appears completely diffusive (Hurst ≈ 0.5), drowning out the core flow.
Low Frequencies (Hours)
The high-frequency "noise" of reaction trades cancels out. The persistent signal of the Core Flow becomes visible, driving the estimated Hurst exponent up towards 0.75.
Endogenous Rough Volatility
One of the most profound contributions is proving that "rough" volatility is not an exogenous assumption, but an endogenous necessity. If the core order flow is highly persistent (H₀ > 1/2), a naive price response would create predictable, exploitable trends.
To maintain market efficiency and prevent statistical arbitrage, the price impact must scale to exactly compensate for the flow's persistence. This compensatory scaling generates the hyper-jagged, rough paths of volatility.
Volatility Hurst Parameter
The mathematical relationship linking the Hurst parameter of volatility to the persistence of the core order flow.
Calculating the Roughness
2(0.75) - 1.5 = 0.0This perfectly matches empirical observations where H_vol ranges from 0.0 to 0.15, explaining why volatility appears so much rougher than the price process itself!
Reconciling the Square-Root Law
The "square-root law" of market impact—which states that the price impact of a large order grows as the square root of its size—is one of the most universal empirical laws in finance.
The Muhle-Karbe framework proves that this concave impact is not a random artifact, but the necessary consequence of processing persistent order splitting efficiently.
Market Impact Power-Law Exponent
The power-law exponent (δ) of market impact derived from core flow persistence.
Plugging in our universal constant H₀ ≈ 0.75:
An exponent of 0.5 is exactly the Square-Root Law! The model seamlessly transitions from high-frequency linear impact of individual child orders to macro-scale concavity for aggregate metaorders.
The Fractal Nature of Traded Volume
While much of the literature focuses strictly on price, the unified theory demonstrates that the traded volume itself (the unsigned magnitude of activity) is also a rough process. This establishes a deep symmetry between the roughness of trading intensity and the roughness of price fluctuations.
Unsigned Traded Volume Hurst Index
Overview of Scaling Parameters (Assuming H₀ ≈ 0.75):