Square-Root Law of Price Impact

• Square-Root Law of Price Impact is a universal scaling law indicating that market impact grows as the square-root of metaorder size.
• Empirical studies use log–log regression and high-precision statistical methods to confirm an invariant exponent (δ ≈ 0.5) across diverse markets.
• Theoretical frameworks based on latent order-book dynamics and reaction-diffusion models provide insights for liquidity modeling, optimal execution, and risk management.

The square-root law of price impact is a universal empirical scaling observed across financial markets, expressing a robust, sublinear relationship between the average market impact of large orders and the size of the executed quantity. Formally, it states that the average (dimensionless) price impact $I(Q)$ induced by executing a metaorder of size $Q$ scales as $I(Q)\propto Q^{\delta}$, with $\delta\simeq1/2$. This law has been validated with high statistical precision for equities, futures, options, cryptocurrencies, and has a well-developed theoretical underpinning rooted in the microstructure of order-driven markets. The law’s universality, robustness, and mechanical origin have far-reaching implications for liquidity modeling, optimal execution, risk management, and regulatory assessment.

1. Mathematical Definition and Empirical Estimation

A metaorder is an aggregated execution of total signed volume $Q$, composed of $L$ successive market orders (child orders) of signed sizes $q_1,\ldots,q_L$, with sign $\epsilon=+1$ for buys and $-1$ for sells. The market impact $I(Q)$ is defined as the conditional expectation of the total signed normalized price change $\Delta p$ induced from the mid-price before the first to after the last child order:

$$I(Q) := \mathbb{E}[\epsilon\,\Delta p\,|\,Q]\,,\qquad Q \leftarrow Q/V,\quad I \leftarrow \Delta p/\sigma$$

where $V$ is the stock’s daily volume and $\sigma$ its daily volatility. Empirically, for all sufficiently large $Q$,

$$I(Q) \approx c\, Q^\delta\,,\quad \delta \simeq 1/2$$

For the Tokyo Stock Exchange (TSE), exhaustive analysis over 2,299 stock–periods and 1,293 traders established with SEM $0.0020$ that $\delta=0.500\pm0.0020$ and mean prefactor $c\approx0.842$ at the stock level, confirming the universality of the exponent at $<0.1$ accuracy (Sato et al., 2024).

2. Universality: Empirical Evidence Across Markets

The square-root law is observed to hold with high precision and with no systematic deviation across disparate market structures or asset classes, including:

• Major equity markets (TSE, U.S. equities, European, etc.)
• Cryptocurrency markets (e.g., Bitcoin/USD) (Donier et al., 2014)
• Options (in vega– and volatility-of-volatility–normalized coordinates) (Toth et al., 2016)
• Microscopically detailed market datasets, with identification at the individual account or trading-desk level.

On the TSE, curves for $I(Q)$ across hundreds of stocks and traders collapsed onto a $\sqrt{Q}$ master curve after normalization. The histogram of exponents $\delta$ for individual stocks and traders has dispersion matching finite-sample noise, with no observable dependence on stock liquidity, microstructure, or trader strategy. Identical scaling is found for "synthetic" metaorders constructed by scrambling market order IDs, demonstrating invariance to agent identity (Maitrier et al., 22 Feb 2025).

3. Microscopic Origins and Theoretical Frameworks

The robustness of $\delta \approx 1/2$ is rooted in the latent order-book (LLOB) theory, which posits that latent liquidity density vanishes linearly close to the mid-price and grows with distance from the price. This "critical" liquidity profile follows from:

• Price diffusion: market price executes an unbiased random walk.
• Order flow: arrivals and cancellations of buy/sell intentions as independent Poisson processes.

Absorbing a metaorder of volume $Q$ requires moving the price to a level where integrated latent liquidity equals $Q$, yielding:

$$Q = \int_{p^*}^{p^*+\Delta p} \rho(u)\,du \simeq (\ell/2)\,(\Delta p)^2 \implies \Delta p \propto \sqrt{Q}$$

This mechanical, reaction–diffusion interpretation is supported by detailed agent-based models and analytic solutions (Toth et al., 2011, Mastromatteo et al., 2013, Donier et al., 2014). Empirically, even for highly granular and complex asset classes, such as options, the scaling holds without parameter tuning, indicating its universality (Toth et al., 2016).

Mechanistically, every market order, after sufficient "digestion," imparts a $\sqrt{q}$ price shock, which then decays as $t^{-1/2}$. Summed over $N$ child orders, the impact aggregates as $\sqrt{Q}$ via a "double" square-root effect (impact per child order and propagator decay in time) (Maitrier et al., 22 Feb 2025, Maitrier et al., 5 Sep 2025, Maitrier et al., 9 Jun 2025).

4. Statistical Procedures and Rejection of Non-Universal Models

Estimation of $\delta$ is performed by log–log binning of non-dimensionalized $Q$ and $I(Q)$, fitting the relative least-squares regression, and extracting standard errors. For the TSE data, the SEM of the mean exponent across stocks is $\approx0.0015$, with between-stock standard deviation $0.071$, matching null-simulation expectations for purely sampling dispersion.

Two leading non-universality models, Gabaix–Gopikrishnan–Plerou–Stanley (GGPS) and Farmer–Gerig–Lillo–Waelbroeck (FGLW), conjectured stock-specific exponents $\delta_\text{GGPS} = \beta-1$, $\delta_\text{FGLW} = \alpha-1$, where $\beta$, $\alpha$ are the power–law exponents of metaorder size and run-length. However, Clauset–KS fits to TSE data revealed no correlation between the empirically measured $\delta$ and $\beta-1$ or $\alpha-1$; Pearson $\rho\sim0$, falsifying both models (Sato et al., 2024).

5. Regimes, Crossover, and Model Extensions

While the square-root law dominates for moderate to large $Q$ (relative to daily volume), a linear regime $I(Q)\propto Q$ emerges at low participation rates or for small $Q$, reflecting fast liquidity replenishment (Bucci et al., 2019, Saddier et al., 2023). Explicit theoretical crossover functions interpolate between these limits. Generalization to multi-timescale, finite-memory liquidity and fractional reaction–diffusion models confirm persistence of the $\sqrt{Q}$ scaling under a broad range of assumptions (Benzaquen et al., 2017, Benzaquen et al., 2017).

A crucial implication is that, even in the presence of persistent (long-memory) order flow—originating from metaorder-splitting and heavy-tailed size distributions—the square-root law precisely "heals" the market, preserving the diffusive ($\mathbb{E}[\Delta p^2]\propto t$) nature of prices. Analytical mapping to Lévy-walk processes shows that only the square-root scaling ensures market efficiency in this sense (Sato et al., 25 Feb 2025).

6. Practical Consequences and Implications

The square-root law provides the quantitative foundation for:

• Forecasting execution impact and transaction costs for large orders.
• Developing optimal execution algorithms beyond the traditional linear Kyle paradigm.
• Benchmarking markets' liquidity and systemic risk assessment.
• Designing proxy measurement algorithms for metaorder impact from anonymized tape data, enabling market-impact estimation without account-level identification (Maitrier et al., 5 Sep 2025).

For practitioners, the universal form

$I(Q)\approx Y\,\sigma\sqrt{\frac{Q}{V}}$

with $Y\sim0.5$–$1.5$ suffices for robust cost estimation across equities, options, and cryptocurrencies. The prefactor $c$ (or $Y$) is market- and stock-specific and encodes liquidity differences, but the exponent remains invariant.

7. Limitations, Deviations, and Future Directions

Empirical studies have identified that while the square-root fit is robust over 2–4 decades of $Q$, at very small or large participation rates deviations may appear. In some datasets, the logarithmic law $I(Q)\sim \log(1+\lambda Q)$ fits data better over a wider dynamic range (Zarinelli et al., 2014). Nonetheless, the square-root law is the dominant scaling in real-world conditions, with observed deviations primarily attributed to overlapping metaorders, herding, and execution profile choices. Understanding transient vs. permanent impact components, the relaxation dynamics post–metaorder, and the formation of impact "surfaces" dependent on both participation rate and execution duration remain ongoing areas of research.

---

Summary Table: Empirical Exponent Estimates for Square-Root Impact

Market / Asset	Mean $\delta$	Reference
Tokyo Stocks	$0.500 \pm 0.0020$	(Sato et al., 2024)
Bitcoin (USD)	$0.50 \pm 0.02$	(Donier et al., 2014)
US Equity (meta)	$0.47$	(Zarinelli et al., 2014)
Options (LT/ST)	$0.40,\ 0.43$	(Toth et al., 2016)
CFM Futures	$0.5$ (small-tick)	(Toth et al., 2011)

Estimates are consistent with $\delta = 1/2$; observed dispersion is accounted for by sampling variation.

---

Key Functional Expression

$$I(Q) = c\,Q^{1/2} \qquad\text{with} \qquad c \approx Y\,\sigma/\sqrt{V}$$

where: $I(Q)$ = average normalized price move, $Q$ = metaorder size, $Y$ = market-specific constant, $\sigma$ = daily volatility, $V$ = daily volume.

---

The square-root law stands as a universal, mechanically-derived scaling for market impact, with theoretically and empirically validated invariance to market specifics, firmly constraining plausible models of liquidity, execution, and price formation.