Square-Root Law Overview

• Square-root law is a universal scaling principle where key observables grow proportionally to the square root of an underlying measure, illustrating the interplay of fluctuations, geometry, and collective behavior.
• In market microstructure, it precisely describes how price impact scales with order volume through a √(Q) relationship, grounded in empirical studies and reaction-diffusion liquidity models.
• Across disciplines, the law provides actionable insights into fair voting systems, sharp operator estimates in functional analysis, and bounded error behavior in analytic number theory.

The square-root law denotes a universal scaling regime in which a key observable (impact, deviation, cancellation, or power index) grows as the square root of an underlying measure (volume, sample size, time, etc.). The law emerges in diverse fields: market microstructure, number theory, probability, voting theory, spectral graph theory, functional analysis, and empirical growth models. It signals an interplay between underlying fluctuations, geometric constraints, and collective behaviour, and its mathematical form and universality class are active areas of contemporary research.

1. Square-Root Law in Market Microstructure

In financial markets, the square-root law governs the average price impact $\mathcal{I}(Q)$ of large orders ("metaorders") as a function of executed volume $Q$. Empirical evidence, across equities, futures, options, FX, and digital assets, robustly confirms

$\mathcal{I}(Q) = Y\,\sigma\,\sqrt{Q/V}$

where $\sigma$ is the asset volatility, $V$ the daily volume, and $Y$ is an order-one constant. The implementation shortfall for metaorders is given by

$S(Q) = \frac{2}{3}Y\sigma Q\sqrt{Q/V}.$

Option markets require recasting in terms of net vega $Q_\nu$, daily vega $V_\nu$, and volatility-of-volatility $\sigma_\sigma$, with analogous scaling and robust estimation of empirical exponents and prefactors (Toth et al., 2016).

Mechanistically, the law is traced to a locally linear ("V-shaped") latent liquidity profile around the midprice and is theoretically realized in reaction-diffusion models of supply and demand. Major large-scale studies, e.g., on the Tokyo Stock Exchange, show that the scaling exponent $Q$0 is universal across both stocks and individual traders, with any observed spread fully accounted for by statistical noise, and decisively contradict system-specific models (Sato et al., 2024).

The so-called "double" square-root law refines the microscopic origin: the impact of a single child order is itself $Q$1 (volume law), and this impact subsequently decays in time as $Q$2 (time law). Thus, the overall impact of a metaorder arises from the combination of both effects (Maitrier et al., 22 Feb 2025).

Key Equations (market impact)

Context	Impact Formula	Exponent
Stocks/Futures	$Q$3	$Q$4
Option Vega	$Q$5	$Q$6 (empirical)
Double sqrt (child)	$Q$7	$Q$8

2. Square-Root Law in Number Theory

The square-root law in analytic number theory classically denotes the bound

$Q$9

for oscillatory functions $\mathcal{I}(Q) = Y\,\sigma\,\sqrt{Q/V}$0 (e.g., Dirichlet characters, Möbius, Liouville, exponential sums); this expresses square-root cancellation of partial sums and is a key phenomenon underlying the distribution of primes and zeros of $\mathcal{I}(Q) = Y\,\sigma\,\sqrt{Q/V}$1-functions. Harper’s "beyond square-root" conjecture posits (conditionally on deep conjectures) that the Liouville sum twisted by characters, $\mathcal{I}(Q) = Y\,\sigma\,\sqrt{Q/V}$2, achieves a stronger bound, breaking the classical square-root barrier under the Ratios Conjecture and GRH: $\mathcal{I}(Q) = Y\,\sigma\,\sqrt{Q/V}$3 where $\mathcal{I}(Q) = Y\,\sigma\,\sqrt{Q/V}$4 arbitrarily slowly (Wang et al., 2024).

The square-root law is also central to exponential sum and Fourier analysis over finite fields and rings. For example, for the character sum over the hyperbola $\mathcal{I}(Q) = Y\,\sigma\,\sqrt{Q/V}$5,

$\mathcal{I}(Q) = Y\,\sigma\,\sqrt{Q/V}$6

Such cancellation is equivalent to Salem-set behaviour and is only possible for finite fields or certain exceptions (Boolean rings); the presence of zero divisors obviates the square-root law for large families (Iosevich et al., 2014, Kingsbury-Neuschotz, 2024).

3. Square-Root Law in Voting Theory

The Penrose-Banzhaf square-root law arises in two-tier voting systems as a prescription for fair voting weights $\mathcal{I}(Q) = Y\,\sigma\,\sqrt{Q/V}$7 for regions of population $\mathcal{I}(Q) = Y\,\sigma\,\sqrt{Q/V}$8. Under the assumption of independent voter behaviour, the Banzhaf power index is maximized for

$\mathcal{I}(Q) = Y\,\sigma\,\sqrt{Q/V}$9

ensuring proportional citizen power. Variants for correlated voting, such as the collective bias model and Curie–Weiss spin models, lead to higher exponents up to $\sigma$0, reflecting that strong intra-group (or global) correlation amplifies the voting block effect and violates the square-root scaling. In the critical regime, intermediate exponents (e.g., $\sigma$1) emerge (Kirsch et al., 2012).

4. Square-Root Law in Probability and Stochastic Processes

In the context of Lévy's arcsine law for occupation times, geometric perturbations produce corrections to the classic CDF of order $\sigma$2: $\sigma$3 where $\sigma$4 is Lévy's arcsine law, and $\sigma$5 is proportional to the mean curvature of the hypersurface and the expected local time of Brownian motion. This geometric deviation is universal and determined by mean curvature, with higher-order effects suppressed (Hsu et al., 2020).

5. Square-Root Law in Functional Analysis

For divergence-form elliptic operators $\sigma$6 with bounded measurable or BMO anti-symmetric coefficients, the square-root law refers to sharp two-sided $\sigma$7 estimates: $\sigma$8 valid for $\sigma$9, quantifying the equivalence between first-order and square-root operators, with essential dependence on off-diagonal kernel estimates and functional calculus (Hofmann et al., 2019).

6. Square-Root Law in Empirical Growth Phenomena

When growth is governed by Gibrat’s law (size-independent relative growth) and restricted (by, e.g., resource limits or interventions), evolution is well-described by a log-normal process. The ratio of the time of maximum value $V$0 to the time of maximum growth rate (inflection point) $V$1 is universally

$V$2

under the maximum entropy production principle. Empirical data from epidemic outbreaks, droplet size experiments, and human growth curves are consistent with this prediction (Wu, 2013).

7. Limitations, Contingency, and Universality

The applicability and universality of the square-root law depend on latent structural features: lack of zero divisors (arithmetic), linear order-book liquidity profiles (finance), or independence/homogeneity (voting). In number theory, structural barriers (such as large ideals in rings or zero divisors) or specific combinatorial structures (e.g., Hamming varieties, coordinate axes unions) can cause sharp failures of the law. In financial markets, universality of the $V$3 exponent is empirically confirmed to high precision; deviations (e.g., exponents near 0.4–0.43) are explained as statistical artifacts, and alternate microstructural models not reproducing $V$4 are ruled out empirically (Sato et al., 2024, Toth et al., 2016).

In summary, the square-root law is a paradigmatic universal scaling law in systems dominated by random fluctuations, tightly connected to geometric, combinatorial, and collective features of the underlying process or system. Its universality and exceptions are central active areas in both theoretical and empirical research across physics, finance, probability, combinatorics, and social choice.