Gibrat's Law¶

The claim that a firm's proportional growth rate is independent of its current size — which, iterated as multiplicative iid noise, makes log-size a random walk and drives the cross-sectional size distribution toward log-normal.

Version: v4 · 2026-09-28 · History
Domain-specific #: 447
Domain group: Social Sciences
Origin domain: Economics & Finance
Subdomain: Industrial Organization and Firm Dynamics → Economics & Finance

Core Idea¶

Gibrat's law (1931) claims a firm's growth rate is statistically independent of its current size: the growth factor S(t+1)/S(t) is a random variable whose distribution does not vary with S(t). Iterating this size-independent proportional growth makes log-size an additive random walk, so by the central limit theorem the cross-section converges to approximately log-normal. The construct thus poses both an empirical regularity and the generative mechanism — multiplicative iid noise — that produces the log-normal distribution.

How would you explain it like I'm…

Same Luck, Any Size

Imagine every shop grows each year by a lucky-dip amount, like adding a random slice of what it already has. A big shop and a tiny shop get the same kind of lucky dip. After many years, most shops end up medium-ish and a few end up really huge. Gibrat's Law says companies grow like that, though it doesn't fit tiny or giant ones perfectly.

Size-Blind Growth

Gibrat's Law says how fast a company grows, as a percentage, doesn't depend on how big it already is. A small company and a big one have the same chances of growing by, say, 10% or shrinking by 5%. If you repeat these random percentage changes year after year, you get a pattern where most companies are modest and a few become enormous. In real life this fits big established companies fairly well, but small companies tend to grow faster than it predicts, and the very biggest ones follow a different pattern.

Size-Independent Proportional Growth

Gibrat's Law (Robert Gibrat, 1931) claims a firm's growth ratio S(t+1)/S(t) is random and independent of its current size S(t). Because growth multiplies size, taking logarithms turns it into adding random steps. So log(size) does a random walk, and by the central limit theorem it becomes roughly normally distributed across firms. Firm sizes therefore become roughly log-normal. Empirically the law fits large firms approximately but fails for small firms, which grow proportionally faster. It also fails at the very top, where sizes look more like a Pareto distribution. The same multiplicative-noise idea is used for city sizes, word frequencies and asset prices.

Gibrat's Law (Robert Gibrat, 1931) states that S(t+1)/S(t) is an independent random variable whose distribution does not depend on S(t). Iterating this proportional random growth makes log S an additive random walk, and the central limit theorem then implies log S is approximately normal across firms, so S is approximately log-normal. The law therefore serves both as an empirical hypothesis about firm growth and as a generative mechanism (multiplicative iid shocks) for log-normal size distributions. As an empirical claim it holds roughly for large established firms but fails for small firms, which grow proportionally faster, and at the upper tail, where Pareto behavior suggests preferential-attachment or merger dynamics. As a mechanism it recurs in city sizes, body sizes within taxa, word frequencies, and asset prices under geometric Brownian motion, the continuous-time analogue used in Black–Scholes. Its main methodological role is as a null model: alternatives like size-dependent growth or minimum efficient scale are framed as deviations from Gibrat.

Scope of Application¶

As an empirical claim it is bounded to firm dynamics; as a generative mechanism it applies literally wherever proportional-random-growth genuinely obtains.

Industrial organization — the home turf: firm-size evolution and tests against size-dependent growth.
Urban economics — city-size distributions, log-normal in the body and Zipf at the top.
Quantitative linguistics — word-frequency distributions in the Estoup-Zipf lineage.
Biology — within-taxon body-size distributions with a log-normal bulk.
Finance — asset prices under geometric Brownian motion, the continuous-time Gibrat analogue.

Clarity¶

Naming the law gives industrial organization a sharp answer to a foundational question — does being large confer a growth advantage? — by positing a falsifiable null with a clean test: regress log-growth on log-size. Deviations become named departures pointing to minimum efficient scale or preferential attachment. Its second move links micro dynamics to macro shape, disentangling three routinely conflated claims: size-independent growth, a log-normal distribution, and equal growth rates.

Manages Complexity¶

The alternative is tracking a whole population of firms each entering, exiting, and growing through its own history. Gibrat collapses that to two scalars and an initial condition: once growth is size-independent and multiplicative, the process is a log-additive walk governed by drift and dispersion, and the cross-sectional shape is derivable. It also converts the field's puzzles into a short menu of signed departures read off a single fitted slope and the tail shape.

Abstract Reasoning¶

It licenses diagnostic reading from cross-sectional shape back to the growth process (and vice versa, mutually constraining); interventionist reasoning about how a shock to growth moments reshapes the population; boundary-drawing on where the null is licensed (large, surviving firms over a long horizon); and temporal reasoning that reads dispersion as elapsed process-time, distinguishing a false law from an unfinished one.

Knowledge Transfer¶

The two levels travel oppositely. As an empirical claim about firms it transfers as mechanism within industrial organization, where minimum efficient scale and survival selection are load-bearing and stay home. As a generative mechanism it is portable mathematical machinery that transfers literally wherever multiplicative iid noise obtains — cities, words, body sizes, asset prices — so the boundary is instrument-reach versus over-reading. The transferable substance sits with Central Limit Theorem, preferential attachment, and power_law families.

Relationships to Other Abstractions¶

Current abstraction Gibrat's Law Domain-specific

Parents (1) — outgoing typed relations

Gibrat's Law is a decomposition of Multiplicative Random Growth Prime

Stripping Gibrat's firm-growth framing leaves Multiplicative Random Growth's size-independent proportional shocks, log-additive random walk, and conditional log-normal limit.

Hierarchy path (1) — routes to 1 parentless root

Gibrat's Law → is a decomposition of Multiplicative Random Growth → includes Random Walk → is a kind of Stochastic Process

Neighborhood in Abstraction Space¶

Gibrat's Law sits in a sparse region of the domain-specific corpus (79^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Scaling Laws & Growth Patterns (13 abstractions)

Nearest neighbors

Allee Effect — 0.83
Big O Notation — 0.83
Gambler's Fallacy — 0.82
Divisia Index — 0.82
Business Cycle — 0.82

Computed from structural-signature embeddings · 2026-10-25