Guide

Grinold-Kahn fundamental law of active management explained

Harbor Capital's systematic equity team delivered a respectable information coefficient (IC) of 0.045 on its value-momentum composite — well above allocator screens. Yet the live sleeve's information ratio (IR) stalled at 0.28, below the 0.5 “worth paying active fees” hurdle. PMs blamed “a bad macro year.” Risk analytics applied the Grinold-Kahn fundamental law of active management and found a different story: realized breadth was only 12 independent bets per year after capacity limits and sector caps, not the 50+ the pitch deck implied. With IR ≈ IC × √BR, even excellent IC cannot produce high IR when breadth collapses. The committee widened the investable universe, relaxed redundant sector constraints, and cut overlapping factor sleeves — raising effective breadth to 28 without changing the signal model. IR improved to 0.52 over the next four quarters. The law, from Grinold and Kahn's Active Portfolio Management (1999), is the bridge between signal quality (IC) and portfolio outcomes (IR). This guide states the formula, defines breadth and the transfer coefficient, contrasts independent vs correlated bets, works the Harbor refactor, provides a lever decision table, lists pitfalls, and ends with an allocator checklist alongside our active share guide.

The fundamental law in one equation

In its simplest form, expected (ex ante) information ratio scales with forecast skill times the square root of breadth:

IR ≈ IC × √BR

where IC is the correlation between forecasts and subsequent returns (cross-sectionally, per rebalance), and BR is the number of independent active bets per year. The square root matters: doubling skill (IC) doubles IR, but quadrupling breadth only doubles IR. Managers with modest IC but many uncorrelated bets can outperform geniuses who concentrate on three correlated ideas.

A fuller version introduces the transfer coefficient (TC) — how efficiently portfolio weights reflect forecasts after constraints, risk limits, and transaction costs:

IR ≈ TC × IC × √BR

TC ranges from 0 (forecasts ignored) to 1 (weights proportional to forecasts with no binding constraints). Closet indexers often have low TC even when IC is positive: they know what to buy but hug the benchmark for career risk. High turnover and costs also drag TC because optimal weights are scaled down to stay tradable.

Ex ante vs ex post

The law is primarily a planning tool: given expected IC and planned breadth, what IR is achievable? Ex post, realized IR will differ because IC varies by regime, breadth is hard to measure precisely, and TC slips when liquidity dries up. Use the law to diagnose which lever to pull, not as a precise forecast.

Defining breadth: independent bets, not position count

Breadth is the most abused term in the law. A fund holding 200 stocks does not automatically have BR = 200. If all 200 names load on the same value factor and move together, effective breadth may be closer to 3–5 independent themes.

Grinold and Kahn define breadth as the number of independent active wagers per year whose outcomes add variance in proportion to 1/BR rather than stacking. Practical proxies allocators use:

Independent factor sleeves: value, momentum, quality, and low-volatility signals with low cross-correlation each contribute breadth if sized separately.
Rebalance frequency × names with material active weight: monthly rebalance across 80 names with non-trivial active weights suggests higher BR than quarterly on 20 names — if stock-specific risk dominates.
Effective bets (Menchero / Boudet): eigenvalue decomposition of the active covariance matrix; sum of explained variance across uncorrelated factors.
Strategy count with low P&L correlation: multi-strategy pods whose daily returns correlate below 0.3 add breadth at the fund level.

Correlated bets subtract effective breadth. A PM running value in financials, value in energy, and value in staples may have three positions but one bet. Sector neutrality constraints that force offsets can reduce TC without increasing BR.

IC, IR, and where each metric lives

Metric	Level	Question answered	Grinold-Kahn role
Information coefficient (IC)	Signal / cross-section	Do forecasts rank stocks correctly?	Skill input to the law
Breadth (BR)	Portfolio construction	How many independent bets per year?	Scaling factor (√BR)
Transfer coefficient (TC)	Implementation	Are weights allowed to reflect forecasts?	Efficiency multiplier
Information ratio (IR)	Fund / time series	Risk-adjusted active return vs benchmark?	Output the law explains

Numeric intuition: suppose IC = 0.05 (strong for equities) and BR = 16 independent bets per year. Then IR ≈ 0.05 × 4 = 0.20 before TC. With TC = 0.85 after mild constraints, expected IR ≈ 0.17. To reach IR = 0.50 with the same IC, you need BR ≈ (0.50 / (0.05 × TC))² — roughly 125 independent bets at TC = 1, or fewer if TC is low. That math explains why many stock-picking mutual funds with 40-name portfolios and IC near 0.03 struggle to clear IR = 0.5: 0.03 × √40 ≈ 0.19.

Harbor Capital systematic sleeve refactor

Harbor's Q2 2025 review decomposed the systematic sleeve:

Lever	Before	After	Action
Mean IC (monthly)	0.045	0.044	No model change
Effective breadth (est.)	12	28	Add mid-cap sleeve; remove duplicate value tilt in two pods
Transfer coefficient	0.62	0.78	Relax 2% single-name cap; improve algo execution
Implied IR (≈ TC × IC × √BR)	0.21	0.51	Aligned with realized IR within noise
Realized IR (rolling 12m)	0.28	0.52	Fees now justified vs hurdle

The committee did not chase higher IC with more complex machine-learning features — the law showed that was the wrong bottleneck. Instead they increased independent bets (mid-cap extension, uncorrelated quality sleeve) and raised TC by reducing constraint stacking. Turnover rose modestly; TCA showed implementation shortfall stayed flat because VWAP slicing improved.

Lever decision table

Situation	Likely bottleneck	First move	Avoid
High IC, low IR, concentrated portfolio	Low breadth	Widen universe, add uncorrelated sleeves, increase rebalance count	More complex signals with same 30-name book
Low IC, high breadth	Weak signal	Feature research, combine orthogonal factors, extend history	Adding correlated “factors” that do not raise IC
High IC, high BR, low IR	Low TC (constraints/costs)	Audit binding constraints, reduce overlap with benchmark, cut fees	Blaming macro without measuring TC
Discretionary PM, few high-conviction names	Inherently low BR	Require very high IC or accept lower IR; size fund smaller	Marketing as “diversified” with 8-stock portfolios
Multi-manager fund of funds	BR at allocator level	Hire low-correlation managers; measure pod correlation	Five growth managers = one bet

Common pitfalls

Counting stocks as bets — 100 correlated value names are not BR = 100.
Using portfolio IR to infer IC without breadth — a lucky year can mask weak signals.
Ignoring TC — risk parity overlays, ESG exclusions, and benchmark hugging silently cap IR.
Assuming IC is constant — regime shifts collapse IC; stress-test breadth and TC too.
Chasing IC with overfit models — in-sample IC of 0.12 that live-delivers 0.02 destroys the law's inputs.
Confusing BR with turnover — churning 500 names monthly without edge adds costs, not breadth.
Forgetting capacity — large AUM forces TC down when alphas cannot be scaled into weights.

Allocator checklist

Estimate or request ex ante IC, BR, and TC from systematic managers — not just backtest Sharpe.
Compute implied IR = TC × IC × √BR and compare to realized IR and fee hurdle.
Decompose active covariance to count effective independent bets, not just position count.
Pair with active share and tracking error to detect closet indexing (low TC).
Stress IC and BR in bear markets and liquidity crises before committing capital.
When IR disappoints despite strong marketing, diagnose breadth before firing the signal team.
Document which constraints (sector, ESG, risk) bind and by how much they reduce TC.
For multi-strategy funds, measure cross-pod return correlation quarterly.
Revisit the law after major AUM flows — capacity often crushes BR and TC together.
Teach investment committees the √BR scaling so they do not expect IR = 1 from IC = 0.05 alone.

Key takeaways

The Grinold-Kahn law links signal skill (IC), breadth (independent bets), and portfolio success (IR): IR ≈ TC × IC × √BR.
Breadth is about independent wagers, not headcount — correlated factors do not multiply IR.
The square root means breadth helps, but IC and TC often dominate for concentrated managers.
Low IR with decent IC usually points to low breadth or low transfer coefficient, not bad luck alone.
Use the law to choose levers: widen bets, improve signals, or remove constraints — in that order when IC is already strong.