The information coefficient, often abbreviated IC, measures the relationship between an investment signal or forecast and the subsequent realized outcome it was intended to predict.
In the fundamental-law framework, it is the primary measure of forecasting skill.
A simple implementation may use a correlation between forecasted active returns and later realized active returns. In cross-sectional research, rank correlation is also common when the strategy cares more about ordering securities than predicting exact return magnitudes.
A simple example
Suppose a model ranks 100 stocks each month by expected benchmark-relative return. If the stocks receiving higher scores tend to produce higher subsequent active returns, the model has a positive information coefficient over that sample.
An IC near zero indicates little observed association. A negative IC means the forecasts and realized outcomes moved in opposite directions over the measured sample.
The exact number depends on the forecast horizon, universe, return definition, weighting method, correlation convention, and treatment of overlapping observations.
Ex ante versus ex post IC
An ex ante information coefficient is an assumption about future forecasting skill. An ex post information coefficient is an estimate from realized historical data.
Those are not interchangeable. A backtest may show a positive historical IC without establishing that the same forecast accuracy will persist out of sample.
The Fundamental Law of Active Management uses expected skill as an input to expected active performance, so overstating IC can mechanically overstate the strategy's prospective information ratio.
IC is not the information ratio
The information coefficient is different from the Information Ratio.
IC measures forecast accuracy. The information ratio measures portfolio-level active return per unit of Tracking Error.
A manager can have useful forecasts but still realize a weak information ratio if constraints, transaction costs, poor sizing, crowding, or implementation errors prevent those forecasts from becoming efficient portfolio positions.
That implementation gap is one reason the Transfer Coefficient appears separately in the generalized fundamental law.
Signal frequency does not guarantee independent evidence
A strategy that generates thousands of forecasts does not automatically have thousands of independent skill observations.
Signals can be highly correlated across securities, dates, factors, sectors, and overlapping holding periods. The effective Breadth in Active Management can therefore be far smaller than the raw number of scores or trades.
Estimation choices matter
Observed IC can be distorted by look-ahead bias, survivorship bias, data revisions, universe changes, multiple testing, transaction-cost omission, and tuning the model to a particular historical sample.
The signal must also be aligned with the outcome it claims to predict. A forecast of one-month relative returns should not be evaluated casually against a different horizon or an unrelated performance measure.
What the information coefficient cannot establish
A positive historical IC does not prove durable alpha or future profitability. It does not incorporate position constraints, turnover, market impact, capacity, fees, or active risk.
It is evidence about forecast quality under a specified measurement design, not a standalone trading recommendation or guarantee of investment skill.
Sources
- CFA Institute, Analysis of Active Portfolio Management, 2026
- CFA Institute, Active Equity Investing: Portfolio Construction, 2026
Continue Research
Continue from the concept into the Grizzly Bulls research surface that best matches the next question. These links are research continuations, not recommendations or required steps.
Explore systematic models
Continue into model research without treating historical information coefficient as guaranteed out-of-sample forecasting skill.
Review market indicators
Use broader indicators as context without inferring a current IC or live signal quality estimate.
Explore more topics in the Financial Research Encyclopedia.