Definition
The information coefficient (IC) is the correlation between a factor's predicted returns — typically the cross-sectional factor scores at the start of a period — and the realised returns of those assets over the period. It is the standard measure of how well a forecast signal lines up with what actually happens. By convention the rank-based Spearman IC is preferred over the Pearson form because it is robust to outliers and matches how the scores are used in practice: as a ranking rather than a cardinal forecast.
A daily IC of 0.05 is excellent for an equity factor. A monthly IC of 0.10 is the threshold institutional teams chase. Anything above 0.20 across a broad universe is either an undiscovered edge or a bug in the backtest — and the second is far more common than the first.
Why it matters
How it works
The computation is straightforward. For each rebalance date, rank every asset in the universe by its factor score, rank the same assets by their realised forward return, and compute the Spearman correlation between the two rankings. Stack the per-date ICs into a time series. The mean of that series is the headline IC; the standard deviation is the IC dispersion; the ratio is the factor's IC information ratio.
What the headline number hides is conditioning. ICs typically vary by regime — a value factor IC may average 0.04 but be -0.05 during growth-led rallies and +0.10 during reversals. Skilled teams therefore look at IC conditional on volatility, sector dispersion, or macroeconomic state, not just the average. They also look at IC decay: a factor whose IC is 0.08 at one-day horizon but 0.01 at one-month horizon is fast-decaying and only viable if turnover and costs allow daily trading.