The Information Coefficient (IC) is a statistic that assesses how well a model’s predicted rankings of securities align with their actual subsequent returns. In practice, IC is typically computed as a Spearman rank correlation, though some practitioners report Kendall’s tau or other rank-based measures. A higher IC suggests that better-ranked securities tend to deliver higher returns, while a value near zero indicates little or no alignment.
IC is used to evaluate scoring models and factor signals in quantitative investing. By measuring the historical alignment between a model’s rankings and realized outcomes, analysts gauge the quality of the signal across a universe of securities and over time. IC can be tracked across time windows to diagnose drift in a model’s predictive ability, compared across factors, or used to set expectations for a ranking-based strategy. In portfolio construction, an information-based ranking might inform allocation decisions when the objective is to exploit systematically better-rated securities.
To compute IC, one collects predicted scores for each security and its realized return over a defined period, then computes the correlation between the ranks (or standardized scores) of the two series. IC is inherently sample-dependent and can be volatile; reporting an average IC over multiple periods can smooth short-term noise. Practitioners distinguish IC from other metrics like the Information Ratio, which compares active return to tracking error, and use IC alongside other performance measures.
In a backtest of a factor model, the IC for the model’s stock rankings averaged 0.18 over 12 months, indicating a measurable alignment between predicted ranks and realized returns.
Spearman correlation · Kendall's tau · Rank correlation · Information ratio · Factor model · Backtesting