The error is in the tracking error

Higher tracking error does not necessarily mean that a diversified, high-conviction portfolio has become riskier, because benchmark concentration in AI-related stocks has reduced its correlation with the broader market.

• Tracking error measures the difference between portfolio and benchmark returns, while the information ratio measures excess return per unit of tracking error; a portfolio that perfectly tracks its benchmark only matches it before fees.
• The tracking error of a hypothetical portfolio holding the five largest companies in each of the 11 S&P 500 sectors was stable for the past decade but rose sharply from mid-2025, a pattern also seen in concentrated active portfolios.
• Since mid-2025, the portfolio's volatility declined while its correlation with the benchmark also fell, causing tracking error to nearly double.
• The authors link this divergence to the benchmark's growing correlation with AI-related stocks: the share of non-AI stocks outperforming the market fell from 80% between June 2017 and June 2025 to 25% thereafter.
• Despite higher tracking error, the portfolio's beta declined because both volatility and benchmark correlation fell, indicating lower sensitivity to broad market movements.

Takeaway: Investors should not interpret rising tracking error alone as rising portfolio risk; as benchmark leadership narrows around AI-related stocks, diversified active portfolios may diverge more from the index while retaining a lower overall beta.