Selection Strategy

Understanding cross-sectional regression in security selection

This is a medium-difficulty question on the mechanics of cross-sectional regression models used to evaluate stock-picking strategies. It tests whether you understand how forecasting variables like valuation ratios predict excess returns, and how to interpret statistical significance when the regression is run across many securities at a single point in time rather than over time.

The question probes a common pitfall in strategy evaluation: the distinction between time-series and cross-sectional inference. When you regress returns on a forecasting variable across the universe of available stocks in a single period, the number of observations is the number of securities, not the number of time periods. This affects how you interpret the t-statistic and what it tells you about the strategy's predictive power.

• Cross-sectional vs. time-series regression setups
• Degrees of freedom and sample size in strategy backtests
• Statistical significance in the context of market microstructure and diversification