Portfolio sorting
Turn point-in-time characteristics into test portfolios
Portfolio sorting groups securities by observed characteristics and compares subsequent portfolio returns. Independent double sorts use separate breakpoints; dependent sorts form a second characteristic sort within first-sort groups. Breakpoint universe, weighting, rebalancing and accounting availability define the estimand. Long-short spreads summarise characteristic-associated returns but do not isolate causal effects of changing that characteristic.
Choose this to visualise characteristic-return patterns, construct factor portfolios or test asset-pricing models when point-in-time security and accounting histories support implementable formation rules and subsequent return measurement.
Strengths
- Makes nonlinear characteristic patterns visible without imposing a regression slope
- Produces interpretable test portfolios and return spreads
Limitations
- Results can depend on breakpoints, weighting and microcaps
- Many candidate sorts create multiple-testing and data-mining risks
Know the boundary
A high-minus-low return spread is an association under a construction rule, not the causal return to changing a firm attribute.