Size

Definition

Size is implemented as point-in-time market capitalization. Lower values correspond to smaller firms, while higher values correspond to larger firms.

The longstanding idea is that smaller firms can earn different expected returns than larger firms because of neglectedness, limits to arbitrage, and implementation frictions that scale with market capitalization.

Inside the finance research stack, the representative implementation sorts the Russell 1000 cross-section on market cap and compares the smallest names against the largest names in a market-neutral spread.

Headline Summary

The first pass on this signal starts with a headline comparison across the sorted signal portfolios and the Russell 1000 benchmark.

Top Quintile contains the lowest-ranked names in the representative sort, while Bottom Quintile contains the highest-ranked names.

The baseline return path shows how the top-ranked bucket, bottom-ranked bucket, and long-short spread evolved through time in the representative Russell 1000 formation.

Start-Date Sensitivity

This section checks whether the signal depends too heavily on when the strategy begins. We restart the same baseline long-short construction every six months and compare how the excess return profile changes across those staggered entry dates.

Each row uses the same monthly market-weighted portfolio construction as the baseline sort, but starts the sample at the stated month and carries it through the final available month. Excess returns are measured relative to the 3-month U.S. Treasury bill rate, proxied by the FRED 'DTB3' series.

Sector Results

Sector results show whether the signal's long-short behavior is broad across the Russell 1000 or concentrated in a smaller set of industries.

The plot shows average monthly excess returns for sector-specific long-short implementations using the same baseline portfolio construction within each sector.

Path-Dependency Distribution

This section tests how sensitive the turnover-constrained implementation is to the path of portfolio formation. We run 1,000 different paths that each try to maximize the signal while allowing 10% turnover at each monthly rebalance.

The realized path starts from the actual T-B portfolio on the first initialization date. The other paths use random initializations on day one, then follow the same monthly turnover budget through time. The plot shows the distribution of average monthly excess returns across those random-start paths, with the realized path marked separately for comparison.