Diversified but Crowded: Evidence from 13F Holdings

Abstract

Can an institution diversify internally while becoming more similar to its peers? Using filing-resolved SEC Form 13F holdings for the 500 largest reported portfolios each quarter from 2013Q2 through 2025Q4, we measure manager crowding as mean pairwise minimum-weight overlap. In regressions with quarter fixed effects, portfolio controls, and manager-clustered standard errors, entropy is positively associated with crowding (β = 0.067, SE = 0.003). A one-standard-deviation entropy difference in the estimation sample corresponds to 0.082 greater crowding, or 50.6% of the mean. Within the quarterly reselected top-500 sample, mean crowding rises from 0.129 to 0.230, while the random-breadth benchmark rises by 0.019; this aggregate trend combines portfolio changes with changes in sample composition. Security-popularity and same-quarter peer-consensus diagnostics sharply attenuate the association; exact large-manager exclusions leave it nearly unchanged. These descriptive results concern reported long positions, not complete portfolios or causal effects. They show that within-portfolio diversification and cross-institution distinctiveness are separate dimensions.