Passive Drift, Active Rebalancing, and the Structure of Institutional Selling

Abstract

This thesis examines movements in institutional portfolio weights and investigates how much of that movement is passive mechanical drift versus active reallocation. Portfolio weights change automatically when a stock appreciates in price, even if an institution takes no action. Relationships between lagged returns and portfolio-weight changes are therefore contaminated by passive drift. To isolate true, discretionary rebalancing activity, we decompose stock-level changes in portfolio weights into passive and active components. The passive component is the change in weight implied by a no-trade counterfactual. The active component is defined as the residual change in portfolio weight beyond the passive benchmark. We use stock-level institutional holdings data to first document that the relation between lagged returns and total portfolio-weight changes is almost entirely driven by passive drift. Active rebalancing, on the other hand, nets to much closer to zero after the passive benchmark is netted out. Next, we document that meaningful active selling does not occur symmetrically around zero. Instead, much of the structure in institutional selling is found in the lower tail of large active sells. Large active sells are more likely to occur in important holdings and are different across institution types. To investigate the sell tail directly, we model large active sell events as a classification problem. Using logistic regression and gradient boosted trees, we find that large active sell events are substantially more predictable than large active buys. Furthermore, their predictability is driven by lagged position size, within-portfolio rank, and institution characteristics like turnover and manager style, as opposed to simple lagged-return variables. Together, these results suggest that institutions impose a lot of structure on the sell tail, which is more heavily influenced by portfolio hierarchy and institution type than return chasing. Lastly, we incorporate full exits into the analysis as a separate margin of institutional selling activity. Unlike large active sells among continuing positions, full exits are concentrated among peripheral holdings. Overall, these results support the idea that institutions lighten their exposure on two margins. They trim their important positions and sell their unimportant ones outright. More generally, this thesis documents that institutional portfolio changes are much more understandable once passive drift is accounted for and the relevant margin of adjustment is isolated.