Imputing Daily Mutual Fund Trades
58 Pages Posted: 17 Jan 2024 Last revised: 18 Oct 2024
Date Written: March 31, 2023
Abstract
We propose a novel method to impute daily mutual fund trades in individual stocks from daily stock prices and returns and quarterly fund holdings, monthly total net assets and daily fund returns - so the method can be applied to standard CRSP mutual fund data. We set up an (underidentified) system of linear equations and solve the underidentification issue with hierarchical preferences and an iterative method that applies random and adaptive constraints on trade incidence. The method produces daily, stock-level trade estimates with associated confidence levels. Validation analyses using proprietary daily fund trading data show good accuracy, especially for larger trades.
Keywords: Mutual funds, optimization, linear programming, daily frequency, Thompson sampling
JEL Classification: C44, C50, C61, C63, C81, G23
Suggested Citation: Suggested Citation