Using High-Frequency Data to Estimate Mutual Fund Holdings

Summary

This research summary describes using market microstructure data to estimate the share of individual stocks held by public mutual funds. It reports that a model trained across the full stock universe has some out-of-sample predictive ability, and that restricting the universe can improve predictions. Grouping stocks by broad index yields relatively better out-of-sample performance within the CSI 300; grouping by sector performs relatively better for technology, consumer, and industrial stocks.

The proposed practical application is to track changes in institutional ownership at higher frequency for selected stocks, sectors, or investment styles. The available summary provides qualitative findings but no model specification, data details, accuracy measures, or test period, so it is not possible to judge the size or robustness of the predictive gains. The estimates should therefore be treated as a monitoring signal rather than a direct or complete measure of actual holdings.

Key ideas

• Microstructure data can help estimate public mutual fund ownership in individual stocks.
• The full-market model has some out-of-sample predictive ability, according to the summary.
• Restricting the universe by index or sector can improve predictive performance.
• The reported results are relatively stronger for CSI 300 stocks and for technology, consumer, and industrial sectors.
• High-frequency estimates may help track institutional positioning, but the summary omits model and accuracy details.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.