This paper delves into the impact of FOMC policy announcements and meeting minutes on the reallocation of US assets at the fund level. We address the challenge of uncertain reallocation scenarios through employing a data imputation technique with random forest forecasts and fund-level features. We then quantify the most predictive information from FOMC policy statements and meeting minutes through constructed diffusion indices using a supervised learning approach. By conducting predictive fixed-effect regression analyses, we unveil the significant role of incorporating textual predictors derived from FOMC statements. Our findings reveal that positive and negative signals in FOMC announcements have contrasting effects on changes in US asset allocation. Additional exercises demonstrate that net assets and reallocation channels matter for the responses of funds to FOMC statements. We also find substantial evidence of the increased attention paid to FOMC announcements after the 2008 crisis. This heightened attention notably impacts funds that adjust US holdings using their own capital. Furthermore, we examine the presence of home bias among US and non-OECD funds in comparison to other OECD-based funds. Our analysis suggests the potential presence of a home bias among US and non-OECD funds.