We use novel monthly security-level data on U.S. household portfolio holdings, flows, and returns to analyze asset demand across an extensive range of asset classes, including both public and private assets. Our dataset covers a diverse spectrum of households across the wealth distribution, notably including 372 portfolios exceeding $1 billion in assets. This ensures representation of ultra-high-net-worth (UHNW) households that are typically not well covered in survey data. With these unique data, we study the portfolio rebalancing behavior of households and ask whether (and, if so, which) households play an important stabilizing role in financial markets. Our findings reveal a stark contrast: less affluent households sell U.S. equities amid market downturns, while UHNW households take the opposite side. This behavior is more pronounced among households who rebalance their portfolios more frequently. However, the sensitivity of flows to returns is generally quite small and as the trades of dierent wealth groups partly oset each other, the aggregate household sector plays a limited role to absorb financial fluctuations. To understand the contrasting trading behavior across households, we show that flows to U.S. equities are negatively correlated with “active returns” (the dierence between an investor’s return and the market return) for all wealth groups. However, the flows to U.S. equities of less auent households are also positively correlated with broad market returns – perhaps due to shifts in risk aversion, sentiment, or perceived macroeconomic risk – leading this group of households to act pro-cyclically. Across all asset classes, three factors with intuitive economic interpretations explain 75% of all variation in portfolio rebalancing. Those factors bet on the long-term equity premium, the credit premium, and the premium on municipal bonds. In sum, our framework and data paint a coherent picture of U.S. households’ rebalancing behavior across the wealth distribution and across broad asset classes.