This article investigates the liquidity risk of U.S. mortgage-backed securities (MBS) in comparison with government and agency securities from an empirical perspective. An empirical evaluation of liquidity risk is performed by negative tails from the historical changes of daily total transactions. A set of bond market risk factors is applied to control the pure market effects on transaction changes. Once market effects are controlled, negative tails are evaluated from the underlying distributions for residual transaction changes. During the recent five years, liquidity-driven MBS transaction changes show fat-tail, as well as high-sample volatility. This suggests that a sudden, large drop from the normal level of transactions is more likely for MBS than for government and agency bonds, in the case that there is negative liquidity shock.