This paper explores the theory of saving when consumers are restricted from borrowing and examines how this theory can explain certain stylized facts about saving behavior. The analysis focuses on consumers who are relatively impatient and whose labor income is independently and identically distributed over time. In such cases, assets act as a buffer stock, protecting consumption against bad income draws. The precautionary demand for saving interacts with borrowing constraints to provide a motive for holding assets. If the income process is positively autocorrelated but stationary, assets are still used to buffer consumption, but less effectively and at a greater cost in terms of foregone consumption. In the limit, when labor income is a random walk, it is optimal for impatient liquidity-constrained consumers to simply consume their incomes. This implies that a liquidity-constrained representative agent cannot generate aggregate U.S. saving behavior if that agent receives aggregate labor income. However, in reality, microeconomic income processes are more complex, and it is possible to construct a model of microeconomic saving under liquidity constraints that reproduces many stylized facts in actual data. While many households are not liquidity constrained and do not behave as described, the models presented in the paper account for important aspects of reality that are not explained by traditional life-cycle models.This paper explores the theory of saving when consumers are restricted from borrowing and examines how this theory can explain certain stylized facts about saving behavior. The analysis focuses on consumers who are relatively impatient and whose labor income is independently and identically distributed over time. In such cases, assets act as a buffer stock, protecting consumption against bad income draws. The precautionary demand for saving interacts with borrowing constraints to provide a motive for holding assets. If the income process is positively autocorrelated but stationary, assets are still used to buffer consumption, but less effectively and at a greater cost in terms of foregone consumption. In the limit, when labor income is a random walk, it is optimal for impatient liquidity-constrained consumers to simply consume their incomes. This implies that a liquidity-constrained representative agent cannot generate aggregate U.S. saving behavior if that agent receives aggregate labor income. However, in reality, microeconomic income processes are more complex, and it is possible to construct a model of microeconomic saving under liquidity constraints that reproduces many stylized facts in actual data. While many households are not liquidity constrained and do not behave as described, the models presented in the paper account for important aspects of reality that are not explained by traditional life-cycle models.