If individuals have self-control problems, they may take up commitment contracts that restrict their spending. We experimentally investigate how contract design affects the demand for commitment contracts. Each participant divides money between a liquid account, which permits unrestricted withdrawals, and a commitment account with withdrawal restrictions that are randomized across participants. When the two accounts pay the same interest rate, the most illiquid commitment account attracts more money than any of the other commitment accounts. We show theoretically that this pattern is consistent with the presence of sophisticated present-biased agents, who prefer more illiquid commitment accounts even if they are subject to uninsurable marginal utility shocks drawn from a broad class of distributions. When the commitment account pays a higher interest rate than the liquid account, the relationship between illiquidity and deposits is flat, suggesting that agents without present bias and/or naïve present-biased agents are also present in our sample