This paper examines whether biased income expectations due to overconfidence lead to higher levels of debt-taking. In a lab experiment, participants can purchase goods by borrowing against their future income. We exogenously manipulate income expectations by letting income depend on relative performance in hard and easy quiz tasks. We successfully generate biased income expectations and show that participants with higher income expectations initially borrow more. Overconfident participants scale back their consumption after feedback. However, at the end of the experiment they remain with higher debt levels, which represent real financial losses. To assess the external validity, we nd further evidence for the link between overcondence and borrowing behavior in a representative survey (GSOEP-IS).