We test the interest rate sensitivity of subprime credit card borrowers using a unique panel data set from a UK credit card company. We were given details of a randomized interest rate experiment conducted by the lender between October 2006 and January 2007. Access to such information is rare. We first calibrate an intertemporal consumption model to show that the experimental design has sufficient statistical power to detect economically plausible responses among borrowers. We then find that individuals who tend to utilize their credit limits fully do not reduce their demand for credit when subject to increases in interest rates as high as 3 percentage points. This finding is naturally interpreted as evidence of binding liquidity constraints. We also demonstrate the importance of isolating exogenous variation in interest rates when estimating credit demand elasticities. We show that estimating a standard credit demand equation with the nonexperimental variation in the data leads to severely biased estimates. This is true even when conditioning on a rich set of controls and individual fixed effects.