This project tests a novel theoretical mechanism for low take-up of collateralized credit. Using a field experiment with a lender in Kenya, we vary whether already-owned assets serve as collateral, or whether instead the new asset being financed by the loan itself serves as collateral, as in car loans. The experimental design holds everything except existing ownership of the collateral constant. We find that borrowers display an endowment effect over their existing assets, and thus dislike loans that place existing assets at risk of loss by providing them as collateral. They are willing to pay 8% per month higher interest to instead collateralize using the new asset being financed by the loan itself. We next show that borrowers systematically under-predict their future attachment to new assets. After taking a loan collateralized with a new asset, borrowers become attached to the new asset, and exert similar repayment effort to loans collateralized with existing assets. We develop and estimate a structural model to provide behavioral parameters and to better understand the welfare consequences of such loans.