This paper studies qualitative characteristics of accounting systems that are used in debt financing. We consider a financially constrained firm that provides to lenders information on the value of assets that serve as collateral in a financing contract for a risky investment project. We find that the investor prefers an accounting system that provides biased signals about the value of assets. This bias adjusts the information content of the signals to maximize the probability of undertaking the project. Under fair value accounting, low book values are more precise measures of actual value than high book values, which is consistent with conditional conservatism. Next, we study accounting risk to study the effect of institutions that govern the financial reporting policy based on the optimal precision. We find that fair value measurement introduces greater accounting risk and is preferred by financially constrained firms to measurement at historical cost.