We examine the credit risk-choices of the public sector banks (PSBs) in India with a novel dataset that is able to trace the borrowers to their banks. We determine the likelihood of the ownership type of the lender bank associated with every firm, using a lender type prediction model with a set of observable risk proxies such as the ex-ante credit ratings. The analysis indicates that the PSBs are more likely to lend to observably risky firms compared to the private banks (PBs). The observed likelihood of lending to riskier firms is significantly higher among the smaller PSBs. The set of firms that majorly contribute to the higher credit-risk choice of the PSBs include the riskier service-sector firms, firms that borrow by pledging promoter shares, and firms that are likely to be impacted by the change of political regime.