Using supervisory loan-level data on corporate loans, we show that banks facing high levels of non-performing loans relative to their capital and provisions were more likely to grant forbearance measures to the riskiest group of borrowers. More specifically, we find that risky borrowers are more likely to get an increase in the overall limit or the maturity of a loan product from a distressed lender. As a second step, we analyse the effectiveness of this practice in reducing the probability of default. We show that the most common measure of forbearance is effective in the short run but no forbearance measure significantly reduces the probability of default in the long run. Our evidence also suggests that forbearance and new lending are substitutes for banks, as high shares of forbearance are negatively correlated with new lending to the same group of borrowers.