Credit Risk, Underwriting and Loss Forecasting

Lending is the only business in which the cost of goods sold is unknown when the product is sold. A lender books the interest today and learns over the next one to thirty years how much of the principal it will get back. Credit losses are the biggest single cost in most lending businesses, larger than funding in many consumer products and larger than operating costs in subprime, and they arrive late, in clusters, and at the same time for everyone. US commercial banks charged off about 4% of credit card balances in late 2025 (Federal Reserve, Q4 2025: 4.11% seasonally adjusted); at banks outside the largest 100 the rate was nearer 8% in mid-2026 (Federal Reserve, reported). This section explains how lenders decide whom to lend to (underwriting, data, scorecards and cutoffs), how they price for risk, how they forecast and measure losses (expected loss, vintage curves, roll rates, stress tests), how accounting rules turned forecasting into reported profit (CECL in the US, with IFRS 9 as the international contrast), how commercial and consumer credit differ, how collections and recoveries work, and what the landmark failures teach. A worked example builds a stylized scorecard, chooses a cutoff, and shows how CECL reserves change reported profit when a lender grows and when a recession is forecast.