Credit Cycle: a Worked Example

Credit decisions are made by people and calibrated on data, and both have short memories. A portfolio with few defaults for six years produces low loss estimates and committees that have not lately sat through a bad loan. Competitors offering easier terms turn a strong borrower's request into a contest. Each decision is reasonable on the evidence in front of it, and together they lower the standard.

The turn feeds itself in the other direction. Asset prices fall, so security is worth less. Forward-looking loan-loss provisions rise as forecasts worsen. Capital tightens, so lending shrinks, and the shrinkage hurts the borrowers who need to refinance. Part of the good news at the top is also timing: a loan written last year has had little chance to go wrong, so a fast-growing book flatters its own loss rate.

Terms often move before volumes do. On the way up a lender gives ground on margin first, then on structure: longer tenor, less amortisation, fewer or looser covenants, more add-backs in the definition of EBITDA. Volumes and asset prices follow. On the way down, structure is often the first thing to tighten.

Readers outside the file use a few gauges. The credit-to-GDP gap compares private credit with its long-run trend, and the Basel Committee recommends it as a guide for setting the countercyclical capital buffer, a layer of up to 2.5% of risk-weighted assets in the Basel text, built when credit runs hot and released when the cycle turns. Regulators set their own: APRA's range runs to 3.5%, with a 1% setting in place since January 2023. The Reserve Bank of Australia has cautioned that the trend is hard to measure in real time. Lender surveys, such as the Bank of England's quarterly Credit Conditions Survey, ask banks directly about availability. Inside a bank the gauge is the new-lending book itself: average leverage, covenant packages, and how many approvals are exceptions to policy.

A caravan maker, constructed for this page like every figure in it, earns EBITDA of US$8.0m in year 3, when lenders are competing for its sector. It borrows at 4.0 times on a five-year loan, repaying US$0.4m a year. In year 8 the loan matures, and lenders in the market will now go to 2.5 times. Both multiples are the example's, not market levels.

The third block is the one people expect. The second is the one that catches them. The business has returned to exactly what it earned when the loan was written, and it is still US$10m short, because the debt was sized on a multiple the market has stopped offering. Nothing about the caravan maker changed between years 3 and 8 except the year. The refinancing risk existed on the day the loan was signed; nothing in the year-3 figures forced the file to look at it.

Five years later the same earnings would support a smaller, amortising loan. A lender writing that loan is exposed to a different and milder version of the problem, because less has to be refinanced. It is the same borrower with a different vintage.

The economic cycle is about output and jobs. The credit cycle is about the supply and terms of borrowing. They overlap and need not move together. Research at the Bank for International Settlements finds that the financial cycle, read through credit and property prices, runs for considerably longer than the usual business cycle. A sector can run on a clock of its own as well: dairy prices or a construction slowdown can tighten credit to one industry while the rest of the book is easy.

Inside credit models the same distinction appears as point-in-time and through-the-cycle. A point-in-time default estimate moves with current conditions; a through-the-cycle estimate is held steadier across good and bad years. Provisions under expected-credit-loss accounting lean towards the first and capital estimates at internal-ratings banks towards the second, so reading the two together shows how much of a rating is the borrower and how much is the year.

DEFINED IN ONE PARAGRAPH → · DEBT TO EBITDA → · THE COVENANT RESET →

At the top, a credit paper is reading earnings at or near a high. The questions that matter are what those earnings look like normalised over the sector's last downturn, whether covenant headroom has been set against a peak, and what the loan looks like at maturity if terms have reverted. None of those needs a forecast of when the cycle turns. They ask whether the loan depends on it not turning.

Near the bottom the same file reads the other way. A lender that has stopped lending may decline a borrower whose earnings have already taken the worst of it. For a bank with a loan already on its book, the same cycle drives problem loan and workout activity: waivers, resets and refinancings arrive together, at the moment fewer lenders are willing to take the debt out.

REFINANCING A BUSINESS LOAN → · HOW BANKS MANAGE CREDIT RISK →