Downside Case: a Worked Example

A borrower's forecast is an argument for a plan. It is usually honest, and it is usually built by people who believe in the plan. A lender does not need to disprove it. The bank's money is at risk mainly in the years the plan does not happen, so it needs to see the loan in those years as well.

The order in many credit papers runs from the borrower's own management case, to a bank case in which the lender trims assumptions it finds generous, to a downside built on the bank case. Supervisors expect the discipline in their own terms. In Australia, APRA's prudential standard on credit risk requires authorised deposit-taking institutions to assess credit risk primarily on repayment capacity and to assess exposures under stressed conditions.

Start with the handful of drivers that make most of the difference: volume, price or margin, the time customers take to pay, the interest rate if the debt floats. Change each one alone first. Those single changes are sensitivities, and they show which line the loan depends on. Then change several at once, because a bad year seldom arrives one line at a time.

Two disciplines separate a useful downside from a decorative one. It has to run through to cash and the balance sheet, since a fall in sales works through tax, working capital and the revolver as well as through EBITDA. And its size needs a reason. The borrower's own history, its worst past season, or a stated event such as losing its largest customer, carries more weight than a round percentage chosen because it is round. Anything management says it would do in response, such as cutting costs or deferring capital spending, is better shown as a separate line so the reader can see how much depends on it.

A maker of roller garage doors that sells into new-house building, constructed for this page like every figure in it, forecasts sales of US$50m at a contribution margin of 30%, with fixed costs of US$8m. It owes US$20m at 7%, repays US$2m of principal a year and has US$2.5m undrawn on its revolver. Its covenants are cover of at least 1.20 times and debt of at most 3.50 times EBITDA. Tax is taken at 15% of EBITDA to keep the sums short, and capital spending is US$1.2m. These levels belong to the example and are no market norm.

The base case has cover of 1.40 times and leverage of 2.86 times. Sales are clearly the line the loan depends on. With all four together, leverage reaches 3.91 times before any further drawing, and both covenants fail. The year's shortfall of about US$1.2m would fit within the revolver, leaving about US$1.3m, but a covenant breach commonly lets a lender refuse further drawings. So at best the loan is serviced that year out of liquidity rather than earnings, and only if the bank chooses to go on lending into a breach. Six extra debtor days are taken as that case's sales ÷ 365 × 6, and lower sales are not credited with releasing any debtors.

Now run the same model backwards, with one variable moving and the rest held at base.

The business's sales fell 13% from peak to trough in the last housing downturn. Set beside these figures, that history matters more than any chosen shock: a repeat would take the business past the point where a year's cash pays its debt service.

The terms overlap and banks use them loosely. A sensitivity changes one thing. A downside case or scenario changes several, usually with a story attached. A stress test, in the sense regulators use it, is often run by the bank across a portfolio or the whole balance sheet to see what a severe economy does to its losses and capital, though the word is also used for a single borrower. A reverse stress test fixes the bad outcome first and solves for what produces it.

The backward reading is more useful to the borrower too. It replaces a debate about whether 8% is the right number with a figure both sides can hold: the room is 5.3% before the covenant, and 10.6% before the year's cash no longer covers debt service. How banks test borrower forecasts sets out the full sequence.

DEFINED IN ONE PARAGRAPH → · HOW BANKS TEST FORECASTS → · COVENANT HEADROOM →

A failed downside rarely ends a file; it usually reshapes it. Here the weakness is principal. If the term were lengthened so that US$1.2m is repaid each year, debt service falls to US$2.6m, base cover rises to 1.83 times, and cover of 1.00 times moves out to a sales fall of 16.9%, beyond what the business has lived through. In the all-four case cover improves to 0.85 times. The cost is that the debt stays outstanding longer.

The same work informs the other terms. Where the rate line bites, a lender looks at fixing or hedging part of the debt. The covenant level is set with the downside in view, so the bank hears about trouble before the money runs out. And the file records what was tested and why, since a credit committee reads the choice of shock as carefully as the answer.

DEBT SERVICE COVER → · OPERATING LEVERAGE → · HOW TO WRITE A CREDIT PAPER →