Cash Flow at Risk (CFaR)

Cash flow at risk (CFaR) is a risk measure that estimates how far a company's cash flow over a future period, such as a quarter or a year, could fall below the expected amount, at a stated confidence level. A statement such as "our one-year CFaR at 95 per cent is $60 million" means that, under the model, there is a 5 per cent chance that cash flow will be at least $60 million below plan. It is used mainly by non-financial companies to understand how currency, commodity and interest rate movements affect their ability to fund operations and pay debts.

CFaR adapts the idea of value at risk (VaR) to companies whose main concern is cash generation rather than the market value of a trading portfolio. It is closely related to earnings at risk, which applies the same approach to reported profit.

A farmer planning next year knows the harvest might be good or bad, and grain prices might rise or fall. Instead of planning for the average year only, the farmer asks: in a bad year, one of the worst in twenty, how much less money would come in? If that shortfall would mean missing a loan payment, the farmer needs a bigger savings cushion or some protection against price falls.

Cash flow at risk asks the same question for a company. It looks at the things outside the company's control, such as exchange rates, fuel prices and interest rates, and works out how badly cash flow could miss the plan in a rare but realistic bad year. Managers then decide whether that risk is acceptable.

CFaR is the gap between expected cash flow and a low percentile of the simulated or estimated distribution:

• $E[\text{CF}]$ is the expected cash flow for the period.
• $\text{CF}_{1-\alpha}$ is the cash flow at the $(1-\alpha)$ percentile, for example the 5th percentile when $\alpha$ is 95 per cent.
• $\alpha$ is the confidence level.

Suppose a manufacturer expects operating cash flow of $200 million next year. A simulation of exchange rates, metal prices and interest rates, run 10,000 times, shows that in the worst 5 per cent of outcomes cash flow is $140 million or less. The 95 per cent CFaR is $200 million − $140 million = $60 million. If the model assumes cash flow is normally distributed with a standard deviation of $36 million, the shortcut is 1.645 × $36 million ≈ $59 million, close to the simulated figure.

• Bottom-up simulation. The company maps each cash flow to the risk factors that drive it, such as foreign exchange risk on export sales, commodity price exposure on raw materials and interest rate risk on floating-rate debt, and simulates those factors over the period, usually with Monte Carlo methods.
• Top-down or comparables approach. Cash flow volatility is estimated from the historical behaviour of the company or a group of similar companies, which captures business risks that are hard to model but gives less detail.
• Scenario overlays. Specific stress cases, such as a sharp currency devaluation, are run alongside the statistical figure.

The model's results depend on cash forecasting assumptions as much as on market data.

Treasurers use CFaR to judge whether the company can absorb bad outcomes without breaching loan covenants or cutting investment. It helps answer questions such as how large a liquidity buffer or committed credit line to hold, how much of an exposure to hedge, and whether a debt service coverage ratio could fall below a covenant level. Comparing CFaR with and without a proposed hedging programme shows how much the hedge reduces downside risk, and boards may express part of their risk appetite as a CFaR limit.

CFaR is only as good as the model of how cash flows respond to risk factors, and those links are often uncertain over long horizons. Correlations between factors can change in a crisis, distributions may have fatter tails than assumed, and the measure says nothing about how bad outcomes beyond the confidence level could be. It also tends to leave out operational and competitive risks unless they are added deliberately, so it is best used with stress tests and judgement rather than as a single target.