The recent global financial crisis has served as a stark reminder of the crucial role of systematic stress testing of financial institutions' portfolios, particularly, their lending books. In response to the regulatory deficiencies thus revealed, Basel III is seeking to achieve the broader macroprudential goal of protecting the banking sector from periods of excess credit growth by requesting longer horizon default probabilities, downturn loss-given-default measures and improved calibration of risk models.

The calibration of models that translate credit ratings and/or market data into default probabilities has direct implications for the determination of the risk-adjusted capital that banks need to hold to back their loans and safeguard solvency.

A Mixture of Markov Chains (MMC) approach is proposed to estimate credit rating migration risk that controls for the business-cycle evolution during the relevant time horizon in order to ensure adequate capital buffers both in good and bad times. The approach allows the default risk associated with a given credit rating to change as the economy moves through different points in the business cycle.

The findings suggest that ignoring business cycles significantly understates default risk during economic contraction. The MMC approach yields more reliable default risk measures than a naïve unconditional cyclical estimator, especially in contraction. The MMC approach is more efficient and provides superior out-of-sample credit rating migration risk predictions at long horizons the naïve approach.

An application to economic capital attribution suggests that the buffers prescribed by the MMC and naïve cyclical approaches are higher than those from classical through-the-cycle estimators, particularly in economic contraction. However, default risk during contraction (expansion) is statistically and economically overestimated (underestimated) by the naïve cyclical approach relative to the MMC approach and more so for longer prediction horizons.

Banks using the MMC estimator would counter-cyclically increase capital by 6% during economic expansion and free up to 17% capital for lending during downturns relative to the naïve estimator. The excess cyclicality in capital requirements associated to the naïve model would make lending very costly for banks in troubled times imposing too great a cost on economic growth and potentially aggravating a contraction.

This analysis has important implications for the ongoing financial regulatory reforms. The properties of the MMC estimator here documented are well aligned with the Basel III macroprudential initiative to dampen procyclicality by reducing the recession-versus-expansion gap in capital buffers and to lengthening the credit risk horizon.