A credit rating transition matrix shows the probability of an obligor (issuer) transitioning from one credit rating to another within a specific period (typically one year). Squaring the one-year transition matrix essentially applies the matrix twice. Each application represents a potential rating change in a single year.

A is incorrect. The geometric mean of one-year probabilities does not correctly model the compounding effect required for multi-year transitions, as it typically underestimates the probability of remaining in the same state or moving to worse states.

B is incorrect. Exponential smoothing is used for forecasting trends rather than extending the matrix for a transition model, which requires a specific compound probability calculation.

D is incorrect. Updating the one-year matrix with recent default and recovery rates adjusts the matrix for current conditions but does not accurately project it over a second year without proper mathematical compounding, as required for transition matrices.

Things to Remember

• Extending the time horizon of a ratings transition matrix requires mathematical operations reflective of compound transitions, not simple averages or linear estimates.
• When dealing with transition matrices, it is crucial to apply the appropriate mathematical manipulation to accurately reflect the timeframe of interest.
• The squaring of a one-year matrix is a method based on the assumption of Markov chains, where the transition is only dependent on the current state and not past states.
• Potential for underestimating extreme migrations: Squaring the matrix might underestimate the probability of very large rating migrations (e.g., from BBB to default).