Faculty of Actuarial Science and Insurance Seminar with Christian Hilpert

We study a dynamic signalling game where a firm, by its decision to stay solvent, signals its quality to a rating agency with the rating feeding back into the firm’s cost of capital. Observing the firm’s true cash flow blurred by a persistent measurement error, the error-minimizing rating agency learns dynamically through the firm’s solvency decision. Firms observed with higher measurement error default earlier, inducing directional learning by successively eliminating measurement errors which are too high to be feasible. In a partially separating perfect Bayesian equilibrium in Markov strategies, the firm employs a measurement-error dependent cut-off strategy. We discuss the extensive economic consequences of such a learning mechanism.