Monetary Policy under Balance Sheet Uncertainty

A group of developing countries bear high rates of financial dollarisation. Under this circumstance, monetary-policy makers are uncertain about the presence and scale of potentially harmful effects that might appear because of balance sheet mismatches arising from high and unexpected depreciations of the domestic currency. We build a setup whereby central bankers have two competing models in mind. Model A is a standard model for a small open economy whereas Model B has a builtin non-linear balance sheet effect. Whether the balance sheet mismatch problem exists or not, a Bayesian optimization procedure that assigns a positive probability to Model B, perpetuates model-indeterminacy. This happens because the optimal Bayesian regulator does not allow sizeable exchange rate swings (dirty floating), and therefore blurs the information to distinguish among models. We call this effect the Balance Sheet Trap. We show that, given the presence of the Balance Sheet Trap, introducing the learning dynamics into the central banker’s problem is optimal. Thus, we argue that intentional policy experimentation is highly desirable since it provides for an escape to the Balance Sheet Trap