An estimated Markov-switching DSGE modeling framework that allows for parameter shifts across regimes is employed to test the hypothesis of regime-dependent credibility of Hong Kong's linked exchange rate system. The baseline model distinguishes two regimes with respect to the time-series properties of the risk premium. Regime-dependent impulse responses to macroeconomic shocks reveal substantial differences in spreads. To test the sensitivity of the results, a number of robustness checks are performed. The findings contribute to efforts at modeling exchange rate regime credibility as a nonlinear process with two distinct regimes.