Stock market bubbles arise as a joint monetary and financial phenomenon. We assess the potential of monetary policy in mitigating the onset of bubbles by means of a Markov-switching Bayesian Vector Autoregression model estimated on US 1960–2019 data. Bubbles are detected and dated from the regime-specific interplay among asset prices, fundamental values, and monetary policy shocks. We rationalize the empirical evidence with an Overlapping Generations model, able to generate a bubbly scenario with shifts in monetary policy, and where agents form beliefs over transition dynamics. By matching the VAR impulse responses, we find that procyclicality and financial instability align with high equity premia and the presence of asset price bubbles. Monetary policy tightening, by increasing real rates, is ineffective in deflating bubble episodes.

This study investigates if the reaction function of the Federal Reserve switches between two distinct policy rules. Using a time-varying transition probability framework, we also determine if forward-looking macroeconomic or financial covariates signal an impending monetary regime switch. We find that US monetary policy is best described by a Markov-switching model with two regime processes, one of which controls for heteroskedasticity in the shocks to the policy rule. We find that the Fed switches between an aggressive regime with a relatively high weight on inflation and a dovish regime that is less responsive to inflationary pressures. We find that an increase in private forecasters’ expectations of an impending recession signals a switch from the more aggressive policy regime to the less aggressive regime. A recovery in equity returns signals a return back to the more aggressive regime.

Output Euler equations (OEE) for the US deliver slope estimates that are not significantly different from zero. This finding is counterintuitive as it implies a zero elasticity of intertemporal substitution (EIS) and aggregate demand movements that are nonresponsive to the short-term real interest rate. This paper shows that failure to account for regime changes in the dynamics of the real interest rate is responsible for this result. Based on a joint specification for the OEE and the real interest rate in an unobserved components model framework with Markov-switching parameters, the means, variances, and degrees of persistence of the real interest rate are different for the periods 1966–1980, 1980–1985, and 1985–2015. Once these regime changes are taken into account, the EIS estimate is 0.1 and no longer statistically insignificant. This finding is robust to alternative measures of the output gap as well as different specifications for the natural real interest rate.