Seung Joo Lee, UC Berkeley Graduate Student, UCB Economics
Abstract: We develop a New-Keynesian framework with endogenous financial markets that features interesting new dynamics and potentials for second-order sunspot equilibria arising from the interactions among aggregate financial volatility, wealth, and aggregate demand. Based on a continuous-time framework, our analytical approach allows us to study the importance of higher order moments tied to financial uncertainty. We re-examine the optimality of conventional monetary policy rules and show the Taylor principle no longer guarantees model determinacy, and self-confirming financial uncertainty shocks cannot be ruled out by aggressive targeting of inflation and output gap alone. This pitfall of traditional rules lies in their inability to adequately target the economy’s risky rate, the relevant rate in a stochastic environment. We then propose a generalized Taylor rule that incorporates risk-premium and asset price targeting, providing necessary conditions that restore determinacy and achieve ultra-divine coincidence: the joint stabilization of inflation, output gap and risk-premium. Even when central banks do not perfectly follow our rule, the inclusion of financial targets improves welfare and speeds up the convergence towards economic and financial stabilization. Our study of financial stability at the zero lower-bound (ZLB) and the use of forward guidance and several other fiscal macroprudential tools reveal two interesting results: (i) By credibly committing to stabilize business cycle in the future, central banks can contain the volatility of financial markets at the ZLB. (ii) By credibly committing to NOT stabilize the future business cycle, central banks can boost the business cycle and asset prices today. To our knowledge, this constitutes a novel result that adds to the interesting trade-offs ingrained in forward guidance interventions. Finally, we provide empirical evidence on the importance of uncertainty shocks as drivers of the business cycle using US data, and show that our calibrated model performs reasonably well in matching the estimated impulse responses.