1. "Information Flows in a New Keynesian DSGE Model: An Empirical Investigation" (submitted)
In this paper, I formulate and estimate a new Keynesian model under four different information structures in monetary policy. The first informational structure is one where agents have no foresight and only adhere to current shocks. In the second structure, agents receive news about changes in future monetary actions. The third structure assumes that agents observe noisy signals about current structural shocks. In the final structure, agents receive noisy signals about future monetary policy shocks as well as current shocks. Analysis of U.S. data from 1967:Q1 to 2008:Q1 shows that data substantially prefers the noise-shock information structures. In addition, the last information structure generates inflation persistence consistent with data, with no reliance on backward price indexation. This finding suggests that information structures can substitute the inertial mechanisms commonly used in new Keynesian models by matching key dynamic properties of the data.
2. "Understanding the Differences in VAR- and DSGE-Based Government Spending Multipliers" (submitted)
In this paper, I compare government spending multipliers emerged from two conventional modeling approaches---estimated DSGEs and structural BVARs---by employing Bayesian Monte Carlo methods. Although similar data is used for both approaches, the resulting impact output multipliers for the BVARs are substantially larger than those of the estimated DSGEs. The DSGE multipliers, measured by simulation experiments, reveal that utilizing the structural VAR method does not directly impact the gap between both distinct sets of multipliers. This gap, as it turns out, is heavily influenced by the use of a DSGE model. These results are robust across various identification assumptions for government spending shocks, used in the structural VAR models.
3. "Monetary Policy Regime Shifts and the Effects of
This paper estimates a Markov-switching DSGE model augmented with the financial accelerator mechanism à la Bernanke et al. (1999), in which regime shifts are allowed in the monetary policy rule coefficients and volatilities of structural shocks. I find that the monetary policy responded more actively to inflation and output growth prior to the 1970s and after the early 1980s, whereas a less-active monetary regime was in place in between the periods. Several findings emerge based on the identified regimes. First, I show that the effects of the structural shocks depend critically upon the monetary regime. Technology, preference, and investment-efficiency shocks are more expansionary when the monetary regime is less-active, whereas a monetary policy shock becomes more contractionary under the regime. The presence of financial frictions amplifies the differences in impulse responses across the monetary regimes, particularly for preference shocks. Second, the contribution of investment-efficiency shocks to output and investment fluctuations rises significantly under the less-active monetary regime. This tendency is more pronounced when the financial accelerator mechanism is present. Third, allowing for regime shifts in monetary policy makes the model-implied risk premium series more consistent with actual time series frequently used as proxies of the premium. Finally, counterfactual analyses reveal that the trade-off between inflation and output is more profound in a DSGE model with financial frictions.
4. "No News is Good News" joint with Eric M. Leeper, Giacomo Rondina, and Todd B. Walker
We estimate a standard dynamic stochastic general equilibrium model under three dfferent information structures to assess the importance of these informational assumptions. In the first information structure, agents receive news about future structural shocks, as in Beaudry and Portier (2006) and Schmitt-Grohé and Uribe (2012); in the second structure, agents observe noisy signals about current structural shocks; in the third structure, agents do not observe either news or noise. Data overwhelming support the noise-shock information structure. News (noise) shocks shift spectral power from the lower (higher) end to the higher (lower) end of the spectrum, which forces internal propagation mechanisms to work harder (less hard) in models with news (noise) shocks. That data prefer noise shocks and the reallocation of spectral power to the lower end connects to Granger's (1969) "typical spectral shape" of macroeconomic variables. As a byproduct, the paper develops a novel estimation methodology for models with incomplete information.
This paper is accepted by the three economic conferences, the National Bureau
of Economic Research summer institute 2013 (Cambridge, MA), the Southern Economic Association annual meetings 2013 (Tampa, FL), and the American Economic Association annual meetings
2014 (Philadelphia, PA). A draft will be available soon.
5. "Fiscal Financing and The Effects of Government Spending: A VAR Approach"
Rational expectations imply that the macroeconomic responses to deficit-financed government spending critically hinge on economic agents' beliefs about how debt innovations are financed by fiscal instruments. This paper quantifies the government spending multipliers with consideration of the intertemporal aspect of government budget behavior and distortionary fiscal financing, using vector autoregressions. To this end, I establish a novel identification strategy of government spending shocks financed by future tax hikes. By applying the method to U.S. data from 1947 to 2007, I find that distorting fiscal financing substantially dampens the stimulus effects of government spending in both the short and medium runs. In addition, the identification strategy reveals that spending reversals---initial government spending increases financed by medium-run spending cuts---do not seem to happen in U.S. data.
6. "Testing the Expectations Hypothesis in Continuous-Time"
I propose and apply a new approach to test the expectations hypothesis. In particular, I test the standard excess bond return regressions in continuous-time to examine the empirical evidence for the expectations hypothesis across an infinitesimal time period. For our model's estimates, I use the martingale regression based on time change for inference on continuous-time conditional mean models. This method is quite intuitive in that it identifies the true parameter value simply by imposing the martingale condition for the error process. I find results in favor of the expectations hypothesis if it is tested in a continuous-time setup. This finding suggests that the empirical evidence for the expectations hypothesis depends to a remarkable degree on the sampling frequency of the data.
7. "An Examination of Macroeconomic Fluctuations in Korea Exploiting a Markov Switching DSGE Approach" joint with Jinho Choi
8. "The Differences between Estimated New Keynesian Models: A Nonlinear Bayesian Investigation" joint with with
Alexander W. Richter and Nathaniel A. Throckmorton