Ryohei Oishi

Welcome to my webpage!

I am a PhD student in Economics at University College London (UCL) and a cemmap PhD scholar

My research interests are in macroeconomics and macroeconometrics

You can find my CV here

Here are some tips to pronounce my first name

Working papers

Robust Inference on Macro Equations with Shock Proxies

I propose weak-instrument-robust inference methods on macroeconomic equations using shock proxies as instruments. Empirical applications include the Phillips curve and fiscal multipliers.

Abstract Paper Slide Example SSRN

This paper proposes weak-instrument-robust inference methods that achieve accurate size in finite samples for macroeconomic equations identified by structural shock proxies. The methods control the size of the Anderson-Rubin and Kleibergen's LM test statistics with Almon-parameterized instruments by using a restricted heteroskedasticity-and autocorrelation-robust long-run variance, computed under the null hypothesis. Applying these methods to the Phillips curve highlights the empirical importance of size accuracy: unlike existing methods that may introduce size distortion, the proposed methods do not reject either a completely flat or a very steep Phillips curve. Turning to fiscal policy rules, I find that the tax multiplier is statistically significantly positive, and approximately one.

Structural Estimation with Unstructured Data

with Sara Casella, Jesús Fernández-Villaverde, Stephen Hansen, and Minchul Shin

We develop a framework that integrates unstructured data, such as text, into DSGE model estimation. The method is applied to a medium-scale DSGE model with FOMC transcripts.

Abstract

Standard macroeconomic data do not cleanly separate the systematic and nonsystematic components of monetary policy; we show that incorporating unstructured text data into the structural estimation of a DSGE model sharpens this distinction by augmenting a standard state-space model with a non-core measurement block linking structural shocks to time series derived from FOMC transcripts, using a spike-and-slab prior to let the data select which series are informative; in a medium-scale New Keynesian model for the U.S., incorporating text improves predictive performance and materially alters structural inference, with estimated Taylor-rule coefficients falling and the Calvo parameter dropping, implying a steeper Phillips curve.

On Simulation Pseudo-Bias and Truncation in the Modified Harmonic Mean Estimator

The modified harmonic mean estimator exhibits a smaller simulation pseudo-bias than the original. The pseudo-bias-correction method provides a computational rationale for the tail truncation commonly used in practice. Empirical applications include the reassessment of Smets and Wouters (2007).

Abstract Paper Slide SSRN

The modified harmonic mean estimator developed by Gelfand and Dey (1994) for the marginal likelihood exhibits a smaller computational bias—known as simulation pseudo-bias—than the original estimator proposed by Newton and Raftery (1994). Simulation pseudo-bias occurs in importance sampling when the simulated samples from the proposal distribution have narrower empirical support than the target distribution, even if their theoretical supports are identical. The modified estimator mitigates this pseudo-bias by changing the importance sampling target from the prior density to a weighting density parameterized by posterior samples. Furthermore, the pseudo-bias correction method provides a computational rationale for the truncation embedded in commonly used weighting densities. Empirically, this paper revisits the finding of Smets and Wouters (2007) that the marginal likelihood of their DSGE model is comparable to that of a VAR model. Using various harmonic-mean-type estimators, I confirm the robustness of their original finding, which was based on the local Laplace approximation. I also demonstrate that relying on pseudo-biased estimators leads to an erroneous preference for the DSGE model.

Working papers at the Bank of Japan

Sticky Information Versus Sticky Prices Revisited: A Bayesian VAR-GMM Approach

with Takushi Kurozumi and Willem Van Zandweghe

We compare Phillips curves with sticky information, sticky prices, and unchanged prices in each period by using Bayesian VAR-GMM and find that all are essential to describe US inflation.

Abstract Paper

Several Phillips curves based on sticky information and sticky prices are estimated and compared using Bayesian VAR-GMM. This method derives expectations in each Phillips curve from a VAR and estimates the Phillips curve parameters and the VAR coefficients simultaneously. Quasi-marginal likelihood-based model comparison selects a dual stickiness Phillips curve in which, each period, some prices remain unchanged, consistent with micro evidence. Moreover, sticky information is a more plausible source of inflation inertia in the Phillips curve than other sources proposed in previous studies. Sticky information, sticky prices, and unchanged prices in each period are all needed to better describe inflation dynamics.

A Comparison of Japanese and US New Keynesian Phillips Curves with Bayesian VAR-GMM

with Takushi Kurozumi

We find that Japan's inflation dynamics are best described by variable demand elasticity with higher persistence in inflation expectations and lower trend inflation than in the US.

Abstract Paper

We compare Japanese and US inflation dynamics during the post-Global Financial Crisis period by utilizing Bayesian VAR-GMM to estimate several specifications of the New Keynesian Phillips curve. With the estimation method, we derive expectations in the Phillips curve from a VAR and analyze the formation of inflation expectations explicitly. We select the specification with variable elasticity of demand for Japan and that with sticky information for the US, using quasi-marginal likelihood. The selected specifications show that the persistence of inflation expectations formation is higher and trend inflation is lower in Japan than in the US. These findings account for persistently weak inflation developments in Japan: in the presence of firms' cautious price-setting behavior that reflects the purchasing attitude of consumers who are sensitive to price increases, inflation remains low and induces, through the highly persistent formation of inflation expectations, low expected future inflation and hence low trend inflation, which in turn put downward pressure on present inflation through the Phillips curve.

Some Work in Progress

Fiscal adjustments and optimal monetary policy in HANK

Ryohei Oishi

Welcome to my webpage!

Working papers

Working papers at the Bank of Japan

Some Work in Progress

Other articles

Education

Professional Experience

Email