Reconciling narrative monetary policy disturbances with structural VAR model shocks?

Abstract

The history of monetary policy disturbances is important and widely used for policy analysis. This time series is essential for the historical decomposition of key macroeconomic variables such as output and prices since it uncovers the historical contribution of monetary policy to the business cycle. The sequence of monetary policy disturbances is moreover important for conducting counterfactual analyses to explore the role of monetary policy. Yet while the importance of the time series of monetary policy shocks is widely recognized, there is strong disagreement about its composition in the empirical literature. In the present paper, we attempt to identify and to quantify this discrepancy, which will be helpful to identify monetary policy shocks better in the future.The existent literature is divided into two prominent strands about identification of monetary policy shocks. One the one hand, structural vector autoregressive models (VAR) are used to shape the endogenous relationship between the policy rate and different economic variables, and monetary policy shocks are identified within the model. On the other hand, monetary policy shocks are identified outside an econometric model as narrative time series, e.g., by Romer and Romer (2004). We attempt to reconcile the monetary policy shock identified with a common structural VAR model and the narrative measure. To achieve this, we incorporate the narrative monetary policy shock account into the VAR model by treating it as a proxy for the structural monetary policy shock. Moreover, we quantify the extent to which the discrepancy still applies and identify two explanations for the disagreement. Alongside the potential measurement error in the narrative time series, as pointed out by the literature, we determine a potential misspecification of the VAR model as a second explanation. Nicht-technische Zusammenfassung University of Bonn AbstractStructural VAR studies disagree with narrative accounts about the history of monetary policy disturbances. We investigate whether employing the narrative monetary shock account as a proxy variable in a VAR model aligns both shock series. We quantify the extent to which the disagreement still applies and identify two explanations for the disagreement. One explanation is measurement error in the narrative time series, another is a misspecification of the VAR model.

Time-Varying Volatility, Financial Intermediation and Monetary Policy

“…The right column of Figure 2 shows that the impulse response functions in the two regimes for the main variables of interest are significantly different from each other, suggesting that the non-linear model specification is supported by the data. Mertens and Ravn (2013) show that under certain distributional assumptions about the measurement error it is possible to compute the correlation coefficient between the monetary policy shock instrument and the unobserved "true" monetary policy shock (see also Kliem and Kriwoluzky 2013). In our setting, this correlation equals 0.41 in the low volatility regime and 0.69 in the high volatility regime, which indicates that the shock instrument contains substantial information for identification.…”

Macroeconomic Shocks and Their Propagation

“…The proxy SVAR is another method that can be used to relax the recursiveness assumption. Kliem and Kriwoluzky (2013) Stock and Watson's (2012) and Mertens and Ravn's (2013) proxy SVAR method (see Section 2 for a description). Figure 3.2C shows the results for the original sample (1969 -1996) thereafter.…”

Identifying Structural VARs with a Proxy Variable and a Test for a Weak Proxy

“…As examples, Mertens and Ravn (2013) follow the narrative approach of Romer and Romer (2009) to construct proxy variables for tax shocks, Gertler and Karadi (2015) follow the high frequency approach of Gürkaynak, Sack, and Swanson (2005) to construct proxy variables for monetary policy shocks, and Montiel Olea, Stock, and Watson (2012) and Stock and Watson (2012) use 18 different proxies to identify shocks to oil, monetary policy, productivity, uncertainty, liquidity and financial risk, and fiscal policy. Finally, Mertens and Ravn (2014) show that it can be used to reconcile the differences between structural VAR and narrative estimates of tax multipliers; however, Kliem and Kriwoluzky (2013) argue that it is not able to reconcile structural VAR and narrative estimates of monetary policy shocks.…”