Unconventional Monetary Policy and the Great Recession: Estimating the Macroeconomic Effects of a Spread Compression at the Zero Lower Bound Christiane Baumeister Luca Benati Bank of Canada University of Bern Fourth Annual IJCB Fall Conference Central Bank of Chile, 7-8 September 1 The views expressed only reflect those of the authors, and should not be attributed to the Bank of Canada.
How effective have central banks unconventional monetary policy actions been in the context of the financial crisis? How powerful is monetary policy at the zero lower bound, once all traditional ammunition has been exhausted? Two components: Two questions: How effective are asset purchase programs at compressing the yield spread at the zero lower bound? Gagnon et al. (IJCB 1), D Amico and King (1), Hamilton and Wu (JMCB 11), Neely (1) for US Bean (9), Meier (9), Joyce et al. (IJCB 1) for UK What is the macroeconomic impact of a compression in the yield spread for a constant nominal rate? Chung, Laforte, Reifschneider and Williams (JMCB 11) Lenza, Pill and Reichlin (EP 1)
Bayesian -varying structural VARs: Instability of macroeconomic series Structural relationships unaltered in face of crisis? Monetary VAR augmented with bond yield spread We identify: Empirical strategy three standard shocks (monetary policy, demand nonpolicy, supply) via sign restrictions a pure spread shock which, by construction, leaves the short (policy) rate unchanged Impact of spread shock on inflation and output growth for unchanged short rate which is what you have at ZLB is what we are interested in
Model Bayesian -varying parameter VAR with stochastic volatility (Cogley et al., AEJ Macro 1) All the parameters of the VAR are -varying They evolve according to random walks subject to reflecting barriers, to keep the VAR stable
Model (continued) The VAR features stochastic volatility for the VAR s reduced-form innovations the innovations to the VAR s random-walk coefficients
Identification Mix of sign restrictions and zero restriction on impact SHOCKS VARIABLE Monetary policy shock Spread shock Demand shock Supply shock Short rate + +? Spread?? Inflation + + GDP growth + + + Rubio-Ramirez, Waggoner and Zha (REStud 1) Deterministic rotation to impose a single zero restriction
First question: Have unconventional monetary policy actions averted catastrophic outcomes?
First question: Have unconventional monetary policy actions averted catastrophic outcomes? Our answer: Most likely, yes Gagnon et al. (IJCB 1): Impact of FED s asset purchase programs on 1-year government bond yield spread between 38 and 8 basis points. We take the average of their estimates basis points and we re-run the U.S. Great Recession with our SVAR, keeping the SVAR s structure and all shocks except the one to the spread unchanged; re-computing the spread shocks for 9 in such a way that the path for the 1-year spread is basis points higher than it has historically been.
3 1-1 - -3 - Inflation, actual and counterfactual Actual inflation 8Q Q 9Q Q 5-5 -1-15 - Real GDP growth, actual and counterfactual Actual real GDP growth Median 8th ile 1th ile 8Q Q 9Q Q Results for the U.S. Very high probability of deflation Significant risks of large output contraction In 199Q annualized output growth was -17.5% From 193Q1 to 193Q U.S. average output growth was -1.%
These results very likely understate true impact of FED s asset purchase program Why? If FED had done nothing, confidence would have collapsed, and demand non-policy shocks would have been much worse Output rebound in counterfactual simulation might not have been there
3.5 3.5 Inflation, actual and counterfactual corridor Real GPD growth, actual and counterfactual corridor 1 11 Unemployment rate, actual and counterfactual 1 1.5 1 - - 9.5 -.5 - -8 8 7-1 -1-1.5-1 8Q Q 9Q Q 8Q Q 9Q Q 8Q Q 9Q Q
- - - -8 Actual inflation Similar results for the United Kingdom: Inflation, actual and counterfactual -1 8Q Q 9Q Q - - - -8-1 -1-1 -1-18 Real GDP growth, actual and counterfactual Actual real GDP growth Median 8th ile 1th ile 8Q Q 9Q Q Charlie Bean (9): Our actions are compressing spreads by about ½ a age point. Conditional on Bean s guesstimate, the counterfactual suggests significant risks of both large output contraction and strong deflation Again, we think these results are probably optimistic
Second question: How powerful is a compression in the yield spread for an unchanged short rate?
Second question: How powerful is a compression in the yield spread for an unchanged short rate? In principle, this should be tackled by zeroing-out all coefficients in SVAR s monetary rule; shocking the spread; and tracing out the evolution of the economy Problem with this: It s vulnerable to Sargent s (Minneapolis FED Review 1979) critique: You can t manipulate the SVAR s policy rule! Indeed, Benati and Surico (AER 9), based on estimated DSGE model: Sargent s point is indeed empirically relevant So we show these results, but we also use an alternative approach, which is routinely used to compute constantinterest-rate projections
U.S.: median IRFs to a negative 1% shock to the 1-year Treasury bond spread, 195Q-11Q Interest rate Spread.5 basis points 3 1 1 197 198199 1 -.5-1 197 198 199 1 1 Inflation Output growth.5 3 1.5 1.5 1 199 198 197 1 1 1 199 198 197 1 Time variation for all IRFs, especially output growth and inflation
U.S.: IRFs to a negative 1% shock to the 1-year Treasury bond spread, selected quarters 197Q 1981Q 199Q 1998Q 9Q GDP deflator inflation 1 1 1 1 1 1 1 1 1 1 Real GDP growth 5 5 5 5 5 1 Quarters after shock 1 Quarters after shock 1 Quarters after shock 1 Quarters after shock 1 Quarters after shock Large impact on both inflation and output growth
U.K.: median IRFs to a negative 1% shock to the long-term bond yield spread, 1975Q-11Q Interest rate Spread basis points - - - 1 198 199 1 -.5-1 198 199 1 1 Inflation Output growth 1.5 1.5 -.5 1 199 198 1 1.5 1.5 -.5 1 199 198 1 Strongest reaction of inflation and output growth in most recent past...
Alternative strategy we shock the spread; we keep estimated VAR s structure unchanged that is, no manipulation of the SVAR s monetary rule; and for 8 quarters after impact, we choose a monetary policy shock such as to keep short rate unchanged However: potential shortcoming: ignores impact on agents expectations monetary policy shocks are all of the same sign practical problem: for some quarters, results are very volatile, due to feedback loop intrinsic in the exercise
U.S.: median IRFs to a negative 1% shock to the 1-year Treasury bond spread, 195Q-11Q Interest rate Spread basis points 15 1 5 1 197 198 199 1 197 198 199 1 1 Inflation Output growth 1 199 198 197 1 1 199 198 197 1
U.K.: median IRFs to a negative 1% shock to the long-term government bond yield spread, 1975Q-11Q Interest rate Spread 5 basis points -5 -.5-1 1 198 199 1-1.5 198 1985 199 1995 5 1 1 Inflation Output growth 1 5 199 1 198 15 1 199 198 1
Normal s vs. ZLB episode: U.S. 1 U.S. prices U.S. real GDP 1 15 8 1 5 197 198 199 1 197 198 199 1 without ZLB constraint with ZLB constraint U.S. output and prices react more strongly to a spread compression when the policy rate is held fixed...
Normal s vs. ZLB episode: U.K. 8 U.K. prices 1 U.K. real GDP 8 198 199 1 198 199 1 without ZLB constraint with ZLB constraint In the U.K., transmission of spread shocks pre-crisis not affected much by zero lower bound BUT macroeconomic consequences greatly amplified from early 9 onwards...
Conclusion Two things: a compression in the long-term yield spread exerts a powerful effect on both output growth and inflation counterfactual simulations suggest that U.S. and U.K. unconventional monetary policy have averted significant risks of both deflation and large output collapses