Adverse Effects of Monetary Policy Signalling Jan FILÁČEK and Jakub MATĚJŮ Monetary Department Czech National Bank CNB Research Open Day, 18 th May 21
Outline What do we mean by adverse effects of monetary policy signalling? Motivation Why it may happen? Simple theoretical model How large is it? Empirical testing 2
What do we mean by AEMPS? Melosi (213), Rousakis (212) and others suggest that in the case of imperfect information among private agents, transmission of monetary policy (shocks) may be weaker or even reversed Reversed transmission empirically observed in some VAR models, Sims (1992), Eichenbaum (1992), Uhlig 2) price puzzle Similar concerns have arisen in practical policymaking close to ZLB IR cuts close to the ZLB are typically small in magnitude, but have large signalling potential But does this type of signalling achieve the desired targets? 3
Why it may happen? Framework: simple NK model with two types of agents fully and partially rational Specific type of bounded rationality, where agents behave optimally according to their respective FOCs, but are not able to form rational expectiations according to the general equilibrium model structure (do not know the model) Similar structure used by Massaro (213, JEDC) Fully rational -observe fundamentals -optimize given expectations -form rational expectations (know the model) Partially rational -observe fundamentals -optimize given expectations -unable to form rational expectations (do not know the model) -update expectations based on observations 4
Simple two-sector NK model IS curves of the fully and partially rational sectors: y i t = E i i y t+1 1 i σ t E i i π t+1 r n D t + ε t i F, P Phillips curves of the fully and partially rational sectors: π i t = E i i π t+1 + κy AG CP t + ε t i F, P Aggregation of the two sectors: y t AG = Ωy t F + (1 Ω)y t P π t AG = Ωπ t F + (1 Ω)π t P Shocks ε j t = ρ j j ε t 1 + ν t j i D, CP, MP
MP rule and expectations formation Monetary policy rule is forward-looking and known to all agents : i t = ρ + φ π E F AG {π t+1 } + φ y E F y AG t+1 + ε t MP Expectations of partially rational agents are updated using inverted MP rule E P P π t+1 E P P y t+1 = γ π i t ρ φ y E P AG y t+1 φ π = γ y i t ρ φ π E P AG π t+1 φ y + (1 γ π ) π t P + (1 γ y ) y t P Expectations of fully rational agents are fully rational: E F X = E{X} 6
Adding habit persistence Under habit persistence, the IS curves change to y F t = 1 1+χ EF F y t+1 + χ y F 1+χ t 1 1 χ σ(1+χ) i t E F F π t+1 r n D t + ε t y P t = 1 1+χ EP P y t+1 + χ y P 1+χ t 1 1 χ σ(1+χ) i t E P P π t+1 r n D t + ε t 7
Responses to D and CP shocks Demand shock (ν t D ) Cost-push shock (ν t CP ) 3 aggregate output 1 aggregate inflation aggregate output 3 aggregate inflation 2 1 -. 2 1 3 2 4 6 8 1 12 output of fully rational sector 1 2 4 6 8 1 12 inflation of fully rational sector -1 2 4 6 8 1 12 output of fully rational sector 3 2 4 6 8 1 12 inflation of fully rational sector 2 1 -.2 2 1 3 2 4 6 8 1 12 output of partially rational sector 1 2 4 6 8 1 12 inflation of partially rational sector -.4 2 4 6 8 1 12 output of partially rational sector 3 2 4 6 8 1 12 inflation of partially rational sector 2 1 -. 2 1 2 4 6 8 1 12 2 4 6 8 1 12-1 2 4 6 8 1 12 2 4 6 8 1 12 8
Reponses to MP shock Monetary policy shock (ν t MP ) aggregate output.1 aggregate inflation -.2 -.4 2 4 6 8 1 12 output of fully rational sector -.1 2 4 6 8 1 12 inflation of fully rational sector -.1 -. -.2.2 2 4 6 8 1 12 output of partially rational sector -.1.2 2 4 6 8 1 12 inflation of partially rational sector -.2 2 4 6 8 1 12 -.2 2 4 6 8 1 12 9
inflation MP shock, sensitivity to share of rational agents Ω output.6..4.2 -. -.1 -.2 1.9.8.7.6..4.3.2.1 omega time 1 1 -.1 1.9.8.7.6..4.3.2.1 omega time 1 1 1
inflation MP shock, sensitivity to learning parameters γ output.2..1 -. -.1 -.1 -.2.9.8.7.6..4.3.2.1 gammas time 1 1 -.1.9.8.7.6..4.3.2.1 gammas time 1 1 11
How large is it? Estimated equation: E t,i π t+h = δ i + λ PUB E t 1,i π t+h,i + λ CB CB E t 1,i + τ r TI t r t 1,i + τ E CB TI t E t 1,i π t+h,i π t+h,i SURP + φr t 1,i + τti t + β X t,i + ν t,i Control variables X for contemporaneous macro news (ER, inflation, GDP) If updating channel in place, φ should be signicantly positive If transparency mitigates the updating channel, τ r should be significantly negative similar equation for GDP expectations 12
Data FE regression on panel data of 12 IT countries (Canada, Czech Rep., Euro area, Hungary, Japan, Norway, Poland, Sweden, Switzerland, Turkey, UK and US) Private expectations from Consensus Forecasts Survey, covering time period 21-213, calendar years Central bank transparency calculated according to Eijfinger, Geraats(24) Approx. 11 observations Three alternative definitions of interest rate surprise: r t SURP = r t r t 1 E t 1 r t r t 1 r t SURP = r t r TAYLOR t where r TAYLOR t stands for backward-looking and forward-looking estimates of the Taylor rule 13
Results: expected inflation Effects of central bank decisions and forecasts on expected inflation Consensus F. Surprise Taylor BW Surprise Taylor FW Surprise L.Change of CPI forecast, CF.11 *** (6.7).929 *** (.2).17 *** (7.26) L.Change of CPI forecast, CB.972 *** (.17).922 *** (4.93).11 *** (.4) L.Policy rate surprise.432 *** (3.9).11 *** (4.99).92 *** (3.93) Transparency.836 (.13).318 (.6).31 (.1) Transp. # L.Policy rate surprise.614 (.78).111 (.92).347 (.29) Transp. # L.Change of CPI forecast, CB.274 (.28).124 (.12).46 (.46) LD.CPI Inflation.837 *** (9.21).8 *** (8.86).727 *** (7.26) D.NEER.24 (1.7).216 (.91).72 (.31) D.GDP growth vintage -.26 (-.6) -.66 (-1.39) -.19 (-.3) Constant.1 (.8).868 (.92) -.164 (-.3) Observations 117 111 939 R 2.197.1637.1823 t statistics in parentheses * p <.1, ** p <., *** p <.1 14
Results: expected GDP growth Effects of central bank decisions and forecasts on expected GDP growth Basic Surprise Taylor BW Surprise Taylor FW Surprise L.Change of GDP forecast, CF.49 *** (13.4).439 *** (14.21).432 *** (13.93) L.Change of GDP forecast, CB.776 *** (3.63).81 *** (3.83).728 *** (3.41) L.Policy rate surprise -.734 *** (-3.1) -.172 *** (-.33) -.179 *** (-.4) Transp. -.284 (-.86) -.28 (-.87) -.2 (-.76) Transp. # L.Policy rate surprise -.162 (-.16).319 ** (1.97).24 (1.3) Transp. # L.Change of GDP forecast, CB -.18 (-.92) -.117 (-1.) -.883 (-.7) LD.CPI Inflation.23 * (1.83).24 * (1.78).234 * (1.68) D.NEER.137 *** (4.24).14 *** (4.37).141 *** (4.37) D.GDP growth vintage.97 (1.2) 1.376 * (1.83) 1.133 (1.1) Constant -.38 *** (-.32) -.296 *** (-4.44) -.314 *** (-4.66) Observations 94 97 941 R 2.197.272.273 t statistics in parentheses * p <.1, ** p <., *** p <.1 1
Conclusions Adverse effects of monetary policy signalling (updating channel) can be established on both theoretical and empirical grounds Strength of updating channel depends on key model parameters (share of partially rational agents in the economy, their learning parameters) Empirical estimations suggest the presence of updating channel - inflation expectations tend to rise after interest rate hike GDP expectations react in opposite way corresponds to model and economic intuition 16
Thank you for your attention! jakub.mateju@cnb.cz
inflation output Monetary policy shock, sensitivity to habit persistence parameter χ.2.1.1 -.1 -.1 -.2 -.2 -.3.9.9.8.8.7.7.6.6.. chi time 1 1 -.3.9.9.8.8.7.7.6.6.. chi time 1 1 18
Summary statistics of key variables Mean sd min max Change of CPI forecast, CF -.9297.32191-3.1 2. Change of CPI forecast, CB.2763.37938-1.8 3.7 Policy rate surprise, CF.174932.4146-2.8 9.9 Policy rate surprise, Taylor BW -1.67e-9.382877-3.666882.3472 Policy rate surprise, Taylor FW 2.9e-9.243474-2.236 3.42 19