Did the introduction of Inflation Targeting represent a Regime - - PowerPoint PPT Presentation

Introduction Model Estimation and Results Conclusions

Did the introduction of Inflation Targeting represent a Regime Switch of Monetary Policy in Latin America?

Sebastián Cadavid Sánchez CEMLA Alberto Ortiz Bolaños CEMLA and EGADE XXII Annual Meeting of the Central Bank Researchers Network November 5, 2017

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Outline

1

Introduction

2

Model

3

Estimation and Results

4

Conclusions

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Inflation, central bank reforms, exchange rate flexibility and inflation targeting regime

Table: Inflation and central banks changes in selected countries of Latin America

Average inflation 1980-1989 1990-1999 2000-2009 2010-2015 Year of new CB legislation Exchange rate flexibility Year of IT introduction Sample starts

Brazil 121.7 147.1 6.6 6.2 No change since 1964 1999 1999 1996 Chile 19.9 11.8 3.5 3 1992 1999 1999 1996 Colombia 20.8 19.9 6.1 3.1 1992 1999 1999 1995 Mexico 53.1 18.3 5.1 3.6 1993 1995 2001 1981 Peru 111 78.5 2.6 3 1993 2002 2002 1995

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Monetary policy and inflation determination

Goals:

• i. analyze if these institutional changes were associated with

shifts in the monetary policy implementation of Brazil, Chile, Colombia, Mexico and Peru;

• ii. determine if the observed reduction of inflation is explained by

changes in the policy stance;

• iii. characterize the inflation determination process (expectations,

inertia and price rigidity). Challenge: the analysis of the policy stance and inflation process is complex as they are jointly determined with other macroeconomic variables. Strategy: estimate Markov-Switching open-economy DSGE models with monetary factors to:

Measure policies and exogenous shocks. Perform counterfactuals under different monetary policy stances.

Introduction Model Estimation and Results Conclusions

A Monetary Small Open Economy General Equilibrium Model

Model based on Gali and Monacelli (2005) and latter estimated by Lubik and Schorfheide (2007) for Commonwealth countries, and by Ortiz and Sturzenegger (2007) a set of large emerging market economies. Alstadheim et al. (2013) use a similar approach to test if some CB respond to exchange rate movements. Open-economy IS curve: yt = Etyt+1 [τ +α (2α)(1τ)](Rt Etπt+1 ρaat +αEt∆qt+1) +α (2α) 1τ τ Et∆y⇤

t+1

(1) Open-economy Phillips curve: πt = β 1+βχp,ξ sp

t

Etπt+1 + χp,ξ sp

t

1+βχp,ξ sp

t

πt1 +βα∆qt+1 α∆qt + κξ sp

t

τ +α (2α)(1τ) (yt yt) (2)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

A Monetary Small Open Economy General Equilibrium Model

Interest rate rule: Rt = ρR,ξ sp

t Rt1 +

⇣ 1ρR,ξ sp

t

⌘h ψπ,ξ sp

t πt +ψy,ξ sp t yt +ψ∆e,ξ sp t ∆et

i +σR,ξ vo

t εR,t

(3) Nominal exchange rate ⇣

]of LCU 1USD

⌘ determination: πt = ∆et +(1α)∆qt +π⇤

t

(4)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Model: External Sector and Technology

AR(1) process for the terms of trade ⇣

Pexports Pimports

⌘ : ∆qt = ρq∆qt1 +σq,ξ vo

t εq,t

(5) Evolution of foreign output : y⇤

t = ρy ⇤y⇤ t1 +σy ⇤,ξ vo

t εy ⇤,t

(6) Evolution of foreign inflation : π⇤

t = ρπ⇤π⇤ t1 +σπ⇤,ξ vo

t επ⇤,t

(7) Evolution of technology : at = ρaat1 +σa,ξ vo

t εa,t

(8)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Empirical Strategy

We estimate the previous model using macroeconomic data on inflation, interest rates, output growth, nominal exchange rate depreciation and changes in terms of trade from Brazil, Chile, Colombia, Mexico and Peru. We allow for endogenous structural breaks and classify regimes according to the relative weight of inflation in an interest rate reaction function. Observable Measurement Equation Shocks Output growth yt yt1 +at εa,t Inflation 4πt εy ⇤,t Nominal interest rate 4rt εR,t Nominal exchange rate depreciation ∆et επ⇤,t Changes in terms of trade ∆qt εq,t

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Solving and estimating the MS-DSGE model

To solve the system we use the Newton methods developed in Maih (2015) which extend the one proposed by Farmer et al. (2011)and concentrates in minimum state variable solutions of the form: Xt = Ω⇤ (ξ sp,θ sp,H)Xt1 +Γ ⇤ (ξ sp,θ sp,H)Zt (ξ vo,θ vo) (9) The presence of unobserved variables and unobserved Markov states of the Markov chains implies that the standard Kalman filter cannot be used to compute the likelihood, so we use the Kim and Nelson (1999) filter.

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Solving and estimating (cont.)

We use the Bayesian approach to estimate the model:

1 Use Maih (2015) algorithm to compute the likelihood

introducing non-linearities and unobserved chains employing the Kim et al. (1999) filter to compute the likelihood with prior distribution of the parameters.

2 Construct the posterior kernel with the estimates from the

Bee-gate optimizer routine.

3 We use the posterior mode as the initial value for the

Metropolis Hastings algorithm with 100,000 iterations.

4 Utilize mean and variance of the last 50,000 iterations from

(3) to run the main Metropolis Hastings algorithm.

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Brazil

Low inflation policy response πt = 0.69E {πt+1}+0.31πt1 0.054qt +1.64(yt yt1) rt = 061rt1 +(10.61)(1.04πt +0.88yt +0.044et) High inflaiton policy response πt = 0.85E {πt+1}+0.15πt1 0.054qt +2.95(yt yt1) rt = 0.76rt1 +(10.76)(3.49πt +0.30yt +0.044et)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Brazil

Strong shift in monetary policy during 1999 Q3 with the introduction of inflation targeting and the greater exchange rate flexibility after a 35% Real depreciation in 1999 Q1. This was accompanied by: a lower persistence of inflation, or equivalently higher importance of expected inflation, in current inflation determination, and a stepper Phillips curve with inflation more responsive to output gap reflecting smaller price stickiness. The analysis captures the 2002 depreciation and the Cardoso - da Silva government transition as a transitory change of the monetary policy regime from 2002 Q4 to 2003 Q3.

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Chile

Low inflation policy response πt = 0.66E {πt+1}+0.34πt1 0.054qt +0.25(yt yt1) rt = 0.49rt1 +(10.49)(0.87πt +0.43yt +0.074et) High inflaiton policy response πt = 0.83E {πt+1}+0.17πt1 0.054qt +0.31(yt yt1) rt = 0.92rt1 +(10.92)(2.73πt +0.56yt +0.084et)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Chile

Chile fully adopted inflation targeting in 1999, but as stated in Corbo et al. (2002) the scheme started to be implemented since the 1990s. The estimation captures a high response to inflation since the start of the sample in 1996 Q2. The exception was from 2008Q1 to 2009 Q4 were there was a marked shift in policy with smaller weight on inflation and larger weight on output during a stagflationary period. When moving from a high interest rate response to a low one, Chile had an increase in inflation inertia without changes in the slope of the Phillips curve. This relatively small slope reflects large price stickiness.

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Colombia

Low inflation policy response πt = 0.75E {πt+1}+0.25πt1 0.104qt +4.92(yt yt1) rt = 0.71rt1 +(10.71)(0.97πt +0.74yt +0.054et) High inflaiton policy response πt = 0.58E {πt+1}+0.42πt1 0.104qt +1.44(yt yt1) rt = 0.73rt1 +(10.73)(3.29πt +0.38yt +0.144et)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Colombia

Colombia experienced a strong shift in monetary policy during 2000Q1 shortly after the introduction of inflation targeting and the greater exchange rate flexibility. When moving to a high interest rate response to inflation, inertia became higher and the slope of the Phillips curve became flatter indicating larger price stickiness.

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Mexico

Low inflation policy response πt = 0.62E {πt+1}+0.38πt1 0.224qt +3.71(yt yt1) rt = 0.63rt1 +(10.63)(0.62πt +0.83yt +0.344et) High inflaiton policy response πt = 0.54E {πt+1}+0.46πt1 0.224qt +2.29(yt yt1) rt = 0.46rt1 +(10.46)(1.85πt +0.73yt +0.114et)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Mexico

The estimation assigns a high probability of responsive interest rate from 1988Q2 to 1988Q3, from 1992Q1 to 1994Q4 and from 1997Q2 onwards. As in Colombia, when moving to a high interest rate response to inflation, inertia became higher and the slope of the Phillips curve became flatter indicating larger price stickiness.

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Peru

Low inflation policy response πt = 0.86E {πt+1}+0.14πt1 0.024qt +0.29(yt yt1) rt = 0.63rt1 +(10.63)(0.92πt +0.56yt +0.154et) High inflaiton policy response πt = 0.87E {πt+1}+0.13πt1 0.024qt +2.54(yt yt1) rt = 0.70rt1 +(10.70)(1.91πt +0.41yt +0.174et)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Peru

After brief episodes of monetary tightening in 1997 Q4 to 1998 Q1 and in 1998 Q4, the monetary policy towards greater responsiveness to inflation started in 2002 Q1 which coincides with the adoption of the inflation targeting regime. Inflation is largely determined by expectations with small weight of the autoregressive component. The previous relatively flat Phillips curve became steeper in the high policy response regime indicating lower price stickiness.

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Counterfactuals

Table: Summary statistics

Variable Series Brazil Chile Colombia Mexico Peru Average SD CV Average SD CV Average SD CV Average SD CV Average SD CV Output growth Observed 0.64 1.26 1.97 3.85 4.21 1.09 3.44 4.27 1.11 2.26 5.73 2.53 4.65 3.31 0.71 High response 0.99 3.28 3.30 3.75 2.78 0.74 3.42 4.11 1.22 1.77 4.84 2.73 4.97 2.74 0.55 Low response 1.00 3.63 3.62 3.65 4.75 1.30 3.37 4.47 1.31 3.46 8.85 2.56 5.38 5.71 1.06 Inflation Observed 6.31 3.72 0.59 3.06 2.51 0.82 9.84 7.43 1.12 20.15 24.78 1.23 3.62 3.18 0.88 High response 3.89 2.73 0.70 2.93 1.90 0.65 7.13 4.93 1.04 11.08 7.65 0.69 3.34 2.12 0.64 Low response 15.73 5.80 0.37 3.14 3.88 1.23 17.13 12.26 0.89 26.83 23.69 0.88 6.89 8.68 1.26 Interest rate Observed 16.49 7.00 0.42 4.59 2.04 0.44 12.56 10.50 2.33 25.38 26.36 1.04 6.91 6.03 0.87 High response 10.69 3.84 0.36 4.76 1.44 0.30 9.01 6.41 1.61 18.90 13.06 0.69 5.43 2.61 0.48 Low response 12.59 4.17 0.33 4.76 3.02 0.63 12.58 8.48 1.57 30.81 27.84 0.90 4.42 5.26 1.19 Real interest rate Observed 10.18 7.58 0.74 1.53 2.15 1.40 2.72 5.13 2.43 5.23 9.08 1.74 3.29 5.73 1.74 High response 6.80 3.49 0.51 1.83 2.04 1.11 1.89 2.80 3.64 7.82 7.94 1.01 2.09 2.61 1.25 Low response

• 3.14

7.36 2.34 1.63 1.48 0.91

• 4.56

8.94 1.07 3.98 15.61 3.92

• 2.47

13.33 5.40 Nominal depreciation Observed 1.64 9.13 5.58 0.67 4.87 7.25 1.36 6.60 0.89

• 0.59

4.81 8.10 0.45 2.70 6.02 High response

• 0.60

8.61 14.26 0.72 2.86 4.00

• 3.15

8.44 1.34

• 5.59

9.92 1.77 0.25 3.29 13.40 Low response 11.24 9.23 0.82 1.28 7.50 5.88 6.33 9.15 0.77 5.74 14.88 2.59 3.77 8.15 2.16

Introduction Model Estimation and Results Conclusions

Robustness: Brazil

2 Markov Chains 3 Markov Chains (κ) 3 Markov Chains (ιp)

Low inflation rt = 0.61rt1 +(10.61)(1.04πt +0.88yt +0.044et) rt = 0.58rt1 +(10.58)(0.98πt +0.67yt +0.094et) rt = 0.72rt1 +(10.72)(3.34πt +0.42yt +0.084et) policy response πt = 0.69E {πt+1}+0.31πt1 0.054qt +1.64(yt yt1) High inflation rt = 0.76rt1 +(10.76)(πt +yt +4et) rt = 0.69rt1 +(10.69)(3.7πt +0.38yt +0.044et) rt = 0.63rt1 +(10.63)(0.71πt +0.31yt +0.014et) policy response πt = 0.85E {πt+1}+0.15πt1 0.054qt +2.95(yt yt1) Low Phillips curve slope πt = 0.71E {πt+1}+0.28πt1 0.094qt +2.29(yt yt1) High Phillips curve slope πt = 0.79E {πt+1}+0.19πt1 0.094qt +3.61(yt yt1) Low price indexation πt = 0.84E {πt+1}+0.15πt1 0.034qt +1.32(yt yt1) High price indexation πt = 0.68E {πt+1}+0.31πt1 0.034qt +1.32(yt yt1)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Robustness: Chile

2 Markov Chains 3 Markov Chains (κ) 3 Markov Chains (ιp) Low inflation rt = 0.49rt1 +(10.49)(0.87πt +0.43yt +0.074et) rt = 0.48rt1 +(10.48)(0.86πt +0.61yt +0.084et) rt = 0.47rt1 +(10.62)(0.93πt +0.55yt +0.284et) policy response πt = 0.66E {πt+1}+0.34πt1 0.054qt +0.25(yt yt1) High inflation rt = 0.92rt1 +(10.92)(2.73πt +0.56yt +0.084et) rt = 0.91rt1 +(10.91)(2.46πt +0.55yt +0.094et) rt = 0.47rt1 +(10.47)(2.63πt +0.57yt +0.094et) policy response πt = 0.83E {πt+1}+0.17πt1 0.054qt +0.31(yt yt1) Low Phillips curve slope πt = 0.88E {πt+1}+0.10πt1 0.044qt +0.27(yt yt1) High Phillips curve slope πt = 0.81E {πt+1}+0.18πt1 0.044qt +0.47(yt yt1) Low price indexation πt = 0.88E {πt+1}+0.11πt1 0.054qt +0.47(yt yt1) High price indexation πt = 0.69E {πt+1}+0.31πt1 0.054qt +0.47(yt yt1)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Colombia

2 Markov Chains 3 Markov Chains (κ) 3 Markov Chains (ιp)

Low inflation rt = 0.71rt1 +(10.71)(0.97πt +0.74yt +0.054et) rt = 0.69rt1 +(10.69)(0.58πt +0.85yt +0.044et) rt = 0.65rt1 +(10.65)(1.16πt +0.43yt +0.094et) policy response πt = 0.75E {πt+1}+0.25πt1 0.104qt +4.92(yt yt1) High inflation rt = 0.73rt1 +(10.73)(3.29πt +0.38yt +0.144et) rt = 0.75rt1 +(10.75)(3.38πt +0.26yt +0.154et) rt = 0.70rt1 +(10.70)(2.78πt +0.32yt +0.114et) policy response πt = 0.58E {πt+1}+0.42πt1 0.104qt +1.44(yt yt1) Low Phillips curve slope πt = 0.66E {πt+1}+0.33πt1 0.104qt +0.61(yt yt1) High Phillips curve slope πt = 0.70E {πt+1}+0.29πt1 0.104qt +3.12(yt yt1) Low price indexation πt = 0.80E {πt+1}+0.18πt1 0.074qt +1.45(yt yt1) High price indexation πt = 0.54E {πt+1}+0.45πt1 0.074qt +1.45(yt yt1)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Robustness: Mexico

2 Markov Chains 3 Markov Chains (κ) 3 Markov Chains (ιp) Low inflation rt = 0.63rt1 +(10.63)(0.62πt +0.83yt +0.344et) rt = 0.77rt1 +(10.77)(0.88πt +0.59yt +0.294et) rt = 0.64rt1 +(10.64)(0.93πt +0.54yt +0.604et) policy response πt = 0.62E {πt+1}+0.38πt1 0.224qt +3.71(yt yt1) High inflation rt = 0.46rt1 +(10.46)(1.85πt +0.73yt +0.114et) rt = 0.52rt1 +(10.52)(2.07πt +0.41yt +0.244et) rt = 0.76rt1 +(10.76)(2.94πt +0.48yt +0.164et) policy response πt = 0.54E {πt+1}+0.46πt1 0.224qt +2.29(yt yt1) Low Phillips curve slope πt = 0.53E {πt+1}+0.47πt1 0.134qt +1.09(yt yt1) High Phillips curve slope πt = 0.53E {πt+1}+0.47πt1 0.134qt +5.65(yt yt1) Low price indexation πt = 0.81E {πt+1}+0.18πt1 0.154qt +2.10(yt yt1) High price indexation πt = 0.50E {πt+1}+0.50πt1 0.154qt +2.10(yt yt1)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Robustness:Peru

2 Markov Chains 3 Markov Chains (κ) 3 Markov Chains (ιp) Low inflation rt = 0.63rt1 +(10.63)(0.92πt +0.56yt +0.154et) rt = 0.60rt1 +(10.60)(0.80πt +0.63yt +0.124et) rt = 0.79rt1 +(10.79)(0.75πt +0.93yt +0.154et) policy response πt = 0.86E {πt+1}+0.14πt1 0.024qt +0.29(yt yt1) High inflation rt = 0.70rt1 +(10.70)(1.91πt +0.41yt +0.174et) rt = 0.72rt1 +(10.72)(2.04πt +0.44yt +0.174et) rt = 0.66rt1 +(10.66)(2.20πt +0.49yt +0.174et) policy response πt = 0.87E {πt+1}+0.13πt1 0.024qt +2.54(yt yt1) Low Phillips curve slope πt = 0.91E {πt+1}+0.09πt1 0.494qt +1.61(yt yt1) High Phillips curve slope πt = 0.90E {πt+1}+0.10πt1 0.494qt +3.11(yt yt1) Low price indexation πt = 0.85E {πt+1}+0.13πt1 0.134qt +0.09(yt yt1) High price indexation πt = 0.86E {πt+1}+0.12πt1 0.134qt +0.09(yt yt1)

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Final remarks

The introduction of inflation targeting is associated with a marked regime switch towards a more reactive interest rate policy. The counterfactual analysis allows us to say that the regime switches towards more responsive interest rate reaction functions helped to avoid many inflationary runs, several large nominal exchange rate depreciations and large volatility of the nominal variables. In Colombia and México, when moving from a low to a high interest rate response to inflation, inertia became higher and the slope of the Phillips curve became flatter indicating larger price stickiness.

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

Introduction Model Estimation and Results Conclusions

Final remarks

Brazil had the opposite behavior, perhaps linked to the quasi-fixed exchange rate regime prevailing from 1994 to 1999 following the Plano Real that tried to combat inflation inertia. Meanwhile, Chile exhibits a high degree of price stickiness. Peru registers low persistence of inflation.

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

References References

Alstadheim, R., Bjørnland, H. C., and Maih, J. (2013). Do central banks respond to exchange rate movements? a markov-switching structural investigation. Farmer, R. E., Waggoner, D. F., and Zha, T. (2011). Minimal state variable solutions to markov-switching rational expectations

• models. Journal of Economic Dynamics and Control,

35(12):2150–2166. Gali, J. and Monacelli, T. (2005). Monetary policy and exchange rate volatility in a small open economy. The Review of Economic Studies, 72(3):707–734. Kim, C.-J., Nelson, C. R., et al. (1999). State-space models with regime switching: classical and gibbs-sampling approaches with

• applications. MIT Press Books, 1.

Lubik, T. A. and Schorfheide, F. (2007). Do central banks respond to exchange rate movements? a structural investigation. Journal

• f Monetary Economics, 54(4):1069–1087.
• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?

References References

Maih, J. (2015). Efficient perturbation methods for solving regime-switching dsge models. Ortiz, A. and Sturzenegger, F. (2007). Estimating sarb’s policy reaction rule. South African Journal of Economics, 75(4):659–680.

• S. Cadavid Sánchez and A. Ortiz Bolaños

Did the introduction of IT represent a RS of MP in LA?