Banque nationale de Belgique SA · EURONEXT:BNB

Banque nationale de Belgique SA — Technical Presentation (June 2018)

37 slides from Banque nationale de Belgique SA's technical presentation, Jun 21, 2018

  1. Slide 1: Monetary Policy and Macroeconomic Stability Revisited

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Monetary Policy and Macroeconomic Stability Revisited Yasuo Hirose Takushi Kurozumi Willem Van Zandweghe Keio University Bank of Japan FRB Kansas City National Bank of Belgium June 21, 2018 1 / 37

  2. Slide 2: Outline

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Outline 1 I NTRODUCTION 2 T HE M ODEL 3 E STIMATION S TRATEGY 4 R ESULTS 5 C ONCLUSION 2 / 37

  3. Slide 3: Background

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Background What led to U.S. macroeconomic stability after the Great Inflation of the 1970s? A large literature: shift from indeterminacy to determinacy of equilibrium Achieved by the Fed’s policy change from a passive to an active

  4. Slide 4: Objective

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Objective Revisit the view on the shift from indeterminacy to determinacy by estimating a staggered price model with trend inflation. Even when the trend inflation rate is non-zero, a fraction of prices is kept unchanged in each pe

  5. Slide 5: Strategy

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Strategy The model is estimated during two periods, before 1979 and after 1982, allowing for both determinacy and indeterminacy. Bayesian method of Lubik and Schorfheide (2004). To evaluate the empirical performance of the model, i

  6. Slide 6: Strategy (cont.)

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Strategy (cont.) A difficulty in the method of Lubik and Schorfheide: The likelihood function is possibly discontinuous at the boundary of determinacy and indeterminacy regions of the parameter space. The RWMH algorithm can get stu

  7. Slide 7: New Findings

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION New Findings 1 The model empirically outperforms its canonical NK counterpart during both the pre-1979 and post-1982 periods. Justifies the use of the model instead of the NK counterpart. That some prices are unchanged in each peri

  8. Slide 8: New Findings (cont.)

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION New Findings (cont.) 3 The rise in the policy response to inflation alone does not suffice for explaining the shift to determinacy. Unless accompanied by either the fall in trend inflation or the change in the policy responses to t

  9. Slide 9: Comparison with Coibion and Gorodnichenko (2011)

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Comparison with Coibion and Gorodnichenko (2011) Estimated policy rule System estimation + calibrated model of entire model Canonical NK Clarida, Galí, and Gertler (2000) Lubik and Schorfheide (2004) Trend inflation Coibion and Gor

  10. Slide 10: Comparison with Coibion and Gorodnichenko (2011) (cont.)

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Comparison with Coibion and Gorodnichenko (2011) (cont.) Our paper offers another alternative view: the shift can be explained by a decrease in the policy response to the output gap and an increase in the response to output growth,

  11. Slide 11: Outline

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Outline 1 I NTRODUCTION 2 T HE M ODEL 3 E STIMATION S TRATEGY 4 R ESULTS 5 C ONCLUSION 11 / 37

  12. Slide 12: Household

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Household The representative household maximizes the utility function ∞ ( Z 1 ) X t 1 1+ η1 E0 β exp(zu,t ) log(C̃t − hCt−1 ) − lt (i) di , t=0 1 + η1 0 subject to the budget constraint Z 1 Pt C̃t + Bt = Pt Wt (i)lt (i)di + rt−1 Bt

  13. Slide 13: Final-Good Firm

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Final-Good Firm The representative final-good firm produces output Yt by choosing a combination of intermediate inputs {Yt (i)} to maximize profit Z 1 Pt Yt − Pt (i)Yt (i) di, 0 subject to the CES production technology Z 1 θ θ−1 θ−

  14. Slide 14: Intermediate-Good Firms

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Intermediate-Good Firms Each intermediate-good firm i produces one kind of differentiated good Yt (i) subject to the production function Yt (i) = At lt (i), where At is the technology level and follows the stochastic process log At

  15. Slide 15: Intermediate-Good Firms (cont.)

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Intermediate-Good Firms (cont.) Set prices on a staggered basis as in Calvo (1983). In each period, a fraction λ ∈ (0, 1) of firms keeps prices unchanged, while the remaining fraction 1 − λ sets prices in the following two ways: 1

  16. Slide 16: Central Bank

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Central Bank Monetary policy rule: log rt = φr log rt−1 " # + (1 − φr ) πt − log π) log r + φπ (log + zr,t +φx log xt + φ∆y log YYt−1 t − log a xt = YYtn is the output gap, where Ytn is the natural output. t Monetary policy shock:

  17. Slide 17: Log-Linearized Equilibrium Conditions

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Log-Linearized Equilibrium Conditions hη π̂t = γb π̂t−1 + γf Et π̂t+1 + κ ŷt + (ŷt − ŷt−1 + za,t ) + ψt , (a − h)(1 + η) (1) θ−1 ψt = βλπ Et ψt+1 + κf (Et ŷt+1 − ŷt + θEt π̂t+1 − r̂t ) , (2) h a ŷt = (ŷt−1 − za,t ) + (Et ŷt

  18. Slide 18: Canonical New Keynesian Model

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Canonical New Keynesian Model To evaluate the empirical performance of the model, its canonical NK counterpart is also estimated. Firms that would keep prices unchanged update prices using indexation to trend inflation as in Yun (1

  19. Slide 19: Outline

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Outline 1 I NTRODUCTION 2 T HE M ODEL 3 E STIMATION S TRATEGY 4 R ESULTS 5 C ONCLUSION 19 / 37

  20. Slide 20: Rational expectations solutions under indeterminacy

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Rational expectations solutions under indeterminacy A full set of LRE solutions (Lubik and Schorfheide, 2003): st = Φx (ϑ)st−1 + Φε (ϑ, M̃ )εt + Φζ (ϑ)ζt , 1 ζt ∼ N (0, σζ2 ): Sunspot shock 2 M̃ : Arbitrary matrix represents multip

  21. Slide 21: Bayesian Inference

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Bayesian Inference Estimate LRE model with full-information Bayesian approach of Lubik and Schorfheide (2004). Likelihood function is constructed for the indeterminacy region and determinacy region of the parameter space: p(X T |ϑ,

  22. Slide 22: Sequential Monte Carlo Algorithm

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Sequential Monte Carlo Algorithm The likelihood function is possibly discontinuous at the boundary of determinacy and indeterminacy regions. The posterior distribution is possibly multimodal. The RWMH algorithm can get stuck near a

  23. Slide 23: Data

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Data Data: real GDP growth rate; inflation rate of the GDP price deflator; federal funds rate Observation equations:       100∆ log Yt ā ŷt − ŷt−1 + za,t  100 log πt  =  π̄  +  π̂t , 100 log rt r̄ r̂t where ā = 100(

  24. Slide 24: Priors

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Priors Parameter Distribution Mean St. dev. ā Normal 0.370 0.150 π̄ Normal 0.985 0.750 r̄ Gamma 1.597 0.250 h Beta 0.700 0.100 ω Beta 0.500 0.150 λ Beta 0.500 0.050 φr Beta 0.750 0.100 φπ Gamma 1.500/1.100 0.750 φx Gamma 0.125 0.1

  25. Slide 25: Outline

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Outline 1 I NTRODUCTION 2 T HE M ODEL 3 E STIMATION S TRATEGY 4 R ESULTS 5 C ONCLUSION 25 / 37

  26. Slide 26: Posterior Estimates: Pre-1979 Period

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Posterior Estimates: Pre-1979 Period Baseline model NK counterpart Parameter Mean 90% interval Mean 90% interval ā 0.353 [0.156, 0.572] 0.387 [0.172, 0.619] π̄ 1.512 [1.189, 1.836] 1.349 [0.900, 1.794] r̄ 1.663 [1.395, 1.941] 1.58

  27. Slide 27: Posterior Estimates: Post-1982 Period

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Posterior Estimates: Post-1982 Period Baseline model NK counterpart Parameter Mean 90% interval Mean 90% interval ā 0.399 [0.211, 0.584] 0.410 [0.223, 0.576] π̄ 0.701 [0.537, 0.880] 0.679 [0.491, 0.873] r̄ 1.442 [1.168, 1.741] 1.3

  28. Slide 28: Model Selection: No Inflation Inertia?

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Model Selection: No Inflation Inertia? Examine whether abstracting from some properties of the model can improve the fit of the model even further. Cogley and Sbordone (2008): No empirical support for intrinsic inertia of inflation

  29. Slide 29: Posterior Estimates: ω = 0

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Posterior Estimates: ω = 0 Pre-1979 Post-1982 Parameter Mean 90% interval Mean 90% interval ā 0.379 [0.193, 0.555] 0.392 [0.221, 0.560] π̄ 1.447 [1.116, 1.768] 0.700 [0.560, 0.839] r̄ 1.641 [1.359, 1.920] 1.446 [1.173, 1.722] h 0.

  30. Slide 30: Model Selection: No Response to Output Gap?

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Model Selection: No Response to Output Gap? Policy response to the output gap φx decreased considerably. Examine whether this decrease suggests virtually no response to the output gap. The model with ω = 0 is further estimated by f

  31. Slide 31: Posterior Estimates for Subsequent Analysis

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Posterior Estimates for Subsequent Analysis Pre-1979 Post-1982 ω=0 ω = φx = 0 Parameter Mean 90% interval Mean 90% interval ā 0.379 [0.193, 0.555] 0.404 [0.231, 0.578] π̄ 1.447 [1.116, 1.768] 0.699 [0.556, 0.841] r̄ 1.641 [1.359,

  32. Slide 32: Sources of Shift from Indeterminacy to Determinacy

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Sources of Shift from Indeterminacy to Determinacy (a) Pre-1979 estimates of all model parameters 4.5 3 ° $ ?: 1.5 # Determinacy Indeterminacy 0 0 2 4 6 8 4π̄ (b) Post-1982 estimates of ? x and ? " ypost82 ×: (4π̄ pre79 , φpre79 π

  33. Slide 33: Sources of Shift from Indeterminacy to Determinacy (cont.)

    I NTRODUCTION Determinacy T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Indeterminacy 0 Sources0of Shift from 2 Indeterminacy 4 to Determinacy 6 8 (cont.) (b) Post-1982 estimates of ? x and ? " y 4.5 3 ° $ ?: 1.5 # 0 0 2 4 6 8 4π̄ (c) Post-1982 estimates of all model parameters ×: (4π̄ pre7

  34. Slide 34: Sources of Shift from Indeterminacy to Determinacy (cont.)

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION 0 Sources0of Shift from 2 Indeterminacy 4 to Determinacy 6 8 (cont.) (c) Post-1982 estimates of all model parameters 4.5 3 ° $ ?: 1.5 # 0 0 2 4 6 8 4π̄ Annualized trend inflation rate (%) ×: (4π̄ pre79 , φpre79 π ); ∗: (4π̄ pre79 ,

  35. Slide 35: Outline

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Outline 1 I NTRODUCTION 2 T HE M ODEL 3 E STIMATION S TRATEGY 4 R ESULTS 5 C ONCLUSION 35 / 37

  36. Slide 36: Conclusion

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Conclusion Revisited the view that US macroeconomic stability after the Great Inflation was achieved by the Fed’s policy change from a passive to an active response to inflation. Estimated a staggered price model with trend inflati

  37. Slide 37: Conclusion (cont.)

    I NTRODUCTION T HE M ODEL E STIMATION S TRATEGY R ESULTS C ONCLUSION Conclusion (cont.) The policy response to inflation was active even during the pre-1979 period, in addition to the post-1982 period. The rise in the response to inflation from the pre-1979 estimate to the post-1982 one alone does n

Banque Nationale De Belgique — Technical Presentation (June 2018) on 6ix