Banque nationale de Belgique SA · EURONEXT:BNB

Banque nationale de Belgique SA — National Bank of Belgium — Technical Presentation (October 2018)

42 slides from Banque nationale de Belgique SA's technical presentation, Oct 25, 2018

  1. Slide 1: Can inflation expectations in business or consumer surveys improve inflation forecasts?

    Can inflation expectations in business or consumer surveys improve inflation forecasts? Raı̈sa Basselier 1,* David de Antonio Liedo 2,∗ Jana Jonckheere 1,∗ Geert Langenus 1,∗ 1 Economics and Research Department, NBB 2 Research and Development (Statistics), NBB ∗ Prepared for the NBB Colloquium 2018:

  2. Slide 2: Table of contents

    Table of contents 1. Motivation 2. Data 3. Model 4. X-models at work 5. Out-of-sample validation 6. Conclusion NBB October 25, 2018 2 / 32

  3. Slide 3: Section 1: Motivation

    Section 1 Motivation NBB October 25, 2018 3 / 32

  4. Slide 4: Inflation in the euro area is consistently overestimated

    Inflation in the euro area is consistently overestimated Figure: ECB core inflation forecasts updates over time 1.8 1.6 1.4 1.2 1 𝒄𝒐𝒓𝒆 𝝅 0.8 𝟐𝟎𝟏𝟕 core 𝝅 𝒄𝒐𝒓𝒆 𝝅 𝒄𝒐𝒓𝒆 𝝅 𝟐𝟎𝟏𝟒 𝟐𝟎𝟏𝟓 𝟐𝟎𝟏𝟔 0.6 Dec 2013 June 2014 Dec 2014 June 2015 Dec 2015 June 2016 Dec 2016 June 2017 Dec…

  5. Slide 5: In Belgium, we have the opposite case

    In Belgium, we have the opposite case Figure: NBB core inflation forecasts updates over time 2.1 1.9 𝒄𝒐𝒓𝒆 𝝅 𝒄𝒐𝒓𝒆 𝝅 1.7 𝟐𝟎𝟏𝟓 𝟐𝟎𝟏𝟔 core 𝝅 1.5 𝟐𝟎𝟏𝟒 𝒄𝒐𝒓𝒆 𝝅 1.3 𝟐𝟎𝟏𝟕 1.1 0.9 Dec 2013 June 2014 Dec 2014 June 2015 Dec 2015 June 2016 Dec 2016 June 2017 Dec 2017 NBB October…

  6. Slide 6: Section 2: Data

    Section 2 Data NBB October 25, 2018 6 / 32

  7. Slide 7: Dataset for the Large Model

    Dataset for the Large Model Name Description Source Total inflation (πt ) HICP, all items, SA, quarterly growth Eurostat Core inflation (πtcore ) HICP, all items excluding energy and food Eurostat Import deflator (πtM ) Quarterly growth rate of the import deflator Eurostat Oil (in euros) (PtOil ) Br

  8. Slide 8: Dataset for the Large Model

    Dataset for the Large Model Name Description Source Total inflation (πt ) HICP, all items, SA, quarterly growth Eurostat Core inflation (πtcore ) HICP, all items excluding energy and food Eurostat Import deflator (πtM ) Quarterly growth rate of the import deflator Eurostat Oil (in euros) (PtOil ) Br

  9. Slide 9: Dataset for the Large Model

    Dataset for the Large Model Name Description Source Total inflation (πt ) HICP, all items, SA, quarterly growth Eurostat Core inflation (πtcore ) HICP, all items excluding energy and food Eurostat Import deflator (πtM ) Quarterly growth rate of the import deflator Eurostat Oil (in euros) (PtOil ) Br

  10. Slide 10: Section 3: Model

    Section 3 Model NBB October 25, 2018 8 / 32

  11. Slide 11: Modeling inflation

    Modeling inflation Trend component πt = τtπ + λπ δt + ζπ ϑt + ηtπ Cyclical component Oil effects ut = τtu + κu δt + ηtu Ptoil = τtoil + λoil δt + ζoil ϑt + ηtoil NBB October 25, 2018 9 / 32

  12. Slide 12: Modeling inflation

    Modeling inflation Trend component πt = τtπ + λπ δt + ζπ ϑt + ηtπ Cyclical component Oil effects ut = τtu + κu δt + ηtu Ptoil = τtoil + λoil δt + ζoil ϑt + ηtoil NBB October 25, 2018 9 / 32

  13. Slide 13: Modeling inflation

    Modeling inflation Trend component πt = τtπ + λπ δt + ζπ ϑt + ηtπ Cyclical component Oil effects ut = τtu + κu δt + ηtu Ptoil = τtoil + λoil δt + ζoil ϑt + ηtoil NBB October 25, 2018 9 / 32

  14. Slide 14: Modeling inflation

    Modeling inflation Trend component πt = τtπ + λπ δt + ζπ ϑt + ηtπ Cyclical component Oil effects ut = τtu + κu δt + ηtu Ptoil = τtoil + λoil δt + ζoil ϑt + ηtoil NBB October 25, 2018 9 / 32

  15. Slide 15: Modeling inflation

    Modeling inflation Trend component πt = τtπ + λπ δt + ζπ ϑt + ηtπ Cyclical component Oil effects ut = τtu + κu δt + ηtu Ptoil = τtoil + λoil δt + ζoil ϑt + ηtoil NBB October 25, 2018 9 / 32

  16. Slide 16: Modeling inflation

    Modeling inflation Trend component πt = τtπ + λπ δt + ζπ ϑt + ηtπ Cyclical component Oil effects ut = τtu + κu δt + ηtu Ptoil = τtoil + λoil δt + ζoil ϑt + ηtoil NBB October 25, 2018 9 / 32

  17. Slide 17: Unobserved Components Model (UCM)

    Unobserved Components Model (UCM) πt = τtπ + λπ δt + ζπ ϑt + ηtπ (1) πtcore = τtπ + λcore δt−4 + ηtcore (2) πtm = τtm + λm δt + ζm ϑt + ηtM (3) Ptoil = τtoil + λoil δt + ζoil ϑt + ηtoil (4) ut = τtu + κu δt + ηtu (5) yt = τty + κy δt + ηty (6) StY = mS + κS (δt − δt−4 ) + ηtS (7) µt = mµ + κµ (δt −

  18. Slide 18: Unobserved Components Model (UCM)

    Unobserved Components Model (UCM) πt = τtπ + λπ δt + ζπ ϑt + ηtπ (1) πtcore = τtπ + λcore δt−4 + ηtcore (2) πtm = τtm + λm δt + ζm ϑt + ηtM (3) Ptoil = τtoil + λoil δt + ζoil ϑt + ηtoil (4) ut = τtu + κu δt + ηtu (5) yt = τty + κy δt + ηty (6) StY = mS + κS (δt − δt−4 ) + ηtS (7) µt = mµ + κµ (δt −

  19. Slide 19: Unobserved Components Model (UCM)

    Unobserved Components Model (UCM) πt = τtπ + λπ δt + ζπ ϑt + ηtπ (1) πtcore = τtπ + λcore δt−4 + ηtcore (2) πtm = τtm + λm δt + ζm ϑt + ηtM (3) Ptoil = τtoil + λoil δt + ζoil ϑt + ηtoil (4) ut = τtu + κu δt + ηtu (5) yt = τty + κy δt + ηty (6) StY = mS + κS (δt − δt−4 ) + ηtS (7) µt = mµ + κµ (δt −

  20. Slide 20: Linking surveys to expected factors: X-model

    Linking surveys to expected factors: X-model Key contribution: our framework connects the surveys (and any variable reflecting expectations) with the factors derived from the model. Remember, forecasts for inflation can be expressed as: π E πt+h |Ωt = E τt+h |Ωt + λπ E δt+h |Ωt + ζπ E ϑt+h |Ωt If St

  21. Slide 21: Linking surveys to expected factors: X-model

    Linking surveys to expected factors: X-model Cyclical components follow a stationary AR(2) process: δt = α1 δt−1 + α2 δt−2 + ζtδ ϑt = ρ1 δt−1 + ρ2 δt−2 + ζtϑ The trends related to inflation follow a random walk. Output and unemployment have smoother I(2) trends: τtu (1 − L)2 = ut τty (1 − L)2 = yt T

  22. Slide 22: Linking surveys to expected factors: X-model

    Linking surveys to expected factors: X-model Surveys may not be fully rational expectations forecasts Coibion et al. (2018)’s rational inattention proposal for survey respondents could result in: Stinat = E πt+h |Ωt + ηtinat ηtinat = ρηt−1 inat + inat t 4 X Since StC ∝ inat St+h the discrepancy ( ηt

  23. Slide 23: Set of models described in the paper

    Set of models described in the paper Variables included Small X-Small Large X-Large Total inflation (πt ) x x x x Core inflation (πtcore ) x x x x Import deflator (πtM ) x x Oil (in euros) (PtOil ) x x Real GDP (yt ) x x x x Unemployment (ut ) x x x x Mark-up (µt ) x x Activity surveys (Sty ) x x Bu

  24. Slide 24: Set of models described in the paper

    Set of models described in the paper Variables included Small X-Small Large X-Large Total inflation (πt ) x x x x Core inflation (πtcore ) x x x x Import deflator (πtM ) x x Oil (in euros) (PtOil ) x x Real GDP (yt ) x x x x Unemployment (ut ) x x x x Mark-up (µt ) x x Activity surveys (Sty ) x x Bu

  25. Slide 25: Section 4: X-models at work

    Section 4 X-models at work NBB October 25, 2018 15 / 32

  26. Slide 26: X-model does not reproduce the missing inflation puzzle for the euro area

    X-model does not reproduce the missing inflation puzzle for the euro area Figure: X-model for the euro area: core HICP 1.8 1.6 1.4 1.2 X-Model 1.0 𝒄𝒐𝒓𝒆 𝝅 𝟐𝟎𝟏𝟕 0.8 𝒄𝒐𝒓𝒆 𝝅 𝒄𝒐𝒓𝒆 𝝅 𝟐𝟎𝟏𝟓 𝟐𝟎𝟏𝟔 0.6 Dec 2013 June 2014 Dec 2014 June 2015 Dec 2015 June 2016 Dec 2016 June 2017 Dec…

  27. Slide 27: X-model emulates the NBB forecasts

    X-model emulates the NBB forecasts Figure: X-model for Belgium: core HICP 1.8 1.7 1.6 1.5 𝒄𝒐𝒓𝒆 𝝅 𝟐𝟎𝟏𝟓 1.4 1.3 X-Model 1.2 1.1 1.0 0.9 Dec 2013 June 2014 Dec 2014 June 2015 Dec 2015 June 2016 Dec 2016 June 2017 Dec 2017 NBB October 25, 2018 17 / 32

  28. Slide 28: X-model emulates the NBB forecasts

    X-model emulates the NBB forecasts Figure: X-model for Belgium: core HICP 1.9 1.8 1.7 𝒄𝒐𝒓𝒆 𝝅 1.6 𝟐𝟎𝟏𝟔 1.5 1.4 X-Model 1.3 1.2 1.1 1.0 0.9 Dec 2013 June 2014 Dec 2014 June 2015 Dec 2015 June 2016 Dec 2016 June 2017 Dec 2017 NBB October 25, 2018 18 / 32

  29. Slide 29: X-model emulates the NBB forecasts

    X-model emulates the NBB forecasts Figure: X-model for Belgium: core HICP 1.9 1.8 X-Model 1.7 1.6 1.5 1.4 𝒄𝒐𝒓𝒆 𝝅 1.3 𝟐𝟎𝟏𝟕 1.2 1.1 1.0 0.9 Dec 2013 June 2014 Dec 2014 June 2015 Dec 2015 June 2016 Dec 2016 June 2017 Dec 2017 NBB October 25, 2018 19 / 32

  30. Slide 30: X-model emulates the NBB forecasts

    X-model emulates the NBB forecasts Figure: X-model for Belgium: core HICP 2.1 1.9 X-Model 1.7 1.5 core 𝝅 𝟐𝟎𝟏𝟖 still unknown 1.3 1.1 0.9 Dec 2013 June 2014 Dec 2014 June 2015 Dec 2015 June 2016 Dec 2016 June 2017 Dec 2017 NBB October 25, 2018 20 / 32

  31. Slide 31: Section 5: Out-of-sample validation

    Section 5 Out-of-sample validation NBB October 25, 2018 21 / 32

  32. Slide 32: Value added of surveys for Belgium is mostly clear for core inflation

    Value added of surveys for Belgium is mostly clear for core inflation Figure: Belgium: relative RMSE for six models for total and core inflation Total inflation Core inflation 1.0 1.0 0.9 0.9 0.8 0.8 0.7 0.7 0.6 0.6 0Q 1Q 2Q 3Q 4Q 0Q 1Q 2Q 3Q 4Q Large Large X Large X2 Large X_noise Source: NBB. Note

  33. Slide 33: X-model has a more positive contribution from the business cycle

    X-model has a more positive contribution from the business cycle Figure: Belgium: one-step ahead forecasts for core inflation 1.0 0.25 0.20 0.8 0.15 0.10 0.6 0.05 Δ 0.4 0.00 -0.05 0.2 -0.10 -0.15 0.0 -0.20 -0.2 -0.25 2007 2009 2011 2013 2015 2017 2007 2009 2011 2013 2015 2017 Core inflation Large La

  34. Slide 34: X-model's forecasts are more connected with the business cycle

    X-model’s forecasts are more connected with the business cycle Figure: Belgium: one-step ahead forecasts for core inflation 1.0 4 3 smooth output gap 0.8 2 1 0.6 0 0.4 -1 -2 0.2 -3 -4 0.0 -5 -0.2 -6 2007 2009 2011 2013 2015 2017 2007 2009 2011 2013 2015 2017 Core inflation Large LargeX Large Large X

  35. Slide 35: Information from price-setters contributes most to these forecast improvements

    Information from price-setters contributes most to these forecast improvements Figure: Belgium: relative RMSE of business versus consumer survey-augmented models Total inflation Core inflation 1.0 1.0 0.9 0.9 0.8 0.8 0.7 0.7 0.6 0.6 0Q 1Q 2Q 3Q 4Q 0Q 1Q 2Q 3Q 4Q Large Large X BE_Large X_Bonly BE_Lar

  36. Slide 36: Using surveys deteriorates forecasts only during one year of the Great Recession

    Using surveys deteriorates the forecasts only during one year over the Great Recession, but the X2-model is the best over the recent sample Figure: Euro area: 4-steps ahead inflation forecasts 4.0 2.5 3.5 2.0 3.0 2.5 1.5 2.0 1.5 1.0 1.0 0.5 0.5 0.0 0.0 -0.5 -1.0 -0.5 2008 2009 2010 2011 2012 2013 20

  37. Slide 37: Relaxing the rationality assumption seems to help in recent period

    Relaxing the rationality assumption (X2 model and persistent noise) seems to help in recent period Figure: Euro area: relative RMSE for six models for total and core inflation Total inflation (EA) Core inflation (EA) 0.8 1.2 1.1 0.7 1.0 0.9 0.6 0.8 0.7 0.5 0.6 0.4 0.5 0Q 1Q 2Q 3Q 4Q 0Q 1Q 2Q 3Q 4Q L

  38. Slide 38: Surveys reduce the short-term impact of oil

    Surveys reduce the short-term impact of oil Figure: Euro area: one-step ahead forecasts for total inflation 2.0 0.8 1.5 0.6 0.4 1.0 0.2 0.5 Δ 0.0 0.0 -0.2 -0.5 -0.4 -1.0 -0.6 -1.5 -0.8 -2.0 -1.0 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

  39. Slide 39: According to the X-model, the gap closes earlier

    According to the X-model, the gap closes earlier (but the trend inflation is lower) Figure: Euro area: one-step ahead forecasts for total inflation 2.0 12 smooth output gap 1.5 9 1.0 6 0.5 3 0.0 0 -0.5 -3 -1.0 -6 -1.5 -9 -2.0 -12 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2007 2007 2008

  40. Slide 40: Section 6: Conclusion

    Section 6 Conclusion NBB October 25, 2018 30 / 32

  41. Slide 41: Concluding remarks (1/2)

    Concluding remarks (1/2) This paper explores the potential of consumer and business (price setters) surveys for forecasting inflation. Assuming surveys are model consistent is a strong assumption, but we show that business surveys do help to improve the forecasts for Belgian core inflation For the e

  42. Slide 42: Concluding remarks (2/2)

    Concluding remarks (2/2) The proposed methodology provides a parsimonious way to incorporate variables reflecting inflation expectations (cummulative or not) for any forecasting horizon R codes needed to specify and estimate this kind of models with the state-space methods of JDemetra+ have been mad

National Bank of Belgium — Technical Presentation (October 2018) on 6ix