Model-based risk analysis at the ECB: Some illustrations - De Nederlandsche Bank
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Frank Smets Model-based risk analysis at
Directorate General Economics
With contributions from M. Darracq Pariès, M. Kühl, G. Müller,
the ECB: Some illustrations
F. Brenna, I. Moutachaker, J. Paredes, C. Montes
(DGE/FPM)
30 September 2019
The views expressed are my own and not necessarily those of the ECBModel-based
Rubric approaches to risk analysis: incorporating expert judgment
• Statistical risk metrics: combining predictive densities with judgment
Combine a range of models (including non-linear ones) to derive a tilted predictive
distribution (with the (B)MPE baseline as a median/mode)
Survey participants on second and/or third moment of the distribution: qualitative or
quantitative input (e.g. QRA)
Derive new tilted predictive distribution, imposing the surveyed moments
• Event-based risk analysis: assessing macroeconomic contingencies
Fully-fledged scenario analysis of selected risk events (potentially using a range of models)
Survey participants on the probability of each risk event
Derive new tilted predictive distribution, imposing the mean effect of the risk events
• Potential combination of the two approaches for the final risk balance indicators
2 www.ecb.europa.eu ©Predictive
Rubric densities around the growth and inflation projections (Sep 2019 MPE)
2020 Real GDP – Euro area 2020 HICP– Euro area
(annual growth rate, in %) (annual growth rate, in %)
0.8 0.8
QRA
0.7 0.7
NAWM II
Sep 2019 MPE
0.6 0.6
BVAR
BVAR median
0.5 0.5
BVAR skewed
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
-3 -2 -1 0 1 2 3 4 5 6 -3 -2 -1 0 1 2 3 4 5
Source: ECB projections database and ECB staff calculations based on: NAWM II, BVAR (density forecast from a combination of seven 3-variables BVAR models, with weights obtained by optimizing
the combined predictive likelihoods), BVAR median (same as BVAR, with median tilted to match the baseline), BVAR skewed (same as BVAR median, tilted to match also the skewness of QRA), QRA
(quantitative risk assessment of the September 2019 MPE baseline). For NAWM II, PCD is used instead of HICP.
3 www.ecb.europa.eu ©Model-based
Rubric risk sensitivity: predictive densities imposing QRA risk balance
Real GDP – Euro area HICP– Euro area
(y-o-y growth rates, in %) (y-o-y growth rates, in %)
3.5 3.5
September 2019 MPE
3.0 3.0
2.5
2.5
2.0
2.0
1.5
1.5
1.0
1.0
0.5
0.5
0.0
0.0
-0.5
-1.0 -0.5
-1.5 -1.0
2018 2019 2020 2021 2018 2019 2020 2021
Source: ECB projections database and ECB staff calculations. The shaded areas represent the 50% (dark grey) and the 90% (light grey) of a forecast density obtained from a combination of seven 3-variables BVAR
models with optimal weights, tilted so that the mean and the skewness match those of the Quantitative Risk Assessment (QRA).
5 www.ecb.europa.eu ©Model-based
Rubric risk sensitivity (NAWM II): Macroeconomic contingencies
Scenario Description
lower world demand and weaker external competitiveness consistent with the Brexit scenario included in
Subdued international the Special Feature 1 of the September 19 MPE Forecast Report and an escalation of the trade dispute
environment which drops exports and raises import tariffs similar to Special Feature “The resurgence of protectionism:
potential implications for global financial stability” in Financial Stability Review November 2018.
lower world demand and weaker external competitiveness consistent with the Brexit scenario included in
Brexit
the Special Feature 1 of the September 19 MPE Forecast Report
an escalation of the trade dispute which drops exports and raises import tariffs similar to Special Feature
Trade dispute “The resurgence of protectionism: potential implications for global financial stability” in Financial Stability
Review November 2018.
reflects the counterfactual paths for the short- and the long-term interest rates which are designed to bring
Financial set-back the financial conditions index (FCI) back to its historically tightest level since the APP announcement in
February 2016. See Annex 3 of September 19 MPE Forecast Report
Weaker bank lending pass-
reflects a scenario with a widening of credit spreads amounting to 15bp
through
reflects the counterfactual paths for the short- and the long-term interest rates which are designed to bring
Financial set-back and de- the financial conditions index (FCI) back to its historically tightest level since the APP announcement in
anchoring February 2016. See Annex 3 of September 19 MPE Forecast Report and a reduction in long-term inflation
expectations by 0.5 p.p.
reflects a scenario with a further 40bp drop in the long-term rates mimicking conditions by which all the
Stronger financial easing revisions in the yield curve since the June 19 BMPE would be explained by favourable interest rate shocks
in the NAWM
Gradual increase in government expenditures according to fiscal package calibration (2pp increase up to
Supportive fiscal policy
2021)
www.ecb.europa.eu ©Model-based
Rubric risk sensitivity (NAWM II) : Subdued international environment
Real GDP – Euro area HICP– Euro area
(y-o-y growth rates, in %) (y-o-y growth rates, in %)
3.5 3.5
September 2019 MPE
3.0 3.0
Subdued international environment
2.5
2.5
2.0
2.0
1.5
1.5
1.0
1.0
0.5
0.5
0.0
0.0
-0.5
-1.0 -0.5
-1.5 -1.0
2018 2019 2020 2021 2018 2019 2020 2021
Source: ECB projections database and ECB staff calculations based on NAWM II. The shaded areas
represent the 50% (dark grey) and the 90% (light grey) of a forecast density obtained from a combination of GDP growth Inflation
seven 3-variables BVAR models with optimal weights, tilted so that the mean and the skewness match those
of the QRA. Difference to baseline (percentage points) 2020 2021 2020 2021
Subdued international environment: lower world demand and weaker external competitiveness consistent
with a Brexit-type risk event and an escalation of the trade dispute which drops exports and raises import
tariffs similar to Special Feature “The resurgence of protectionism: potential implications for global financial Subdued international environment -0.6 -0.3 -0.1 -0.3
stability” in Financial Stability Review November 2018.
7 www.ecb.europa.eu ©Model-based
Rubric risk sensitivity (NAWM II) : Financial risk events
Real GDP – Euro area HICP– Euro area
(y-o-y growth rates, in %) (y-o-y growth rates, in %)
3.5 3.5
September 2019 MPE
3.0 Financial set-back
Stronger financial easing
3.0
2.5
2.5
2.0
2.0
1.5
1.5
1.0
1.0
0.5
0.5
0.0
0.0
-0.5
-1.0 -0.5
-1.5 -1.0
2018 2019 2020 2021 2018 2019 2020 2021
Source: ECB projections database and ECB staff calculations based on NAWM II. The shaded areas represent the
50% (dark grey) and the 90% (light grey) of a forecast density obtained from a combination of seven 3-variables GDP growth Inflation
BVAR models with optimal weights, tilted so that the mean and the skewness match those of the QRA.
Financial set-back reflects the counterfactual paths for the short- and the long-term interest rates which are Difference to baseline (percentage points) 2020 2021 2020 2021
designed to bring the financial conditions index (FCI) back to its historically tightest level since the APP
announcement in February 2016. Financial set-back -0.4 -0.5 -0.1 -0.2
Stronger financial easing reflects a scenario with a further 40bp drop in the long-term rates mimicking conditions
by which all the revisions in the yield curve since the June 19 BMPE would be explained by favourable interest rate
shocks in the NAWM II.
Stronger financial easing 0.4 0.5 0.0 0.1
8 www.ecb.europa.eu ©Model-based
Rubric risk sensitivity (NAWM II): Supportive fiscal policy
Real GDP – Euro area HICP– Euro area
(y-o-y growth rates, in %) (y-o-y growth rates, in %)
3.5 September 2019 MPE 3.5
Supportive fiscal policy
3.0 3.0
2.5
2.5
2.0
2.0
1.5
1.5
1.0
1.0
0.5
0.5
0.0
0.0
-0.5
-1.0 -0.5
-1.5 -1.0
2018 2019 2020 2021 2018 2019 2020 2021
Source: ECB projections database and ECB staff calculations based on NAWM II. The shaded areas GDP growth Inflation
represent the 50% (dark grey) and the 90% (light grey) of a forecast density obtained from a combination
of seven 3-variables BVAR models with optimal weights, tilted so that the mean and the skewness match Difference to baseline (percentage points) 2020 2021 2020 2021
those of the QRA.
Supportive fiscal policy reflects a scenario in which government expenditures increase gradually over
2020 and 2021 eventually reaching a 2% higher level compared to the baseline. This implies a fiscal Supportive fiscal policy 0.3 0.3 0.0 0.1
package of around 0.5% of GDP in line with available Fiscal space in DE and NL
9 www.ecb.europa.eu ©Model-based
Rubric risk sensitivity (NAWM II): Macroeconomic contingencies
Real GDP – Euro area HICP– Euro area
(y-o-y growth rates, in %) (y-o-y growth rates, in %)
3.5 3.5
September 2019 MPE
Financial set-back
3.0 Subdued international environment 3.0
Stronger financial easing
2.5 Supportive fiscal policy
2.5
2.0
2.0
1.5
1.5
1.0
1.0
0.5
0.5
0.0
0.0
-0.5
-1.0 -0.5
-1.5 -1.0
2018 2019 2020 2021 2018 2019 2020 2021
Source: ECB projections database and ECB staff calculations based on NAWM IIThe shaded areas represent the 50% (dark grey) and the 90% (light grey) of a forecast density obtained from a combination of seven 3-
variables BVAR models with optimal weights, tilted so that the mean and the skewness match those of the QRA.
Financial set-back: higher short- and long-term interest rates.
Subdued international environment: lower world demand and weaker external competitiveness consistent with the Brexit scenario included in the Special Feature 1 of the September 19 MPE Forecast Report and an
escalation of the trade dispute which drops exports and raises import tariffs similar to Special Feature “The resurgence of protectionism: potential implications for global financial stability” in Financial Stability Review
November 2018.
Stronger financial easing: lower long-term interest rates.
Supportive fiscal policy: increase in government consumption. 10 www.ecb.europa.eu ©Model-based
Rubric approaches to risk analysis: incorporating expert judgment
• Statistical risk metrics: combining predictive densities with judgment
Combine a range of models (including non-linear ones) to derive a tilted predictive
distribution (with the (B)MPE baseline as a median/mode)
Survey participants on second and/or third moment of the distribution: qualitative or
quantitative input (e.g. QRA)
Derive new tilted predictive distribution, imposing the surveyed moments
• Event-based risk analysis: assessing macroeconomic contingencies
Fully-fledged scenario analysis of selected risk events (potentially using a range of models)
Survey participants on the probability of each risk event
Derive new tilted predictive distribution, imposing the mean effect of the risk events
• Potential combination of the two approaches for the final risk balance indicators
11 www.ecb.europa.eu ©You can also read
