Real Time Early Warning Indicators Alessi Detken for Costly Asset - PowerPoint PPT Presentation

‘Real Time’ Early Warning for Costly Boom/Bust Cycles ‘Real Time’ Early Warning Indicators Alessi Detken for Costly Asset Price Boom/Bust Cycles: Introduction Methodology A Role for Global Liquidity Boom identification Data Evaluation Results Best indicators Lucia Alessi Carsten Detken Mid-2000s boom Conclusions European Central Bank Milan, 10 May 2010

Motivation The need for Early Warning Models ‘Real Time’ Early Warning The set up of an Early Warning System is one of the for Costly Boom/Bust key tasks of the European Systemic Risk Board Cycles Alessi Detken With the aim of identifying threats to financial stability in Introduction a timely manner, and allow for the adoption of targeted Methodology macro-prudential regulatory measures Boom identification Data Evaluation Results Implications of financial stability for price stability in the Best indicators Mid-2000s boom medium to long run Conclusions EWMs are necessary input for “leaning against the wind” New generation of EWMs

Introduction Literature and contributions ‘Real Time’ Early Warning Signalling approach of Early Warning Indicator Models for Costly Boom/Bust (e.g. Kaminsky/Reinhart 1999, AER) Cycles Alessi Detken Application to predict costly aggregate asset price boom/bust cycles Introduction Real vs financial variables Methodology Boom identification Global vs domestic financial variables Data Evaluation Money vs credit Results Best indicators Prediction whether mid-2000s asset price boom wave Mid-2000s boom would be costly Conclusions Pseudo Real Time Ranking according to policy maker loss function, not noise to signal ratio (similar to Bussière/Fratzscher 2008, JPM), and focus on usefulness

Country-specific boom identification Recursive detrending ‘Real Time’ Boom quarter if at least 3 consecutive quarters where Early Warning for Costly Boom/Bust QAAPR i > recursive HP trend i + 1 . 75 ∗ recursive stdev i Cycles Alessi Detken Introduction Methodology Boom identification Data Evaluation Results Best indicators Mid-2000s boom Conclusions

Country-specific boom identification Recursive detrending ‘Real Time’ Boom quarter if at least 3 consecutive quarters where Early Warning for Costly Boom/Bust QAAPR i > recursive HP trend i + 1 . 75 ∗ recursive stdev i Cycles Alessi Detken Introduction Methodology Boom identification Data 300 Evaluation 250 Results 200 Best indicators Mid-2000s boom 150 Conclusions 100 50 0 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 Asset Prices

Country-specific boom identification Recursive detrending ‘Real Time’ Boom quarter if at least 3 consecutive quarters where Early Warning for Costly Boom/Bust QAAPR i > recursive HP trend i + 1 . 75 ∗ recursive stdev i Cycles Alessi Detken Introduction Methodology Boom identification Data 300 Evaluation 250 Results 200 Best indicators Mid-2000s boom 150 Conclusions 100 50 0 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 Asset Prices Ex-Post Trend

Country-specific boom identification Recursive detrending ‘Real Time’ Boom quarter if at least 3 consecutive quarters where Early Warning for Costly Boom/Bust QAAPR i > recursive HP trend i + 1 . 75 ∗ recursive stdev i Cycles Alessi Detken Introduction Methodology Boom identification Data 300 Evaluation 250 Results 200 Best indicators Mid-2000s boom 150 Conclusions 100 50 0 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 Asset Prices Ex-Post Trend Rec. Trend

Country-specific boom identification Recursive detrending ‘Real Time’ Boom quarter if at least 3 consecutive quarters where Early Warning for Costly Boom/Bust QAAPR i > recursive HP trend i + 1 . 75 ∗ recursive stdev i Cycles Alessi Detken Introduction Methodology Boom identification Data 300 Evaluation 250 Results 200 Best indicators Mid-2000s boom 150 Conclusions 100 50 0 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 Asset Prices Ex-Post Trend Rec. Trend

Country-specific boom identification Recursive detrending ‘Real Time’ Boom quarter if at least 3 consecutive quarters where Early Warning for Costly Boom/Bust QAAPR i > recursive HP trend i + 1 . 75 ∗ recursive stdev i Cycles Alessi Detken Introduction Methodology Boom identification Data 300 Evaluation 250 Results 200 Best indicators Mid-2000s boom 150 Conclusions 100 50 0 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 Asset Prices Ex-Post Trend Rec. Trend

Country-specific boom identification Recursive detrending ‘Real Time’ Boom quarter if at least 3 consecutive quarters where Early Warning for Costly Boom/Bust QAAPR i > recursive HP trend i + 1 . 75 ∗ recursive stdev i Cycles Alessi Detken Introduction bridged periods if less than 3 quarters in between Methodology Boom identification booms artificially ended if QAAPR drops by more than Data Evaluation 35% Results Best indicators Mid-2000s boom 60 booms identified Conclusions results robust to classification

Country-specific boom identification High cost vs Low cost booms ‘Real Time’ Early Warning for Costly Boom/Bust High Cost Booms Cycles Alessi Detken if real GDP growth 1 pp p.a. lower than potential growth on Introduction average over 3 post boom years Methodology Boom identification Data Evaluation Results 45 classifiable booms: Best indicators Mid-2000s boom 29 are HC Conclusions 16 are LC (control group) costly banking crises (FI ’91-’94, IT ’90-’95, SE ’91-’94) follow HC booms

Boom/bust cycles 3 waves: mid-late 80s, 90s, 00s ‘Real Time’ Early Warning for Costly Boom/Bust Cycles Number of Countries with Aggregate Asset Price Booms Alessi Detken 14 Introduction 12 Methodology 10 Boom identification Data 8 Evaluation Results 6 Best indicators Mid-2000s boom 4 Conclusions 2 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 7 7 7 7 7 7 7 7 7 7 8 8 8 8 8 8 8 8 8 8 9 9 9 9 9 9 9 9 9 9 0 0 0 0 0 0 0 0 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 High Cost Booms Low Cost or Unclassified Booms

Data 18 variables for 18 countries ‘Real Time’ Early Warning for Costly Boom/Bust Cycles Real: GDP , private consumption, total investment, Alessi Detken housing investment, consumer prices Introduction Financial: equity, private housing, aggregate asset Methodology prices (including also commercial housing), long rates, Boom identification Data short rates, term spread, M1, M3, private credit, Evaluation Results domestic credit, real effective exchange rates, global Best indicators Mid-2000s boom M1, global M3, global short rates, global private credit, Conclusions global domestic credit Sample is 1970:Q1 to 2007:Q4

Indicators 89 indicators ‘Real Time’ Early Warning for Costly Boom/Bust (Up to) 6 transformations per variable: Cycles Alessi Detken ratios to GDP Introduction annual growth rates Methodology 6 quarters cumulated growth rates Boom identification Data recursively HP detrended (constant and variable Evaluation Results history) Best indicators Mid-2000s boom ratios to GDP recursively HP detrended (constant and Conclusions variable history) cumulated shocks from recursive VAR in growth rates (for money and credit variables)

Global private credit gap and optimal threshold Evaluation period 1979-2002 ‘Real Time’ Early Warning for Costly Boom/Bust Cycles 1.5 Housing/ dot.com Credit Alessi Detken Savings and Loans 1 Introduction Methodology 0.5 Boom identification Data Evaluation 0 Results Best indicators Mid-2000s boom -0.5 Conclusions -1 -1.5 1979Q1 1983Q1 1987Q1 1991Q1 1995Q1 1999Q1 2003Q1 2007Q1

Signalling approach Policy Maker’s Loss Function ‘Real Time’ Early Warning for Costly Costly No Costly Boom/Bust Cycles Boom Boom Alessi Detken Signal A B No signal C D Introduction Methodology Boom identification Data C B Evaluation L = θ +( 1 − θ ) Results A + C B + D Best indicators � �� � � �� � Mid-2000s boom booms not called false alarms Conclusions θ = policy maker’s relative aversion of missing a call versus receiving a false alarm Usefulness = min [ θ ; 1 − θ ] − L

The trade-off between missed crises and false alarms Optimal thresholds for the Global M1 Gap ‘Real Time’ Early Warning for Costly Boom/Bust Cycles Alessi Detken Introduction 1 0.9 Methodology 0.8 Boom identification 0.7 Data Frequency 0.6 Evaluation 0.5 Results 0.4 Best indicators 0.3 0.40 0.95 0.90 0.85 0.10 Mid-2000s boom 0.2 Conclusions 0.1 0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Type I errors (missing crises) Type II errors (false alarms)