Predicting economic turning points by professional forecasters -Are they useful?

Similar documents
Nowcasting gross domestic product in Japan using professional forecasters information

Approximating Fixed-Horizon Forecasts Using Fixed-Event Forecasts

FISCAL MULTIPLIERS IN JAPAN

Approximating fixed-horizon forecasts using fixed-event forecasts

Record date Payment date PID element Non-PID element. 08 Sep Oct p p. 01 Dec Jan p 9.85p

Introduction to Forecasting

The Case of Japan. ESRI CEPREMAP Joint Workshop November 13, Bank of Japan

TIGER: Tracking Indexes for the Global Economic Recovery By Eswar Prasad, Karim Foda, and Ethan Wu

Research Brief December 2018

TIGER: Tracking Indexes for the Global Economic Recovery By Eswar Prasad and Karim Foda

GAMINGRE 8/1/ of 7

Short-term forecasts of GDP from dynamic factor models

Lucrezia Reichlin London Business School & Now-Casting Economics Ltd and Silvia Miranda Agrippino, Now-Casting Economics Ltd

How Well Are Recessions and Recoveries Forecast? Prakash Loungani, Herman Stekler and Natalia Tamirisa

BUSI 460 Suggested Answers to Selected Review and Discussion Questions Lesson 7

Lecture Prepared By: Mohammad Kamrul Arefin Lecturer, School of Business, North South University

The Blue Chip Survey: Moving Beyond the Consensus

Outlook vs. Futures: Three Decades of Evidence in Hog and Cattle Markets. Evelyn V. Colino and Scott H. Irwin

REPORT ON LABOUR FORECASTING FOR CONSTRUCTION

Euro-indicators Working Group

Warwick Business School Forecasting System. Summary. Ana Galvao, Anthony Garratt and James Mitchell November, 2014

Chapter 7 Forecasting Demand

ESRI Research Note Nowcasting and the Need for Timely Estimates of Movements in Irish Output

NOWCASTING REPORT. Updated: May 5, 2017

CAN SOLAR ACTIVITY INFLUENCE THE OCCURRENCE OF RECESSIONS? Mikhail Gorbanev January 2015

NSP Electric - Minnesota Annual Report Peak Demand and Annual Electric Consumption Forecast

Error Statistics for the Survey of Professional Forecasters for Industrial Production Index

Error Statistics for the Survey of Professional Forecasters for Treasury Bond Rate (10 Year)

SYSTEM BRIEF DAILY SUMMARY

Design of a Weather-Normalization Forecasting Model

Korea Chemicals Management Association 1/54

Error Statistics for the Survey of Professional Forecasters for Treasury Bill Rate (Three Month)

Applicability of wind power forecasting models in Japan

Application of Real-Time Rainfall Information System to CSO control. 2 October 2011 Naruhito Funatsu METAWATER Co., Ltd.

Chapter 3. Regression-Based Models for Developing Commercial Demand Characteristics Investigation

NOWCASTING REPORT. Updated: July 20, 2018

NOWCASTING REPORT. Updated: August 17, 2018

Are 'unbiased' forecasts really unbiased? Another look at the Fed forecasts 1

A look into the factor model black box Publication lags and the role of hard and soft data in forecasting GDP

Abram Gross Yafeng Peng Jedidiah Shirey

TIME SERIES ANALYSIS AND FORECASTING USING THE STATISTICAL MODEL ARIMA

Determine the trend for time series data

A MACRO-DRIVEN FORECASTING SYSTEM FOR EVALUATING FORECAST MODEL PERFORMANCE

NOWCASTING REPORT. Updated: September 23, 2016

GDP forecast errors Satish Ranchhod

The Measurement and Characteristics of Professional Forecasters Uncertainty. Kenneth F. Wallis. Joint work with Gianna Boero and Jeremy Smith

LODGING FORECAST ACCURACY

Outage Coordination and Business Practices

DROUGHT INDICES BEING USED FOR THE GREATER HORN OF AFRICA (GHA)

SYSTEM BRIEF DAILY SUMMARY

Rob J Hyndman. Forecasting using. 3. Autocorrelation and seasonality OTexts.com/fpp/2/ OTexts.com/fpp/6/1. Forecasting using R 1

= observed volume on day l for bin j = base volume in jth bin, and = residual error, assumed independent with mean zero.

The Spectrum of Broadway: A SAS

Seasonality in macroeconomic prediction errors. An examination of private forecasters in Chile

NOWCASTING REPORT. Updated: October 21, 2016

PJM Long-Term Load Forecasting Methodology: Proposal Regarding Economic Forecast (Planning Committee Agenda Item #5)

Jackson County 2018 Weather Data 67 Years of Weather Data Recorded at the UF/IFAS Marianna North Florida Research and Education Center

Salem Economic Outlook

How Useful Are Forecasts of Corporate Profits?

STAT 115: Introductory Methods for Time Series Analysis and Forecasting. Concepts and Techniques

Multiple Regression Analysis

YACT (Yet Another Climate Tool)? The SPI Explorer

Annual Average NYMEX Strip Comparison 7/03/2017

NOWCASTING REPORT. Updated: February 22, 2019

Lecture Prepared By: Mohammad Kamrul Arefin Lecturer, School of Business, North South University

Jackson County 2013 Weather Data

In Centre, Online Classroom Live and Online Classroom Programme Prices

NOWCASTING REPORT. Updated: January 4, 2019

Forecasting using R. Rob J Hyndman. 1.3 Seasonality and trends. Forecasting using R 1

Time Series Analysis

Chapter 3 ANALYSIS OF RESPONSE PROFILES

AUTO SALES FORECASTING FOR PRODUCTION PLANNING AT FORD

Are Forecast Updates Progressive?

Discussion Papers In Economics And Business

CWV Optimisation. TWG 12 th February 2014

Cost of Inflow Forecast Uncertainty for Day Ahead Hydropower Production Scheduling

Climatography of the United States No

STE-V-xx. Operation Manual for. Table/Floor Top Autoclave

5 Medium-Term Forecasts

DAILY QUESTIONS 28 TH JUNE 18 REASONING - CALENDAR

INTERNATIONAL ORGANISATIONS VS. PRIVATE ANA- LYSTS GROWTH FORECASTS: AN EVALUATION*

Technical note on seasonal adjustment for M0

2003 Water Year Wrap-Up and Look Ahead

Forecasting Practice: Decision Support System to Assist Judgmental Forecasting

Forecasting Inflation and Growth: Do Private Forecasts Match Those of Policymakers? William T. Gavin and Rachel J. Mandal

Time series and Forecasting

NOWCASTING REPORT. Updated: September 7, 2018

Lower Tuolumne River Accretion (La Grange to Modesto) Estimated daily flows ( ) for the Operations Model Don Pedro Project Relicensing

Changes in spatial distribution of chub mackerel under climate change: the case study using Japanese purse seine fisheries data in the East China Sea

NOWCASTING REPORT. Updated: September 14, 2018

Project Appraisal Guidelines

Demand Forecasting Reporting Period: 19 st Jun th Sep 2017

A NEW APPROACH FOR EVALUATING ECONOMIC FORECASTS

AREP GAW. AQ Forecasting

Can a subset of forecasters beat the simple average in the SPF?

Project No India Basin Shadow Study San Francisco, California, USA

2018 Annual Review of Availability Assessment Hours

Diploma Part 2. Quantitative Methods. Examiner s Suggested Answers

BRADSHAW'S RAILWAY GUIDE : accessible copies

WORKING PAPER NO DO GDP FORECASTS RESPOND EFFICIENTLY TO CHANGES IN INTEREST RATES?

Transcription:

Predicting economic turning points by professional forecasters -Are they useful? 21 Feb 2013 Nobuo Iizuka KANAGAWA University nobuo iizuka 0915@kanagawa u.ac.jp

Motivation Evaluating Economic forecasting has long history Debated vigorously about the best measure of inaccuracy Directional analysis is used in many papers recently Ash et al(1998) : Forecasts by OECD Pons(2000) : Comparing forecasts by IMF and OECD Ashiya(2003) : Forecasts by IMF Ashiya(2006) : Forecasts by Japanese institutional forecasters Sinclair et al (2010) : Forecasts by FRB Tsuchiya(2013) : Comparing forecasts by IMF and Japanese Government

Motivation On the other hand, not so many literature analyzed the accuracy of economic turning points prediction by professional economist Since 2004, ESPF has been gathering predictions of economic turning points, comparable with ESRI s Index of Business Condition Using ESPF data, this article evaluates accuracy of economic turning point predictions by professional forecasters in Japan through directional analysis

Objective Question1: Do professional forecasters have accurate predictions for the economic turning points? Question2: Are those predictions consistent with other predictions, such as real GDP growth?

Results Question1 ESPF s consensus Business Condition forecast is useful for 0~4 months ahead prediction More useful than Leading Index in ESRI s Index of Business Condition Only a few forecaster s have accurate predictions for 5 months ahead As useful as ESPF s GDP Forecast Consensus Business Condition forecast is slightly more useful than GDP forecast Looking into Individual forecaster, GDP consensus may be more useful

Results Question2 Most of Business Condition forecasts are consistent with GDP growth forecasts Business Condition and GDP growth forecast move in the same direction Both consensus and most of individual forecasters Forecasters who are good at Business condition forecast may be different from who are good at GDP forecast Individual DI forecast s accuracy rates vary widely than GDP forecast s

Outline Dataset Stylized Facts Methodology Results Conclusion

Dataset Actual data ESRI "Index of Business Condition Coincident Index, Leading Index Diffusion Index, Quarterly averaged Both historical data and real time data ESRI "Quarterly Estimates of GDP Real GDP Growth Quarter to quarter, Seasonally adjusted, annualized Both historical data and real time data

100 90 80 70 60 50 40 30 20 10 0 % 15 10 5 0 5 10 15 2012Ⅱ 2004Ⅰ 2004Ⅱ 2004Ⅲ 2004Ⅳ 2005Ⅰ 2005Ⅱ 2005Ⅲ 2005Ⅳ 2006Ⅰ 2006Ⅱ 2006Ⅲ 2006Ⅳ 2007Ⅰ 2007Ⅱ 2007Ⅲ 2007Ⅳ 2008Ⅰ 2008Ⅱ 2008Ⅲ 2008Ⅳ 2009Ⅰ 2009Ⅱ 2009Ⅲ 2009Ⅳ 2010Ⅰ 2010Ⅱ 2010Ⅲ 2010Ⅳ 2011Ⅰ 2011Ⅱ 2011Ⅲ 2011Ⅳ 2012Ⅰ 2012Ⅲ Business Condition and GDP Growth Rate (Historical data) Business Condition Index (coincident DI) GDP growth rate (annualized,rhs) DI>50 23 GDP growth>0 22 DI=50 0 GDP growth=0 1 DI<50 12 GDP growth<0 12

100 90 80 70 60 50 40 30 20 10 0 % 10 5 0 5 10 15 2004Ⅰ 2004Ⅱ 2004Ⅲ 2004Ⅳ 2005Ⅰ 2005Ⅱ 2005Ⅲ 2005Ⅳ 2006Ⅰ 2006Ⅱ 2006Ⅲ 2006Ⅳ 2007Ⅰ 2007Ⅱ 2007Ⅲ 2007Ⅳ 2008Ⅰ 2008Ⅱ 2008Ⅲ 2008Ⅳ 2009Ⅰ 2009Ⅱ 2009Ⅲ 2009Ⅳ 2010Ⅰ 2010Ⅱ 2010Ⅲ 2010Ⅳ 2011Ⅰ 2011Ⅱ 2011Ⅲ 2011Ⅳ 2012Ⅰ 2012Ⅱ 2012Ⅲ Business Condition and GDP Growth Rate (Real time data) Business Condition Index (Coincident DI) GDP growth rate (annualized,rhs) DI>50 23 GDP growth>0 25 DI=50 0 GDP growth=0 0 DI<50 12 GDP growth<0 10 Business Condition and GDP growth tend to move in the same direction!

Forecast Data JCER "ESP Forecast Dataset Forecasting Business Condition Choose from 3 options Expansion= Coincident DI will be over 50 Flat= Coincident DI will be (around) 50 or difficult to judge Contraction= Coincident DI will be below 50 Consensus= % of expansion+% of flat*0.5 Forecasting Real GDP Growth Quarter to quarter, Seasonally adjusted, annualized

Forecast Data Sample Dataset From 2004:1 to 2012:3, 35 quarters 12 forecasts for each quarter Forecast Horizon is 11 ~0 months ahead 32 forecasters Exclude 19 forecasters whose participate rate is fewer than 70% (concerning Business Condition forecast)

Stylized Facts Forecasting Business Condition Consensus Tend to forecast Expansion 83.2% of all forecasts ( 327 in 393 ) Over 90% in ft 6 ~ f t 11 Accuracy may not be a proper gauge for Usefulness Larger than 65.7% in Actual( 23 in 35 quarters ) Forecasting may be affected by actual Coincident DI Contraction prediction began in Feb08 Real time Coincident DI of Nov07(released Jan09) was below 50 But Contraction prediction ended in Apr09 Real time Coincident DI began to be over 50 in Jul09(released Sep09)

Stylized Facts Forecasting Business Condition (cont.) Consensus Turning Point prediction Contraction from 08Q1 Forecast for 08Q1= below 50 from (Feb08) Forecast for 08Q2=below 50 from f t 4 (APR08) Turning Point predicted in APR08 4 months Earlier than judgment by Index of Business Condition ( coincident CI ) Expansion from 09Q2 f t 3 Forecast for 09Q2=over 50 from (May09) Forecast for 09Q3=over 50 from f t 6 (May09) Turning Point predicted in May09 7 months Earlier than judgment by Index of Business Condition ( coincident CI) f t 3

Stylized Facts Forecasting Business Condition(cont.) Individual Forecasters Tend to forecast Flat Average ratio in predictions is 20% ~ 30% Max ratio is around 50 % and Over Also tend to forecast Expansion Turning Point Prediction Contraction from 08Q1 Need to care for flat answers Forecast for 08Q1 =Contraction Ratio around 50% from f t 3 (Mar08) Forecast for 08Q2 =Contraction Ratio over 50% from f t 2 (Jul08) Expansion from 09Q2 Forecast for 09Q2 f t 3 =Expansion Ratio over 50% from (Jun09) Forecast for 09Q3 =Expansion Ratio over 50% from f t 6 (Jun09)

Distribution of Forecasts (Business Condition (coincident DI )) Consensus Individual Forecaster

Turning Point Prediction(contraction) <Individual Forecaster> Forecast for 2008Q1 Forecast for 2008Q2 100% 100% 90% 90% 80% 80% 70% 70% 60% 60% 50% contraction flat 50% contraction flat 40% Expansion 40% Expansion 30% 30% 20% 20% 10% 10% 0% t 11 t 10 t 9 t 8 t 7 t 6 t 5 t 4 t 3 t 2 t 1 No data in t 0% t 11 t 10 t 9 t 8 t 7 t 6 t 5 t 4 t 3 t 2 t 1 t

Turning Point Prediction(Expansion) <Individual Forecaster> Forecast for 2009Q2 Forecast for 2009Q3 100% 100% 90% 90% 80% 80% 70% 70% 60% 60% 50% contraction flat 50% contraction flat 40% Expansion 40% Expansion 30% 30% 20% 20% 10% 10% 0% t 11 t 10 t 9 t 8 t 7 t 6 t 5 t 4 t 3 t 2 t 1 t 0% t 11 t 10 t 9 t 8 t 7 t 6 t 5 t 4 t 3 t 2 t 1 No data in t 1

Methodology Directional Analysis The null hypothesis The direction of change in a forecast and that in the actual are independent Two type test Fisher's Exact Test Pesaran and Timmermann s(1992) Test Sample excluding 50 or flat answer is also tested Because there is no just 50 in Actual data but many 50 or flat answer in Forecast data

Results Forecast Performance ESPF Business Condition All Sample Consensus 0 ~4 months ahead forecasts are Useful No significant difference between historical and realtime data Individual Forecasters Results are mixed and hard to interpret No significant difference between historical and realtime data Partly because there is no just 50 in actual DI data but many flat answer in forecast DI data

Table 1 1 Results(Consensus) Historical data Real time data Forecast Horizon (months) a>50 f>50 a>50 f 50 a 50 f>50 a 50 f 50 Accuracy rates(%) ft 19 1 3 8 87.1 0.00 ** 16.33 ** ft-1 21 2 3 9 85.7 0.00 ** 16.56 ** ft-2 20 1 4 8 84.8 0.00 ** 15.22 ** ft-3 21 1 5 7 82.4 0.00 ** 12.86 ** ft-4 20 2 6 6 76.5 0.01 * 7.44 ** ft-5 19 2 9 3 66.7 0.24 1.47 ft-6 20 1 10 2 66.7 0.29 1.35 ft-7 19 2 11 1 60.6 0.71 0.01 ft-8 17 3 11 1 56.3 0.52 0.31 ft-9 18 2 12 0 56.3 0.38 1.32 ft-10 19 1 12 0 59.4 0.63 0.64 ft-11 17 3 11 0 54.8 0.25 1.89 Notes: a= actual data,f=forecast data,fe=fisher's Exact Test PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance FE PT Forecast Horizon (months) a>50 f>50 a>50 f 50 a 50 f>50 a 50 f 50 Accuracy rates(%) ft 18 1 4 8 83.9 0.00 ** 13.91 ** ft-1 20 2 4 9 82.9 0.00 ** 14.12 ** ft-2 19 1 5 8 81.8 0.00 ** 13.09 ** ft-3 20 1 6 7 79.4 0.00 ** 11.08 ** ft-4 19 2 7 6 73.5 0.02 * 6.17 * ft-5 18 2 10 3 63.6 0.29 1.08 ft-6 19 1 11 2 63.6 0.34 1.06 ft-7 18 2 12 1 57.6 0.66 0.05 ft-8 17 3 11 1 56.3 0.52 0.31 ft-9 18 2 12 0 56.3 0.38 1.32 ft-10 19 1 12 0 59.4 0.63 0.64 ft-11 17 3 11 0 54.8 0.25 1.89 Notes: a= actual data,f=forecast data,fe=fisher's Exact Test PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance FE PT

Table1 2 Results(Individual Forecasters) Historical data Forecast Horizon (months) Accuracy rates(%) PT (Number of forecasters) Mean Max Min ** * ft 73.2 96.2 57.7 5 7 ft-1 75.0 90.9 58.8 8 9 ft-2 74.0 86.2 57.6 3 6 ft-3 66.8 78.8 47.6 0 2 ft-4 62.5 78.8 45.2 0 1 ft-5 56.5 75.0 40.6 0 3 ft-6 51.7 66.7 30.0 0 6 ft-7 48.7 64.5 23.3 2 11 ft-8 49.6 65.5 29.0 2 6 ft-9 49.4 65.6 26.1 3 8 ft-10 48.9 65.6 27.6 3 10 ft-11 48.7 67.7 36.7 3 10 Notes: PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance Real time data Forecast Horizon (months) Accuracy rates(%) PT (Number of forecasters) Mean Max Min ** * ft 74.5 93.8 45.0 8 4 ft-1 73.0 83.9 55.9 6 5 ft-2 72.2 87.9 56.3 3 3 ft-3 63.0 75.8 48.4 0 1 ft-4 60.0 75.8 41.9 0 0 ft-5 54.4 71.4 34.4 3 1 ft-6 50.9 69.7 33.3 0 9 ft-7 48.2 64.5 23.3 5 6 ft-8 49.6 65.5 29.0 5 5 ft-9 48.4 71.9 29.0 4 9 ft-10 48.3 68.8 25.8 5 9 ft-11 48.5 73.3 23.3 5 9 Notes: PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance

Results Forecast Performance ESPF Business Condition Sample excluding F=50 or flat Consensus 0 ~4 months ahead forecasts are useful No significant difference between historical and realtime data Same as all sample analysis Individual Forecasters Most of 0~3 months ahead forecast are useful Around half of 4 months ahead forecast are useful Only 1 or 2 forecaster s 5 months ahead forecast are useful

Table 1 3 Results(consensus) excluding f=50 or flat answer Historical data Real time data Forecast Horizon (months) a>50 f>50 a>50 f<50 a<50 f>50 a<50 f<50 f=50 Accuracy rates(%) ft 19 1 3 8 0 87.1 0.00 ** 16.33 ** ft-1 21 0 3 9 2 90.9 0.00 ** 22.33 ** ft-2 20 1 4 8 0 84.8 0.00 ** 15.22 ** ft-3 21 1 5 7 0 82.4 0.00 ** 12.86 ** ft-4 20 2 6 6 0 76.5 0.01 * 7.44 ** ft-5 19 2 9 3 0 66.7 0.24 1.47 ft-6 20 1 10 2 0 66.7 0.29 1.35 ft-7 19 2 11 1 0 60.6 0.71 0.01 ft-8 17 2 11 1 1 58.1 0.52 0.04 ft-9 18 2 12 0 0 56.3 0.38 1.32 ft-10 19 1 12 0 0 59.4 0.63 0.64 ft-11 17 2 11 0 1 56.7 0.25 1.28 Notes: a= actual data,f=forecast data,fe=fisher's Exact Test PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance FE PT Forecast Horizon (months) a>50 f>50 a>50 f<50 a<50 f>50 a<50 f<50 f=50 Accuracy rates(%) ft 18 1 4 8 0 83.9 0.00 ** 13.91 ** ft-1 20 0 4 9 2 87.9 0.00 ** 19.63 ** ft-2 19 1 5 8 0 81.8 0.00 ** 13.09 ** ft-3 20 1 6 7 0 79.4 0.00 ** 11.08 ** ft-4 19 2 7 6 0 73.5 0.02 * 6.17 * ft-5 18 2 10 3 0 63.6 0.29 1.08 ft-6 19 1 11 2 0 63.6 0.34 1.06 ft-7 18 2 12 1 0 57.6 0.66 0.05 ft-8 17 2 11 1 1 58.1 0.67 0.04 ft-9 18 2 12 0 0 56.3 0.38 1.32 ft-10 19 1 12 0 0 59.4 0.63 0.64 ft-11 17 2 11 0 1 56.7 0.39 1.28 Notes: a= actual data,f=forecast data,fe=fisher's Exact Test PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance FE PT

Table1 4 Results(Individual Forecasters) excluding f=50 or flat answer Historical data Forecast Horizon (months) Accuracy rates(%) PT (Number of forecasters) Mean Max Min ** * ft 86.8 96.0 71.4 29 1 ft-1 87.9 96.2 78.3 31 1 ft-2 85.0 100.0 75.0 26 6 ft-3 80.3 93.3 66.7 20 8 ft-4 73.6 84.6 57.1 5 9 ft-5 64.7 76.2 47.6 0 2 ft-6 59.3 75.0 38.1 0 0 ft-7 54.6 72.2 29.4 0 1 ft-8 54.4 76.9 31.6 0 1 ft-9 54.8 69.6 28.6 0 2 ft-10 53.9 69.6 23.8 1 0 ft-11 54.5 69.0 26.3 1 0 Notes: PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance Real time data Forecast Horizon (months) Accuracy rates(%) PT (Number of forecasters) Mean Max Min ** * ft 86.4 100.0 64.3 28 3 ft-1 85.1 100.0 71.4 31 1 ft-2 82.3 100.0 67.9 23 7 ft-3 75.6 87.0 60.9 13 12 ft-4 69.7 86.4 47.6 2 9 ft-5 60.3 76.2 40.0 0 1 ft-6 56.8 73.1 35.7 0 0 ft-7 52.1 66.7 25.0 0 3 ft-8 53.7 68.4 34.8 0 3 ft-9 53.6 73.9 30.4 0 1 ft-10 52.9 73.9 29.4 0 1 ft-11 53.5 77.8 33.3 0 0 Notes: PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance

Comparing Forecast Performance vs ESPF GDP Consensus ESPF Business condition forecast may be slightly more useful than ESPF GDP Individual Forecaster ESPF GDP is as useful as ESPF Business condition forecast vs Business Condition Index s Leading DI ESPF Business condition forecast is more useful than Leading DI

Results Forecast Performance ESPF GDP(Historical data and Real time data) Consensus 0 ~3 months ahead forecasts are Useful All sample and Sample excluding F=0 and A=0 Individual Forecasters Most of 0~3 months ahead forecasts are useful one third of 4 months ahead forecasts are useful Also some forecaster s 5 months ahead forecasts are useful Results of sample excluding actual=0 and forecast=0 is almost the same See the appendix

Table 2 1 Results(Consensus) Historical data Real time data Forecast Horizon (months) a>0 f>0 a>0 f 0 a 0 f>0 a 0 f 0 Accuracy rates(%) ft 21 1 6 7 80.0 0.00 ** 15.37 ** ft-1 22 0 6 7 82.9 0.00 ** 20.28 ** ft-2 20 1 8 5 73.5 0.02 * 10.73 ** ft-3 21 0 9 4 73.5 0.02 * 7.54 ** ft-4 20 1 10 3 67.6 0.14 2.67 ft-5 19 2 11 1 60.6 0.71 0.01 ft-6 20 1 11 1 63.6 0.60 0.18 ft-7 19 2 12 0 57.6 0.40 1.25 ft-8 17 3 12 0 53.1 0.23 2.05 ft-9 19 1 12 0 59.4 0.63 0.64 ft-10 19 1 12 0 59.4 0.63 0.64 ft-11 20 0 11 0 64.5 1.00 NA Notes: a= actual data,f=forecast data,fe=fisher's Exact Test PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance FE PT Forecast Horizon (months) a>0 f>0 a>0 f 0 a 0 f>0 a 0 f 0 Accuracy rates(%) ft 25 0 2 8 94.3 0.00 ** 26.69 ** ft-1 25 0 3 7 91.4 0.00 ** 22.52 ** ft-2 23 1 5 5 82.4 0.00 ** 10.51 ** ft-3 24 0 6 4 82.4 0.00 ** 11.21 ** ft-4 23 1 7 3 76.5 0.07 4.68 * ft-5 21 2 9 1 66.7 0.68 0.01 ft-6 22 1 9 1 69.7 0.52 0.40 ft-7 21 2 10 0 63.6 0.48 0.95 ft-8 19 3 10 0 59.4 0.31 1.55 ft-9 21 1 10 0 65.6 0.69 0.48 ft-10 21 1 10 0 65.6 0.69 0.48 ft-11 22 0 9 0 71.0 1.00 NA Notes: a= actual data,f=forecast data,fe=fisher's Exact Test PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance FE PT

Table 2 2 Results(Individual Forecasters) Historical data Forecast Horizon (months) Accuracy rates(%) PT (Number of forecasters) Mean Max Min ** * ft 82.7 89.7 76.5 32 0 ft-1 77.3 86.7 66.7 17 13 ft-2 71.9 82.1 59.3 12 12 ft-3 71.8 78.6 62.5 11 11 ft-4 66.0 75.0 58.6 1 3 ft-5 63.0 70.8 56.3 0 1 ft-6 60.1 69.7 50.0 0 0 ft-7 56.5 63.6 48.0 0 0 ft-8 56.5 66.7 40.0 0 1 ft-9 58.7 71.0 45.8 0 1 ft-10 57.4 67.7 45.8 0 0 ft-11 58.9 67.9 50.0 0 0 Notes: PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance Real time data Forecast Horizon (months) Accuracy rates(%) PT (Number of forecasters) Mean Max Min ** * ft 94.3 100.0 85.3 32 0 ft-1 87.4 100.0 77.1 31 1 ft-2 82.0 90.3 72.7 24 6 ft-3 80.6 85.7 71.9 22 7 ft-4 74.5 84.4 67.6 4 7 ft-5 68.9 79.2 62.5 0 4 ft-6 66.1 75.8 55.0 0 1 ft-7 63.0 69.7 54.8 0 0 ft-8 63.0 70.8 48.0 0 0 ft-9 64.8 77.4 47.4 1 0 ft-10 63.4 71.9 50.0 0 0 ft-11 65.1 74.2 54.2 0 0 Notes: PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance

Results Forecast Performance Business Condition Index Leading DI Leading DI are processed as below 1 month ahead forecast=(t 1+t 2+t 3)/3 2 month ahead forecast=(t 2+t 3+t 4)/3 12 month ahead forecast=(t 12+t 13+t 14)/3 Only 1 2 months ahead forecast are useful

Table3 Results(Leading Index) Forecast Horizon (months) a>50 l>50 a>50 l 50 a 50 l>50 a 50 l 50 Accuracy rates(%) 1 21 2 3 9 85.7 0.00 ** 16.56 ** 2 22 1 3 9 88.6 0.00 ** 19.86 ** 3 16 6 7 5 61.8 0.31 0.76 4 16 6 8 4 58.8 0.50 0.14 5 17 5 8 4 61.8 0.39 0.46 6 15 6 7 5 60.6 0.35 0.61 7 14 7 9 3 51.5 0.46 0.26 8 16 5 8 4 60.6 0.42 0.36 9 14 6 7 5 59.4 0.38 0.47 10 14 6 8 4 56.3 0.57 0.04 11 19 1 12 0 59.4 0.63 0.10 12 11 9 9 2 41.9 0.14 2.30 Notes: a= actual data,l=leading Index,FE=Fisher's Exact Test PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance FE PT

Results Consistency Comparing directions of Business condition forecasts and GDP growth forecasts by directional analysis Most of consensus DI forecasts are consistent with GDP forecasts Most of individual DI forecasts are also consistent with individual GDP forecasts As for 6 ~ 11 months ahead forecast, some forecast s PT test static couldn t be calculated

Table 4 1 Results(Consensus) All sample Excluding F=50(F=0) Forecast Horizon (months) d>50 g>0 d>50 g 0 d<50 g>0 d<50 g 0 Coincide rates(%) ft 22 3 0 6 90.3 0.00 ** 18.79 ** ft-1 24 4 0 7 88.6 0.00 ** 15.77 ** ft-2 24 3 0 6 90.9 0.00 ** 20.17 ** ft-3 26 4 0 4 88.2 0.00 ** 15.18 ** ft-4 26 4 0 4 88.2 0.00 ** 15.18 ** ft-5 27 3 1 2 87.9 0.05 7.03 ** ft-6 30 1 0 2 97.0 0.01 ** 21.96 ** ft-7 30 1 0 2 97.0 0.01 ** 21.96 ** ft-8 27 2 1 2 90.6 0.03 * 9.17 ** ft-9 30 1 0 1 96.9 0.06 15.98 ** ft-10 31 0 0 1 100.0 0.03 * 33.03 ** ft-11 28 3 0 0 90.3 1.00 NA Notes: d= DI forecast data, g=gdp forecast data,fe=fisher's Exact Test PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance FE PT Forecast Horizon (months) d>50 g>0 d>50 g<0 d<50 g>0 d<50 g<0 Coincide rates(%) ft 22 3 0 6 90.3 0.00 ** 18.79 ** ft-1 24 3 0 6 90.9 0.00 ** 20.17 ** ft-2 24 3 0 6 90.9 0.00 ** 20.17 ** ft-3 26 4 0 4 88.2 0.00 ** 15.18 ** ft-4 26 4 0 4 88.2 0.00 ** 15.18 ** ft-5 27 3 1 2 87.9 0.05 7.03 ** ft-6 30 1 0 2 97.0 0.01 ** 21.96 ** ft-7 30 1 0 2 97.0 0.01 ** 21.96 ** ft-8 27 1 1 2 93.5 0.02 * 12.75 ** ft-9 30 1 0 1 96.9 0.06 15.98 ** ft-10 31 0 0 1 100.0 0.03 * 33.03 ** ft-11 28 2 0 0 93.3 1.00 NA Notes: d= DI forecast data, g=gdp forecast data,fe=fisher's Exact Test PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance FE PT

Table4 2 Results(Individual Forecasters) All sample Forecast Horizon (months) Coincide rates(%) PT (Number of forecasters) Mean Max Min ** * ft 72.9 92.6 48.3 12 10 ft-1 72.1 88.6 51.4 11 13 ft-2 71.6 89.7 48.3 10 10 ft-3 67.2 88.2 41.4 6 9 ft-4 66.6 84.4 46.4 7 7 ft-5 63.4 85.7 37.5 2 8 ft-6 64.0 87.9 44.0 0 7 ft-7 66.2 87.9 45.5 3 2 ft-8 69.0 93.3 46.9 3 3 ft-9 68.9 96.7 46.9 3 3 ft-10 71.3 100.0 46.9 5 4 ft-11 72.1 93.5 48.4 2 5 Notes: PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance Excluding F=50(F=0) Forecast Horizon (months) Conincide rates(%) PT (Number of forecasters) Mean Max Min ** * ft 91.3 100.0 71.4 31 0 ft-1 90.4 100.0 68.2 30 1 ft-2 90.7 100.0 68.2 30 1 ft-3 91.1 100.0 75.0 31 1 ft-4 91.1 100.0 72.7 28 3 ft-5 89.0 100.0 75.0 22 7 ft-6 89.8 100.0 71.4 19 4 ft-7 91.4 100.0 65.0 23 6 ft-8 91.3 100.0 61.9 23 2 ft-9 92.3 100.0 66.7 18 4 ft-10 94.2 100.0 73.7 20 3 ft-11 94.0 100.0 70.0 23 3 Notes: PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance

Results consistency Comparing Accuracy rates of individual forecasters As for 0~4 months forecast 13 forecasters are useful in GDP growth forecast 11 forecasters are useful in Business condition forecast Only 7 forecasters are useful in both Business condition and GDP growth forecast As for 0~5 months forecast Only 1 forecaster is useful in Business condition forecast 3 forecasters are useful in GDP growth forecast No forecaster is useful in both Business condition and GDP growth forecast Most of GDP forecast accuracy rates are better than DI forecast s in each forecasting period Business Condition forecast accuracy rates vary widely

Table 4 3 Results (Individual Forecasters, Real time data, excluding F=50(F=0) ) Categorized individual forecasters Whether each individual forecasts are useful in all 0~4 months ahead or not DI GDP sum 7 4 11 6 15 21 sum 13 19 32

Comparing Accuracy (real time data, excluding F=50(F=0)) f t f t 1 100 100 Accyracy rate of GDP forecast(%) 90 80 70 60 50 40 40 50 60 70 80 90 100 Accuracy rate of DI forecast (%) Accyracy rate of GDP forecast(%) 90 80 70 60 50 40 40 50 60 70 80 90 100 Accuracy rate of DI forecast (%)

Comparing Accuracy (real time data, excluding F=50(F=0)) f t 2 f t 3 100 100 Accyracy rate of GDP forecast(%) 90 80 70 60 50 40 40 50 60 70 80 90 100 Accuracy rate of DI forecast (%) Accyracy rate of GDP forecast(%) 90 80 70 60 50 40 40 50 60 70 80 90 100 Accuracy rate of DI forecast (%)

Comparing Accuracy (real time data, excluding F=50(F=0)) f t 4 f t 5 100 100 Accyracy rate of GDP forecast(%) 90 80 70 60 50 40 40 50 60 70 80 90 100 Accyracy rate of GDP forecast(%) 90 80 70 60 50 40 40 50 60 70 80 90 100 Accuracy rate of DI forecast (%) Accuracy rate of DI forecast (%)

Conclusion Do professional forecasters have accurate predictions for the economic turning points? Yes for 0~4 months ahead prediction Both Consensus and Individual forecasters Only a few individual forecasters are useful for 5 months ahead More useful than Business Condition Index s Leading DI As useful as ESPF s GDP Forecast

Conclusion Are those predictions consistent with other predictions, such as real GDP growth? Most of Business Condition forecasts move in the same direction with GDP forecasts Both Consensus and Individual forecasters Forecasters who are good at DI forecast may be different from who are good at GDP forecast Individual Business Condition forecasts accuracy rates vary widely than GDP forecasts

Future Research Treating Flat answers in Business Condition Forecast In this article, treat as their answer mean difficult to judge The way of calculation for ESPF s consensus appropriate? Consensus= % of expansion+% of flat*0.5 The reason for varying in individual forecaster s accuracy rates for Business Condition forecast Especially in f t Are Business condition forecast helpful for GDP growth forecast?, or vice versa

Appendix

Consensus Forecast for Business condition coincident DI

Target, Horizon and Release month (ESPF DI forecast)

Consensus Forecast for Business condition coincident DI

Target, Horizon and Release month (ESPF DI forecast)

Stylized Facts Forecasting GDP growth rate Consensus Accuracy may not be a proper 89.7% of all prediction ( 357 in 398 ) gauge for Over 90% in from ft 5 to f t 11 Usefulness Larger than 62.9% in Actual(Historical: 22 in 35 Qtr) Larger than 71.4% in Actual(Real time: 25 in 35 Qtr ) Larger than Business Condition forecast Tend to forecast Expansion Individual Forecasters Tend to forecast Expansion From ft 5 to f t 11, some forecasters predict no negative growth Not necessarily the same forecaster

Consensus Distribution of Predictions (GDP Growth rate) f f f f f f t t 1 t 2 t 3 t 4 t 5 t 6 t 7 t 8 t 9 t 10 t 11 f f f f f f Sum f>0 27 28 28 30 30 30 31 31 29 31 31 31 357 f<0 8 7 6 4 4 3 2 2 3 1 1 0 41 f=0 0 0 0 0 0 0 0 0 0 0 0 0 0 sum 35 35 34 34 34 33 33 33 32 32 32 31 398 Individual Forecaster % ft ft 1 ft 2 ft 3 ft 4 ft 5 ft 6 ft 7 ft 8 ft 9 ft 10 ft 11 mean 75.1 77.5 79.7 84.0 85.7 89.7 90.9 91.6 91.4 93.9 93.3 93.3 f>0 max 87.5 88.2 88.2 93.3 94.1 100.0 97.0 100.0 100.0 100.0 100.0 100.0 min 63.0 55.6 55.6 61.5 65.4 69.2 76.0 78.1 80.0 84.4 79.2 79.2 mean 24.2 21.0 18.6 14.8 13.4 9.0 7.8 7.7 7.5 5.3 6.0 6.3 f<0 max 37.0 44.4 37.0 38.5 34.6 30.8 24.0 21.9 20.0 12.9 20.8 20.8 min 12.5 8.8 11.8 5.9 3.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 mean 0.7 1.5 1.6 1.3 0.9 1.3 1.3 0.7 1.0 0.8 0.7 0.4 f=0 max 3.3 8.6 7.4 4.8 6.3 7.1 10.7 6.1 6.7 7.1 6.3 4.0 min 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

Table 2 3 Results(consensus) GDP growth forecast excluding F=0 and A=0 Historical data Real time data Forecast Horizon (months) a>0 f>0 a>0 f<0 a<0 f>0 a<0 f<0 f=0 a=0 Accuracy rates(%) ft 21 1 5 7 0 1 82.4 0.00 ** 12.86 ** ft-1 22 0 5 7 0 1 85.3 0.00 ** 16.65 ** ft-2 20 1 7 5 0 1 75.8 0.02 * 7.21 ** ft-3 21 0 8 4 0 1 75.8 0.01 * 8.21 ** ft-4 20 1 9 3 0 1 69.7 0.13 3.03 ft-5 19 2 11 1 0 0 60.6 0.71 0.01 ft-6 20 1 11 1 0 0 63.6 0.60 0.18 ft-7 19 2 12 0 0 0 57.6 0.40 1.25 ft-8 17 3 12 0 0 0 53.1 0.23 2.05 ft-9 19 1 12 0 0 0 59.4 0.63 0.64 ft-10 19 1 12 0 0 0 59.4 0.63 0.64 ft-11 20 0 11 0 0 0 64.5 1.00 NA Notes: a= actual data,f=forecast data,fe=fisher's Exact Test PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance FE PT No 0% data exists in Realtime GDP growth data

Table 2 4 Results(Individual Forecasters) GDP growth forecast excluding F=0 and A=0 Historical data Forecast Horizon (months) Accuracy rates(%) PT (Number of forecasters) Mean Max Min ** * ft 85.0 90.9 78.8 32 0 ft-1 79.5 88.9 66.7 23 7 ft-2 74.5 84.6 60.0 19 7 ft-3 74.0 81.5 64.5 15 9 ft-4 67.8 75.9 60.6 1 5 ft-5 63.2 70.8 56.3 0 1 ft-6 60.4 69.7 47.4 0 0 ft-7 56.7 63.6 48.4 0 0 ft-8 56.8 66.7 40.0 0 1 ft-9 58.8 71.0 45.8 0 1 ft-10 57.6 67.7 45.8 0 0 ft-11 59.2 67.9 50.0 0 0 Notes: PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance Real time data Forecast Horizon (months) Accuracy rates(%) PT (Number of forecasters) Mean Max Min ** * ft 94.6 100.0 85.3 32 0 ft-1 87.9 100.0 77.1 31 1 ft-2 82.8 90.3 72.7 28 2 ft-3 80.9 87.5 71.9 22 8 ft-4 74.7 83.3 67.6 5 10 ft-5 69.2 79.2 62.5 0 4 ft-6 66.4 75.8 52.6 0 1 ft-7 63.2 69.7 54.8 0 0 ft-8 63.4 70.8 48.0 0 0 ft-9 65.1 77.4 50.0 1 0 ft-10 63.8 71.9 50.0 0 0 ft-11 65.4 74.2 54.2 0 0 Notes: PT=Pesaran and Timmermann's(1992)test * indicates rejection at 5% significance ** indicates rejection at 1% significance