Challenges for estimating and forecasting macroeconomic trends during financial crises: implications for counter-cyclical policies

1
Challenges for estimating and forecasting
macroeconomic trends during financial crises
implications for counter-cyclical policies

• Pingfan Hong
• Chief for Global Economic Monitoring
• UN/DESA
• International Seminar at Ottawa, Canada
• 27-29 May 2009
• Views expressed here are solely those of the
  speaker and they do not necessarily represent
  those of the United Nations

2
Outline

• Introduction
• Forecasting performance of UN/LINK global
  modeling system
• High Frequency Modeling for Rolling estimation
  and forecast
• turning point Over-year-ago (oya) Quarterly
  GDP growth versus Seasonally Adjusted Annual Rate
  (SAAR) of Quarterly GDP growth
• The importance of correctly estimating potential
  output

3
Introduction

• Estimating versus forecasting
• Estimating
• Forecasting
• Importance of estimating and forecasting for
  counter-cyclical macroeconomic policy
  timeliness, consistent, accuracy, turning
  point, and correct estimate of the potential gap
  

4
Forecasting performance of UN/LINK global
modeling (1)
5
Forecasting performance of UN/LINK global
modeling (2)
6
Forecasting performance of UN/LINK global
modeling (3)
7
Forecasting performance of UN/LINK global
modeling (4)
world developed economies developed economies developing countries developing countries
Mean 0.02 0.04 -0.36
Median 0.05 0.05 -0.1
Standard Deviation Standard Deviation 0.7 0.76 1.25
Fraction of positive errors Fraction of positive errors Fraction of positive errors 0.52 0.5 0.42
Serial correlation Serial correlation -0.2 -0.1 0.29

Source DESA Source DESA
8
High Frequency Modeling for rolling estimating
quarterly GDP

• Collecting weekly data stream
• Principle Component
• ARIMA
• Weekly rolling estimate and forecast of quarterly
  GDP
• Sources for slides 8-12 L.R. Klein and W. Mak,
  University of Pennsylvania Current Quarter Model
  of the United States Economy
• Y. Inada, Konan University Current
  Quarter Model Forecast For the Japanese Economy

9
Example US weekly data stream

• Date Economic Indicator for Latest and Prior
  Month
• Apr 01 Construction Spending February -0.9 -3.5
• Apr 01 Auto Sales March 9.9 Million 9.1 Million
• Apr 02 Manuf Ships, Inv, Orders February -0.1,
  -1.2, 1.8 -2.6, -1.1, -3.5
• Apr 03 Nonfarm Payroll Employment March -663,000
  -651,000
• Apr 07 Consumer Credit Outstanding February -7.5
  billion 8.1 billion
• Apr 09 Export/Import Price Index March -0.6,
  0.5 -0.3, -0.1
• Apr 09 Trade Balance February -26.0 billion
  -36.2 billion
• Apr 15 Producer Price Index, Total Core March
  -1.2, 0.0 0.1, 0.2
• Apr 14 Retail Sales, Total Ex-Auto March -1.1,
  0.9 0.3, 1.0
• Apr 15 Industrial Production March -1.5 -1.5
• Apr 14 Business Inventories February -1.3 -1.3
• Apr 15 Consumer Price Index, Total Core March
  -0.1, 0.2 0.4, 0.2
• Apr 16 Housing Starts February 510,000 572,000

10
Example indicators used in US model for
estimating quarterly GDP

• Industrial Production Index
• Manufacturers orders, deflated by producer price
  index
• Manufacturers shipments, deflated by producer
  price index
• Manufacturers unfilled orders, deflated by
  producer price index
• Yield spread between 6-month commercial paper and
  6-month treasury bills
• Real interest rate (6-month commercial paper
  yield adjusted by consumer price index)
• Real M1, adjusted by consumer price index
• Real retail sales, adjusted by consumer price
  index
• Real personal income, adjusted by consumer price
  index
• Real 10-year treasury yield
• Yield spread between 10- and 1-year treasury
  bills
• Nonfarm payrolls
• Average weekly hours, production workers total
  private
• Trade-weighted value of the US dollar, nominal
  broad dollar index

11
Example Equations for GDP and PGDP in US model

• Dlog (QGDP) 0684 0.954 Dlog C1
• 0.304 Dlog C2
• -0.0661 Dlog C6
• 0.295 Dlog C7
• 0.581
  AR(1)
• 0.677
  MA(1)
• Dlog (QPGDP) 0.817 2.463 Dlog C1 0.925 Dlog
  C2
• 1.383 Dlog C3 5.113 Dlog C4
• 4.189 Dlog C5 2.233 Dlog C6
• 0.908 MA(4)

12
Example Japan H-F model forecast versus
consensus forecast

• Source Y. Inada

13
Convergence in the rolling forecast of the US H-F
model
M1 M2 M3 M4
Mean error -2.625 -0.5475 -0.665 -0.8375
RMSE 3.607652 1.853126 1.144312 1.4058
14
Convergence in the rolling forecast of the Japan
H-F model
M1 M2 M3 M4
Mean error -7.3 -6.7 -3.2 -0.7
RMSE 8.4 7.2 4.5 2.1
15
turning point oya versus saarExample of
Chinas GDP

• Sources China NBS, JPM

16
Importance of correct estimate of potential
output for counter cyclical macroeconomic policy

• Taylor rule
• Hodrick-Prescott filter for estimating potential
  GDP growth
• Production function for estimating potential GDP
  growth

17
Estimate of output Gap for the US economyby H-P
filter
18
Estimate output Gap for the US economyby
production function

• Source Business Week

19
Are these Output GAPs corrected estimated?
Output gap of GDP
Record levels of spare capacity
11
10
09
Source World Bank.
20
Concluding remarks

• Its a big challenge to make a timely and
  consistent estimate and forecast for economic
  trends during financial crisis
• But they are crucial for macroeconomic policies
• We can make improvement