Transcript Title Arial Bold 36 - Philippine Statistics Authority

ENHANCEMENT OF THE LEADING
ECONOMIC INDICATOR SYSTEM
OF THE PHILIPPINES
By
Dennis S. Mapa, Divina Gracia L. del Prado,
Plenee Grace S. Castillo, Al-Ahmadgaid B. Assad,
and Ibarra Aaron R. Poliquit
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Background
•
In 1993, the Leading Economic Indicator System
(LEIS) of the Philippines was developed to short-term
forecast the movement of the Philippine economy.
•
The latest revision of the LEIS was in 2002.
(Bersales, Reyes, and De Guia, 2004)
•
The current LEI system has difficulty anticipating the
upturns and downturns of the economy.
(Tabunda, 2013)
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Composite Leading Indicator
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Current LEI System
• Reference series is Non-Agriculture Gross Value Added
• Eleven (11) indicators:
(1) Consumer Price Index
(2) Electric Energy
Consumption
(3) Peso/Dollar Exchange
Rate
(4) Hotel Occupancy Rate
(5) Money Supply
(6) Number of New
Business Incorporations
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
(7) Stock Price Index
(8) Terms of Trade Index
(9) Total Imports
(10) Tourist/Visitor Arrivals
(11) Wholesale Price Index
Current LEI System
• Challenges
Timeliness of Data
Only four (4) indicators have
available data during the
scheduled computation of
the quarterly composite LEI
Imputation of data that are
not available on time
• Four (4) indicators imputed
using Growth Rate
• Three (3) indicators
forecasted using ARIMA
Models (X-11 DOS based)
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Current LEI System
• Composite Leading Economic Indicator
11
CLEIt 
w
i 1
z
i , li i , t
11
,
t  li  1,...,
i  1,2,..., 11 li  1,2,...,10
where
li - refers to the lead period of the ith indicator
wi ,li - refers to the weight which is the correlation
coefficient of the non-agriculture GVA and the
leading economic indicator i at lead period li
zi ,t - standardized cycle of the leading economic
indicator i at time t
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Computation of Composite LEI
Method
Current
Proposed
Seasonal
Adjustment
(Trend-Cycle
Estimation)
DOS-based X-11
ARIMA
TRAMO-SEATS of
DEMETRA+
software
Detrending
Procedure
(Cycle Extraction)
1. Polynomial trend
model using MS
Excel ( wi ,li )
2. Hodrick-Prescott
using EViews
(zi ,t )
Hodrick-Prescott
Filter using Eviews
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Computation of Composite LEI
Method
Current
Proposed
Weights used in the
aggregation of the
LEIs
Double weighting
scheme ( wi ,li and
the simple average)
Normalization of
Composite LEI
Composite LEI is not Composite LEI is
standardized
standardized
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Standardized Partial
Correlation
Standardized Cycles of GDP and Non-Agriculture GVA, 1991 Q1 - 2015 Q4
(using Hodrick-Prescott Filter Method)
103
102
101
100
99
98
97
1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015
GDP Standardized Cycle
Non-Agriculture GVA Standardized Cycle
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Indicator Series
Candidate Indicator
Aggregation
(1) Business Confidence Index (Current Quarter)
(2) Business Outlook Index (Next Quarter)
(3) Export of Goods
Sum
(4) Government Final Consumption Expenditure (2000 =
100)
(5) Government Final Consumption Expenditure (Current)
(6) Gross International Reserves
Average
(7) Lending Rate
Average
(8) London Inter-Bank Offered Rate
Average
(9) Meralco Sales (GwH)
(10) Money Supply (M2)
(11) National Government Expenditures
Average
Average
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Indicator Series
Candidate Indicator
(12) National Government Revenues
(13) OFW Remittances
(14) Savings Deposit Rate
(15) Singapore Inter-Bank Offered Rate
(16) Time Deposit Rates (Long-Term Rates)
(17) Time Deposit Rates (Short-Term Rates)
(18) Treasury Bill Rates (364-Day Tbill Rates)
(19) Treasury Bill Rates (91-Day Tbill Rates)
(20) Universal and Commercial Bank Loan Outstanding
(21) Volume of Palay
(22) Volume of Production Index
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Aggregation
Average
Sum
Average
Average
Average
Average
Average
Average
Average
Sum
Average
RAW DATA
OF EACH
VARIABLE
(33 TOTAL)
OUTLIER DETECTION AND
CORRECTION, AND ESTIMATON
OF TREND-CYCLE USING
TRAMO-SEASTS OF DEMETRA+
NORMALIZATION
Standardize each cycle
series using the formula:
cycle - mean(cycle)
standard deviation(cycle)
CORRELATION
Correlate each
standardized
cycle to GDP
standardized
cycle
(Cut-off: |0.25|)
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
CYCLE EXTRACTION
USING HODRICKPRESCOTT OF
EVIEWS
GRANGER CAUSALITY
"Granger causality measures
whether one thing happens
before another thing and helps
predict it." (Sorensen, 2005)
Tested at α = 0.10
Proposed Indicator
Correlation
Granger
Causality
Timeliness
Peso/US Dollar
Exchange Rate
-0.5184
0.0181(2)
Every 2nd workday after the
end of the reference month
(final data)
Stock Price Index
0.2711
0.0344(4)
30 days after the end of the
reference month (final data)
0.0728(3)
Last Friday of the month
before the reference quarter
(final data)
Business Expectation
Survey (Confidence
Index Next Quarter)
0.2858
Gross International
Reserves
-0.3695
0.0008(1)
Every 19th day of the month
after the end of the reference
month (final data)
National Government
Revenues
0.2546
0.0826(4)
21 days after the end of the
reference month (final data)
Universal and
Commercial Bank Loan
Outstanding
-0.4950
0.0006(1),
0.0035(3)
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
30 days after the end of the
reference month (preliminary
data)
Comparison of Different Weighting Schemes
• Simple Average
6
CLEIt 
z
i 1
6
i ,t
, t  1,...,
i  1, 2,..., 6
(equation 1)
where
zi ,t - standardized cycle of the leading economic
indicator i at time t
• Standardized Partial Correlation
The individual regression coefficients of the LEIs when fitting a
regression line for the reference series given values of the indicator
series.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Comparison of Different Weighting Schemes
• Correlation at One Quarter Lead Period
6
CLEIt   wi zi ,t ,
t  1,...,
i  1, 2,..., 6
(equation 2)
i 1
where
wi - correlation of the standardized cycles of the
reference series and the indicator series at one
quarter lead period
• Correlation at One to Four Lead Period
- Choose the highest correlation coefficient within one to
four lead quarters
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Summary of Quarter-to-Quarter Forecast of GDP Movement at Different
Weighting Schemes: 1st Quarter 1991 to 4th Quarter 2014
Current LEIs
Weighing Procedure
Indicator
CrossCross
Standardized
Correlation (1 Correlation (1
Partial
Qtr. Lead
to 4 Qtr. Lead
Correlation
Period)
Periods)
Current
Method
Simple
Average
Number of
Correct
Predictions
53
53
62
61
65
Number of
Quarters
92
92
92
92
92
Percentage
of Correct
Predictions
58%
58%
67%
66%
71%
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Summary of Quarter-to-Quarter Forecast of GDP Movement at Different
Weighting Schemes:
Training Dataset (1st Quarter 2002 to 4th Quarter 2011)
Proposed LEIs
Weighing Procedure
Simple Average
CrossCorrelation (1
Qtr. Lead
Period)
Cross
Correlation (1 to
4 Qtr. Lead
Periods)
Standardized
Partial
Correlation
Number of
Correct
Predictions
21
25
25
20
Number of
Quarters
38
38
38
38
55%
66%
66%
53%
Indicator
Percentage of
Correct
Predictions
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Summary of Quarter-to-Quarter Forecast of GDP Movement at Different
Weighting Schemes:
Testing Dataset (1st Quarter 2012 to 4th Quarter 2015)
Proposed LEIs
Weighing Procedure
Simple Average
CrossCorrelation (1
Qtr. Lead
Period)
Cross
Correlation (1 to
4 Qtr. Lead
Periods)
Standardized
Partial
Correlation
Number of
Correct
Predictions
9
11
10
12
Number of
Quarters
16
16
16
16
56%
69%
63%
75%
Indicator
Percentage of
Correct
Predictions
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Estimated GDP Cycle and Composite LEI using Current and Proposed LEIs
throught Standardized Partial Correlation Coefficients as Weights,
1st Quarter 2002 – 4th Quarter 2014
(long-term trend = 100)
103
102
101
100
99
98
97
2002
2004
2006
GDP Cycle
2008
Current LEIs
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
2010
2012
Proposed LEIs
2014
Summary of Possible Turning Points of Standardized Cycles of GDP and
Composite LEI at Different Weighting Schemes
1st Quarter 2002 - 4th Quarter 2014
Current LEIs
Weighing Procedure
Median
Trough
Peak
Simple
Average
CrossCorrelation (1
Qtr. Lead
Period)
Cross
Correlation (1
to 4 Qtr. Lead
Periods)
Standardized
Partial
Correlation
7.5
2
2
2
7
0
0
0.5
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Summary of Possible Turning Points of Standardized Cycles of GDP and
Composite LEI at Different Weighting Schemes
1st Quarter 2002 - 4th Quarter 2015
Proposed LEIs
Weighing Procedure
Median
Trough
Peak
Simple
Average
CrossCorrelation (1
Qtr. Lead
Period)
Cross
Correlation (1
to 4 Qtr. Lead
Periods)
Standardized
Partial
Correlation
2
2
2
0.5
-1.5
-1
-1
-2.5
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
Conclusion
• For seasonal adjustment and detrending methods, TRAMOSEATS of DEMETRA and Hodrick-Prescott (HP) of Eviews,
respectively, were found better than the current methods.
• The best weights were the standardized partial correlations
which were obtained by regressing the reference series and
the indicators series.
• The 6 proposed indicators showed good prediction
performance on the movement of GDP especially in recent
periods. These also have the advantage of being timely and
free from imputations compared to the 11 current indicators.
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City
THANK YOU!
13th National Convention on Statistics
October 3-4, 2016, EDSA Shangri-La Hotel, Mandaluyong City