GDP flash estimates based on ESI: Does it work

1
GDP flash estimates based on ESI Does it work?

• An econometric approach
• using real data for Slovakia

Ján Haluka Institute of Informatics and
Statistics (INFOSTAT) Bratislava
2
Content

• Business and consumer tendency surveys (BCTS) in
  Slovakia
• Methodology of GDP flash estimates based on
  econometric approach
• Results of econometric modelling
• Conclusion

3
Motivation

• since 2005 the Statistical Office of the Slovak
  Republic (SR) has been obliged to compute and
  publish flash estimates of GDP (and total
  employment) always within 45 days after the end
  of each quarter

• it is 15 days earlier than preliminary data about
  economic development in a given quarter are
  released

• main objective is to create a specific model
  framework based on BCTS results supporting
  preparation of GDP flash estimates

4
Motivation

• BCTS results are published the last working day
  of each reference month (quarter) while GDP
  figure is published on a quarterly basis 60 days
  after the end of each reference quarter

• Economic Sentiment Indicator (ESI) is the most
  popular composite indicator primarily used to
  anticipate or forecast the performance of key
  economic variables

• ESI is being used as a reference (explanatory)
  variable in econometric model for GDP flash
  estimates

5
BCTS in SLOVAKIA

• in Slovakia BCTS have been conducted on the
  monthly basis by the Statistical Office of the SR
  for industry, construction and retail trade since
  1993 and for services since 2002

• fully harmonised form with the methodology
  recommended by the European Commission was
  reached in 1998

• ESI follows a common methodological approach
  developed by the U.S. National Bureau of Economic
  Research (NBER) for U.S. indicator

6
BCTS in SLOVAKIA

• ESI summarizes information gained from BCTS among
  economic actors

• respondents have the choice of fixed set of
  answers for their assessment of the current or
  future economic situation (positive, neutral and
  negative)

• ESI facilitates the interpretation of BCTS
  results as it summarizes the answers for
  different variables in a single number and in a
  simple time series

7
BCTS in SLOVAKIA

• ESI4 is calculated as a weighted average of four
  confidence indicators

• in industry (40), construction (20), retail
  trade (20) and consumer confidence indicator
  (20)

• ESI4 started in Slovakia in January 1996, i.e. 52
  observations exist on the quarterly basis

8
BCTS in SLOVAKIA

• ESI5 is calculated as a weighted average of five
  confidence indicators

• in industry (40), services (30), construction
  (5), retail trade (5) and consumer confidence
  indicator (20)

• ESI5 started in Slovakia in January 2002, i.e. 24
  observations exist on the quarterly basis

9
BCTS in SLOVAKIA

• BCTS results for Slovakia can be found on the
  website of the Statistical Office of the SR

www.statistics.sk
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
Methodology of GDP flash estimates based on
econometric approach

• ESI should be compared with the reference
  variable recording movements in the economy as a
  whole, i.e. real GDP growth (compared to the same
  period of the previous year)

• the initial hypothesis it is assumed that there
  exists statistically significant dependency
  between percentage growth rate of GDP (compared
  to the same quarter of the previous year) and ESI

• quarterly time series of ESI created from its
  original, i.e. monthly time series (simple
  arithmetic mean)

16
Coefficient of correlation 0.734 (40
observations)
17
Methodology of GDP flash estimates based on
econometric approach

• time series of both GDP and ESI are I(1), i.e.
  non-stationary using OLS regression provides
  incorrect conclusions (spurious regression)

• the starting hypothesis real GDP is assumed to
  grow at a constant rate, however, changes in ESI
  are supposed to make this rate variable

• construction of the ECM relationship based on two
  steps the Engle-Granger approach

18
Methodology of GDP flash estimates based on
econometric approach

• long-term equilibrium (LTE) between the
  non-stationary variables is estimated

GDP ? e b TIME c ESI or log (GDP) a
b TIME c ESI

• ECM relationship is estimated using the
  stationary time series of residuals derived from
  LTE

19
Methodology of GDP flash estimates based on
econometric approach

• methodology applied in BUSY model has been used
  by the European Commission since 1996

20
Results of econometric modelling

• two ECMs created and estimated for GDP flash
  estimates using original quarterly time series
  covering the period from the 1st quarter 1997 to
  the 4th quarter 2007, i.e. 44 observations in
  combination with seasonal dummies

• ECM with broken linear long-term trend

• ECM with quadratic long-term trend

21
LTE with broken linear trend
22
(No Transcript)
23
ECM with broken linear long-term trend
24
(No Transcript)
25
LTE with quadratic trend
26
(No Transcript)
27
ECM with quadratic long-term trend
28
(No Transcript)
29
(No Transcript)
30
Conclusion

• ESI can be considered as a statistically
  significant indicator of real GDP from a
  long-term point of view

• strictly speaking, ESI can be considered as a
  statistically significant indicator of real GDP
  deviations from its long-term trend, which can be
  approximated by either broken linear trend or
  quadratic trend

31
Conclusion

• short-term changes of the indicator of expected
  external demand (IEED) can be considered as
  statistically significant indicator of real GDP
  in short-term period

• both ESI and IEED can be used as explanatory
  factors for construction of model relationship in
  ECM form for flash estimates of real GDP

32
THANK YOU

• FOR YOUR ATTENTION