THE GOVERNMENT OF THE REPUBLIC OF SLOVENIA INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT Iasi, 26 SEPTEMBER 2008 Forecasting macroeconomic Forecasting macroeconomic variables with dynamic factor variables with dynamic factor models – The case of Slovenia models – The case of Slovenia Marko Glažar Marko Glažar
21
Embed
Forecasting macroeconomic variables with dynamic factor models – The case of Slovenia
Iasi , 26 SEPTEMBER 2008. Forecasting macroeconomic variables with dynamic factor models – The case of Slovenia. Marko Glažar. Outline. Introduction Theoretical background Data Results Pseudo out-of-sample analysis Past forecasts compared to realization Conclusion. Introduction. - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
THE GOVERNMENT OF THE REPUBLIC OF SLOVENIAINSTITUTE OF MACROECONOMIC ANALYSIS AND
DEVELOPMENT
Iasi, 26 SEPTEMBER 2008
Forecasting macroeconomic Forecasting macroeconomic variables with dynamic factor variables with dynamic factor models – The case of Sloveniamodels – The case of Slovenia
Marko GlažarMarko Glažar
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENTOutline
Introduction Theoretical background Data Results
– Pseudo out-of-sample analysis– Past forecasts compared to realization
Conclusion
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
– Allowed weak serial correlation of the idiosyncratic errors– Idiosyncratic errors may be cross-correlated and heteroscedastic– Allowed weak correlation among factors and idiosyncratic components
Forecasting models:
h – forecast horizon
hhttt
tht fLyLy )()(
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
Relative mean squared error is the measure for comparison of the models
hT
Tt
htht
hht
hT
Tt
hthti
hht
YY
YYMSE
2
1
2
1
2
|,0
2
|,
ˆ
ˆ
MSE of the factor models is compared to the MSE of the AR model in the pseudo out-of-sample analysis
Theoretical backgroundTheoretical background Altogether we have 158 different models. Differentiated by:
• Number of factors, unbalanced or balanced panel• Inclusion of the AR component• Inclusion of the factor lags• Inclusion of the intercept correction
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
DataData Dataset consists of 80 quarterly series, from 1994:
– National account data– Survey data – confidence indicators– Prices– Foreign trade– Production indices– Labour market– Financial variables
Sources: Eurostat, Statistical Office of the Republic of Slovenia, Centre for European Economic Research, Bank of Slovenia, Ministry of Finance,…
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
In sample forecastIn sample forecastinging performance performance In sample forecasts for GDP growth one horizon ahead, performance of the
best factor model (relative MSE = 0.48) and AR model
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
In sample forecastIn sample forecastinging performance performance In sample forecasts for GDP growth one horizon ahead, performance of the
best factor model (relative MSE = 0.48) and AR model
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
In sample forecastIn sample forecastinging performance performance In sample forecasts for GDP growth one horizon ahead, performance of the
best factor model (relative MSE = 0.48) and AR model
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
In sample forecastIn sample forecastinging performance performance In sample forecasts for GDP growth 4 horizons ahead, performance of the
best factor model (relative MSE = 0.29) and AR model
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
In sample forecastIn sample forecastinging performance performance In sample forecasts for GDP growth 4 horizons ahead, performance of the
best factor model (relative MSE = 0.29) and AR model
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
In sample forecastIn sample forecastinging performance performance In sample forecasts for GDP growth 4 horizons ahead, performance of the
best factor model (relative MSE = 0.29) and AR model
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
In sample forecastIn sample forecastinging performance performance In sample forecasts for INDUSTRIAL PRODUCTION growth one horizon
ahead, performance of the best factor model (relative MSE = 0.69) and AR model
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
In sample forecastIn sample forecastinging performance performance In sample forecasts for INDUSTRIAL PRODUCTION growth one horizon
ahead, performance of the best factor model (relative MSE = 0.69) and AR model
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
In sample forecastIn sample forecastinging performance performance In sample forecasts for INDUSTRIAL PRODUCTION growth one horizon
ahead, performance of the best factor model (relative MSE = 0.69) and AR model
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
ForecastForecastinging performance for performance for annual GDP growthannual GDP growth
Forecasts for the year 2007
Forecasts for the year 2008
Data to Q406 Q107 Q207 Q307DFM forecasts 6.1 6.4 6.3 6.4official IMAD forecasts 4.7 5.8realization 6.1realization after revision 6.8
Data to Q207 Q307 Q407 Q108 Q208DFM forecasts 5.4 5.8 4.2 6.0 5.4official IMAD forecasts 4.6 4.4
INSTITUTE OF MACROECONOMIC ANALYSIS AND DEVELOPMENT
Forecasts with DFM for the year 2007 compared to the realization and IMAD official forecasts
ForecastForecastinging performance performance of the of the growth of GDP componentsgrowth of GDP components