Pooled Time Series CrossSection Estimation of Demand for

Pooled Time Series Cross-Section Estimation of Demand for Gasoline and Diesel in G 7 Countries 2009 International Energy Workshop 17 -19 June Venice Italy Mehdi Asali (Ph. D. ) Petroleum Studies Department OPEC

Outline • Introduction • Modeling Approach • Disaggregation of data • Structure of the Models and Estimation Results • Concluding Remarks

Introduction: Crude and Products Prices

Introduction: Per Car Gasoline Consumption

Introduction: Real Consumer Prices of Gasoline

Introduction: Real Per Capita GDP

Modeling Approach: VAR, ARDL and Pooled Time Series Cross-Section • VAR and ARDL models are used for individual country estimates. Pooled time series cross-section models are estimated for all G 7 countries data combined • Pooled time series cross-section is a method of studying a particular subject (e. g. demand for products) within multiple sites (e. g. countries) periodically observed over a time frame • The combination of time series with cross-section can enhance quality and quantity of estimations

Structural of the VAR Model

General Structural of the Pooled Model (ARDL)

Individual Country Specific and Polled Estimation • Two VAR and ARDL type econometric models are estimated for each G 7 country individually and a pooled time series cross-section model is estimated for all countries as a whole • Here we have used about (80*7*6) 3360 observations for our panel estimations and compared the findings with that of individual countries’ estimation results

Disaggregation of Data • • Different frequency of time series: annual car ownership, quarterly demand for gasoline Lack of time series with high frequency for developing countries (China, India, . . ) Temporal disaggregation of data in applied econometrics: Mathematical, Statistical and State Space Aproach Handbooks by IMF and EC Eurostat, ECOTRIM

State Space Algorithm of Temporal Disaggregation. Process 12

State Space Approach

Statistical Approach, Fernandez and Litterman

Statistical Approach, Fernandez and Litterman

De-seasonalisation of the Quarterly Time Series

De-seasonalisation of the Quarterly Time Series

Estimation Results • Individual Country Estimates (Unrestricted VAR Models and Auto. Regressive Distributed Lag ARDL Models) • Pooled Time Series Cross-Section Estimates

Individual Country Estimates • • • VAR and ARDL(1, 1, 1, 1) models are estimated for each one of the G 7 countries The main results include (estimates of) the short and long-run price and income elasticities of demand for gasoline and diesel and speed of adjustment in economies of concerned This allows comparison of elasticities of demand for gasoline and diesel in these countries and G 7 as a whole

Individual Country Estimates: Canada

Individual Country Estimates: France

Individual Country Estimates: Germany

Individual Country Estimates: Italy

Individual Country Estimates: Japan

Individual Country Estimates: UK

Individual Country Estimates: USA

Price and Income Elasticities of (Per- Car) Demand for Gasoline (G 7) Per GDP Elasticity (εy) Price Elasticity (εP) Short-run Long-run Short-run Canada 0. 12 0. 387 -0. 02 -0. 06 0. 32 France -0. 14 -0. 95 -0. 03 -0. 29 0. 15 Germany -0. 21 -2. 8 -0. 03 -0. 45 0. 08 Italy -0. 19 -1. 9 - - 0. 10 Japan -0. 11 -0. 50 - - 0. 22 UK -0. 10 -1. 9 -0. 013 -0. 26 0. 05 USA 0. 11 0. 73 -0. 03 -0. 20 0. 15 Long-run Speed of Adjustment

Pooled Model Estimation Results

Pooled Model Estimation Results

Pooled Model Estimation Results Variable Coefficient Short-run Effect t-statistic Long-run Effect Per Capita GDP -0. 024 -0. 20 (-1. 1*) Gasoline Prices -0. 019 -0. 16 (-5. 63) coefficient Lagged Demand for Gasoline 0. 87 (50. 23) Demand for Diesel (G 7) 0. 10 (4. 5) Lagged Demand for Diesel (Ca) -0. 11 (-5. 34) Lagged Demand for Diesel (Fr) -0. 12 (-6. 54) Lagged Demand for Diesel (Ge) -0. 13 (-6. 45) Lagged Demand for Diesel (It) -0. 14 (-6. 80) Lagged Demand for Diesel (Ja) -0. 12 (-6. 24) Lagged Demand for Diesel (UK) -0. 11 (-5. 42) Lagged Demand for Diesel (US) -0. 08 (-4. 04)

Estimating a Partial Adjustment Model of Demand for Oil in G 7 Countries

Summary of Estimation Results for Demand for Oil in G 7 Countries Price Elasticity Short-run GDP Elasticity Long-run Canada 0. 05 0. 14 0. 54 France -0. 02 -0. 07 0. 52 Germany -0. 04 -0. 17 0. 54 Italy -0. 05 -0. 20 0. 54 UK -0. 04 -0. 15 0. 52 US -0. 01 -0. 03 0. 59 Range -0. 01 to -0. 06 -0. 03 to -0. 21 0. 52 to 0. 59

Quarterly Changes of GDP Elasticity of Demand for Oil in G 7 Countries

Summary and Concluding Remarks 1 of 2 • • In this report (per car) demand for gasoline and diesel in G 7 countries for the period of 1990 -2009 is investigated using VAR, ARDL and Pooled Time Series Cross 0 Section methods Models are of quarterly frequency and we had to disaggregate some of the time series that were only available annually (car ownership) to arrive at quarterly data

Summary and Concluding Remarks • • 2 of 2 Statistically significant negative relations between (per car) demand for gasoline and increases in per capita income in 5 out of 7 countries under study Only for USA and Canada a positive relation between per capita income growth was found Gasoline prices, and particularly its ratio to prices of diesel appear to exert significant negative impact on demand for gasoline An increase in demand for diesel reduces demand for gasoline with one period lag and not at the same period

Thank you