Anne Neumann Agenda 1. Introduction 2. Data and Methodology 3. - PowerPoint PPT Presentation

The Dynamics of Global Crude Oil Production INFRADAY 9 October, 2010 Aleksandar Zaklan Georg Zachmann Anne Neumann

Agenda 1. Introduction 2. Data and Methodology 3. Results and Discussion 4. Conclusions References - 2 -

Motivation • Crude oil production and prices are key economic variables, but understanding of their interaction over time and for different producers is incomplete • We analyze the dynamic response of crude oil output to past prices and other important control variables • Disentangle determinants of global oil production by major country groups and on individual country level • OPEC, OECD and non-OECD/non-OPEC production • Individual country analysis - 3 -

State of the Literature • Two major streams: models evaluating consequences of physical attributes/exhaustibility and tests of strategic producer behavior • Physical attributes: • Hotelling (1931): resource exhaustibility • Hubbard (1956): oil depletion • Strategic behavior: • Griffin (1985): empirical testing of popular hypotheses: Cartel behavior, competitive behavior, revenue targeting • No thorough dynamic analysis exists • Methodological challenges call into question validity of results - 4 -

Agenda 1. Introduction 2. Data and Methodology 3. Results and Discussion 4. Conclusions References - 5 -

Descriptive Analysis: Graphical 40 150 35 Constant 2005 U.S. dollars ds Million barrel per day (bpd) dddd 100 30 25 50 20 OPEC OECD Non-OECD/non-OPEC Real WTI price (right axis) 15 0 Jan-94 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 OECD: stable and slightly declining trend, with sharp movements OPEC: appears to mirror development in oil price, with a certain delay Non-OECD/non-OPEC: smoothly increasing, broadly in line with trend in price - 6 -

Descriptive Analysis: Principal Components Analysis (PCA) • PCA of oil production (first differences), various samples: 1994-2009 1994-1998 2002-2006 0.5 0.5 0.5 NOR NGA 0.4 0.4 0.4 GBR SYR EGY SAU KAZ IND IND 0.3 0.3 KWT 0.3 ECU GBR USA YEM QAT IDN ARE NGA 0.2 ECU 0.2 0.2 Second PC Second PC AGO Second PC GAB AZE EGY SYR KAZ SYR OMN NOR RUS 0.1 0.1 IRN 0.1 ARG QAT AUS RUS GBR CAN ECU OMN VEN EGY IRN 0.0 0.0 0.0 CAN KWT IND YEM AUS QAT BRA BRA AUS AZE DZA DZA USA KWT VEN NOR OMN DZA NGA RUS MEX CAN MEX IRN CHN IDN VEN -0.1 -0.1 -0.1 GAB BRA USA ARG ARE SAU SAU YEM IDN MEX ARE -0.2 CHN -0.2 -0.2 GAB ARG CHN AGO KAZ AZE -0.3 -0.3 -0.3 AGO -0.4 -0.4 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 First Principal Component First Principal Component First Principal Component Production behavior seems to differ according to: • Country groups • Individual countries • Time - 7 -

Hypotheses Hypothesis 1: Crude oil output responds to prices and other control variables over a range of lags from the short to the long term. Hypothesis 2: The response is heterogeneous among the three main groups of countries, OPEC, OECD and non-OECD/non-OPEC, as well as on the level of individual countries. - 8 -

Methodology • Address stationarity issue by taking first differences of non-stationary variables • Estimation strategy: • Determine effects of both global and local factors • Specify rich lag structure to analyze dynamics 11 K L M                    Q D INST P REAL RIG   t i , i j t j , i t i , k i , t k m i , l i , t l i ,  t m     j 1 k 0 l 0 m 1 N P           SD ( P ) EX USD ( )   k i , t t k k i , t k t i ,   n 0 p 0 • Optimize model structure using a data-driven algorithm based on the Bayesian Information Criterion (BIC) - 9 -

Agenda 1. Introduction 2. Data and Methodology 3. Results and Discussion 4. Conclusions References - 10 -

Role of Prices Oil supply (1) (2) (3) OPEC OECD Non-OECD/non-OPEC Real WTI Price, monthly average -21.50 0.750 -0.963 (0.171) (0.925) (0.575) Real WTI Price, monthly average (-1) 21.97** -10.46* -0.679 (0.031) (0.072) (0.791) Counter-cyclical Real WTI Price, monthly average (-2) 26.14*** -20.32** -2.894 price response (0.001) (0.015) (0.521) Real WTI Price, quarterly average (-1) 33.15 -111.5*** 4.783 (0.517) (0.000) (0.497) Real WTI Price, quarterly average (-2) -69.26** 9.495 Pro-cyclical (0.034) (0.135) short-term Real WTI Price, quarterly average (-3) -68.47* (0.074) response Real WTI Price, yearly average (-1) 37.33 -123.9 (0.856) (0.529) Pro-cyclical Real WTI Price, yearly average (-2) -86.42 -152.0 174.2** Price response (0.722) (0.454) (0.028) Real WTI Price, yearly average (-3) -471.6 174.1* (0.144) (0.067) Real WTI Price, yearly average (-4) -683.0** 36.63 (0.020) (0.515) Counter- Real WTI Price, yearly average (-5) -1.560 (0.975) cyclical longer Real WTI Price, yearly average (-6) 0.297 term response (0.994) Real WTI Price, yearly average (-7) Real WTI Price, yearly average (-8) -792.0*** -123.5 -20.87 (0.000) (0.547) (0.394) Real WTI Price, yearly average (-9) -300.6*** -90.74 (0.000) (0.425) Observations 156 156 156 Robust p-values in parentheses; *, **, *** indicate significance at the 10% , 5% and 1% level, respectively - 11 -

Role of Real Global Activity Oil supply (1) (2) (3) OPEC OECD Non-OECD/non-OPEC Kilian real activity, quarterly average -19.87*** 25.37*** 3.394** (0.003) (0.001) (0.023) Kilian real activity, quarterly average, (-1) 3.282 -16.31 0.581 (0.623) (0.197) (0.714) Kilian real activity, quarterly average, (-2) 7.926* 1.957 2.744 (0.076) (0.876) (0.245) Kilian real activity, quarterly average, (-3) 23.88** -16.80*** 1.447 (0.027) (0.006) (0.161) Kilian real activity, yearly average, (-1) 30.72 -1.927 6.185 (0.260) (0.847) (0.101) Pro-cyclical Kilian real activity, yearly average, (-2) 28.93** -1.340 -0.806 (0.020) (0.964) (0.877) demand effects Kilian real activity, yearly average, (-3) 14.66 -13.55 8.078** (0.409) (0.655) (0.035) Kilian real activity, yearly average, (-4) -9.552 18.20** -1.861 (0.312) (0.027) (0.565) Kilian real activity, yearly average, (-5) -4.019 -18.99 3.670 (0.729) (0.487) (0.267) Kilian real activity, yearly average, (-6) 13.62 -10.14 3.537 (0.503) (0.865) (0.716) Kilian real activity, yearly average, (-7) 42.67** -27.83** 14.81*** (0.020) (0.031) (0.001) Kilian real activity, yearly average, (-8) 62.24*** 39.68 4.362 (0.005) (0.110) (0.300) Kilian real activity, yearly average, (-9) 14.99 -39.89 -2.171 (0.364) (0.376) (0.735) Observations 156 156 156 Robust p-values in parentheses; *, **, *** indicate significance at the 10% , 5% and 1% level, respectively Mostly pro-cyclical Mostly counter-cyclical global global demand effects demand effects - 12 -

Agenda 1. Introduction 2. Data and Methodology 3. Results and Discussion 4. Conclusions References - 13 -

Conclusions • Dynamic structure of our main model is important • We find significant responses across all lags • Static models may only capture part of the effect • In some cases the signs switch along the lag spectrum • Static models may be misleading since net effect may differ • Substantial heterogeneity across both country groups (and individual countries) • Next steps: • Model optimization • Multiple Equations, e.g. SVAR - 14 -

Thank you for your attention! Contact: azaklan@diw.de German Institute for Economic Research