Gas Does Affect Oil. Evidence from Intraday Prices and Inventory Announcements. Marketa W. Halova & Robert Rosenman. Washington State University

Transcription

1 Gas Does Affect Oil Evidence from Intraday Prices and Inventory Announcements Marketa W. Halova & Robert Rosenman Washington State University 06/19/13

2 Overview Question: Do events in the natural gas market cause spillovers into the crude oil market? Previous empirical studies: Cointegration studies (Villar & Joutz 2006 and others) One-way causality (Pindyck 2004; Asche, Osmundsen & Sandsmark 2007) Our approach: 1. High-frequency price data (10-minute intervals) 2. Include inventory (U.S. weekly inventory announcements, i.e., specific events) Main result: Bi-directional causality between oil and gas markets

3 Data: Oil and gas futures intraday prices Oil and gas NYMEX futures prices Intraday data on individual trades 10-min interval Trading day: Mo-Fr 9:00 am-2:30 pm ET (33 intervals) 06/13/03-09/24/10 (54, min intervals on 1,826 days) Different maturities (1 mo-9 yrs for oil, 1 mo-12 yrs for gas) P k,j is price of k {O, G} at the end of 10-min interval j Dependent variable: R k,j = ln(p k,j ) ln(p k,j 1 ) (Ding, Granger & Engle 1993; Ederington & Lee 1993; Bollerslev, Cai & Song 2000 and others)

4 Data: Oil price volatility intraday pattern

5 Data: Oil and gas weekly inventory reports Assumes efficient markets, i.e., only the unanticipated component of news announcements matters Actuals: U.S. Department of Energy Financial Times 09/21/11: The U.S. is the world s largest oil consumer and its weekly petroleum status reports are closely watched. Oil: We 10:30 am Gas: Th 10:30 am Expectations: Bloomberg Survey of approx. 25 industry experts Median forecast Surprise in percentage terms k,j A k,j E k,j A k,j 100

6 Methodology: Accounting for asymmetry S kj < 0: A kj < E kj (Actual inventory is lower than what analysts expected) Analysts overforecast inventory Shortage S kj > 0: A kj > E kj (Actual inventory is higher than what analysts expected) Analysts underforecast inventory Glut

7 Methodology: Regression Regression: R j = α + Σ I i=1 β i R j i + Σ K k=1 ΣL l=0 γ kl S k,j l + +Σ K k=1 ΣL l=0 δ kl S k,j l I (S k,j l > 0) + Σ M m=1 θm{z m,j } + ε j (Andersen, Bollersev, Diebold & Vega 2003) Newey-West standard errors GARCH(1,1), GARCH(1,1) with t-distribution, GARCH-in-mean(1,1), EGARCH(1,1)

8 Methodology: Control variables Gasoline inventory Distillate inventory Refinery utilization Beginning-of-the-day dummy End-of-the-day dummy First-trading-day dummy Trader composition 3-month Treasury bill Volume

9 Results: Cross-commodity effect Table 1: Price volatility regressions for OIL nearby contract (1) (2) Oil shortage *** *** S O < 0 (.00066) (.00066) Oil glut *** *** S O > 0 (.00079) (.00079) Gas shortage *** S G < 0 (.00075) Gas glut *** S G > 0 (.00053) Gasoline shortage ** ** S Gasol < 0 (.00047) (.00047) Gasoline glut *** *** S Gasol > 0 (.00069) (.00069) Distillate shortage ** ** S D < 0 (.00067) (.00067) Distillate glut * * S D > 0 (.00053) (.00053) Beg-of-day dummy *** *** (.00026) (.00026) End-of-day dummy *** *** (.00009) (.00009) Trader composition *** *** (.00003) (.00003) T-bill rate *** *** (.00001) (.00001) Volume 1st lag *** *** (.00001) R j 1st lag *** *** (.01285) (.01287)

10 Results: Joint model of oil and gas price volatility Table 2: Joint model of OIL and GAS price volatility (SUR) Oil price volatility Gas price volatility Oil shortage *** ** S O < 0 (.00030) (.00042) Oil glut *** ** S O > 0 (.00030) (.00044) Gas shortage *** *** S G < 0 (.00061) (.00087) Gas glut *** *** S G > 0 (.00044) (.00064)

11 Results: Effect across maturities Table 3: Price volatility of OIL contracts with longer maturities Contract 1 Contract 2 Contract 3 Contract 4 Contract 5 Contract 6 Contract 7 Oil shortage *** *** *** ** *** * S < 0 (.00030) (.00026) (.00026) (.00028) (.00030) (.00031) (.00031) Oil glut *** *** *** *** *** *** *** S > 0 (.00031) (.00027) (.00027) (.00029) (.00031) (.00031) (.00032) Gas shortage *** *** ** * *** * e-06 S < 0 (.00062) (.00054) (.00054) (.00058) (.00063) (.00063) (.00065) Gas glut *** *** *** *** *** *** *** S > 0 (.00045) (.00040) (.00039) (.00042) (.00046) (.00046) (.00047) Table 4: Price volatility of GAS contracts with longer maturities Contract 1 Contract 2 Contract 3 Contract 4 Contract 5 Contract 6 Contract 7 Oil shortage ** ** *** *** ** S < 0 (.00043) (.00041) (.00038) (.00035) (.00035) (.00033) (.00034) Oil glut ** ** * S > 0 (.00044) (.00041) (.00038) (.00036) (.00036) (.00034) (.00035) Gas shortage *** *** *** *** *** *** *** S < 0 (.00088) (.00083) (.00077) (.00072) (.00073) (.00069) (.00070) Gas glut *** *** *** *** *** *** *** S > 0 (.00064) (.00061) (.00056) (.00052) (.00053) (.00050) (.00051)

12 Results: Robustness checks Structural breaks Business cycle O and G price returns O and G price volatility O and G inventory Gasoline and distillate inventory Seasonality Lagged cross-commodity terms Chicago Board Options Exchange Volatility Index (VIX)

13 Results: Summary Cross-commodity effect (Two-way causality supports theory) Long-term effect (Effect across futures contracts maturities)

14 Why are these results important? Understand relationship between the two largest energy sources and futures markets Impact on financial institutions Spillovers from oil market to financial markets Increasingly important due to new gas discoveries