Time-Varying Effects of Oil Supply Shocks on the US Economy

1. Time-Varying Effects of Oil Supply Shocks on the US Economy Christiane Baumeister Ghent University Gert Peersman Ghent University

2. Motivation • The dynamic effects of oil supply shocks on the economy have probably changed over time • Negative oil supply shocks are frequently considered as the underlying source of the 1970s stagflation • The second part of the eighties is recognized by significant declines of oil prices without corresponding effects on economic growth • Recent oil prices have never been as high whilst inflation remains stable and output growth reasonable • This paper investigates the importance of oil supply shocks when time variation is accounted for • Influence on oil production, the real price of crude oil, US real GDP growth and consumer price inflation

3. Why time variation? • The oil market has undergone substantial changes over time • Global capacity utilization rates in crude oil production have not been constant over time • Constantly above sustainable capacity since late 1980s as well as in 1973/74 and 1979/80 (Kilian 2006) • Dramatic rise in oil price volatility since 1986 • Transition from a regime of administered oil prices to a market-based system of direct trading in the spot market and collapse of the OPEC cartel • Lee, Ni and Ratti (1995) and Ferderer (1996): increased volatility has led to a breakdown of oil price-macroeconomy relationship • Relative importance of driving forces behind oil price movements has changed • Shifts in composition of oil supply and demand shocks (Barsky and Kilian 2002, Hamilton 2003, Rotemberg 2007) • Kilian (2007): these shifts help explain unstable estimates over time

4. Why time variation? • Macroeconomic structure has changed over time • Improved monetary policy • Bernanke, Gertler and Watson (1997), Blanchard and Gali (2007) • Time-varying mark-ups of firms (Rotemberg and Woodford 1996) • Changes in firm capacity utilization (Finn 2000) • More flexible labor markets (Blanchard and Gali 2007) • Share and role of oil in the economy has varied over time • Declining share in consumption and production (Bernanke 2006) • Changes in the composition of automobile production and declining overall importance of the automobile sector (Edelstein and Kilian 2007)

5. Contribution • Multivariate time-varying parameters bayesian vector autoregression (TVP-BVAR) with stochastic volatility to model time variation • Used in “Great Moderation” literature: Cogley and Sargent (2002, 2005), Canova and Gambetti (2004), Benati and Mumtaz (2007) • Existing evidence • Splitting the sample in two subperiods assuming a break in mid 1980s (Edelstein and Kilian 2007, Herrera and Pesavento 2007) • Bivariate VARs over different time windows (Blanchard and Gali 2007) • Find a more muted impact on the real economy in more recent times • TVP-BVAR should capture smooth transitions in the propagation mechanism of oil shocks without imposing a specific breakpoint • Stochastic volatility models changes in the magnitude of structural shocks and its immediate impact • Multivariate approach to learn more about sources of variation

6. Contribution • New method to identify exogenous oil supply shocks • Most studies: all variations in oil prices are assumed to be exogenous oil supply shocks • Barsky and Kilian (2002), Kilian (2006): only a small fraction can be attributed to exogenous oil production disruptions • Kilian (2007): exact underlying source is crucial for economic consequences • Hamilton (2003), Kilian (2006): measure oil production shortfalls in the wake of political crises and military conflicts • Selection of episodes is crucial and no generic supply shocks are identified • Kilian (2007): oil supply shocks only source of innovations in oil production (demand shocks only affect oil prices immediately) • Less appropriate for a quarterly VAR • We use sign restrictions from a simple supply and demand model of the global oil market • Oil supply shocks the only disturbances that displace the oil supply curve • Supply and demand shocks can affect oil production and prices

7. Empirical model • TVP-BVAR with global oil production, real crude oil price, US real GDP and US CPI • “Great Moderation” literature: Cogley and Sargent (2002, 2005), Canova and Gambetti (2004), Benati and Mumtaz (2007) • First differences, 4 lags • 1947Q1-2006Q2 (first 20 years as a training sample) • Drifting coefficients to capture time variation (smooth transition) in the propagation mechanism

8. Empirical model • Stochastic volatility: time-varying covariance matrix • Allows for heteroskedasticity of the shocks and time variation in the simultaneous relationships between the variables • Approach is particularly useful given instabilities in oil-macro relationship • Up to the data to determine source of time-variation: shocks and contemporaneous impact and/or propagation mechanism • Error terms of transition equations are independent of each other and of the innovations of the observation equation • Estimated with Bayesian methods (MCMC algorithm): Primiceri (2005)

9. Poil S1 D S2 P1 P2 Q1 Q2 Qoil Identification • Sign restrictions from a simple supply and demand model of the global oil market • Faust (1998), Uhlig (2005), Peersman (2005) • Oil supply shocks: disturbances that displace the oil supply curve • Shock which moves oil prices and oil production in opposite direction • Other oil (demand) shocks moves prices and production in same direction • Supply and demand shocks can affect oil production and prices immediately and this impact can change over time

10. Results

11. Results • Considerable time variation which is statistically significant • Typical oil supply shock is characterized by a much smaller impact on world oil production and a greater effect on the real price of crude oil over time

12. Results • Considerable time variation which is statistically significant • Typical oil supply shock is characterized by a much smaller impact on world oil production and a greater effect on the real price of crude oil over time • Complicates the analysis of time-varying dynamic effects • The way the normalization is done becomes very important

13. 10% oil price rise 1% oil production shortfall Results

14. Results • Considerable time variation which is statistically significant • Typical oil supply shock is characterized by a much smaller impact on world oil production and a greater effect on the real price of crude oil over time • Complicates the analysis of time-varying dynamic effects • The way the normalization is done becomes very important • Comparing the impact of a 10 percent rise of real oil prices (Blanchard and Gali 2007): smaller impact on real economic activity over time • Assumes a constant oil demand curve over time • Corresponds to a fall in oil production of currently 1-2 percent whilst 15 percent in 1970s • Comparing the impact of a 1 percent production shortfall: currently substantial stronger impact on oil prices and consequently also on real economic activity and inflation • What does really matter? Oil production or oil prices?

15. Panel A Panel B D’ Poil Poil S1 D D S1 S2’ S2 S2 P1 P1 P2’ P2 P2 P2’ Q1 Q1 Q2’ Q2 Q2’ Q2 Qoil Qoil Interpretation • Why such a change in the oil market over time? • Identification problem: magnitude of shocks versus contemporaneous impact (economic structure) • Oil demand curve must have become steeper (less elastic) over time!

16. Interpretation • Why less elastic oil demand w.r.t. oil price fluctuations? • Transition from a regime of administered prices to a (flexible) market-based system of direct trading in the spot market (mid 1980s) cannot explain this • Could only affect the speed of adjustment, not the long-run effect (correct fundamental price) • Even smaller volatility of oil production detected • Steeper oil demand curve could be a source of increased volatility and transition to flexible system • “Great Moderation” (mid 1980s) cannot explain this • No obvious reason for a steeper oil demand curve • If it is the quantity of oil production what really matters, production disruptions became smaller over time (1/4th) and could be considered as a source of reduced macro volatility

17. Interpretation • Why less elastic oil demand w.r.t. oil price fluctuations? • Structural changes in industrialized economies since id-1980s • High oil prices of 1970s caused industries to switch away from oil to other sources of energy • Smaller cost share of oil in total production (oil intensity) can make firms less reactive to oil price movements • Especially in environment of increased oil price volatility where cost increases are likely to be reversed quickly • Remaining amount of oil is absolute necessity, so less elastic because there are no substitutes anymore (e.g. transportation) • Declining share of industries with high price elasticity of oil demand • Developing economies that are in the process of industrialization (e.g. China, India) heavily rely on oil and are therefore particularly dependent on oil as an input factor • Their increasing (less elastic) share in world oil demand could have contributed to the steepening of the oil demand curve

18. Interpretation • Why less elastic oil demand w.r.t. oil price fluctuations? • Capacity utilization rates of crude oil production: close to full capacity can lead to relative higher share of (less elastic) precautionary oil demand • A production shortfall cannot be replaced somewhere else • Can also explain the increased price elasticity in 1973/74 and 1979/80

19. Other findings • Variance decompositions

20. Other findings • Variance decompositions • Oil supply shocks account for approx. 30% of oil production variance, which is rather constant over time • Contribution to real GDP and consumer price inflation: 10%-20% • Share in output volatility relatively constant over time • Slight increase in explaining inflation (decreased inflation variability combined with higher leverage effect on oil prices) • Contribution to variability of oil price declines over time from 30% to 15% • Current oil price fluctuations are more demand driven • Also the oil supply curve must have become steeper over time • Confirmed in further research (Baumeister and Peersman, 2008b) • Steeper oil supply curve can also be considered as a source of increased oil price volatility

21. Other findings • Baumeister and Peersman (2008b)

22. Other findings • Historical contributions

23. Other findings • Historical contributions • Major oil shocks are not only supply driven • Baumeister and Peersman (2008b): also oil-specific demand shocks and demand shocks resulting from economic activity (e.g. monetary policy) played an important role • Confirmation of Barsky and Kilian (2002, 2004) • Significant (but non-exclusive!) role in 1974/75 and 1990s recessions • Minor importance in 1980/81 and millennium slowdowns • Explain little of the “Great Inflation”

24. Robustness • Very robust for several alternative variables, specifications and implementation sign restrictions • E.g. adding interest rate, using GDP deflator or unemployment • Always a steeper oil demand curve over time • Model properties: time-varying parameter methodology • Splitting sample in two and use sign restrictions

25. Robustness • Model properties: identification stra tegy • TVP-BVAR but using Choleski decomposition (Kilian 2007)

26. Conclusions • Remarkable structural change in the oil market over time • “Typical” oil supply shock is characterized by a much smaller impact on world oil production and a greater effect on the real price of crude oil over time • Steepening of the oil demand curve (less elastic oil demand) is the only possible explanation for this stylized fact • Complicates the comparison over time of specific experiments (because a constant slope is implicitly assumed) • The contribution of oil supply shock to real oil price fluctuations has decreased considerably over time • Current oil prices are more demand driven • Also oil supply curve became steeper over time • The role of supply shocks in explaining the macro-economy • Significant but non-exclusive role in 1974/75 and early 1990s recessions • Minor importance in the 1980/81 and millennium slowdowns • Explains little of the “Great Inflation” • Increased relative importance to explain inflation variability