Not All Oil Price Shocks Are Alike: A Neoclassical Perspective

Not All Oil Price Shocks Are Alike: A Neoclassical Perspective Vipin Arora Pedro Gomis-Porqueras Junsang Lee U.S. EIA Deakin Univ. SKKU December 16, 2013 GRIPS Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 1 / 32

FAQ What is the consequence of oil price changes on macroeconomic performance? or What is business cycle facts of both oil and macroeconomic aggregates? Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 2 / 32

Motivation No 1: RBC literature A seminal work by Kim and Loungani (1992, JME) takes oil price dynamics as an exogenous shock process found that exogenous oil price shocks can explain only a modest component of the variance of US business cycle. Rotemberg and Woodford (1996) introduces mark-up pricing into a standard RBC to replicate some of the observed movements in US GDP and wages with oil price Finn (2000) obtains similar results using variable capacity utilization in a standard RBC Bottom Line: Many papers/researches model the oil price dynamics as an exogenous process. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 3 / 32

Motivation No 2: Empirical literature An seminal empirical work by Kilian (2009) emphasizes that the real price of oil is ultimately determined by market forces and subject to demand and supply shocks. Bottom Line: Endogenized oil price. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 4 / 32

Motivation: RBC literature Some papers took the oil price as an endogenous variable in models. Backus and Crucini (2000) and Bodenstein et al. (2011) Nakov and Pescatori (2009) Nakov and Pescatori (2010) Balke et al. (2010) Bottom Line: Endogenized oil price, But oil price dynamics and co-movements with macroeconomics aggregates not reported. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 5 / 32

Motivation: Business Cycles Statistics 1974-1983 Variables Corr(X t, Y t ) SD X /SD Y Corr(X t, X t 1 ) Y t 1.00 1.00 0.835 C t 0.834 0.773 0.866 I t 0.934 4.682 0.778 N t 0.924 1.136 0.867 p q,t -0.093 2.818 0.817 Q t 0.683 2.136 0.66 Table : Quarterly Real U.S. and Oil Market Summary Statistics 1974-1983 Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 6 / 32

Motivation: Business Cycles Statistics 1974-2011 Variables Corr(X t, Y t ) SD X /SD Y Corr(X t, X t 1 ) Y t 1.00 1.00 0.862 C t 0.864 0.800 0.882 I t 0.915 5.200 0.835 N t 0.889 1.333 0.911 p q,t 0.152 7.133 0.678 Q t 0.336 1.933 0.639 Table : Quarterly Real U.S. and Oil Market Summary Statistics 1974-2011 Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 7 / 32

What we do in this paper Consider Kim and Loungani (1992, JME) as a benchmark RBC model where oil price are exogenous. Extend the benchmark model by introducing an oil exporting region so that oil price and quantities are endogenous. Introduce typical extensions of variable capacity utilization (VCU) as is Finn (2000), and imperfect competition (MC) as in Rotemberg and Woodford (1996). Highlight the key differences and shortcoming of the models considered. What we do not do? Oil exhaustibility Storages / Inventories Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 8 / 32

Our Objectives to provide a framework for studying the propagation of oil shocks through the economy. to account for the standard business cycle facts of both oil and macroeconomic aggregates. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 9 / 32

Quantitative Results with Exogenous prices Correlations with Y t C t I t N t p q,t Q t Data Model 0.864 0.880 0.915 0.916 0.889-0.191 0.152-0.028 0.336 0.518 Standard Deviations Relative to Y t Y t C t I t N t p q,t Q t Data Model 1.00 1.00 0.800 0.634 5.200 2.735 1.333 0.000 7.133 5.023 1.933 1.780 First-Order Autocorrelation Y t C t I t N t p q,t Q t Data 0.862 0.882 0.835 0.911 0.678 0.639 Model 0.954 0.994 0.911 0.965 0.565 0.718 Table : U.S. Business Cycle Predictions with Exogenous Oil Prices Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 10 / 32

Quantitative Results with Endogenous prices Correlations with Y t C t I t N t p q,t Q t Data 0.864 0.915 0.889 0.152 0.336 Exog. 0.880 0.916-0.191-0.028 0.518 Model 0.941 0.925-0.598 0.625 0.651 Standard Deviations Relative to Y t Y t C t I t N t p q,t Q t Data 1.00 0.800 5.200 1.333 7.133 1.933 Exog. 1.00 0.634 2.735 0.000 5.023 1.780 Model 1.00 0.725 2.239 0.491 1.390 0.823 First-Order Autocorrelation Y t C t I t N t p q,t Q t Data 0.862 0.882 0.835 0.911 0.678 0.639 Exog. 0.954 0.994 0.911 0.965 0.565 0.718 Model 0.979 0.995 0.948 0.994 0.652 0.739 Table : U.S. Business Cycle Predictions with Endogenous Oil Prices Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 11 / 32

Key features of the endogenous oil model There are two regions: an oil importing and an oil exporting. Oil is an intermediate input to final goods production, and both regions consume only final goods. Crude oil is not demanded directly by consumers. Capital and labor are not mobile across regions. Representative consumer owns the capital stock. The intermediate goods producers are monopolistically competitive. Technology shocks on final goods production (an oil demand shock), and on oil production (an oil supply shock). Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 12 / 32

Model overview Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 13 / 32

Consumer in Oil importing region subject to max {C t,n t,k t+1,t t } t=0 { ( )} E t β t C 1 σc t N 1 ξ t + ξ 0 1 σ c 1 ξ t=0 C t + I t = w t N t + r t T t K t + π t I t = K t+1 (1 δ t )K t Because of variable capacity utilisation (T t ), the depreciation rate is: δ t = δt t η Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 14 / 32

Intermediate firms problem in Oil importing region Continuum of intermediate goods producers indexed by i [0, 1] that behave as imperfect competitors. max Q t (i),k t (i),n t (i) subject to p y,t (i)y t (i) p q,t Q t (i) r t T t K t (i) w t N t (i) Y t (i) = Z t J t (i) ψ N t (i) 1 ψ where p y,t (i) is the price of firm i s good, Z t is an exogenous (aggregate) total factor productivity (TFP) shock, J t (i) are capital services and p q,t is the price of oil. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 15 / 32

Intermediate firms problem in Oil importing region Capital services are a CES composite of the capital stock and oil: J t (i) = [γq t (i) τ + (1 γ)[t t (i)k t (i)] τ ] 1 τ where γ is an oil share, and τ = (σ qk 1) σ qk, with σ qk the elasticity of substitution between oil and capital. Dynamics Driven by Shocks The total factor productivity shock evolves according to an AR(1): ln Z t = ρ ln Z t 1 + ɛ t where ɛ t i.i.d N(0, σv) 2 which captures demand oil shocks. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 16 / 32

Final good producer s problem in Oil importing region The final goods producing firm behaves competitively and chooses intermediate inputs each period as to maximise profits Y t p y,t (i)y t (i)di max Yt (i)di subject to [ Y t = ] 1 Y t (i) θ θ di These goods are bought by consumers in the oil importing and exporting region. Why? Rotemberg and Woodford (1996) showed that imperfect competition significantly increases the predicted effects of an energy price increase on both output and real wages. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 17 / 32

Firm s problem in Oil exporting region Standard competitive firm that takes prices as given, choosing capital and labour as to maximise profits subject to max K t,n t p q,t Q t rt Kt wt Nt Q t = Z t K (α) t N (1 α) t Dynamics Driven by Shocks Zt follows an AR(1) process as Z t, but may have different persistence and volatility. Captures oil supply shocks. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 18 / 32

Definition of A Competitive Equilibrium A list of prices (capital, labour, oil, and intermediate goods), allocations (consumption, labour, final goods, intermediate goods, oil, investment, and capital), a utilisation rate, and exogenous technology processes such that: 1 Households in both regions maximise utility, taking prices as given. 2 The final goods and oil firms maximise profits, taking prices as given. 3 The intermediate goods firm maximises profits, taking all prices but its own as given. 4 The oil, labour, and final goods markets clear. 5 Each consumer s budget constraint is met. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 19 / 32

Some observations: In equilibrium we have p q,t = θψγq t τ 1 Y t J t τ r t = θψ(1 γ)(t tk t ) τ 1 Y t J t τ w t = θ(1 ψ)y t N t Due to monopolistic competition, there is a markup ( 1 θ ) associated with the price of final goods which reduces the marginal products of all factors relative to the perfectly competitive case (θ=1). Macroeconomic conditions in the oil importing region directly affect the price of oil by shifting the demand for oil. Oil prices respond to economic forces, as any other input price. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 20 / 32

Data Consider quarterly U.S. data for the period 1974 2011: Variables Description Data Source Y t Real GDP FRED C t Real Consumption FRED I t Real Investment FRED N t Hours FRED p q,t Deflated US refiner acquisition costs U.S. EIA of oil imports Q t Non-US Oil Production U.S. EIA Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 21 / 32

Calibration Parameters are chosen to match quarterly data. σ c, β, and capital shares (ψ and α) are given standard values for a quarterly calibration of 2, 0.99, and 0.36. Choose δ in the oil consuming region to set steady state K Y =12. Choose γ, the oil share in capital services, to match average value of U.S. oil imports (0.015). Choose ξ 0 so that labour supply in both regions is 0.33. Choose ξ so that Frisch elasticity is 2. Choose η to get steady state T at 0.80. Choose θ = 0.9, to have a mark-up of 1.1. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 22 / 32

Calibration: Shock Processes z and z The following parameters (σ q k, ρ, ρ, σ ν, σν) are calibrated by the simulated method of moments as in Bodenstein, Erceg, and Guerrieri(2011;JIE). the elasticity of substitution between oil and capital: (σq k) in consuming region the first-order autocorrelation on each shock process: (ρ, ρ ) the volatilities of each shock: (σν, σν) calibrated by minimizing the square of the distance between simulated model moments and those observed in the sample data. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 23 / 32

Benchmark Model: Exogenous prices Assume that the oil price follows an AR(1) process and that the oil produced is exactly what the oil consuming region demands. Note that in equilibrium ( pq,t J ρ ) 1 ρ 1 t Q t = θψγy t If p q,t J t (capital and oil are complements) Y t. Of course this negative correlation between p q,t and Y t may be off-set somewhat by general equilibrium effects, but this can only work through Y t because p q,t is exogenous. Oil prices cannot respond to changes in macroeconomic conditions in oil consuming region. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 24 / 32

Model with Endogenous oil prices The quantitative results indicate that both of the models give similar results with respect to the standard real business cycle macroeconomic aggregates However, they differ substantially in predictions regarding the oil price and oil demand over the business cycle. The model with an exogenous oil price yields a counterfactual correlation between the real oil price and U.S. GDP. In contrast, it is the model that comes closest to accounting for the correlation between non-u.s. oil production and U.S. GDP. The exogenous oil price model is not suitable for studying the implications of changes in the oil price due to variations in demand. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 27 / 32

Structural VAR exercise as in Kilian (2009) Based on quarterly data for Z t = (Q, G, P o ), where Q, G, P o denotes the percentage deviation from trend of oil production, OECD production of total industry index, the real price of oil, respectively. The sample period is 1975:Q1 to 2011:Q4. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 28 / 32

Structural VAR exercise as in Kilian (2009) SVAR representation: A 0 Z t = α + 4 A i Z i 1 + ε t i=1 where ε t denotes the vector of serially and mutually uncorrelated structural innovations. A 1 0 has a recursive structure such that the reduced-form errors e t can be decomposed according to e t = A 1 0 ε t: e t e Q t e G t e P 0 t = a 11 0 0 a 21 a 22 0 a 31 a 32 a 33 oil supply shock εt aggregate demand shock εt oil specific-demand shock εt Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 29 / 32

Impulse Responses with Endogenous oil prices IR of price of oil to the shock that induces 1% increase in Oil Production. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 30 / 32

Impulse Responses with Endogenous oil prices IR of price oil to the shock that induces 1% increase in Global Industry Production. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 31 / 32

Conclusions Exogenous oil prices into a standard real business cycle model (and variants) accounts poorly for the oil business cycle facts and yields counterfactual predictions with respect to oil business cycle facts. When oil is endogenous, we substantially improve oil dynamics and co-movements and slightly improve standard business cycle predictions for consumption and investment. The model falls short of accounting the relative volatilities of the oil price with respect to output. Junsang Lee (SKKU) Oil Price Dynamics in RBC Model December 16, 2013 GRIPS 32 / 32