Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
April 9, 2014 OPEC’S KINKED DEMAND CURVE Marc H. Vatter∗ Economist Economic Insight, Incorporated 9 Underhill Street, Nashua, NH 03060-4060 603.402.3433; 503.227.1994; [email protected] Abstract I estimate world demand for crude oil, non-OPEC supply, and the effects of changes in price on world GDP using quarterly data covering 1973 to 2010. World GDP falls more with increases in price than it rises with decreases in price. Therefore, instability in price negatively impacts GDP. World demand and net demand to OPEC are kinked due to the asymmetric effects. The kink implies a vertical discontinuity in OPEC’s marginal revenue curve. Shifts in cost and horizontal shifts in demand are less destabilizing with the kink. However, vertical shifts cause larger changes in price, and this accentuates destabilizing feedback between changes in price and GDP. The kink also implies a range of equilibrium prices and quantities of production for the cartel. If OPEC’s long run marginal cost, inclusive of marginal user cost, is $35/bl in 2014, I estimate that any price between $99/bl and $106/bl would provide OPEC no incentive to change under its long run demand schedule. Using a demand curve applicable to a twelve month period, an increase (shock) to prices above $142/bl would be profitable for OPEC. Net demand to OPEC is quite inelastic in the short run, and estimated contemporaneous effects of price on GDP are negative. Therefore, unstable prices generate countercyclic profits for OPEC, which have hedging value in financial markets, further incenting OPEC to set unstable prices. JEL Classification Numbers: Q34; Q41; Q43 Keywords: OPEC; Crude Oil; Price Shock I thank Sophocles Mavroeidis, Harumi Ito, Talbot Page, Sam Van Vactor, John Driscoll, Andres Rosas, Chun Chung Au, and David Weil for helpful comments. I thank my mother, Barbara Hudson, Elizabeth Ferreira, and Micki Spenst for personal support. I thank Barbara and Marshall Hudson for financial support. I thank my father, Harold Vatter, for his support, including many discussions about the U.S. economy after the 1970’s oil shocks. I am grateful that I am someone who can “rejoice in his labor, this is a gift of God” (Ecclesiastes 5:19; Young’s Literal Translation). Errors and views expressed are mine and do not necessarily reflect those of Economic Insight, Inc. ∗ This project began as dissertation research in Economics at Brown University and has continued in affiliation with Economic Insight, Inc., P.O. Box 2295, Sisters, OR 97759; www.econ.com. Contact information for Marc Vatter: vatter.econ.com; [email protected]; 603.402.3433; 503.227.1994; 9 Underhill Street, Nashua, NH 03060-4060. 1 1 Introduction The price of crude oil has been less stable, and marked by upward shocks, and world economic growth has been slower, since the Organization of Petroleum Exporting Countries first wielded its market power assertively in 1973.1 Figure 1 shows the log real price of West Texas Intermediate crude oil and the real rate of growth of the world economy from 1951 to 2010. Figure 1: Log Price of WTI and Annual Growth Rate of World GDP; 1951-2010 8 7 6 5 4 3 2 1 0 -1 GDP 2008 2006 2003 2001 1998 1996 1993 1991 1988 1986 1983 1981 1978 1976 1973 1971 1968 1966 1963 1961 1958 1956 1953 1951 WTI Sources: Federal Reserve Bank of St. Louis; Angus Maddison Project; IMF From the 1930’s through the 1960’s, major international oil companies known as the “seven sisters”2 colluded through periodic agreements whose terms effectively stabilized the price of crude oil above marginal cost. Moran (1993) wrote Taken together, the IPC and the As Is Agreements, operating in tandem with the Texas Railroad Commission in the United States, stabilized 1935-40 world prices at a level ten to eleven times the marginal cost of production in the Persian Gulf. (p. 176; emphasis added) Beyond the automatic execution of standing agreements, much of the short term collusion among these firms had to be secret or tacit. 1 See Greenhouse (1987) for journalistic observations. Their names changed with time. Moran (1993) listed them as Exxon, British Petroleum, Royal Dutch-Shell, Gulf, Mobil, Texaco, and Chevron. 2 In 1954, sales agreements became the principal focus of the Justice Department's anti-trust case against the oil companies. In 1960, the corporations signed a consent decree promising not to engage in any further explicit market-division practices. (p. 178; emphasis added) This also stabilized price because the cost of coordinating a change in price secretly or tacitly was high. The sisters risked uncoordinated changes in price being seen by one another as violations of contracted, secret, or tacit agreements, rather than as price leadership. OPEC was also founded in 1960. Unlike the sisters, OPEC can meet openly to discuss and agree on changes in price and output. Individual members can have much more confidence that other members will raise price (cut output) when they themselves do as part of a price increase that is profitable to the cartel, more confidence than if the price increase had to be accomplished through secret or tacit cooperation. ...the challenge of maintaining an oligopoly should have been and should continue to be easier for the Organization of Petroleum Exporting Countries (OPEC) than it was for the international oil companies: OPEC can meet and negotiate openly, while the companies had to be furtive and wary of public attack..." (p. 159) But maintaining an oligopoly and maintaining price are not the same thing. OPEC generally maintains price above marginal cost in the long run, as the sisters did, but it does not need to accept an unprofitably stable price in the short run. OPEC can effect changes in price ad hoc. OPEC has turned its back on, and relinquished access to, the most important mechanisms of restraint the companies managed to impose on themselves. It has systematically unraveled the corporate structure of supra-sovereign limitations, self-denials, and automatic penalties on the members’ own behavior... (p. 161; emphasis added) Collusion within OPEC in the short run is overt, so it does not have the same stabilizing effect on price that tacit collusion would. Nor is OPEC constrained by non-OPEC producers from changing price. OPEC’s unrivaled market power means that it need not hesitate to increase price for fear that others with market power will quickly raise production and usurp its market share. Non-OPEC production rises significantly with price in the long run, but is insensitive to price in the short run. Here, I argue that OPEC as a whole faces a kinked demand curve, because of asymmetric effects of changes in the price of crude oil on world GDP, not because non-OPEC suppliers have market power. Increases in the price of crude oil lower world GDP, and, therefore, demand for crude oil, more than decreases in price raise them. The kink in OPEC’s demand curve implies a vertical discontinuity in its marginal revenue curve. Within a corresponding range, decreases in production and sales raise 3 price, but reduce revenue by more than they reduce cost, and increases in production and sales lower price, but raise revenue by less than they raise costs. Shifts in cost and horizontal shifts in demand cause less instability in price under a kinked demand curve than under a non-kinked demand curve. With a kinked demand curve, a modest shift in marginal cost will not change the profit-maximizing quantity of production and sales, or price. A proportional horizontal shift in demand will also cause no change in price. A parallel horizontal increase in demand will cause no change or an increase in price, while such a shift always increases price when there is no kink in demand.3 That said, I still argue that the kink likely has de-stabilizing effects on price that exceed the stabilizing effects. First, while the kink gives OPEC stronger incentive not to deviate from any profit-maximizing price/production combination, there is a range of combinations from which OPEC has such a strong incentive not to deviate, rather than the single such combination that would obtain without the kink. Thus, modest “cheating” on quotas, disruptions in production, and the like do not necessarily motivate any stabilizing correction by the cartel. OPEC has been described as “clumsy”.4 The apparent clumsiness may result, in part, from a multiplicity of equilibria in the cartelized market. Second, a vertical shift in demand causes a greater change in price than it would absent the kink. Marginal cost passes through the discontinuity gap in marginal revenue before and after a modest vertical shift in demand, incenting no change in output, leading to a change in price equal to the full vertical shift in demand. I show this in Figure 2. Demand shifts from D to D′ , and price shifts by the same amount, from P to P′ , with no change in quantity produced. 3 4 See Frasco (1993). See Adelman (2004). 4 Figure 2: Vertically Shifting Kinked Demand Curve P’ P MC D’ D Q MR’ MR In contrast, with a non-kinked demand curve, an increase in demand would lead to an increase in price less than the full vertical shift in demand because the firm would increase output as marginal revenue intersected marginal cost at a greater quantity of output. Third, the kink accentuates feedback between the macroeconomy and the price of crude oil. According to Shepherd (1933; pp. 724-5), a change in GDP per capita is best represented by a vertical shift. Exogenous changes in world GDP, then, cause larger changes in the price of crude oil in the presence of the kink. This instability in price will, in turn, further destabilize the macroeconomy. Fourth, the instability in price fostered by the kink produces a countercyclic stream of profits for OPEC, which has hedging value in financial markets. Since changes in price negatively impact the macroeconomy, OPEC’s profits when price is unstable are countercyclic. Demand is inelastic in the short run. Assuming increasing marginal costs, an increase (decrease) in price will raise (lower) revenue, lower (raise) cost, and lower (raise) world GDP. The countercyclic stream of profits can be bundled into a financial instrument that commands a risk premium in financial markets. The premium obtains because such an instrument can be used to smooth out undesirable fluctuations in consumption associated with the macroeconomic instability caused by the changes in price. 5 Fifth and finally, though a change in population is better represented by a horizontal shift, if the macroeconomy is less stable than costs of extracting crude oil and world population, shifts in cost and demand taken together will cause greater instability in the price of crude oil under a kinked demand curve than under a non-kinked curve. OPEC, then, may well find unstable prices more profitable than stable prices. The Seven Sisters as a whole may also have faced a kinked demand curve, but oil prices were more stable before 1973 because the Sisters’ collusion had to be tacit, and because of their greater accountability in general to the government of the United States and other large oil-consuming nations. I estimate world demand for crude oil, non-OPEC supply, the effect of crude oil prices on world GDP, and, therefore, net demand to OPEC. In their survey of literature on energy demand, Atkins and Jazayeri (2004) discuss three major areas of refinement to the traditional model of demand that apply to crude oil: asymmetry; regime change; and changing seasonal patterns. Increases in the price of crude oil affect quantity demanded and GDP differently from decreases. Regarding demand, according to Atkins and Jazayeri (p. 31), “To say that there is an asymmetry of response appears to be observationally equivalent to saying that there is some underlying, longer run decrease in demand due to some kind of energy efficiency of use.” Griffin and Schulman (2005) make a case that a symmetric specification with a trend toward energy saving technical change is superior. Such a trend may include deterministic and stochastic elements. Wing (2008; p. 24) states “Of the changes that occur within industries, disembodied exogenous technical progress is the predominant energy-saving influence.” I model the direct effects of price on demand as symmetric, and I include both a deterministic trend and lagged dependent variables in my regression. I allow for asymmetric effects of crude oil prices on the world economy. In short, I model the market as though all asymmetric impacts of price on demand result from asymmetric impacts of changes in price on GDP. Regression analysis of nonstationary time series can produce spurious estimates. Atkins and Jazayeri argue against the hypothesis of unit roots in oil market data and in favor of multiple structural breaks, or “regime change”. When modeled as dynamic processes with no changes in intercept or trend, world GDP and world production, non-OPEC production, and the price of crude oil are nonstationary. With one major exception, the 1990-91 Gulf War, these apparent nonstationarities are cointegrated by the estimated coefficients. To the considerable extent that the apparent nonstationarities actually 6 reflect structural breaks, the regressions use most of the information the breaks provide in a way that leaves stationary residuals and makes economic sense. The exception, as noted, is that I did model structural breaks in non-OPEC supply around the time of the Gulf War. Changing seasonal patterns observed in the market for crude oil may be the result of global climate change. I make no estimate of such a trend, but I purge the data of sample-wide seasonal regularity, so forecasts based on the estimates do not reflect outdated seasonal patterns in a static sense. I discuss the data in Section 2. I describe the estimates of world demand for and non-OPEC supply of crude oil in Section 3, and the effects of crude oil prices on world GDP in Section 4. In Section 5, I discuss OPEC per se: I calculate net demand to OPEC and present estimated elasticities, marginal revenue gaps, and ranges of profit-maximizing prices. I conclude in Section 6. 2 Data Table 1, at end, shows the basic data, various transformations of which I use to make the estimates. The footnotes to Table 1 explain most of the columns, but Column A is the U.S. refiners’ acquisition cost of imported crude oil, tabulated and defined by the U.S. Energy Information Administration (EIA) as the “world price” of crude oil in its analyses and forecasts. I assume that the world market for crude oil is integrated. According to Adelman (2004; p. 19), “Most oil moves by sea, and ships can be diverted from one destination to another relatively easily. Moreover, much additional oil can be diverted from land shipment to sea. Hence, it is fairly easy to reroute shipments of oil from nations that have a sufficient supply to nations that are experiencing a shortage. It is only a minor exaggeration to say that every barrel in the world competes with every other.” 7 Between early 2011 and mid-2013, oil prices in the North American interior, as measured by the West Texas Intermediate (WTI) New York Mercantile Exchange (NYMEX) benchmark, fell relative to their historic relationship with oil prices elsewhere. Historically high world prices made large quantities of unconventional crude oil economically recoverable. The additional production congested pipelines between Cushing, OK, where the benchmark is priced, and the U.S. Gulf Coast, separating WTI from the world market. The Brent-WTI split stood at $6.31/bl as of March 6, 2014, much narrower than its $27.31/bl average in September, 2011, but still higher than its, slightly negative, historical norm.5 The separation in markets extended to that for retail products. Figure 3 shows the Brent – WTI split (in $/bl) along with the difference between the simple average of retail gasoline prices (in ¢/gallon) across PADDs 1, 3, and 5 (U.S. coastal areas) and that across PADDs 2 and 4 (U.S. interior).6 As the Brent – WTI split became large, so did the difference in gasoline prices. A regression of the difference in gasoline prices on the split, monthly dummies, and twelve lags of the dependent variable gives a coefficient on the split with a t-statistic of 2.14; there is no autocorrelation in the residuals. When the split is small, so is its effect on gasoline prices, but in September 2011, this coefficient implies an elasticity of the difference in gasoline prices with respect to the Brent – WTI split of 0.45, which is significant. I conclude that the North American interior has been a distinct and separate market for crude oil and products since early 2011, and I do not use data from after 2010 in my analysis. Apart from its simplicity, this way of dealing with the fragmentation of the market in recent years means that when I later apply the estimates in a more current (2014) context, I am doing so well out of sample, a good test of the accuracy of a model. 5 More information is available at http://www.eia.gov/todayinenergy/detail.cfm?id=11891, accessed July 15, 2013. 6 “PADD” stands for “Petroleum Administration Defense District”. See http://www.eia.gov/todayinenergy/detail.cfm?id=4890, accessed March 6, 2014, for a map. 8 Figure 3: Brent - WTI Split and U.S. Regional Gasoline Prices 50 40 30 20 10 0 -10 PADDs 1, 3, & 5 less PADDS 2 & 4 2013 2012 2011 2010 2010 2009 2008 2007 2006 2005 2005 2004 2003 2002 2001 2000 2000 1999 1998 1997 1996 1995 1995 1994 1993 -20 Brent less WTI Source: U.S. Energy Information Administration, http://tonto.eia.gov/dnav/pet/hist/LeafHandler.ashx?n=PET&s=EMM_EPM0_PTE_R10_DPG&f=M, accessed March 6, 2014. I estimate the price of crude oil for October-December of 1973 based on monthly data on the rest of this series and free-on-board and landed costs of crude oil imported to the U.S. from January 1974 to April 2009. The prices of crude oil for October-December 1973 used to calculate the quarterly average, highlighted in yellow in Table 1, are predicted values based on the following regression: ∆Pt world = −0.0083+ 0.5037 ∆Pt fob + 0.5435 ∆Pt landed 0.0276 0.0828 0.0856 (1) where t is in months, ∆Pt world is the month-to-month change in the world price, ∆Pt fob the change in the free-on-board price, and ∆Pt landed the change in the landed price. The R 2 = 0.9640 . A dynamic process can be described as I ( d ) , where d = 1 corresponds to a unit root and d = 0 to perfect covariance level stationarity. A generalization of this distinction is to allow d to assume non-integer values. Robinson tests7 estimate the degree of fractional differencing, d, needed to render an I(d) series I(0) using a log-periodogram regression. According to Baillie (1996; p. 21), “For −0.5 < d < 0.5 , the process is covariance stationary, while d < 1 implies mean reversion.” A 7 See Robinson (1995). 9 ( ) Robinson test of the residuals in (1) estimates them to be I −0.2173 , where the number below is 0.0435 the standard error of the estimate. Column H is the percent drop in world output of crude oil caused by war or civil conflict. The episodes of war and civil conflict include the November 1973 Yom Kippur war, the November 1978 onset of the Iranian Revolution, the October 1980 onset of the Iran-Iraq war, the August 1990 Iraqi invasion of Kuwait, civil unrest in Venezuela beginning in December 2002, and the U.S. invasion of Iraq in March of 2003. I derived real crude oil prices using Columns A and B. I derived quarterly world GDP by applying quarterly variation in U.S. GDP, Column F, to annual world GDP, Column E. Both U.S. GDP and U.S. consumption of crude oil declined from a fourth to a fifth of the worldwide totals, and the oil intensity of the U.S. economy was about the same as that of the world economy, throughout the sample. I used world GDP rather than OECD GDP as a measure of income. The non-OECD share of world consumption of crude oil has increased from 25% to 50% since 1970.8 Rapid economic growth has caused demand for oil to grow especially fast in some non-OECD countries such as China and India. I used purchasing power parity (PPP) because market exchange rates are subject to political and speculative influences that do not reflect the real incomes of consumers of crude oil and those affected by the market for it. National GDPs calculated using market exchange rates can deviate from purchasing power parity for extended periods.9 For each series in Table 1, I either used seasonally adjusted data, made adjustments for seasonality, or found no seasonal variation when regressing them on seasonal dummy variables, so I made no seasonal adjustments. I made no adjustments to my methods of estimation to account for the extent to which the data used were estimated or constructed. 8 Source: U.S. Energy Information Administration, http://www.eia.gov/finance/markets/demand-oecd.cfm, accessed March 7, 2014. 9 See Cashin and McDermott (2001). 10 2.1 Seasonal Adjustment I removed seasonal variation in world production, non-OPEC production, and price in the following manner: I regressed each series and its first differences on seasonal dummy variables; I added the residuals from each regression to the mean of the dependent variable to get preliminary seasonally adjusted levels and first differences, respectively; I added the cumulative sum of the preliminarily seasonally adjusted first differences to the initial value of the preliminarily seasonally adjusted levels to get secondarily seasonally adjusted levels; I added a constant to the secondarily seasonally adjusted levels so that the mean of the resulting series was the same as that of the original series; this gave me a final seasonally adjusted series. I found that this extensive process was necessary to purge both the series and their first differences of regular seasonal variation, as measured by the (in)significance of the coefficients when the final series and their first differences were regressed on seasonal dummy variables. 3 World Demand for and Non-OPEC Supply of Crude Oil I assume that the quantity of crude oil demanded equals the quantity supplied, as measured by world production of crude oil. The quantity demanded, then, includes that amount added to inventory. 3.1 Specification of World Demand I estimate quarterly world demand for crude oil as a linear function of a constant, price, world GDP, a linear time trend, one lag of quarterly quantity demanded, and annual quantity demanded lagged one quarter: Dt = δ 0 + δ P Pt + δ G Gt + δ t + δ D Dt −1 + δ Dave Davet −1 + ε tD (2) Dt is quarterly quantity of crude oil demanded worldwide in billion barrels per year, Pt is the real price of crude oil in 2005$/bl, Gt is world gross domestic product (purchasing power parity; normalized to average 1 in 2005), t is time in quarters (t = 0 in 2011:II), Davet −1 is annual demand 11 for crude oil for the four quarters ending with quarter t − 1 , and ε tD is potentially heteroskedastic, autocorrelated, and correlated with ε tS from (3), below. The price term δ P captures the effect of contemporaneous price on the quantity of crude oil demanded worldwide. I treated quantity demanded as a linear function of price so that the price elasticity of demand would increase with price. Numerous substitutes for crude oil, including potential conservation measures, exist, but they have only become competitive as oil prices have reached new highs. These may include ethanol from cellulose as well as corn, biodiesel, coal to liquids, natural gas to liquids, potentially huge reserves of natural gas hydrates below the ocean floor, nuclear power, wind power, photovoltaic solar power, electric cars, and denser urban design. The choice of price rather than log price as a regressor, then, is motivated by the assumption that greater competitiveness of alternatives to crude oil increases the price elasticity of demand for crude oil as the price reaches new highs. If price trends up faster than demand, the associated percentage drop in quantity demanded increases as the price increases. Adelman (1990; p. 11) wrote “…the higher the price, the greater the incentive to consumers to substitute other comparable goods; and to producers, to substitute labor and capital or other inputs. In addition, for both consumers and firms, there is an income effect pushing the same way: the greater the importance of the product in the total budget, the more important the impact of a further increase. Hence the higher the price, the greater the response to a given price change.” Statistically, price performed better than log price in specification tests. I include a time trend to account for increasing efficiency in the use of crude oil. From Table 1, the ratio of world crude oil consumption to GDP in 2009 was a fifth of what it was in 1974. While some of this resulted from substitution of other fuels for oil, most did not. In 2010, worldwide consumption of primary energy of all kinds per unit of GDP was a fourth of what it was in 1980.10 According to Atkins and Jazayeri, inclusion of the time trend obviates the need to model the direct effects of price on demand as asymmetric, since the two are observationally equivalent. According to Griffin and Schulman, the trend is superior. According to Wing, the deterministic trend is important, at least in the U.S. I model persistence in demand for crude oil using both quarterly and annual quantity demanded of crude oil lagged one quarter; Dt −1 and Davet −1 . Rather than using a series of quarterly lags, I use an 10 Source: U.S. Energy Information Administration; http://www.eia.gov/cfapps/ipdbproject/IEDIndex3.cfm?tid=44&pid=44&aid=2, accessed July 15, 2013. 12 annual average to emphasize long term persistence.11 Short term persistence results from rigidity in planning for the use of oil-specific capital and durables, such as travel planning. Long term persistence results from the sunk costs of large amounts of substantially oil-specific physical and human capital and durables, beginning at the refinery and continuing downstream to such things as gasoline-powered vehicles and auto-oriented urban design. The sunk costs make maintenance of existing capital cheaper than replacement with new capital better optimized to a current price of crude oil. Both “short term” and “long term” persistence are short run phenomena, in that there are no sunk costs of oil-specific capital and durables, and, therefore, no persistence, in the long run. Although its coefficient is only statistically significant at the “85% level” in the regression below, including Davet −1 improves the model’s performance in specification tests. 3.2 Specification of Non-OPEC Supply I modeled non-OPEC supply as a linear function of a constant, a dummy variable for the quarters surrounding the Gulf War, log price, cumulative non-OPEC supply lagged one quarter, a one quarter lag in interest on three-month U.S. treasury bills, a one quarter lag in the trade-weighted exchange value of the U.S. dollar, a linear time trend, and quarterly non-OPEC production lagged one quarter. St = η0 + ηGW M t90 III 92 IV + η p pt + ηCS Ct −1 + ηi it −1 + η X X t −1 + η t + η S St −1 + ε tS (3) where St is quarterly non-OPEC quantity supplied, M t90 III 92 IV is a dummy variable equaling 1 from 1990:III through 1992:IV, pt ≡ ln Pt is log price, Ct −1 is cumulative non-OPEC production lagged one quarter, it −1 is interest on three-month U.S. treasury bills lagged one quarter, X t −1 is the trade-weighted exchange value of the U.S. dollar lagged one quarter, t is a linear trend measured in quarters and equaling zero in 2011:II, and ε tS is potentially heteroskedastic, autocorrelated, and correlated with ε tD from (2). Iraqi and Kuwaiti production fell during and after the Gulf War, the International Energy Agency tapped strategic stocks, expectations changed significantly and frequently, and short term volatility in price increased at the time of and following Iraq’s August 1990 invasion of Kuwait. Allowing a 11 This also keeps down the number of regressors and, therefore, the variances of the estimates. 13 temporary shift in intercept during this time improved the results of tests of specification and stationarity. I use log price so that the price elasticity of non-OPEC supply decreases in quantity supplied. While many sources of crude oil may be available, there is substantial variability in the cost of extracting and finding them. The decreasing short run elasticity reflects the effect of increasing costs of extraction as existing sources are used more intensively. Decreasing long run elasticity and lagged cumulative supply reflect increasingly costly sources being exploited. The U.S. interest and exchange rates reflect the importance of the U.S. dollar to commodity markets in general and that for crude oil in particular. Both organized exchanges and contracts for crude oil typically quote prices in dollars. The dollar was the world’s “petro-currency” throughout the sample period and remains so today. The nominal rate of interest on dollar-denominated securities essentially measures the degree of inflationary expectation for the U.S. economy: The “real interest rate” component measures the extent to which the Federal Reserve tightens credit to prevent inflation, and the remaining component represents the extent to which it accommodates inflation. A change in the exchange value of the dollar can affect the incentive to produce under non-indexed lease agreements and forward contracts, but it will only shift demand for crude oil between the U.S. and other countries, without having much affect on world demand, since the U.S. economy is about as oil-intensive as the world economy as a whole. Empirically, if I include it −1 and X t −1 in the demand equation, both when retaining them in and when removing them from the supply equation, they are not at all statistically significant in the demand equation. The time trend captures the effect of advancing technology of exploration, development, and extraction. Lagged quantity supplied reflects persistence in supply resulting from the presence of physical and human capital specific to exploration and extraction of crude oil. I tried including an annual moving average term lagged one quarter, Savet −1 , like Davet −1 in (2), in (3). While including Davet −1 in (2) improved the results of specification tests, excluding Savet −1 improved the results of specification tests. The re-optimization of investment in exploration and extraction in response to changes in the price of crude oil is less complicated than that of the capital, durables, and other factors used in combination with crude oil, so long term persistence does not warrant emphasis in a separate variable in the case of non-OPEC supply, as it does in demand. 14 3.3 Identification and Estimation of World Demand and Non-OPEC Supply Identification of world demand and non-OPEC supply is the most challenging aspect of making these estimates. I treat the price term in (2) as an endogenous regressor because innovations in demand can cause contemporaneous changes in price. I treat log price in (3) as endogenous because it depends on shifts in supply. I also treat GDP, in (2), as endogenous because, for example, a war could affect both demand for crude oil and GDP. Price and log price are codetermined, so there are, effectively, two endogenous regressors in the system. Finding excluded instruments that are both strong and exogenous to a global market with macroeconomic externalities is difficult to do perfectly. I use one- and two-year (i.e. four and eight quarter) lags of disruptions in world supplies of crude oil resulting from war or civil conflict, Qt − 4 and Qt −8 , and one- and two-year lags of world GDP, Gt − 4 and Gt −8 . All of the excluded instruments are lagged and, therefore, not causally impacted by innovations in demand for or supply of crude oil. Hamilton (2003) uses disruptions to crude oil supplies to extract variation in price that was exogenous to U.S. GDP. The variable Qt is the percent drop in world supply of crude oil during Quarter t associated with the episodes listed in the description above of Table 1, all of which occurred in OPEC countries. It is an extension of Hamilton’s instrument. Qt represents changes in OPEC production associated with conflict. Lagged GDP is a strong instrument for current GDP, and it is unlikely that the four- and eight-quarter lags would correlate with current demand without also correlating with lagged demand, which is well controlled for in (2). More generally, because current and lagged market quantities, prices, incomes, and technological change are controlled for as well, the excluded instruments correlate with these and, therefore, need not correlate with the current error term. In a recursive dynamic process such as that followed by demand in (2), lagged values of the regressors affect the dependent variable through its lags in a way that declines geometrically with the length of the lag. When lagged GDP affects lagged price, it will also, in turn, affect current demand through lagged demand. GDP can be modeled as having its own recursive process, but its dynamics are modeled in a manner in the recursive process describing demand, as well. Finally, inasmuch as, beyond these effects, lagged GDP still affects the error term, it will likely have a greater effect on lagged errors than current ones, 15 and these effects will also be captured in the dynamic process followed by demand. The current error term, ε tD , may be unaffected if it is not correlated with lagged errors; Figure 4, below, shows no 95% statistically significant autocorrelation in the residuals from estimation of (2), and the p-value in a Portmanteau’s Q test of the null hypothesis that the residuals associated with (2) are white noise is 0.9985. In sum, Equation (2) substantially provides ways for Gt − 4 and Gt −8 to affect Dt through ε tD−8...1 , Gt −7...0 , Pt −8...0 , Dt −8...1 , and Davet −8...1 , rather than through ε tD , and there is no evidence of correlation between ε tD−8...1 and ε tD . Furthermore, the p-value in a Sargan test of the joint null hypothesis that Gt − 4 , Gt −8 , Qt − 4 , and Qt −8 are uncorrelated with ε tD and ε tS and correctly excluded from (2) and (3) is 0.69. Perhaps most tellingly, in a first stage regression of Dt on all exogenous variables, neither Gt − 4 nor Gt −8 is statistically significant ( t = 0.04 and − 0.77 , respectively), while Dt −1 is highly significant ( t = 6.92 ). 3.4 Results and Testing of Estimated World Demand and Non-OPEC Supply Three stage least squares accounts for correlations in errors across equations. I used ®Stata’s reg3 command to estimate Equations (2) and (3) simultaneously using iterated generalized least squares. 12 The estimates, shown in Equations (4) and (5), converged in five iterations. Dt = -1.7283 − 0.0125 Pt + 0.0441Gt − 0.0715 t + 0.7655 Dt −1 + 0.1526 Davet −1 + etD 1.6155 0.0044 0.0184 0.0362 0.0956 0.1010 St = 8.1301− 0.0750 M t90 III 92 IV + 0.2376 pt − 0.0227 it −1 3.2638 0.0385 0.0764 0.0086 + 0.3520 X t −1 − 0.0416 Ct −1 + 0.2211t + 0.9416 St −1 + etS 0.1479 0.0162 0.0892 (4) (5) 0.0192 The R 2 = 0.99 for demand and 0.98 for non-OPEC supply. All of the coefficients are significantly different from zero at the 99% level except: the Gulf War dummy variable, which is significant at the 90% level; the demand constant, which is not significant; the deterministic trend in demand, which is significant at the 95% level; and Davet −1 , which is “significant at the 85% level”. Portmanteau’s Q 12 I performed my calculations using Stata/SE Version 10.1 for Windows. 16 tests of the null hypothesis that etD and etS are statistical “white noise” give p-values of 0.9985 and 0.5243, respectively. Figure 4 shows the autocorrelations among the residuals from (4) and (5). Figure 4: Autocorrelations _____________________________________________________________________________ Supply Residuals -0.20 -0.20 -0.10 -0.10 0.00 0.00 0.10 0.10 0.20 0.20 Demand Residuals 0 10 20 Lag 30 40 0 Bartlett's formula for MA(q) 95% confidence bands 10 20 Lag 30 40 Bartlett's formula for MA(q) 95% confidence bands ______________________________________________________________________________ Robinson tests estimate the degree of fractional differencing, d, needed to render etD and etS level ( ) stationary I ( 0 ) to be d = −0.0410 and d = 0.0918 , respectively. The residuals are sufficiently 0.0572 0.0872 stationary to dismiss the possibility that the estimates in (4) are spurious because of any nonstationarity in the regressors. Whatever non-stationarity exists in the regressors is cointegrated. 4 The Price of Crude Oil and World GDP The price of crude oil affects world GDP which, in turn, affects demand for crude oil. OPEC must account for the effects of its pricing and production on the world economy when deciding what will be most profitable for the cartel. Figure 5 shows the real (2005$) price of crude oil from 1973:IV to 2011:II. Worldwide recessions in my constructed data are shown in red. Three of the five recessions were preceded by oil price shocks, and none of the oil price shocks failed to precede a recession. Far and away the two largest quarter-to-quarter price increases in the OPEC era were $21.41 between 1973:IV and 1974:I and 17 $22.04 between 2008:I and 2008:II. There was a slowdown in the world economy in 1974:II, a recession beginning in 1974:III, and a recession beginning in 2008:III. The third largest increase in price during the OPEC era of $12.63 occurred between 2007:III and 2007:IV. The large shock in 1973 preceded a long term slowdown in world economic growth, and the 2008 recession has been termed “Great”. Over two quarters, from 1978:III to 1980:I, price increased $39.51. GDP declined at an annual rate of more than 8% from 1980:I to 1980:II and 0.7% the following quarter. 0 20 40 60 80 100 Figure 5: Real (2005$) Price of Crude Oil from 1973:IV to 2011:II 1972 1977 1982 1987 1992 1997 2002 2007 2012 4.1 Specification of World GDP A good deal has been written about asymmetry in the response of the macroeconomy to changes in the price of crude oil.13 Increases in price damage the economy more than decreases help. To illustrate one reason why, suppose wages are sticky in the downward direction. Figure 6 describes the market for labor when the price of crude oil fluctuates. 18 Figure 6: Labor Market with Downwardly Sticky Wages and Fluctuating Price of Crude Oil W S LS W1 W0 S L1 DL S L0 L0 L1 L The price of crude oil changes from P 0 to P1 (not shown), raising the cost of fuel used to travel to work and back, shifting the competitive supply of labor from S L0 to S L1 . The equilibrium wage increases from W 0 to W 1 , and the quantity of labor demanded decreases from L0 to L1 . Next, the price of crude oil returns to P 0 , but workers continue to require wage W 1 , so the supply curve changes to S LS , the quantity of labor supplied remains at L1 , and less output is produced than before the price of crude oil increased and subsequently decreased to its original level. To capture the asymmetric effects econometrically, I specify increases and decreases in the price of crude oil separately, using first differences in log price to explain first differences in log GDP. With this functional form, regression coefficients represent the, constant, elasticities of GDP with respect to the price and lagged prices of crude oil. I allow for response in the first difference in the log of GDP to zero through five quarter lags in increases and decreases in the log price of crude oil. I also include five lagged dependent variables, first differences in log GDP, in the regression. gt − gt −1 = γ 0 + 13 t ∑γ s =t −5 + s max ( ps − ps −1 , 0 ) + t −1 ∑γ s =t −5 − s min ( ps − ps −1 , 0 ) + t −1 ∑ γ (g s =t − 5 g s s − g s −1 ) + ε tg (6) See, for example, Hamilton (2003), Greene and Ahmad (2005), or Gately and Huntington (2002). 19 where gt is the natural logarithm of world GDP in quarter t, ( Gt , (normalized to 100 in 2005), gt ≡ ln Gt ), and ε tg is a potentially heteroskedastic and autocorrelated error. With the lagged dependent variables, changes in price beyond the fifth lag may impact current GDP. The lag structure is just long enough to include the lag with the largest impact. Impacts through the lagged dependent variables decline with their distance in time from t . 4.2 Identification of World GDP I treat contemporaneous changes in log price, max ( pt − pt −1 , 0 ) and min ( pt − pt −1 , 0 ) as endogenous because GDP can affect current price. The excluded instruments include lags of positive and negative differences between the residuals from the demand and non-OPEC supply ( ) equations (unexplained changes in OPEC production), specified separately, max etD−1 − etS−1 , 0 and min ( esD=t −1...3 − esS=t −1...3 , 0 ) . In first-stage regressions, only the first lag of the positive changes in unexplained OPEC production was significant, while the first three lags of negative changes were, so I did not use the longer lags of the positive changes. One interpretation is that the cooperative game among OPEC members in which prices are raised through cuts in production must be repeated to succeed; cooperation is easier to establish in repeated than in one-shot cooperative games. On the other hand, when OPEC wants prices to fall, it may simply allow cooperation to break down, and this can be done quickly and easily. Estimates of coefficients are consistent in OLS regressions with “generated regressors”. In 2SLS, the transformations of etD− s − etS− s are “generated instruments”. According to Wooldridge (2002; p. 117), “…there are practical reasons for using 2SLS with generated instruments rather than OLS with generated regressors.” Whether the generated instruments are included or excluded, inference using heteroskedasticity- and autocorrelation-robust standard errors in a GMM context, a generalization of both OLS and 2SLS done here, is consistent and asymptotically efficient. I assume that OPEC producers are the only strategic decision-makers impacting the price of crude oil, and the price of crude oil impacts world GDP. Thus, OPEC production is a driver of both the price of crude oil and world GDP, through price. This does not mean that OPEC producers do not also 20 respond to GDP, but lagged OPEC production would only depend on current innovations in GDP through accurate expectations of those innovations. The innovations in (6) are independent of lagged changes in price and GDP. I assume that OPEC producers are not able to anticipate these innovations. The excluded instruments for contemporaneous changes in log price also include a dummy variable equaling 1 from the beginning of the dataset through 1985:I, M t1985I . The assumed shift in volatility of price in 1985:II coincides with the maturation of crude oil as an actively traded commodity. The coefficient of determination (standard deviation over mean) of price equals 0.30 through 1985:I and 0.57 thereafter. Verleger (2005; p. 5) explains that as prices dropped in the early 1980’s, the world crude oil market was in transition from a vertically integrated one to one characterized by active commodity markets. In particular, “...the development of North Sea production introduced classic commodity market institutions into the global oil market. A true spot market was created.” The development of the commodity market preceded, and facilitated, the “price collapse of 1986”, which likely resulted from OPEC’s increasing excess capacity associated with falling short term demand. Examples of dummy variables as excluded instruments can be found in Evans and Schwab (1995) and Djankov and Reynal-Querol (2007). 21 4.3 Results and Testing of Estimated World GDP Continuously updated GMM estimates of (6), shown in Table 2, converged after 153 iterations. Table 2: Estimated Effects of Changes in Crude Oil Prices on the First Difference in World GDP, ∆gt ≡ gt − gt −1 __________________________________________________________________________________________ Coefficient Standard. Error Coefficient Standard. Error ∆ + pt -0.086 0.020 ∆gt −1 0.181 0.116 − ∆ pt -0.039 0.016 ∆g t − 2 -0.053 0.042 + 0.012 0.006 ∆gt −3 -0.293 0.063 − 0.016 0.008 ∆gt − 4 0.463 0.121 + -0.009 0.008 ∆g t −5 0.001 0.077 − -0.011 0.007 Constant 0.009 0.002 + -0.006 0.005 − ∆ pt −3 -0.003 0.005 ∆ + pt − 4 ∆ pt −1 ∆ pt −1 ∆ pt − 2 ∆ pt − 2 ∆ pt −3 0.005 0.004 − -0.015 0.010 + -0.021 0.005 − -0.014 0.004 ∆ pt − 4 ∆ pt −5 ∆ pt −5 __________________________________________________________________________________________ ∆g s = g s − g s −1 is the quarterly first difference in the natural logarithm of world GDP. ∆ + ps ≡ max ( ps − ps −1 , 0 ) is the positive change in the natural log price of crude oil between Quarter s − 1 and Quarter s , and ∆ − ps ≡ min ( ps − ps −1 , 0 ) the negative change. Of the 18 regressors, including the constant, six have coefficients that are significantly different from zero at the 99% level, two others at the 95% level, and one more at the 90% level. I retained the remaining nine so as to maintain the contiguity and consistency of the lag structure. The test statistics are robust to heteroskedasticity and autocorrelation of any form. However, A Pagan-Hall test using 22 the predicted values, gˆ t , associated with Table 2 as an indicator variable fails to reject homoskedasticity in the errors ( p = 0.970 ) . There is no statistically significant autocorrelation in the residuals at any specific lag, as shown in Figure 7, and a Cumby-Huizinga test fail to reject a null hypothesis that there is no autocorrelation at any lag up to 25 ( p = 0.985 ) . -0.20 -0.10 0.00 0.10 0.20 Figure 7: Autocorrelations Among GDP Residuals 0 10 20 Lag 30 40 Bartlett's formula for MA(q) 95% confidence bands A Bartlett cumulative periodogram test puts a p-value of 0.857 on a null hypothesis that the residuals were generated by a white noise process. They are stationary; Estimated d = -0.009 in a Robinson 0.075 test. There are five excluded instruments and two endogenous regressors in a dataset of 158 observations, used to perform a regression with five lags in the regressors. The p-value in a Hansen test of a null hypothesis that the excluded instruments are independent of the error term is 0.882. When I omit max ( etD−1 − etS−1 , 0 ) and min ( etD−1 − etS−1 , 0 ) from the set of instruments, the p-value is 0.515, and ( ) ( ) when I test a null that max etD−1 − etS−1 , 0 and min etD−1 − etS−1 , 0 are independent of the error term, assuming the other excluded instruments are also, the p-value is 0.888. The value of a 23 Kleibergen-Paap Wald rk F-statistic is 23.80, which exceeds the highest, 90%, critical value reported by Stata of 4.32, leading to a rejection of a null hypothesis that the excluded instruments are weak. 4.4 The Effect of Changes in Crude Oil Prices on World GDP Oil prices are important to the macroeconomy, but are not all important. A clear example is shown in Figure 8, which plots the residuals associated with the estimates of (6) appearing in Table 2. The deepening of the “Great Recession” in the fall of 2008 stands out, a result of a collapse in private lending, but the oil shock of the previous summer was a contributing factor. -.06 -.04 -.02 0 .02 .04 Figure 8: GDP Residuals 1972 1977 1982 1987 1992 1997 2002 2007 2012 Table 2 indicates that decreases in the price of crude oil raise GDP less than increases lower it. There is some oscillation in the effects of either an increase or a decrease in price, but at no lag is the cumulative effect of an increase (decrease) in price on GDP non-negative (non-positive). t-tests reject negativity of the sum of the coefficients on increases and of those on decreases with 99% confidence. As to asymmetry, a null hypothesis that the effects of decreases are greater than those of increases is rejected with 95% confidence: The p-value associated with a null hypothesis that the sum of the absolute effects of increases in price is greater is 0.963; that on the alternative is 0.037. 24 The asymmetry in Table 2 echoes that in a number of estimates. Hamilton (2003) explains the asymmetry in the relationship as the result of allocative disturbances and uncertainty. An unexpected change in oil prices in either direction changes the optimal mix of industrial equipment and consumer durables that firms and consumers, respectively, desire. If the change makes them uncertain about the future direction of prices, then they are likely to postpone major purchases until the uncertainty is resolved. This would also apply to governments, in particular with regard to transportation infrastructure and urban planning. (Hamilton (2009; pp. 39-40) notes “…house prices in 2007 were likely to rise slightly in the zip codes closest to the central urban areas but fall significantly in zip codes with longer average commuting distances.”) Thus, there is a contractionary element in the effects of either increases or decreases in the price of crude oil, but no corresponding expansionary element in the effects of increases. Another reason for the strong asymmetry is downward stickiness of wages in the short run, as shown in Figure 6, which illustrates effects through the supply of labor. Demand for labor may also shift. When the price of oil rises, to the extent that labor and oil are complements, demand for labor falls, causing only unemployment. To the extent that labor and oil are substitutes, demand for labor rises, raising wages and employment. When the price of oil falls, to the extent that labor and oil are complements, demand for labor rises, raising wages and employment. To the extent that labor and oil are substitutes, demand for labor falls, causing only unemployment. There is no reason before the fact to suppose that the sum of these effects is zero. Most workers in countries that are not poor use oil products or close substitutes, such as natural gas, to travel to and from work, so oil is related to labor in the production of a large majority of goods. Other inputs related to oil in production of goods may also have downwardly sticky prices, and this may contribute to the asymmetry. A third reason for the asymmetry is income and liquidity effects. Since oil products and their substitutes take a large share of many budgets, increases from a given price level reduce spending more than decreases from that level increase spending. These effects can be especially strong in poor countries. Teitenberg (2007; p. 202) writes “The lack of foreign exchange has been exacerbated during periods of high oil prices. Many developing nations must spend large portions of export earnings merely to import energy. Fourth, investors whose wealth and income are sensitive to the price of oil will hold increases therein in liquid form for a time before committing to a less liquid investment. A large change in price in 25 either direction will redistribute wealth and income and, as a result, tend to slow down real investment. Consider the response in GDP in Quarter t to a change in price s quarters earlier. From (6), the short run elasticity of world GDP s quarters hence with respect to a change in price lasting one quarter is t t ∆g t = ∑ γ i+ / − Π γ gj ∆pt − s i =t − s j = 2 t − s −i where γ tg ≡ 1 and γ i+ / − is the coefficient on an increase/decrease in log price in Quarter i and γ gj the coefficient on the change in log GDP in Quarter j. Evaluating this at s = 5 in 2009:III gives an estimate of the impact of the large increase going into summer 2008. Price (2005$) in 2008:II was Pt −5 = $106.31 , and in 2008:I was Pt −5−1 = $84.05 . The estimated elasticity is -0.0203, and the price change was 23.5%, so the estimated effect is that world GDP was 0.48% lower in 2009:III due to that largest ever quarter-to-quarter increase in the real price of crude oil. In my constructed data, world GDP was 2.14% lower in 2009:III than in 2008:II, so Table 2 implies that a fifth of the decrease was caused by the upward shock in 2008:II. The oil shock contributed significantly to the Great Recession, but was not its primary cause. These negative short run effects reduce GDP in the long run because lower current GDP means lower current investment, which is the only way to provide now for future GDP. Short run dips in GDP lower future GDP because investment varies directly and elastically with GDP. Hysteresis also sets in among unemployed workers, lowering their future productive capability; a recession in employment implies less current investment in human capital through accumulation of work experience. The long run elasticity of world GDP with respect to a temporary, one quarter, change in the price of crude oil is ∑ γ + − ∑ i =t −5 γ i− i =t − 5 i t t 1 − ∑ i =t −5 γ ig t −1 ≈ −0.056 using Table 2. 26 A 1% increase in the price of crude oil lasting exactly one quarter lowers world GDP in the long run by 0.056% . Thus, if it had been exactly reversed in 2008:III, the largest-ever, $22.27/bl in 2005$, increase the previous quarter would have caused world GDP to be 1.32% lower in each quarter over the long run. In my constructed data, world GDP grew about 1.96% quarterly. The shock of 2008, then, set the world economy back about two months. 5 OPEC The supply curve of a monopolist does not exist. OPEC is not a monopolist in the literal sense, but its market power in the world market for crude oil is unrivaled, so, as a whole, its profit-maximization problem is like that of a monopolist. OPEC does not interact strategically with non-OPEC suppliers, with the possible exceptions, ignored here for simplicity, of the governments of Mexico, Norway, and Russia; OPEC takes account of its influence on non-OPEC production when deciding its own production, but non-OPEC producers do not take account of their influence on OPEC in deciding their production14. Since OPEC’s market power is substantial and unrivaled, it decides the world price of crude oil as it decides its own production. Celta and Dahl (2000) estimated OPEC’s short run marginal costs in 1995 dollars as ln ( MC ) = 46.3263 + 0.3026 ln ( QOPEC ) − 2.3356 ln ( ROPEC ) (7) where QOPEC is OPEC production in thousand b/d and ROPEC is proven OPEC reserves in thousands of barrels. In 2012, OPEC production was 36.599 million b/d, and proved reserves were 1.113 trillion barrels. (7) implies short run marginal costs in 2012 dollars of $2.76/bl. In that year, the price of Brent crude oil averaged $111.63/bl, and that of WTI averaged $94.05/bl.15 A graphic on page 33 of Van Vactor (2010) puts long run marginal cost in OPEC countries, including Venezuela, between $15/bl and $30/bl in 2009. Though competitive producers whose future marginal costs increase in current production will produce where current marginal cost is less than price, such a large difference between price and marginal cost suggests something other than price-taking behavior. 14 Other overlooked exceptions include non-OPEC governments who subsidize production of crude oil to reduce “dependence on foreign oil”, who can be said to interact strategically with OPEC. 27 5.1 Net Demand to OPEC Since OPEC output is identically the difference between world quantity demanded and non-OPEC quantity supplied, Ot = Dt − St , I calculate net demand to OPEC by subtracting (3) from (2). Ot = δ 0 − η0 + δ P Pt − η p pt + δ G Gt + (δ − η ) t + δ D Dt −1 + δ Dave Davet −1 −ηGW M t90 III 92 IV − ηC Ct −1 − ηi it −1 − η X X t −1 − η S St −1 + ε tD − ε tS Using the estimates in (4) and (5) gives the values in Table 3. Table 3: Net Demand to OPEC, Ot ≡ Dt − St _________________________ Variable Coefficient Pt -0.013 pt 0.238 Gt 0.044 t -0.293 Dt −1 0.766 Davet −1 0.153 M t90 III 92 IV -0.075 Ct −1 -0.042 it −1 -0.023 X t −1 0.352 St −1 0.942 -9.858 Constant 15 Reserve, production, and price data come from EIA. 28 (8) The asymmetric effects of the price of crude oil on world GDP imply asymmetric effects of price on world demand and net demand to OPEC. Increases in price lower quantity demanded more than decreases in price raise quantity demanded. OPEC’s demand curve is concave to the origin. Table 4 shows estimated elasticities of world demand, world GDP, non-OPEC supply, and net demand to OPEC. I assume in my calculations of elasticities that it is 2014:II, price is $100/bl, annual quantity demanded is 27.80 billion barrels per year, GDP flows at an annual rate of 123.1, where GDP in 2005 ≡ 100 , and non-OPEC quantity supplied is 15.79 billion barrels per year, implying OPEC production of 12.00 billion barrels per year. Long run demand is elastic at the price of $100/bl, and net demand to OPEC at time-horizons of twelve months or less is inelastic. The next to last column of Table 4 shows prices at which demand over various time-horizons is unit-elastic for upward changes in price. Net demand to OPEC over a twelve month time-horizon becomes unit elastic for increases in price at $142.41/bl. This suggests that an oil price shock lasting up to a year in which prices would fluctuate around $150/bl is a real possibility. Had private borrowing not collapsed at about the same time, it might well have been profitable for OPEC to sustain the shock of 2008, which went about to this level, longer than it did. At shorter time-horizons, net demand to OPEC only becomes elastic at prices significantly higher than any in the data from which these estimates were derived, so these numbers should be viewed skeptically. I include them for the sake of completeness. The discontinuity gap in marginal revenue is shown in the last column of Table 4. I calculate the gap ( ) as P ∗ 1 ε + − 1 ε − , where ε + is the elasticity with respect to an increase in price, and ε − is the elasticity with respect to a decrease in price. OPEC’s long run marginal cost curve may pass through the discontinuity gap in marginal revenue over a wide range of prices and quantities. Since 1973, historic changes in the world economy, as with the deep recession in the early 1980’s and rapid growth beginning in the late 1990’s, have preceded large lasting changes in the price of crude oil. (See Figure 1.) 29 Table 4: Elasticities and MR Gaps at $100/bl and Upwardly Unit-Elastic Prices in 2014:II (2013$) Price+ Price- Income -0.0557 -0.0863 0.0150 -0.1488 -0.0464 -0.0389 0.0150 -0.1274 0.1955 -0.1019 -0.0934 0.0292 -0.2744 -0.0824 -0.0318 0.0292 -0.2292 0.3526 Upwardly Unit-Elastic Price MR Gap $554.36 -$571.96 -$685.18 = $113.22 $377.36 -$264.49 -$336.24 = $71.75 $240.25 -$115.83 -$158.30 = $42.46 $142.41 -$32.62 -$58.67 = $26.05 $67.95 $36.71 $26.74 = $9.97 3 Month Demand GDP NO Supply OPEC Demand 0.4527 6 Month Demand GDP NO Supply OPEC Demand 0.8166 9 Month Demand GDP NO Supply OPEC Demand -0.1759 -0.0935 0.0425 -0.4633 -0.1430 -0.0368 0.0425 -0.3872 0.4820 1.1164 12 Month Demand GDP NO Supply OPEC Demand -0.2943 -0.0780 0.0551 -0.7541 -0.2408 -0.0325 0.0551 -0.6302 0.5921 Demand -0.7896 GDP -0.1518 NO Supply 0.3347 OPEC Demand -1.5799 Quarter = Price = World Quantity Demanded = World GDP = Non-OPEC Quantity Supplied = -0.6634 -0.0958 0.3347 -1.3649 2014 100.00 27.80 123.10 15.79 2.3875 1.3714 Long Run 30 5.5296 II in 2013$ in bbl/yr ; 2005=100 in bbl/yr In a purely static exercise, one would expect OPEC to produce and price where long run marginal cost, between $15/bl and $30/bl in Van Vactor, fell in the marginal revenue gap for the appropriate time-horizon. However, two (upward) adjustments to these figures are appropriate. First, the numbers in Van Vactor applied in 2009, and one would expect costs to be higher in 2014. Second, they do not consider the effect of present production on future costs, which increases the opportunity cost of present production. OPEC’s marginal costs, inclusive of “marginal user cost”, may be considerably higher than those indicated in Van Vactor. Though interest rates as of early 2014 are low, marginal user cost is increasing in the discount rate, and OPEC is generally thought to have a high discount rate.16 Table 5 shows upper and lower bounds of ranges of prices (UB − LB ) within which OPEC cannot increase profits by change price and production in the long run assuming the long run marginal costs, inclusive of marginal user cost, LRMC , indicated in the column headings. Table 5: Profit-Maximizing Prices at Various LRMC’s in 2014:II (2013$) LRMC: UB: LB: Size of Range: 20.00 95.14 86.57 8.57 30.00 102.30 94.77 7.54 35.00 105.78 98.69 7.09 40.00 109.19 102.51 6.68 50.00 115.85 109.90 5.95 The price of WTI averaged $97.98/bl in 2013, and that of Brent averaged $108.56/bl, while the standard deviation in the price of WTI in 2013 was $5.46/bl, and that of Brent was $4.64/bl, according to EIA. At least recently, the multiplicity of equilibria implied by OPEC’s discontinuous long run marginal revenue curve, associated with the asymmetric effects of oil prices on world GDP, is a reasonable explanation for much of the variation in those prices, and large price shocks can also be explained by applying the model to shorter time-horizons. 16 See, for example, Adelman (1990; pp. 11-13). 31 6 Conclusion Instability in the price of crude oil does not imply that OPEC is unable to use its market power effectively. The asymmetric effects of changes in the price of crude oil on the macroeconomy imply that world demand and demand to OPEC net of non-OPEC production are kinked, that there is a vertical discontinuity in OPEC’s marginal revenue curve. Therefore, there are multiple combinations of price and OPEC production at which an increase in price (decrease in production) lowers revenue more than cost, and at which a decrease in price (increase in production) raises revenue less than cost. The asymmetry results in multiplicity of equilibria. In 2014, using long run demand, the range of equilibrium prices appears to be about $7/bl wide, and an increase in price lasting one year to levels above $142/bl would also appear to be profitable. In the short run, demand to OPEC is quite inelastic, and the contemporaneous effects of changes in price on GDP are negative and statistically significant. From Table 2, a 1% increase (decrease) in the price of crude oil causes a 0.086% decrease (0.039% increase) in world GDP in the same quarter. Assuming non-decreasing marginal costs, OPEC can collect a countercyclic stream of profits by promulgating instability in the price of crude oil. The stream can be used to smooth out undesirable fluctuations in consumption associated with the changes in GDP, so OPEC can sell instruments in financial markets that command a risk premium. Thus, within some range, OPEC has incentive to promulgate unstable prices for crude oil. Because increases in the price of crude oil damage the macroeconomy more than decreases improve it, the instability in price that OPEC has incentive to promulgate damages the macroeconomy. The discontinuity in marginal revenue implies that vertical shifts in demand, associated with changes in world GDP, cause larger changes in the price of crude oil. These, in turn, affect GDP. 32 Table1: Basic Data A Obs 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 Year 1972 1972 1972 1972 1973 1973 1973 1973 1974 1974 1974 1974 1975 1975 1975 1975 1976 1976 1976 1976 1977 1977 1977 1977 1978 1978 1978 1978 1979 1979 1979 1979 1980 1980 1980 1980 1981 1981 Qrtr 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 B C D E Price Implicit Price Deflator NonOPEC Output World Output Annual World GDP (PPP) US$/bl 2005:I=1 bbl/yr bbl/yr bUS$ 1.93 6.51 12.94 12.66 11.99 13.64 14.39 13.29 13.44 15.88 14.70 13.16 12.86 14.64 15.74 14.27 13.88 15.69 15.85 14.21 13.83 15.82 17.28 18.93 23.38 28.04 33.54 33.85 33.81 36.16 40.07 37.49 0.2837 0.2893 0.2949 0.3019 0.3109 0.3202 0.3276 0.3325 0.3386 0.3446 0.3484 0.3521 0.3569 0.3633 0.3694 0.3747 0.3793 0.3876 0.3933 0.4005 0.4072 0.4158 0.4232 0.4336 0.4430 0.4519 0.4614 0.4718 0.4826 0.4959 0.5085 0.5181 9.34 9.58 9.68 9.64 9.59 9.73 9.69 9.76 9.73 9.77 10.11 10.08 10.12 10.27 10.48 10.56 10.77 10.98 11.12 11.29 11.39 11.75 11.81 11.94 11.99 12.16 12.28 12.39 12.42 12.52 12.59 12.51 12.71 12.85 21.33 21.79 22.28 21.19 21.79 22.34 21.38 21.19 20.23 20.26 21.55 20.46 21.26 21.79 22.51 23.67 23.20 23.17 22.95 23.62 22.59 23.28 23.69 24.21 23.74 24.64 24.73 24.68 24.44 23.68 23.18 22.12 23.03 22.64 33 5988 5988 5988 5988 6697 6697 6697 6697 7452 7452 7452 7452 8266 8266 8266 8266 9176 9176 9176 9176 10273 10273 10273 10273 11539 11539 11539 11539 12875 12875 12875 12875 14363 14363 F G H I Quarterly U.S. GDP Nominal Rate on 3Month Treasuries Drop in World Supply TradeWeighted Exchange Value of U.S.$ bUS$ secondary % 2005:I=1 1381 1418 1437 1479 1495 1534 1563 1603 1620 1656 1714 1766 1825 1857 1891 1938 1993 2060 2122 2169 2209 2337 2399 2482 2532 2596 2670 2731 2797 2800 2860 2994 3132 3167 3.44 3.77 4.22 4.86 5.70 6.60 8.32 7.50 7.62 8.15 8.19 7.36 5.75 5.39 6.33 5.63 4.92 5.16 5.15 4.67 4.63 4.84 5.50 6.11 6.39 6.48 7.31 8.57 9.38 9.38 9.67 11.84 13.35 9.62 9.15 13.61 14.39 14.91 0 0 0 0 0 0 0 4.90064 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1.9533 0 0 0 0 0 0 4.54801 0 0 0.30 0.29 0.28 0.29 0.30 0.29 0.30 0.30 0.30 0.30 0.31 0.32 0.32 0.33 0.33 0.33 0.34 0.34 0.34 0.33 0.33 0.33 0.31 0.31 0.32 0.32 0.32 0.33 0.33 0.33 0.33 0.33 0.35 0.37 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 1981 1981 1982 1982 1982 1982 1983 1983 1983 1983 1984 1984 1984 1984 1985 1985 1985 1985 1986 1986 1986 1986 1987 1987 1987 1987 1988 1988 1988 1988 1989 1989 1989 1989 1990 1990 1990 1990 1991 1991 1991 1991 1992 1992 1992 1992 1993 1993 1993 1993 1994 1994 1994 1994 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 35.28 36.91 36.38 32.86 32.48 34.12 31.55 28.30 28.60 30.40 30.24 28.92 28.21 29.56 28.58 27.21 25.92 27.75 20.44 12.59 11.21 14.52 18.25 18.01 18.39 19.00 16.56 15.41 13.66 14.32 18.16 18.70 16.95 19.93 21.10 15.69 22.90 30.59 20.75 17.88 17.95 19.84 17.51 18.39 18.75 19.27 18.70 17.41 14.93 15.11 14.36 15.50 16.03 17.22 0.5273 0.5369 0.5442 0.5508 0.5586 0.5647 0.5693 0.5735 0.5793 0.5836 0.5910 0.5960 0.6008 0.6047 0.6114 0.6148 0.6174 0.6214 0.6246 0.6279 0.6314 0.6357 0.6416 0.6453 0.6503 0.6553 0.6607 0.6669 0.6744 0.6795 0.6872 0.6940 0.6986 0.7032 0.7117 0.7199 0.7266 0.7324 0.7403 0.7455 0.7513 0.7557 0.7595 0.7642 0.7678 0.7721 0.7768 0.7811 0.7847 0.7890 0.7931 0.7969 0.8016 0.8058 12.75 12.68 12.82 13.01 13.11 13.27 13.28 13.39 13.52 13.48 13.74 13.90 13.94 14.01 14.01 14.06 14.15 14.21 14.12 14.09 14.23 14.19 14.16 14.15 14.31 14.26 14.34 14.28 14.17 14.12 14.04 13.94 14.08 14.02 13.96 13.92 13.76 13.85 13.99 13.73 13.70 13.63 13.40 13.25 13.13 13.03 12.93 12.91 12.86 12.96 13.14 13.17 13.23 13.41 21.47 21.33 21.32 20.82 21.10 21.57 20.34 20.86 21.76 21.63 21.86 22.09 21.56 21.46 21.65 21.26 21.16 22.32 22.12 22.55 22.86 22.12 22.01 22.27 23.31 23.05 22.98 23.20 23.45 24.36 23.44 23.74 24.06 24.41 24.64 24.64 23.64 24.06 24.43 23.97 24.21 24.33 24.49 24.18 24.17 24.35 24.71 24.39 24.40 24.46 25.00 25.03 24.92 25.20 34 14363 14363 15387 15387 15387 15387 16464 16464 16464 16464 17883 17883 17883 17883 19111 19111 19111 19111 20235 20235 20235 20235 21599 21599 21599 21599 23357 23357 23357 23357 25135 25135 25135 25135 26812 26812 26812 26812 28119 28119 28119 28119 29324 29324 29324 29324 30640 30640 30640 30640 32380 32380 32380 32380 3261 3284 3274 3331 3367 3408 3480 3584 3692 3796 3913 4015 4087 4148 4237 4302 4395 4453 4516 4555 4620 4669 4736 4821 4901 5023 5091 5208 5300 5413 5527 5628 5712 5763 5891 5975 6030 6023 6055 6144 6218 6279 6381 6492 6587 6698 6748 6830 6904 7033 7136 7270 7352 7477 15.05 11.75 12.81 12.42 9.32 7.91 8.11 8.40 9.14 8.80 9.17 9.80 10.32 8.80 8.18 7.46 7.11 7.17 6.90 6.14 5.52 5.35 5.54 5.66 6.04 5.86 5.72 6.21 7.01 7.73 8.54 8.41 7.84 7.65 7.76 7.75 7.48 6.99 6.02 5.56 5.38 4.54 3.89 3.68 3.08 3.07 2.96 2.97 3.00 3.06 3.24 3.99 4.48 5.28 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4.07159 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.38 0.38 0.40 0.42 0.44 0.45 0.46 0.47 0.49 0.50 0.52 0.53 0.56 0.59 0.62 0.62 0.61 0.59 0.58 0.57 0.56 0.57 0.55 0.55 0.56 0.55 0.54 0.54 0.57 0.56 0.58 0.61 0.62 0.63 0.65 0.67 0.65 0.64 0.65 0.69 0.69 0.68 0.69 0.70 0.69 0.73 0.75 0.75 0.77 0.79 0.83 0.84 0.83 0.83 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 1995 1995 1995 1995 1996 1996 1996 1996 1997 1997 1997 1997 1998 1998 1998 1998 1999 1999 1999 1999 2000 2000 2000 2000 2001 2001 2001 2001 2002 2002 2002 2002 2003 2003 2003 2003 2004 2004 2004 2004 2005 2005 2005 2005 2006 2006 2006 2006 2007 2007 2007 2007 2008 2008 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 18.35 17.96 15.92 17.83 19.74 19.99 20.06 24.08 22.37 17.64 17.10 18.57 14.68 12.07 11.22 11.88 12.22 15.16 19.09 24.08 28.15 26.25 28.46 29.35 25.51 23.59 22.35 17.97 20.54 23.69 25.27 26.50 31.94 25.35 26.71 28.85 32.34 33.55 37.93 40.82 42.36 45.64 56.09 53.09 56.01 63.26 63.25 54.50 54.57 62.14 69.79 83.50 91.08 115.56 0.8104 0.8140 0.8178 0.8220 0.8267 0.8299 0.8325 0.8371 0.8425 0.8445 0.8474 0.8506 0.8520 0.8540 0.8573 0.8599 0.8637 0.8668 0.8700 0.8731 0.8800 0.8845 0.8898 0.8945 0.9005 0.9067 0.9095 0.9123 0.9156 0.9197 0.9236 0.9289 0.9354 0.9382 0.9434 0.9482 0.9564 0.9645 0.9715 0.9787 0.9878 0.9944 1.0046 1.0130 1.0206 1.0295 1.0372 1.0419 1.0538 1.0610 1.0645 1.0696 1.0759 1.0830 13.39 13.38 13.58 13.53 13.67 13.76 13.79 13.93 14.00 14.01 14.03 14.13 14.19 14.15 14.01 14.03 14.12 13.99 14.14 14.27 14.30 14.33 14.48 14.57 14.54 14.42 14.62 14.70 14.76 14.95 14.86 14.90 15.00 14.98 15.16 15.32 15.30 15.41 15.34 15.31 15.26 15.45 15.14 15.16 15.22 15.22 15.27 15.28 15.30 15.29 15.16 15.10 15.08 15.06 25.48 25.66 25.77 25.69 26.14 26.24 26.18 26.44 26.95 27.03 27.06 27.25 28.01 27.90 27.29 27.26 27.82 27.09 27.23 27.14 27.74 28.33 28.69 28.79 28.80 28.23 28.34 28.06 27.99 27.98 28.06 28.38 28.89 28.80 28.91 29.60 30.07 30.27 30.52 30.48 30.81 31.16 30.81 30.72 30.99 30.93 30.99 30.69 30.87 30.91 30.73 30.94 31.42 31.44 35 34186 34186 34186 34186 36226 36226 36226 36226 38351 38351 38351 38351 39790 39790 39790 39790 41827 41827 41827 41827 44729 44729 44729 44729 46866 46866 46866 46866 49015 49015 49015 49015 51824 51824 51824 51824 55655 55655 55655 55655 59560 59560 59560 59560 63420 63420 63420 63420 68710 68710 68710 68710 72112 72112 7545 7605 7707 7800 7893 8062 8159 8287 8402 8552 8692 8788 8890 8995 9147 9326 9450 9562 9719 9932 10036 10284 10364 10475 10513 10642 10644 10703 10837 10938 11040 11106 11231 11371 11628 11819 11991 12184 12369 12564 12816 12976 13207 13383 13650 13803 13911 14068 14235 14425 14572 14690 14673 14817 5.74 5.60 5.37 5.26 4.93 5.02 5.10 4.98 5.06 5.05 5.05 5.09 5.05 4.98 4.82 4.25 4.41 4.45 4.65 5.04 5.52 5.71 6.02 6.02 4.82 3.66 3.17 1.91 1.72 1.72 1.64 1.33 1.16 1.04 0.93 0.92 0.92 1.08 1.49 2.01 2.54 2.86 3.36 3.83 4.39 4.70 4.91 4.90 4.98 4.74 4.30 3.39 2.04 1.63 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.05313 0 0 0 0.3079 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.85 0.82 0.84 0.86 0.88 0.89 0.89 0.90 0.93 0.94 0.96 0.99 1.05 1.05 1.08 1.05 1.06 1.07 1.06 1.05 1.06 1.08 1.10 1.13 1.13 1.16 1.15 1.16 1.18 1.16 1.14 1.15 1.13 1.09 1.09 1.06 1.03 1.06 1.05 1.01 1.00 1.01 1.01 1.02 1.01 0.99 0.99 0.98 0.98 0.96 0.94 0.91 0.89 0.88 147 148 149 150 151 152 153 154 155 156 157 158 2008 2008 2009 2009 2009 2009 2010 2010 2010 2010 2011 2011 3 4 1 2 3 4 1 2 3 4 1 2 111.12 52.89 41.81 57.30 65.74 74.09 76.54 74.09 72.65 81.94 95.42 108.55 1.0916 1.0930 1.0972 1.0959 1.0966 1.0994 1.1036 1.1079 1.1116 1.1164 1.1240 1.1312 14.86 14.95 15.08 15.10 15.22 15.29 15.38 15.41 15.40 15.48 15.45 15.30 31.16 30.82 30.65 30.75 30.88 30.92 31.22 31.46 31.46 31.43 30.35 29.95 72112 72112 72119 72119 72119 72119 76810 76810 76810 76810 81330 81330 14844 14547 14381 14342 14384 14564 14673 14879 15050 15232 15243 15462 1.49 0.30 0.21 0.17 0.16 0.06 0.11 0.15 0.16 0.14 0.13 0.05 0 0 0 0 0 0 0 0 0 0 3.43509 1.3259 A: refiners' acquisition cost of imported crude oil, U.S. dollars per barrel; 1973:III is official price of Saudi Light; from U.S. Energy Information Administration B: U.S. gross domestic product implicit price deflator; quarterly figures from Federal Reserve Bank of St. Louis http://research.stlouisfed.org/fred2/data/GDPDEF.txt, accessed April 1, 2014. C: non-OPEC production of crude oil; from U.S. Energy Information Administration, http://www.eia.doe.gov/emeu/international/oilproduction.html, accessed April 1, 2014. D: world crude oil production; from U.S. Energy Information Administration, also http://www.eia.doe.gov/emeu/international/oilproduction.html, accessed April 1, 2014. E: world gross domestic product; from IMF World Economic Outlook, April 2005, http://www.imf.org/external/pubs/ft/weo/2005/01/data/dbasubm.cfm and IMF World Economic Outlook, April 2009, http://www.imf.org/external/pubs/ft/weo/2009/01/weodata/weoselser.aspx?a=1&c=001&t=1 F: quarterly U.S. gross domestic product; from Federal Reserve Bank of St. Louis http://research.stlouisfed.org/fred2/data/GDP.txt, accessed April 1, 2014. G: nominal rate on 3-Month U.S. treasury securities, secondary market; from http://www.federalreserve.gov/releases/H15/data/Monthly/H15_TB_M3.txt H: percent drop in world output due to violence; from EIA Monthly Energy Review http://www.eia.doe.gov/emeu/international/oilproduction.html, accessed April 1, 2014. I: Trade-Weighted Exchange Value of U.S. Dollar; from Federal Reserve Bank of Saint Louis http://research.stlouisfed.org/fred2/data/TWEXBMTH.txt, accessed April 1, 2014. 36 0.89 0.99 1.01 0.98 0.95 0.92 0.93 0.95 0.93 0.91 0.89 0.87 REFERENCES Adelman, M.A., 1990. OPEC at thirty years: what have we learned? Annual Review of Energy 15, 1-22. Adelman, M.A., 2004. The real oil problem. Regulation, Spring 2004, 16-21. Angus Maddison Project, http://www.ggdc.net/maddison/maddison-project/home.htm, accessed April 3, 2014. Atkins, F.J., Jazayeri, S.M.T., 2004. A Literature Review of Demand Studies in World Oil Markets. University of Calgary, Discussion Paper 2004-07. Bible, The, Young’s Literal Translation of, 1898, 3rd ed. Greater Truth Publishers, 2004, Lafayette, IN. Cashin, Paul and C. John McDermott, 2001. An Unbiased Appraisal of Purchasing Power Parity. IMF working paper WP/01/196. Celta, D.J., Dahl, C.A., 2000. OPEC as a social welfare maximizer. Colorado School of Mines working paper, February. Cumby, R. E., and J. Huizinga. 1992. Testing the autocorrelation structure of disturbances in ordinary least squares and instrumental variables regressions. Econometrica 60(1): 185–195. Djankov, Simeon and Marta Reynal-Querol, 2007. The Causes of Civil War. Working Paper available at http://www.doingbusiness.org/documents/causes_of_civil_wars.pdf. Energy Information Administration, U.S. Department of Energy, http://www.eia.gov/ Evans, William N., and Robert M. Schwab, 1995. Finishing High School and Starting College: Do Catholic Schools Make a Difference? Quarterly Journal of Economics, 110, 947-974. Federal Reserve Bank of St. Louis, http://research.stlouisfed.org/fred2/series/OILPRICE/downloaddata?cid=98. Frasco, Gregg P., 1993. The Kinked Demand Curve When Demand Shifts. Journal of Economic Education 24:2, 137-143. Gately, Dermot and Hillard G. Huntington, 2002. The Asymmetric Effects of Changes in Price and Income on Energy and Oil Demand, Energy Journal 23:1, 19-55. Greene, David L. and Sanjana Ahmad, 2005. Costs of U.S. Oil Dependence: 2005 Update. Oak Ridge National Laboratory, U.S. Dept. of Energy. ORNL/TM-2005/45. Greenhouse, Steven, 1987. When the World’s Growth Slows. New York Times, December 27. 37 Griffin, J.M., Schulman, C.T., 2005. Price asymmetry in energy demand models: a proxy for energy-saving technical change? Energy Journal 26, 1-21. Hamilton, James D., 2003. What is an oil shock? Journal of Econometrics 113, 363-398. Hamilton, James D., 2009. Causes and Consequences of the Oil Shock of 2007-08, Brookings Papers on Economic Activity, Spring 2009 Conference Draft. Moran, Theodore H., January 1993. Managing an oligopoly of would-be sovereigns: the dynamics of joint control and self-control in the international oil industry past, present, and future. In The International Political Economy of Natural Resources (Vol 1), Mark Zacher, ed. Edward Elgar Publishing, Brookfield, VT. Organization of Petroleum Exporting Countries (OPEC), http://www.opec.org/home/ Robinson, P.M., 1995. Log-periodogram regression of time series with long range dependence. Annals of Statistics 23, 1048-1072. Shepherd, Geoffrey, 1933. Vertical and Horizontal Shifts in Demand Curves. Journal of Farm Economics 15:4, 723-729. Stevens, P., 2005. Oil markets. Oxford Review of Economic Policy 21:1, 19-42. Teitenberg, Tom, 2007. Environmental Economics and Policy. 5th ed. Pearson Addison-Wesley. Van Vactor, S.A., 2010. Introduction to the Global Oil and Gas Business. Penn Well Corporation, Tulsa, OK. Verleger, P.K., 2005. Are oil markets entering yet another “new era”? The Petroleum Economics Monthly 22:7. Wing, I.S., 2008. Explaining the declining energy intensity of the U.S. economy. Resource and Energy Economics 30, 21-49. Wooldridge, Jeffrey M., 2002. Econometric Analysis of Cross Section and Panel Data. MIT Press, Cambridge, MA. 38