608 Intemational Organization for which data are available.49 Descriptive statistics are in an appendix.5o Ordinary least squares (OLS)regression is used on the dependent variable, which is the depletion rate in each state-year.5I The models use Huber-White standard errors clustered by state on the premise that standard errors for multiple observations within a state cannot be assumed to be independent of each other. All independent variables are lagged by one year to reduce the potential for endogeneity. Several explanatory variables are used,reflecting the factors just identified.One is the variable oPEC,a dichotomous measure indicating whether the state is a member of OPEC in a given year,which is of crucial interest to this inquiry.The second is WoRLD ECONOMIC GROWTH,measured by that year's annual global GDP growth,as a proxy for global demand for oil that might create incentives for especially high or low depletion rates in a particular year.The third is FISCAL STRENGTH,measured by the natural log of oil reserves per capita.This variable is included because states with large oil reserves per capita can typically meet the fiscal demands of the govern- ment without maximizing production.52 Data on oil production and oil reserves are from the BP Statistical Review of World Energy.Fourth,the state's INVESTMENT RISK affects the ease with which international businesses can operate and the extent to which they invest in oil production capacity.53 It is measured using the (inverse) risk score from the International Country Risk Guide.Fifth,the state's regime type (as measured by Polity IV)is included,because it could affect the state's depletion policy in a variety of ways.54 Sixth,a dichotomous variable,wAR,indicates those state-years in which a state was engaged in a major international war in its territory,such as the Iran-Irag or Irag-Kuwait wars.Seventh,another dichotomous variable,sANCTIoN,indicates observations in which a state was the target of a major international sanction.35 LIFr cosrs of production (that is,costs of getting oil to the ground,including exploration)are included only as a robustness check,as I will discuss.56 Table 2 presents the results of regression analyses.Model 1 shows a simple bivariate model that indicates that OPEC membership is statistically associated with low depletion rates,as expected by the conventional "OPEC-as-cartel" 49.BP Statistical Review of World Energy provides data on forty-seven oil-producing countries,but Brunei,Chad,Equatorial Guinea,Turkmenistan,and Uzbekistan are not included because of data avail- ability for other variables. 50.Available on the joumal's and author's websites. 51.To check robustness,the regressions were also conducted using a tobit model:the results were similar. 52.See Teece 1982:and Cremer and Salehi-Isfahani 1991.My use of this measure follows the convention in previous research.Other measures of fiscal strength such as government debt or expenditure ratios are possible but less preferable because they do not necessarily indicate surplus oil reserves,such as the state's capacity to meet fiscal demands without maximizing production. 53.Jensen and Johnston 2011. 54.See Jensen 2006;and Li 2009. 55.Data from Hufbauer,Schott,Elliott,and Oegg 2007. 56.Data from Waghorn et al.2006
for which data are available.49 Descriptive statistics are in an appendix.50 Ordinary least squares (OLS) regression is used on the dependent variable, which is the depletion rate in each state-year.51 The models use Huber-White standard errors clustered by state on the premise that standard errors for multiple observations within a state cannot be assumed to be independent of each other. All independent variables are lagged by one year to reduce the potential for endogeneity. Several explanatory variables are used, reflecting the factors just identified. One is the variable OPEC, a dichotomous measure indicating whether the state is a member of OPEC in a given year, which is of crucial interest to this inquiry. The second is WORLD ECONOMIC GROWTH, measured by that year’s annual global GDP growth, as a proxy for global demand for oil that might create incentives for especially high or low depletion rates in a particular year. The third is FISCAL STRENGTH, measured by the natural log of oil reserves per capita. This variable is included because states with large oil reserves per capita can typically meet the fiscal demands of the government without maximizing production.52 Data on oil production and oil reserves are from the BP Statistical Review of World Energy. Fourth, the state’s INVESTMENT RISK affects the ease with which international businesses can operate and the extent to which they invest in oil production capacity.53 It is measured using the (inverse) risk score from the International Country Risk Guide. Fifth, the state’s regime type (as measured by Polity IV) is included, because it could affect the state’s depletion policy in a variety of ways.54 Sixth, a dichotomous variable, WAR, indicates those state-years in which a state was engaged in a major international war in its territory, such as the Iran-Iraq or Iraq-Kuwait wars. Seventh, another dichotomous variable, SANCTION, indicates observations in which a state was the target of a major international sanction.55 LIFT COSTS of production (that is, costs of getting oil to the ground, including exploration) are included only as a robustness check, as I will discuss.56 Table 2 presents the results of regression analyses. Model 1 shows a simple bivariate model that indicates that OPEC membership is statistically associated with low depletion rates, as expected by the conventional “OPEC-as-cartel” 49. BP Statistical Review of World Energy provides data on forty-seven oil-producing countries, but Brunei, Chad, Equatorial Guinea, Turkmenistan, and Uzbekistan are not included because of data availability for other variables. 50. Available on the journal’s and author’s websites. 51. To check robustness, the regressions were also conducted using a tobit model; the results were similar. 52. See Teece 1982; and Crémer and Salehi-Isfahani 1991. My use of this measure follows the convention in previous research. Other measures of fiscal strength such as government debt or expenditure ratios are possible but less preferable because they do not necessarily indicate surplus oil reserves, such as the state’s capacity to meet fiscal demands without maximizing production. 53. Jensen and Johnston 2011. 54. See Jensen 2006; and Li 2009. 55. Data from Hufbauer, Schott, Elliott, and Oegg 2007. 56. Data from Waghorn et al. 2006. 608 International Organization
The Limits of OPEC in the Global Oil Market 609 hypothesis.The statistical significance of OPEC membership disappears,however, when other variables are added in the subsequent models.Model 2 shows a baseline model,without taking into account the potential impact of OPEC.As expected, INVESTMENT RISK and FISCAL STRENGTH are negatively correlated with the depletion rate,reflecting the fact that poor investment climates inhibit oil production,and oil-rich states have low fiscal needs and thus long-time horizons for depletion. TABLE 2.Regression analysis on states'oil depletion rates,1980-2009 1) (2) (3) (4) (5) (6) OPEC MEMBER ALL -3.9569 -0.733 -0.81 (-4.08)** (-0.76) (-0.84) SAUDI ARABIA -1.625 -1.734 (-1.59 (-1.26) OPEC MEMBER NON SAUDE -0.683 (-0.71) OPEC CORE MEMBER -0.911 (-0.6) OPEC NONCORE MEMBER -0.627 (-0.62) POLITY SCORE 0.039 0.034 0.03 0.028 0.052 0.73) (0.68) (0.59 (0.5) (0.91) WORLD ECONOMIC GROWTH 0.051 0.044 0.045 0.046 0.047 (0.76 (0.67) (0.68) (0.68) (0.71) FISCAL STRENGTH -0.992 -0.91 -0.905 -0.892 -0.905 (-4.42)*** (-3.18)*** (-3.15)** (-2.68)** (-3.16)** INVESTMENT RISK -0.084 -0.079 -0.08 -0.081 -0.079 (-2.59)** (-2.33)** (-2.36)* (-2.4)** (-2.37** INTERNATIONAL WAR 1.097 1.155 1.112 1.043 1.104 (0.8) (0.86) (0.83) 0.69) (0.86 INTERNATIONAL SANCTIONS -0.649 -0.61 -0.654 -0.677 -0.57 (-1) -0.98) (-1.05)** (-1.1) (-0.92)* LIFT COST -0.17 (-1.09) Observations 1286 993 993 993 993 993 R-squared 0.181 0.373 0.376 0.377 0.377 0.379 Notes:t-scores in parentheses (robust standard errors clustered by state).significant at 10%;**significant at 5%; ***significant at 1%. In Model 3,OPEC membership is reintroduced to the regression.The new variable is not statistically significant and does nothing to improve the explanatory power of the model (the R-squared moves from 0.373 to 0.376).The t-statistic is just-0.76, indicating that one cannot reject the null hypothesis that OPEC membership has no impact on a state's depletion rate. Thus OPEC members produce oil at more or less exactly the same rates that they could be expected to produce in the absence of OPEC.The findings imply that,to the extent that OPEC members underproduce compared with non-OPEC members,they do so because of other factors in the model (for example,fiscal strength,investment risk)that have nothing to do with their OPEC membership.Some OPEC members might restrict their depletion rate as a conscious act of policy,but they appear to do so out of their own self-interest,without institutional support from OPEC.For
hypothesis. The statistical significance of OPEC membership disappears, however, when other variables are added in the subsequent models. Model 2 shows a baseline model, without taking into account the potential impact of OPEC. As expected, INVESTMENT RISK and FISCAL STRENGTH are negatively correlated with the depletion rate, reflecting the fact that poor investment climates inhibit oil production, and oil-rich states have low fiscal needs and thus long-time horizons for depletion. In Model 3, OPEC membership is reintroduced to the regression. The new variable is not statistically significant and does nothing to improve the explanatory power of the model (the R-squared moves from 0.373 to 0.376). The t-statistic is just −0.76, indicating that one cannot reject the null hypothesis that OPEC membership has no impact on a state’s depletion rate. Thus OPEC members produce oil at more or less exactly the same rates that they could be expected to produce in the absence of OPEC. The findings imply that, to the extent that OPEC members underproduce compared with non-OPEC members, they do so because of other factors in the model (for example, fiscal strength, investment risk) that have nothing to do with their OPEC membership. Some OPEC members might restrict their depletion rate as a conscious act of policy, but they appear to do so out of their own self-interest, without institutional support from OPEC. For TABLE 2. Regression analysis on states’ oil depletion rates, 1980–2009 (1) (2) (3) (4) (5) (6) OPEC MEMBER - ALL −3.9569 −0.733 −0.81 (−4.08)** (−0.76) (−0.84) SAUDI ARABIA −1.625 −1.734 (−1.59) (−1.26) OPEC MEMBER - NON SAUDI −0.683 (−0.71) OPEC CORE MEMBER −0.911 (−0.6) OPEC NONCORE MEMBER −0.627 (−0.62) POLITY SCORE 0.039 0.034 0.03 0.028 0.052 (0.73) (0.68) (0.59) (0.5) (0.91) WORLD ECONOMIC GROWTH 0.051 0.044 0.045 0.046 0.047 (0.76) (0.67) (0.68) (0.68) (0.71) FISCAL STRENGTH −0.992 −0.91 −0.905 −0.892 −0.905 (−4.42)*** (−3.18)*** (−3.15)*** (−2.68)** (−3.16)*** INVESTMENT RISK −0.084 −0.079 −0.08 −0.081 −0.079 (−2.59)** (−2.33)** (−2.36)** (−2.4)** (−2.37)** INTERNATIONAL WAR 1.097 1.155 1.112 1.043 1.104 (0.8) (0.86) (0.83) (0.69) (0.86) INTERNATIONAL SANCTIONS −0.649 −0.61 −0.654 −0.677 −0.57 (−1) (−0.98) (−1.05)** (−1.1) (−0.92)** LIFT COST −0.17 (−1.09) Observations 1286 993 993 993 993 993 R-squared 0.181 0.373 0.376 0.377 0.377 0.379 Notes: t-scores in parentheses (robust standard errors clustered by state). * significant at 10%; ** significant at 5%; *** significant at 1%. The Limits of OPEC in the Global Oil Market 609