Lifting the Alaskan Oil Export Ban: An Intervention Analysis.

Frank W. Rusco [*]

W. David Walls [**]

In this paper we examine the price effects on crude oils of removing the U.S. export ban on Alaskan North Slope crude oil in 1996. We estimate the long-run

impact of removing the export ban through the use of a time series intervention

analysis. The results indicate that Alaskan crude oil prices increased between $0.98 and $1.30 on the West Coast spot market relative to prices of comparable crude oils as a result of removing the export ban. However, we find no evidence that West Coast prices for refined oil products--regular unleaded gasoline, diesel fuel, and jet fuel--increased as a result of lifting the ban.

INTRODUCTION

In this paper we examine the price effects resulting from removal of the export ban on Alaskan North Slope crude oil (hereafter ANS). We quantify the direct effects on crude oil prices as well as effects on prices of various refined products on the West Coast. [1]

Congress authorized the trans-Alaska pipeline during the 1973 oil crisis, but in a compromise with maritime unions and environmental groups the exportation of ANS was prohibited. The resulting production of ANS--beginning in 1977--led to a situation in which the supply of oil produced in the West Coast region exceeded refining demand. After reaching a peak in 1988, falling ANS production reduced the glut of oil on the West Coast market, and in 1993 West Coast crude oil production began to fall below refining demand. However, crude oil remains relatively abundant on the West Coast: Currently, Alaska and California each produce about a million barrels of crude oil per day, an amount equivalent to roughly 80 percent of demand. Figure 1 depicts West Coast production and refining volumes from 1989 through 1999 and shows the increasing role of imported crude oil as combined Alaska and California crude production has fallen.

The West Coast is the natural market for ANS because of its close proximity and consequent low shipping costs. However, the relative abundance of oil produced in the region has meant that a significant proportion of ANS produced since the pipeline was built has been shipped to the Gulf Coast and other markets at much greater transportation cost. Because exports were precluded until mid-1996, producers of ANS did not have access to the closest alternative markets of Asia.

In the next section we develop predictions about the effects of removing the export ban on the price of ANS on the West Coast, on prices of other oils and petroleum products, and on the pattern of shipping. Following this we discuss the econometric methodology used to examine the behavior of crude oil prices and present the estimation results. Finally, we interpret the empirical results and conclude with a discussion of recent events that may ultimately have a further impact on ANS prices.

2. REMOVING THE ANS EXPORT BAN: PREDICTED EFFECTS

The relative abundance of ANS, along with the high costs of shipping it to other markets, caused the competitive market price of ANS on the West Coast to be derived from the price of ANS in the market that received the marginal supply. Due to the ban on exports, ANS producers were precluded from selling to the relatively close Asian markets. Therefore, the nearest alternative domestic market was the Gulf Coast, defined for the purposes of this paper as Alabama, Arkansas, Louisiana, Mississippi, New Mexico, and Texas. The incremental cost of shipping ANS to the Gulf Coast or other domestic markets rather than to Asia was between $2.00 and $4.50 per barrel (GAO, 1999); this figure is higher still when compared to West Coast shipments. Due to the many potential buyers and large size of the Asian and Gulf Coast markets relative to the volume of ANS, it is reasonable to assume that ANS sellers are price takers in these markets and that the price of ANS in these markets can be characterized as a competitive world p rice.

Although the market for crude oil and petroleum products is more concentrated on the West Coast than on the Gulf Coast, in this paper we assume that producers and refiners act competitively and that, as a result, the price of crude oil on the West Coast is determined by its netback value from the best alternative market, either the Gulf Coast or the Asian market. Essentially, we assume that ANS sellers equalize the netback price on the marginal barrel of crude oil in the two markets, which implied lower ANS prices on the West Coast prior to the lifting of the export ban. [2]

While refiners enjoyed low ANS prices during this period, the marginal barrels of oil as well as some petroleum products were imported at world prices. As a result, petroleum product prices reflected the cost of these imports rather than the lower cost of ANS, leading to higher refiner margins during this period in the West Coast than in most other regions.

Under the simple hypothesis that the price of ANS on the West Coast is determined by the seller equalizing marginal netbacks between the West Coast and alternative markets, the lifting of the export ban would be equivalent to a decrease in transportation costs to the alternative market. As a result, more ANS would be shipped to the alternative market, causing price to rise in the West Coast until the marginal netbacks were once again equal at the higher level.

Lifting the export ban on ANS should in principle have caused prices of other crude oils produced and largely sold in the West Coast to rise. Van Vactor (1995) and most others predicted that the prices of California crudes--as substitutes for ANS--would rise in response to higher ANS prices.

Sales to the Gulf Coast market should end abruptly with removal of the export ban as all these barrels are transferred to the Asian market. [3] Over time, the proportion of ANS sold in the West Coast market should rise as long as ANS production continues to fall.

Finally, we expect no impact on petroleum product prices because the marginal barrels of oil and some petroleum products will still be imported into the region at world prices. As discussed above, the main impact should be on refiners' margins, which should fall with the lifting of the export ban.

3. INTERVENTION ANALYSIS

To quantify the effect of lifting the export ban on crude oil prices we developed an econometric time series model of these prices. We model the time series of oil prices and estimate how removing the export ban changed the time series. This type of empirical analysis has been successfully employed to model various types of policy interventions. [4] The intervention analysis permits normal testing for a structural change in the behavior of a time series.

Because oil prices are volatile and influenced by many factors other than removing the export ban, we controlled for other factors by modeling the differentials between prices of West Coast oils and the prices of similar oils in other markets. Modeling price differentials between two crude oils controls for factors that impact them similarly--such as changes in global supply and demand--while capturing local market changes. Modeling the spread between prices is also consistent with the way industry participants view the formation of market oil prices. "In oil markets, the process of price formation largely results from the operations of economic agents mostly concerned with price relatives--with spreads between the prices of different crudes, the prices of a crude at different dates or between petroleum products and crude oil prices--rather than absolute levels" (Horsnell 1993, p.5). Thus it also seems appropriate from an institutional perspective to model the same margin on which market participants base the ir trades.

Formally, let [pd.sub.t] represent the price differential between two types of oil at time t, and let z, represent an intervention variable with value zero prior to removal of the export ban and unity after the ban is removed. For the simple case where [pd.sub.t] follows a first-order autoregressive process the intervention model can be written as: [5]

[pd.sub.1] = [alpha] + [beta][pd.sub.t-l] + [yz.sub.t] + [[upsilon].sub.t] (1)

where [[upsilon].sub.t] is a white noise random disturbance and \[beta]\ [less than] 1. The impact of the intervention can be measured by examining the long-run mean of the price series [pd.sub.1] before and after the intervention.

Before the intervention the variable [z.sub.t] is equal to zero. The constant term in the equation is [alpha], and it follows from the autoregressive structure of the model that the long-run mean of the price differential is equal to [alpha]/(1 - [beta]). After the intervention the variable [z.sub.t] takes on a value of unity, so the constant term becomes [alpha] + [gamma] and the new long-run mean becomes ([alpha] + [gamma])/(l - [beta]). The magnitude of the coefficient [gamma] indicates the immediate effect of removing the oil export ban on the price of oil. The long-run effect of removing the oil export ban is [gamma]/(1 - [beta]), which is generated by subtracting the old long-run mean from the new long-run mean.

4. DATA AND ESTIMATION RESULTS

Table 1 lists the crude oils used in the intervention analysis. The benchmark crudes are in the left hand column and those that are expected to be affected by removal of the ban are in the right hand column. We chose Brent Blend and West Texas Intermediate because these are common benchmark crude oils. The other comparison oils were chosen because they have similar physical properties--API gravity and sulfur content, for example--as the West Coast oils. Specifically, Forcados has similar properties to ANS and the blend of California crudes called Line 63, while Shengli (China) and Dun (Indonesia) are heavier oils and are closer in properties to the California heavy oils, Kern River and THUMS.

We used daily spot prices in California as reported by Platt's Oilgram (www.platts.com) and we chose a period of study between January 8 1992 and December 2 1998. This time period encompasses the export ban's removal on May 28, 1996 with sufficient observations on either side of the event.

Before identifying the time series model generating the data, we investigated the data's stationarity properties. [6] It is well known that many financial time series are nonstationary and that nonstationarity invalidates standard statistical procedures for inference. To investigate the stationarity properties of our data, we performed the augmented Dickey-Fuller (1979) and Phillips-Perron (1988) tests for a unit root in the price spread. [7] In the Dickey-Fuller tests we began with 16 lags and tested down to zero lags; in every case the test statistic rejected the null hypothesis of a unit root. For the Phillips-Perron tests, the lag truncation suggested by the Newey-West criterion was used; the results of these tests also rejected the null hypothesis of a unit root for each series of price differentials. The Dickey-Fuller test statistics with four lagged values and the Phillips-Perron test statistics are reported in Table 2. [8]

We then proceeded to identify the time series model that generates the price differentials. To do this, we examined the correlogram and partial correlogram of each series for the time interval preceding removal of the export ban. We chose this interval, rather than the interval after removal of the ban, because it contained more observations with which to identify the data generating process. The correlogram showed a smooth exponential decline as a function of lag length. The partial correlogram showed a notable spike at the first lag and near-zero partial autocorrelations for longer lags: these patterns in the autocorrelation and partial autocorrelation functions correspond to the theoretical patterns for a first-order autoregressive process in equation 1.

We estimated a first-order autoregressive model and performed various diagnostic tests to refine our specification. For each of the oil price differentials we found that the residuals exhibited strong evidence of autoregressive conditional heteroscedasticity according to the ARCH test; in other words, the residuals appear not to have a constant variance but instead are characterized by a process in which there is a distinct pattern in the variance of the random disturbances of the price differential equation; Engle (1982) pioneered the autoregressive conditional heteroscedasticity model to capture this type of behavior in the variance. The serial correlation we found in the variances is common in financial time series where observations with high and low variances are typically clustered together.

To control for the possibility of autocorrelation in the variance of the random disturbances, we explicitly augmented the price differential equation with an equation that controls for the correlation in residual variances: ARCH and GARCH terms were included in the residual variance equation. This specification of the residual variance equation is parsimonious and is commonly used in the estimation of financial time series. To the first-order autoregressive model described in equation 1, we append

[[[sigma].sup.2].sub.t] = [omega] + [phi] [[[sigma].sup.2].sub.t-1] + [theta] [[epsilon].sub.t-1] [2]

where [[[sigma].sup.2].sub.t] is the conditional variance, the one-period ahead forecast variance conditional on information available in period t - 1. The conditional variance has a mean term, [omega] and an autoregressive or ARCH term that captures information about volatility from the immediately preceding period, and the last period's forecast variance also known as the GARCH term. [9]

The revised specification of the intervention model, consists of the price differential equation (1), and the conditional variance equation (2). We estimated this model on the pre-intervention sample of data and found no evidence of serial correlation in the residuals. We also found no evidence of autoregressive conditional heteroscedasticity in the residuals of the model, indicating that the GARCH(1, 1) term adequately controls for the temporal pattern of residual variances. This model was our final specification and it was estimated on each of the oil price differentials for the entire span of our data with the inclusion of a dummy variable to capture the effect of lifting the Alaskan export ban on the price differential.

Table 3 shows the estimates for the intervention model on price differentials between ANS and three benchmark comparison oils, West Texas Intermediate (WTI), Brent Blend, and Nigerian Forcados, and the price differential between WTI and Line 63 oil. In each case, the coefficient on the intervention term is negative and statistically significant. [10] The magnitude of the coefficient on Export indicates that the initial impact of lifting the export ban is about two cents for each price differential equation. The long-run impact on the price differential is obtained from the ratio of the coefficient on the intervention to one minus the coefficient on the lagged price differential. For the price differentials of ANS with WTI, Brent Blend, and Forcados, estimated long-run impacts are -$1.30, -$0.98, and -$1.10, respectively. Lifting the export ban had a negative and statistically significant impact on the WTI-Line 63 differential. [11] For the WTI-Line 63 differential the long-term impact of lifting the export b an is -$1.28.

The estimates for Kern River and THUMS crude oils relative to comparison oils revealed that the impact of the intervention was not statistically different from zero. [12] Overall, the empirical results of the intervention model indicate that the reduction in the price differential between ANS and the comparison oils, and between Line 63 and West Texas Intermediate, is due to rising prices of ANS and Line 63 oils relative to the comparison oils. The reduction of the price differentials is also similar in magnitude for the four affected price differentials. The point estimates of the long-run impact on the prices of ANS and Line 63 oils is consistent with our and other analysts' expectations of the impact of removing the export ban on crude oil prices.

An intervention analysis of several petroleum product spot prices-regular unleaded gasoline, no. 2 diesel fuel, and jet fuel-revealed no statistically significant change from lifting the export ban. To determine the effect on refined product prices of removing the export ban, we compared West Coast prices of regular unleaded gasoline, no. 2 diesel fuel,. and kerosene-type jet fuel with prices of these same products in the markets, namely Chicago and the Gulf Coast.

For the crude oil price differentials showing a statistically significant reduction after removing the export ban-those reported in Table 3-we calculated 95% confidence bands. Calculation of confidence bands for the long-run impact of removing the ban is somewhat problematic because the estimator is the ratio of two non-independent normally distributed random variables. In order to calculate confidence bands, we make use of Fieller's (1944) method. [13] The calculated confidence bands for the long-run impact on price differentials between ANS and all three comparison oils, and between West Texas Intermediate and Line 63, are reported in Table 4.

When the Alaskan export ban was lifted, there was also some movement in the price differential between light and heavy crude oils, with heavy oil prices falling in the face of increased production relative to light oils. To control for the changes in the world light/heavy price gap we included the Brent-Dun price differential in the West Texas Intermediate-THUMS and West Texas Intermediate-Kern River models. Controlling for the price differential between light and heavy crude oils in this way may pick up a generic trend and isolate the effect of lifting the export ban in these intervention equations. However, controlling for changes in the light/heavy price differential did not change the results: We still find that lifting the export ban had no statistically significant impact on the WTI-Thums and WTI-Kern River price differentials.

Even though the specification analysis indicates no misspecification, and even though the estimates of the price impact of lifting the export ban are within the range of previous predictions, the results should still be interpreted cautiously. There is no counterfactual in this analysis and, in particular, the model did not control structurally for the falling level of ANS production during the period studied. It is expected that ANS prices would have risen gradually as production fell below demand, even with the continued existence of the export ban. On the bright side, a dollar decrease in the price differential of Alaskan oil relative to the comparison oils seems reasonable based on previous expectations and the magnitude of shipping cost differences between the Gulf Coast and Asia. It is also reasonable that crude oil prices would not be passed onto consumer products in light of the argument made above that refiners had been earning the rents associated with low crude prices in the West Coast.[14]

While it may seem troubling that California heavy crude prices did not rise along with ANS, there are several mitigating effects that may have impeded or masked the linkage between ANS and California heavies. First, the California heavy crude has an extremely low API gravity and is high in sulfur and heavy metal content. It cannot even be pumped through a pipeline without heating and/or blending with lighter oils or drag resistant agents and is costly to refine into light products. Because there are few delivery mechanisms for this oil and it is impractical to ship it very far, the market is thin. Second, a huge volume of heavy oil from Canada, Mexico and Venezuela comes into play if prices rise much. Third, refineries capable of processing this crude operate at close to capacity. Finally, advances in enhanced recovery and lifting technology have reduced extraction costs, resulting in a global increase in heavy oil production relative to light oil in recent years.

Finally, the actual pattern of sales of ANS has been consistent with expectations. As production has fallen, the proportion sold to West Coast refiners has risen, and sales to alternative markets have declined. Specifically, in 1994 ANS production averaged 1.6 million barrels per day and 84% of this oil was refined in the West Coast. Production has declined every year since then and in 1999 had fallen to just under 1 million barrels, 92% of which was refined in the West Coast market. Figure 2 shows that as the export ban was lifted sales to other markets were replaced by sales to Asia.[15]

5. CONCLUSIONS

We found strong empirical evidence that the prices of ANS and a similar California crude oil (Line 63) rose on the West Coast as a result of removing the export ban. Heavier California oils appear to have been unaffected by the ban's removal, but there were confounding factors that could have masked or mitigated any such effect. Consumer prices did not appear to have been affected by higher ANS prices, which is consistent with expectations--at the time the export ban was removed the marginal barrels of oil sold on the West Coast were imported at world prices and these prices rather than the price of ANS determined the marginal cost of petroleum products.

Since the lifting of the export ban, the BP-ARCO merger with the consequent sale of ARCO's ANS production to Phillips has changed the situation on the West Coast. BP-ARCO will be refining the bulk of their ANS production, and given ARCO's (now Phillips') smaller share of total production, exports should shrink even further. [16] An interesting possibility emerges if the recently proposed merger between Phillips and Tosco is approved. In this case, all the major producers of ANS will be integrated with refiners on the West Coast and the third-party market for ANS will be extremely thin. An intriguing possibility is that the three big producers of ANS, BP-ARCO, Exxon, and Phillips, would all have an incentive to reduce ANS spot prices on the West Coast, because this would reduce the amount of royalties they must pay the state of Alaska. [17]

Looking forward, the most significant impact of lifting the export ban is to increase the value of exploration and new oil field development in Alaska. An increase in the price of ANS will ultimately lead to greater amounts of oil being produced in Alaska from existing and new fields. There is a potential for large increases in ANS production in the future if more of the National Petroleum Reserve is opened to oil development and if the Alaska National Wildlife Refuge comes into play. Higher prices make the value of these future prospects higher and make it more likely that companies would pursue such projects if they are opened for exploration and development. Should there be a significant increase in ANS production, the existence of the Asian market as an alternative to the West Coast will ensure that the price of ANS remains at the world level.

Earlier versions of this paper were presented at the 75th annual conference of the Western Economics Association, Vancouver, July 2000, and at the Department of Economics, University of Calgary. We would like to thank Kevin Banks, Philip Thompson, Frank Atkins, Bob McRae and three anonymous referees for helpful comments. The views expressed in this paper are solely those of the authors and are not to be attributed to their employers.

(*.) Center for Economics, U.S. General Accounting Office, 441 G Street NW, Washington, D.C. 20548, USA. Corresponding author e-mail: ruscof@gao.gov

(**.) Department of Economics, University of Calgary, Calgary, Alberta, T2N 1N4 Canada. E-mail: wdwalls@ucalgary.ca.

(1.) For the purposes of this paper, the West Coast consists of Alaska, Arizona, California, Hawaii, Nevada, Oregon, and Washington.

(2.) Numerous studies have argued that under the export ban, the West Coast price of ANS was lower than the world price for crude oil, adjusted for specific characteristics. See EIA (1990), GAO (1990a), GAO (1990b), DOE (1994), and Sam Van Vactor (1995).

(3.) As discussed above, both the Gulf Coast and Asian markets are large relative to the volume of ANS sold there, and we can assume that the price of that oil will always be at the world level. The world price in the two markets may differ due to differences in refinery configurations but this difference should be relatively small and should be stable over time.

(4.) See, for example, the discussion in Enders (1995) and the references contained therein.

(5.) In the empirical analysis, the time series process is identified by inspecting the data. For ease of exposition, we focus on the simple ar(1) process in this section.

(6.) Because the intervention model is based on price differentials we investigate the stationarity properties below. However, it is also true that the individual price series do show strong evidence of a unit root as is common in most energy price series, even when using Perron's (1989) cointegration test allowing for the structural break when the ANS ban was lifted.

(7.) For these tests, we restricted the time series to the four years leading up to the passage of the Legislation lifting the export ban. We chose these dates to avoid the potential structural break associated with removal of the export ban.

(8.) Since the individual price series contain a unit root, and the price differential does not, the prices are cointegrated. Ewing and Harter (2000) provides recent evidence on this relationship between ANS prices and Brent prices.

(9.) The GARCH(1,1) specification can also be expressed algebraically as a weighted average of the lagged squared residuals where die weights decline geometrically in the lag. Another way of interpreting the GARCH(1,1) specification is that the squared residuals follow an ARMA(1,1) process. See Bollerslev (1986) for a detailed analysis of the GARCH model.

(10.) We also note that in each case the model is consistent with stationarity of the price differential series under study as evidenced by the coefficients on the ARCH and GARCH terms.

(11.) Coefficient estimates for the intervention were negative but not statistically different from zero for Brent-Line 63 and Forcados-Line 63. For brevity, we have only reported the regression results for those price differentials where the intervention was statistically significant.

(12.) For brevity, we have not reported the detailed empirical results for the price differential regressions in which the intervention variable was statistically insignificant. These are available from the corresponding author.

(13.) The following brief description of Fieller's method follows Maddala's exposition (1977, pp. 101-102). A new variable [theta] = [gamma] - [phi](1 - [beta]) is defined, where [phi] = [gamma]/(1 - [beta]) is an unknown constant that we seek to place confidence bands around. The new variable [theta] has an expected value given by E([theta]) = [gamma] - [phi](1 - [beta]) = 0 and a variance given by Var([theta]) = Var([gamma]) + [[phi].sup.2]Var([beta]) + 2[phi]Var([gamma], [beta]). Because the estimators [gamma] and [beta] are normally distributed, it follows that the ratio [[gamma] - [phi](1 - [beta])] [divided by] [square root of]Var([phi]) has a limiting normal distribution. Making use of this fact, the definition of a confidence interval states that Prob [{[[[gamma] - [phi](1 - [beta])].sup.2]} [divided by] Var([phi]) [[less than or equal to] [[Z.sup.2].sub.a/2] = 1 - a where [Z.sub.a/2] is the appropriate percentage point from a standard normal distribution. If the inequality inside the bracketed term is taken to be a strict equality, the resulting quadratic equation can be solved for its roots, [[phi].sub.1] and [[phi].sub.2] which are the lower and upper confidence limits for the long-run impact of the intervention variable.

(14.) There is some evidence that refining margins in the west Coast also fell in the mid 1990s, and this is consistent with our results. However, some refiners' capital costs also rose during this period in order to comply with California fuel restrictions.

(15.) The law legalizing exports of ANS to Asia went into effect May 28, 1996, which explains the still large volume of ANS sold in other markets in that year.

(16.) In fact, since summer 2000, there have been no exports to Asia, consistent with an agreement between the Federal Trade Commission (FTC) on one hand, and BP-ARCO and Phillips, on the other, that these producers of ANS would suspend exports to Asia. According to the FTC, this agreement is not legally binding. Therefore, the authors believe that exports remain a viable future option for ANS producers in the event that production increases significantly.

(17.) Royalties are calculated as a percentage of an index of ANS and other benchmark crude oil spot prices. Hence, if ANS spot prices fall, royalty payments are also reduced.

REFERENCES

Bollerslev, T. (1986). "Generalized autoregressive conditional heteroskedasticity." Journal of Econometrics 31: 307-327.

Dickey, D. A. and W. A. Fuller (1979). "Distribution of the estimators of autoregressive time series with a unit root." Journal of the American Statistical Association 74: 427-431.

Enders, W. (1995). Applied Econometric Time Series. New York: John Wiley & Sons.

Engle, R. (1982). "Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation." Econometrica 50: 391-407.

Ewing, B. T. and C. L. Harter (2000). "Co-movements of Alaska North Slope and UK Brent crude oil prices." Applied Economics Letters 7: 553-558.

Fieller, E. C. (1944). "A fundamental formula in the statistics of biological assay and some applications." Quarterly Journal of Pharmacy, pages 117-123.

Horsnell, P. and R. Mabro (1993). Oil Markets and Prices: The Brent Market and the Formation of World Oil Prices. Oxford: Oxford University Press.

Maddala, G. S. (1977). Econometrics. Singapore: McGraw-Hill, international edition.

Perron, P. (1989). "The great crash, the oil price shock, and the unit root hypothesis." Econometrica 57: 1361-1401.

Phillips. P. and P. Perron (1988). "Testing for a unit root in time series regression." Biometrika 75: 335-346.

U.S. Department of Energy (1994). "Exporting Alaskan North Slope Crude Oil: Benefits and Costs." DOE/PO-0025. U.S. GPO, Washington, D.C.

U.S. Energy Information Administration (1990). "Implications of Lifting the Ban on Exports of Alaskan Crude Oil: Price and Trade Impacts." SR/EMEU/90-3. U.S. GPO, Washington, D.C.

U.S. General Accounting Office (1990a). "Alaskan Crude Oil Exports." GAO/T-RCED-90-59. U.S. GPO, Washington, D.C.

U.S. General Accounting Office (1990b). "Energy Security: Impacts of Lifting Alaskan North Slope Oil Exports Ban." GAO/RCED-91-21. U.S. GPO, Washington, D.C.

U.S. General Accounting Office (1999). "Alaskan North Slope Oil: Limited Effects of Lifting Export Ban on Oil and Shipping Industries and Consumers." GAO/RCED-99-191. U.S. GPO, Washington, D.C.

Van Vactor, S. A. (1995). "Time to end the Alaskan oil export ban." Policy Analysis 227, Cato Institute.

[Graph omitted]

[Graph omitted]

Table 1. Comparison of Oils


Comparison Oils                West Coast Oils

Brent Blend                    ANS, Line 63
Nigerian Forcados              ANS, Line 63
West Texas Intermediate (WTI)  ANS, Line 63. Kern River, THUMS
Shengli, China                 Kern River, THUMS
Duri, Indonesia                Kern River, THUMS
Table 2. Unit Root Tests on Oil Price Differentials


                    Unit Root Test Statistic

Price Differential       Dickey-Fuller        Phillips-Perron

WTI-ANS                      -3.62                 -4.09
Forcados - ANS               -4.17                 -4.67
Brent - ANS                  -4.38                 -4.91
WTI - Line 63               -3.08                 -3.62
Forcados - Line 63           -2.99                 -3.23
Brent - Line 63              -2.90                 -3.09
WTI - Kern                   -3.27                 -3.75
Duri - Kern                   -3.17                 -3.32
Shengli - Kern               -3.03                 -3.12
WTI - THUMS                 -3.15                 -3.59
Duri - THUMS                 -2.95                 -2.99
Shengli - THUMS              -2.89                 -2.92



Note: Dickey-Fuller test statistics reported here correspond to four
lagged differences included in the testing equation. Phillips-Perron
test statistics reported here correspond to the lag truncation
suggested by the Newey-West criterion. The 5% critical value for
each statistic is 2.86.
Table 3. Estimation Results - Comparison Oil Price Differentials


                   Price Differential

Coefficient             WTI-ANS        Brent-ANS  Forcados-ANS

Price Equation

Conscant                  0.037            0.019         0.022
                        (0.003)          (0.005)       (0.006)
Differential(-1)          0.986            0.976         0.982
                        (0.002)          (0.003)       (0.003)
Export                   -0.018           -0.024        -0.020
                        (0.004)          (0.007)       (0.008)

Variance Equation

Conscant                  0.004            0.014         0.009
                        (0.0001)         (0.001)      (0.0006)
ARCH(1)                   0.460            0.504         0.399
                        (0.022)          (0.027)       (0.020)
GARCH(1)                  0.365            0.259         0.465
                        (0.014)          (0.029)       (0.013)
[R.sup.2]                 0.976            0.946         0.952






Coefficient        WTI-Line 63

Price Equation

Conscant                 0.051
                       (0.007)
Differential(-1)         0.986
                       (0.003)
Export                  -0.018
                       (0.006)

Variance Equation

Conscant                7.4E-4
                     (3.33E-5)
ARCH(1)                  0.075
                       (0.004)
GARCH(1)                 0.910
                       (0.007)
[R.sup.2]                0.976



Note: Differential(-1) is the respective lagged price differential
in each price differential equation. Export is a dummy variable
assuming a value of unity after the export ban was lifed.
Estimated standard errors are reported in parentheses.
Table 4. Estimated Long-Term Impact of Lifting Export Ban


                       Estimated Impact on Price
                             Differential

                                WTI-ANS           Brent-ANS

Lower Confidence Band            -2.16               1.66
Point Estimate                   -1.30              -0.98
Upper Confidence Band            -0.64              -0.43







                       Forcados-ANS  WTI-Line 63

Lower Confidence Band     -2.07         -2.59
Point Estimate            -1.10         -1.28
Upper Confidence Band     -0.23         -0.46



Note: The point estimate of the long-term impact of removing the export
ban is calculated as the quotient of the coefficient on Export and
(1 - Differential(-1)). We calculated the long-term impact only for the
models in which the intervention of lifting the export ban was
statistically significant. The confidence bands on the long-term impact
of lifting the export ban were calculated using Fieller's method.