RICE UNIVERSITY National Oil Company Efficiency: Theory and Evidence

Stacy Eller Peter Hartley Kenneth B Medlock III

James A. Baker III Institute for Public Policy RICE UNIVERSITY

1 RICE UNIVERSITY

Theoretical Model

2 Economic model precepts RICE UNIVERSITY

n Intertemporal, optimizing model of a National Oil Company (NOC)

n Contrast a NOC to a shareholder­owned firm

n Capture systematic effects from the NOC institutional arrangement

n Weaker monitoring of a NOC and differing political goals imply that in addition to commercial returns NOC management choices will reflect:

u Increased employment in the firm of labor or other domestic inputs u Domestic consumer surplus from oil product sales u Pressure to increase current relative to future revenue – a high discount rate n Without these concerns, the NOC optimization problem approaches that of a private monopoly firm

n In the efficient case:

u Domestic oil consumers are neither taxed nor subsidized relative to other constituents, and u Domestic consumer surplus is weighed identically to NOC profits

3 NOC versus efficient case: output, inputs, cash flow

RICE n Output shifted forward, lower reserves & cash flow, higher employment UNIVERSITY Output Reserves 0.40 1.20

NOC output NOC reserves 0.35 Efficient output Efficient reserves 1.00

0.30

0.80 0.25

0.20 0.60

0.15 0.40

0.10

0.20 0.05

0.00 0.00 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Years Years

Employment Cash Flow 4.00 0.25

NOC employment NOC cash flow 3.50 0.20 Efficient cash Efficient employment flow

3.00 0.15

2.50 0.10

2.00 0.05

1.50 0.00

1.00 ­0.05

0.50 ­0.10

0.00 ­0.15 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Years Years 4 Effect of excess emphasis on current revenue

RICE n Output & cash flow shifted forward, reduced investment in reserves, UNIVERSITY increased employment except in the long term

Change in output Change in reserves 0.025 0.01

0.020 0.00 v = 0.05 r v = 0.05 0.015 r v = 0.1 ­0.01 r v = 0.1 0.010 r ­0.02 0.005 ­0.03 0.000 ­0.04 ­0.005 ­0.05 ­0.010

­0.015 ­0.06

­0.020 ­0.07 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Years Years

Change in employment Change in cash flow 0.7 0.010

0.6 v = 0.05 0.005 r v = 0.05 r v = 0.1 0.5 r v = 0.1 r 0.000 0.4

0.3 ­0.005

0.2 ­0.010 0.1 ­0.015 0.0

­0.1 ­0.020 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Years Years 5 Effect of increasing the employment incentive

RICE n Output & cash flow shifted forward, reduced investment in reserves, UNIVERSITY increased employment except in the long term

Change in output Change in reserves 0.012 0.005

v = 0.1w 0.010 0.000 L

0.008 ­0.005 vL = 0.2w

0.006 ­0.010

0.004 v = 0.1w ­0.015 L

0.002 v = 0.2w ­0.020 L

0.000 ­0.025

­0.002 ­0.030

­0.004 ­0.035 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Years Years

Change in employment Change in cash flow 0.35 0.006

0.30 0.004 v = 0.1w v = 0.1w 0.25 L L v = 0.2w vL = 0.2w 0.002 L 0.20

0.15 0.000

0.10 ­0.002 0.05 ­0.004 0.00

­0.05 ­0.006 0 5 10 15 20 25 30 0 5 10 15 20 25 30 6 Years Years NOC versus efficient domestic consumption RICE UNIVERSITY n The subsidy raises domestic consumption for the NOC

n Increasing the domestic subsidy also shifts employment forward relative to the efficient case, but the effects are small

0.0112 Change in employment 0.0006

0.0111 v = 1.025 0.0004 c 0.0110 v = 1.05 0.0002 c 0.0109

0.0108 NOC domestic demand 0.0000

0.0107 Efficient domestic demand ­0.0002

0.0106 ­0.0004

0.0105 ­0.0006 0.0104 ­0.0008 0.0103 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Years Years

7 RICE UNIVERSITY

Empirical Analysis

8 Data and methods RICE UNIVERSITY n Sample of 80 firms over 2002­2004 (Energy Intelligence “Ranking the World’s Oil Companies”) with data on:

u revenue, u reserves of natural gas and crude oil, u employment, u production of natural gas and crude oil and crude oil products, and u the government ownership share n Used both non­parametric Data Envelopment Analysis (DEA) and a parametric Stochastic Frontier Approach (SFA)

n Motivated by the theoretical model we use revenue as the output measure

u Political pressure is likely to force a NOC to subsidize domestic consumers u To the extent that NOC’s generate less revenue for given inputs we can conclude that their objectives differ from a private firm n Also in accordance with the theoretical model, we allow for three inputs: employees, oil reserves and natural gas reserves

9 Firms in the sample (statistics for 2004)

RICE Revenue Revenue Revenue Revenue per per Gov ernment per per Gov e rnment UNIVERSITY Compan y Emp lo yee Reserves Ownership Country Compan y Emplo yee Reserves Owne rship Country $/employee $/boe % $/employee $/boe % NOCs Others Adnoc 205 0.20 100% UAE Amerada Hess 1,532 16.07 0% US CNOOC 2,656 2.97 71% China Anadarko 1,838 2.52 0% US EcoPetrol 824 2.26 100% Colombia Apache 2,019 2.71 0% US Eni 1,056 10.50 30% Italy BG 1,547 3.64 0% UK Gazprom 103 0.16 51% Russia Burlington 2,537 2.74 0% US INA 187 11.70 75% Croatia Chesapeake Energy 1,577 3.22 0% US KMG n/a n/a 100% Kazakhastan CNR 4,606 3.85 0% Canada KPC 1,650 0.34 100% Kuwait Devon 2,356 4.33 0% US MOL 635 42.37 25% Hungary Dominion 847 13.81 0% US NIOC 283 0.11 100% Iran EnCana 2,915 4.48 0% Canada NNPC 1,460 0.56 100% Nigeria EOG 1,844 2.38 0% US NorskHydro 673 11.37 44% Norway ForestOil 1,841 4.02 0% US OMV 2,214 8.90 32% Austria HuskyEnergy 2,149 9.53 0% Canada ONGC 298 2.11 84% India Imperial 2,838 35.72 0% Canada PDO 1,591 0.98 60% Oman Kerr­McGee 1,263 4.15 0% US PDVSA 1,985 0.66 100% Venezuela Lukoil 233 1.68 0% Russia Pemex 506 4.01 100% Mexico Maersk 60 2.90 0% Denmark Pertamina 453 0.73 100% Indonesia Marathon 1,757 39.14 0% US Petrobras 773 3.39 32% Brazil Murphy 1,436 21.60 0% US PetroChina 111 2.52 90% China Newfield 2,114 4.45 0% US Petroecuador 1,026 1.25 100% Ecuador Nexen 1,048 4.25 0% Canada Petronas 1,202 1.45 100% Malaysia NipponOil 2,690 131.74 0% Japan PTT 2,896 16.68 100% Thailand Noble 2,433 2.54 0% US QP 1,800 0.10 100% Qatar Novatek 220 0.21 0% Russia Rosneft 86 0.19 100% Russia Occidental 1,577 4.46 0% US SaudiAramco 2,261 0.40 100% Saudi Arabia PennWest 1,577 2.53 0% Canada Sinopec 192 19.76 57% China Petro­Canada 2,370 9.24 0% Canada Socar n/a n/a 100% Azerbaijan PetroKazakhstan 546 4.12 0% Kazakhstan Sonangol 755 1.37 100% Angola Pioneer 1,183 1.76 0% US Sonatrach 688 0.93 100% Algeria Pogo 5,088 4.38 0% US SPC 375 1.71 100% Syriac RepsolYPF 1,561 10.79 0% Spain Statoil 1,910 10.85 71% Norway Santos 789 1.92 0% Australia TPAO 154 1.53 100% Turkey Sibneft 189 1.81 0% Russia Average 1,000.27 5.23 Suncor 1,447 78.50 0% Canada Surgutneftegas 121 1.01 0% Russia Major IOCs Talisman 2,207 3.26 0% Canada BP 2,788 15.68 0% UK TNK 63 1.66 0% Russia Chevron 2,606 12.78 0% US Total 1,406 14.33 0% France ConocoPhillips 3,368 14.03 0% US Unocal 1,259 4.63 0% US ExxonMobil 3,148 12.26 0% US Vintage 1,136 1.76 0% US Shell 2,418 21.67 0% Netherlands Woodside 758 2.11 0% Australia Average 2,865.48 15.28 XTO 1,437 1.94 0% US Average 1,628.94 11.24 10 Simplified representation of DEA

RICE n UNIVERSITY To graph the data in two dimensions, reserves are converted to barrels of oil equivalent and normalized, along with revenue, on the number of employees

n Technical inefficiency in generating revenue from these inputs can be calculated using the vertical distance of a firm from the frontier

4

NOCs Majors Others

e

e 3 oy l p m e

r e p ) $

US 2 on i l l i m (

e u n

ve 1 Re

0 0 0.2 0.4 0.6 0.8

Reserves (mmboe) per employee 11 Other variables in the analysis

RICE n Vertical integration could influence estimated technical efficiency:

UNIVERSITY u A vertically integrated firm captures the value added by the internal sale of crude oil to its refining unit u Without measuring capital employed in the refining, transporting and marketing, a vertically integrated firm would appear to be relatively efficient at generating revenue from employees and reserves alone n Government ownership share is a key variable for our hypothesis:

u Theory implies higher government ownership should give lower efficiency at generating revenue u Excess employment should be a key mechanism for this measured technical inefficiency u Two­tier pricing is another reason a NOC may generate less revenue

Venezuela 0.11 Iran 0.21 Saudi Arabia 0.64 Indonesia 0.85 Algeria 0.89 Kuwait 0.91 UAE 1.06 Oman 1.08 Syria 1.12 Malaysia 1.12 Azerbaijan 1.12 Angola 1.29 Ecuador 1.53 Nigeria 1.59 Kazakhstan 1.70 Thailand 1.72 China 1.72 Russia 1.89 Mexico 1.97 Colombia 2.04 Average pump prices 2004 US 2.10 Brazil 2.52 Canada 2.57 India 2.82 Australia 3.18 Japan 4.18 Spain 4.37 Croatia 4.49 Austria 4.75 Hungary 4.77 Turkey 4.84 France 5.05 Italy 5.37 Netherlands 5.39 Denmark 5.41 Norway 5.77 UK 5.98 ­ 1.00 2.00 3.00 4.00 5.00 6.00 7.00 12 Average DEA scores over 2002­04

P o go 1.00 Nippo nOil 1.00 RICE P TT 0.96 P etro Kaza khs tan 0.87 Exxo nMo bil 0.84 UNIVERSITY Imperial 0.81 C NR 0.77 B P 0.75 Co no co P hillips 0.71 EnC ana 0.68 Shell 0.67 C hevro n 0.67 Newfie ld 0.64 Stato il 0.63 P etro ­C anada 0.62 Sino pec 0.61 n Hus kyEnergy 0.61 Murphy 0.60 Five major IOC’s are Sunco r 0.58 C NOOC 0.57 No ble 0.56 Talis man 0.55 clustered near the frontier OMV 0.54 B urlingto n 0.53 Devo n 0.52 Maratho n 0.50 n EOG 0.49 P ennWes t 0.48 NOC’s tend to be clustered Reps o lYP F 0.46 Apache 0.46 Amerada Hes s 0.45 Occidenta l 0.44 near the bottom Che s apeake Energy 0.44 Fo res tOil 0.43 B G 0.40 To ta l 0.39 Anadarko 0.38 n XTO 0.37 NOCs average TE ≈ 0.27 SaudiAramco 0.36 KP C 0.35 Kerr­McGee 0.34 Uno c al 0.33 QP 0.32 n P DVSA 0.32 Sample average TE ≈ 0.40 Vintage 0.31 Nexen 0.31 P etro nas 0.30 P etro ecuado r 0.30 Eni 0.29 n MOL 0.29 Five major IOCs TE ≈ 0.73 NNP C 0.29 P io neer 0.28 Do m inio n 0.27 P etro bras 0.26 So nango l 0.26 P ertamina 0.22 P DO 0.21 Eco P etro l 0.21 No rs kHydro 0.20 Santo s 0.20 Wo o ds ide 0.19 Sibneft 0.17 So natrach 0.16 P em ex 0.16 TNK 0.12 INA 0.09 TP AO 0.09 ONGC 0.08 NOCs M aers k 0.06 Adno c 0.06 Major IOCs Luko il 0.06 NIOC 0.05 Other No vatek 0.04 P etro China 0.03 Surgutneftegas 0.03 Gazpro m 0.03 Ro s neft 0.02 ­ 0.20 0.40 0.60 0.80 1.00 Model 1 Technical Efficiency Output: Revenue Inputs: Gas Reserves, Oil Reserves, Employees 13 Po g o Nip p o nOil PTT Structural & institutional Pet ro Kazakhstan Exxo nMo b il RICE Imp erial CNR BP Co no co Phillips adjustments UNIVERSITY EnCana Shell Chevro n Newfield St ato il Pet ro ­Canad a Sino p ec HuskyEnerg y M urp hy Sunco r CNOOC No b le Talisman OM V Burlingt o n Devo n If we include: M aratho n EOG PennWest Rep so lYPF u Ap ache a vertical integration Amerad a Hes s Occid ental Chesap eake Energ y Fo rest Oil measure (petroleum BG To tal Anad arko XTO product sales divided by Saud iAramco KPC Kerr­McGee Uno cal total liquids production) QP PDVSA Vintag e Nexen u Pet ro nas and government Pet ro ecuad o r Eni M OL NNPC ownership share Pio neer Do minio n Petro b ras So nang o l Pert amina as “inputs” more firms appear PDO Eco Pet ro l No rs kHyd ro Santo s Wo o d sid e to be on the revised frontier Sib neft So nat rach Pemex TNK INA TPAO ONGC Maers k Ad no c Luko il NIOC No vatek Petro China Surg ut nefteg as Gazp ro m Ro sneft ­ 0.20 0.40 0.60 0.80 1.00 Model 1, 2 and 3 Technical Efficiency Output: Revenue Inputs: Model 1 ­ Gas Reserves, Oil Reserves, Employees Model 2 ­ Model 1 plus Vertical Integration Model 3 ­ Model 2 plus Government Share 14 Stochastic frontier estimation

n RICE Estimated TE is now assumed constant over the three year period UNIVERSITY u Include yearly effects to allow especially for varying oil and gas prices by year u Year effects are not necessary in DEA analysis since TE is calculated for each year separately n Model 2sf includes vertical integration and government share (like DEA model 3)

n Model 3sf includes a dummy for 2­tier pricing

n Model 4sf includes an employment­government share interaction

Nippo nOil Exxo nM o bil a BP Table 4 Š Panel estimation of stochastic frontier S hell C hevro n Co no co P hillips To tal M aratho n Model 1sf Model 2sf Model 3sf Model 4sf Impe rial Sta to il P TT 0.4847*** 0.6459*** 0.5648*** 0.6077*** Reps o lYP F ln L Eni 0.0666 0.0504 0.0637 0.0362 Am erada Hes s Sa udiAram co OMV 0.0463 0.0666 0.1188*** 0.1524*** P e tro ­Canada ln OilRsv So car 0.0415 0.0462 0.0459 0.0396 EnC ana Murphy P e tro bras 0.1695*** 0.2091*** 0.2069*** 0.2035*** P DVSA ln NGRsv Hus kyEne rgy 0.0493 0.0485 0.0471 0.0415 Sunco r Oc cide ntal P em ex C NR ­0.5970*** ­0.3109** 2.7912*** No rs kHydro GovShare KP C 0.1398 0.1607 0.8316 Devo n Do minio n C NOOC 0.0737*** 0.0969*** 0.0824*** P etro nas VertInt P o go 0.0203 0.0198 0.0198 S ino pec S P C Talis m an ­0.5435*** ­0.6654*** NNP C 2TierP Burlingto n 0.1570 0.1382 M OL BG Apa che ­0.3099*** Uno cal GovShare ∗ ln L Anada rko 0.0824 Kerr­McGee KMG P e tro e cua do r 0.3022*** 0.2950*** 0.2877*** 0.2872*** P e rtam ina year2003 Ne xen 0.0307 0.0325 0.0331 0.0335 So nango l No ble So natrach 0.4767*** 0.4626*** 0.4633*** 0.4652*** Newfield year 2004 EOG 0.0312 0.0330 0.0334 0.0339 C hes ape ake P e nnWes t Ec o P e tro l 4.3644*** 1.5483*** 1.9375*** 1.2476*** P etro Kazakhs tan constant QP 0.6561 0.3474 0.4860 0.2894 F o res tOil XTO Luko il P io neer P etro China Vintage 2 ONGC χ (d ) 451.33 1112.72 992.72 1643.43 Sibneft P DO Adno c d 5 7 8 9 Wo o ds ide Santo s INA Log Likelihood ­111.300 ­100.041 ­94.109 ­87.427 TNK NIOC Surgutnefte gas # Observations 236 236 236 236 TP AO Gazpro m Mae rs k R o s neft a Estimated standard errors included beneath each coefficient estimate. No va tek *­ statistically significant at the 1% level; **­ statistically significant at the 5% level; ***­ statistically significant at the 10% level ­0.40 ­0.20 0.00 0.20 0.40 0.60 0.80 1.00 Model 1sf and 4sf Technical Efficiency 15 Other indications of model adequacy

XTO 3 .6 7 % ICE Wo o ds ide 1.8 2 % R Vintage 3 .4 7 % Uno cal 2 .4 4 % TP AO 5 .2 0 % UNIVERSITY To tal 1.3 0 % TNK 1.8 4 % Talis m an 3 .7 5 % Surgutneftegas 2 .3 6 % Sunco r ­ 6 .6 1% Stato il ­11.2 8 % 4 1.0 9 % SP C So natrach 2 .2 5 % So nango l 3 .9 2 % 3 16 .4 5 % So c ar Sino pec ­3 .5 3 % Sibneft 1.2 9 % Shell ­3 .6 4 % SaudiAramco ­14 .6 0 % Santo s 2 .5 0 % R o s neft 0 .3 3 % R eps o lYP F 2 .3 8 % QP 2 .2 2 % P TT ­2 .3 2 % P o go ­9 .4 8 % P io nee r 1.6 4 % P etro nas 0 .8 1% P etro Kazakhs tan 1.3 6 % P etro e cuado r 7 .7 6 % P etro China 2 .4 6 % P etro ­C anada 0 .17 % P etro bras 5 .4 6 % $200,000 P ertamina 0 .18 % P ennWe s t 5 .12 % $175,000 P em ex 1.9 8 % P DVSA ­ 2 5 .8 0 % ­2 .3 3 % P DO $150,000 ONGC 2 .2 1% OMV 6 .2 3 % Occidental 2 .8 3 % $125,000 No vatek ­10 .3 9 % No rs kHydro 2 .4 3 % $100,000 No ble 2 .3 9 % NNP C 3 .13 % Nippo nOil 5 .18 % $75,000 NIOC 0 .9 2 % Nexen 2 .3 6 % Newfield 4 .2 6 % $50,000 Murphy 0 .0 6 % MOL ­0 .0 1% $25,000 Maratho n 1.9 2 % Maers k 14 .3 9 % Luko il 0 .4 6 % $­ KP C 2 .2 1% 5 4 .7 2 % KMG GovShare > 0 2­Tier Pricing Majors Others Kerr­McGee 3 .0 9 % INA 3 .7 2 % Predicted Revenue Observed Revenue Imperial ­19 .9 2 % Hus kyEnergy 1.19 % Gazpro m 1.9 3 % Fo res tOil 0 .10 % Exxo nMo bil ­6 .2 9 % EOG 1.2 7 % Eni 1.3 8 % EnC ana 0 .5 0 % Eco P etro l 2 .2 4 % Do m inio n 2 .6 7 % Devo n 0 .5 9 % C o no co P hillips ­6 .6 4 % C NR ­3 .15 % CNOOC ­ 2 .4 8 % C hevro n ­ 3 .7 4 % C hes apeake ­0 .0 2 % B urlingto n 1.3 5 % BP ­7 .6 5 % BG 1.4 0 % Apache 2 .6 9 % Anadarko 1.7 7 % Amerada Hes s 2 .2 1% Adno c 3 .0 0 % ­25.0% ­20.0% ­15.0% ­10.0% ­5.0% 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% Percent deviation of predicted revenue from observed revenue 16 Interpreting the stochastic frontier results n Vertically integrated firms generate more revenue from inputs of employees and reserves RICE n UNIVERSITY Government ownership reduces the ability of the firm to generate revenue n Domestic price subsidies are one reason government share may reduce revenue

u However, since the government share still has a negative effect, this is not the only reason u The positive and significant coefficient on oil reserves in models 3sf and 4sf suggest that accounting for domestic subsidies gives a better model of the determinants of TE in generating revenue n Since the model implies many consequences of government control should lead to over­ employment, we allow for a government share­employment interaction

u The negative coefficient on the interaction implies that the productivity of labor in generating revenue is lower the higher is government ownership u Furthermore, the overall effect of government ownership remains negative (controlling for 2­tier pricing) for firms with a positive government share

$10,000 Full government ownership with domestic price subsidies $9,000 Full government ownership without domestic price subsidies Publicly traded firm without domestic price subsidies $8,000 Average # employees $7,000

$6,000 Revenue million US$ $5,000 Effect of increasing exercise of government

$4,000

$3,000

$2,000

$1,000

$­ ­ 100,000 200,000 300,000 400,000 500,000 Employees 17 RICE UNIVERSITY

Conclusion

18 Summary remarks RICE UNIVERSITY n The theoretical model implies that government ownership of a NOC will redistribute revenue via over­employment and under­investment in reserves and by subsidizing domestic consumption u Many of the influences reinforced each other in their effects n Evidence confirmed that increased government ownership makes the firm less effective at producing revenue from employment and reserves

n We further found specific evidence that: u Over­employment was a strong common feature of government­owned firms u Domestic price subsidies negatively affect a NOC’s ability to generate revenue n The relative technical inefficiencies of NOC’s, which are observed when one considers only commercial objectives, are largely the result of governments exercising control over the distribution of rents

n The forgone revenue will, however, reduce government spending on other items or require higher taxes u Product subsidies or over­employment in a NOC are generally inefficient compared to taxes and transfers as a way of redistributing income u They are poorly targeted as transfers, and more inefficient than a broadly­ based tax as a means of raising revenue

19