Global LNG Contract Terms and Revisions An Empirical Analysis Mark - - PowerPoint PPT Presentation

Global LNG Contract Terms and Revisions

An Empirical Analysis Mark Agerton

Graduate Fellow Center for Energy Studies, Rice University magerton@rice.edu IAEE International Conference, June 17, 2014

Research question

I How is LNG priced under long-term contracts (LTC)?

– Pricing terms are not public – Rules of thumb? – JKM spot price?

Mark Agerton 2 / 31

How to infer characteristics of LTC terms?

I Exploit public customs data

– Importer + Exporter – Total volume + value

I Characterize statistical relationship I Understand LTC pricing terms better I Overcome econometric issues

– Cointegration – Unknown structural breaks

Mark Agerton 3 / 31

Why focus on JKST, and Japan especially?

50 100 150 200 250 1970 1980 1990 2000 2010

million metric tons

Others Taiwan Spain Korea Japan

Source: Petroleum Economist

Mark Agerton 4 / 31

Asian markets are tight

Atlantic Pacific

50 100 150 2000 2005 2010 2000 2005 2010

million metric tons

LTC Actual Total

Source: GIIGNL and author's calculations

Mark Agerton 5 / 31

Customs data capture LTC, not spot cargoes

Atlantic Pacific

50 100 150 2000 2005 2010 2000 2005 2010

million metric tons

Actual Total Spot + ST

Source: GIIGNL and author's calculations

Mark Agerton 6 / 31

What do LTC pricing formulas look like?

I Alaskan LNG export license I Oil benchmark I S-curve I Regimes j = 1, . . . , m + 1

Oil LNG

yt = cj + j

p+1

X

r=0

✓jrzt−r + ut t = Tj−1 + 1, . . . , Tj

Mark Agerton 7 / 31

Why worry about structural breaks?

I Renegotiation clauses (Weems, 2006) I Changing value of LNG

– Substitutes – Supply and demand shocks

I S-curves =

⇒ nonlinearities

I Heterogeneity in underlying contracts

Mark Agerton 8 / 31

DOLS with structural breaks

yt = cj + δj

p+1

X

r=0

θjrzt−r + ut m yt = cj + δjzt +

p

X

s=0

πjs∆zt−s + ut

I Dynamic OLS I Minimize SSR I UD max, sup Wald and SEQ(k|k + 1) I Bai and Perron (1998, 2003); Kejriwal and Perron (2008b, 2010)

Mark Agerton 9 / 31

Breaks in the LNG–oil relationship

Indonesia Qatar USA Brunei Malaysia Australia UAE Qatar Oman Algeria Nigeria Qatar Indonesia 1988 1992 1996 2000 2004 2008 2012 Japan Korea Spain Taiwan Country of origin Breaks in time or s-curve kinks? Elimination of s-curves is contract revision Fukushima? Mark Agerton 10 / 31

Breaks in the LNG–oil relationship

Indonesia Qatar USA Brunei Malaysia Australia UAE Qatar Oman Algeria Nigeria Qatar Indonesia 1988 1992 1996 2000 2004 2008 2012 Japan Korea Spain Taiwan Country of origin Breaks in time or s-curve kinks? Elimination of s-curves is contract revision Fukushima? Mark Agerton 10 / 31

Breaks in the LNG–oil relationship

Indonesia Qatar USA Brunei Malaysia Australia UAE Qatar Oman Algeria Nigeria Qatar Indonesia 1988 1992 1996 2000 2004 2008 2012 Japan Korea Spain Taiwan Country of origin Breaks in time or s-curve kinks? Elimination of s-curves is contract revision Fukushima? Mark Agerton 10 / 31

Breaks in the LNG–oil relationship

Indonesia Qatar USA Brunei Malaysia Australia UAE Qatar Oman Algeria Nigeria Qatar Indonesia 1988 1992 1996 2000 2004 2008 2012 Japan Korea Spain Taiwan Country of origin Breaks in time or s-curve kinks? Elimination of s-curves is contract revision Fukushima? Mark Agerton 10 / 31

Breaks in the LNG–oil relationship

Indonesia Qatar USA Brunei Malaysia Australia UAE Qatar Oman Algeria Nigeria Qatar Indonesia 1988 1992 1996 2000 2004 2008 2012 Japan Korea Spain Taiwan Country of origin Breaks in time or s-curve kinks? Elimination of s-curves is contract revision Fukushima? Mark Agerton 10 / 31

Japan: LNG vs oil prices

• ●
• ●
• ●
• USA

Indonesia Malaysia Brunei Australia UAE Qatar 5 10 15 20 5 10 15 20 50 100 50 100 50 100

Oil LNG

Regime

• 5

4 3 2 1

Mark Agerton 11 / 31

Focusing in on two series

• ●
• Indonesia

Malaysia 5 10 15 20 50 100 50 100

Oil LNG

Regime

• 4

3 2 1

Mark Agerton 12 / 31

Japanese contract revisions

I S-curve kinks?

– 2000: USA, Brunei, Malaysia ($25–26) – early 2004: Qatar ($31) – late 2004: Indonesia, UAE ($40)

I Eliminate S-curves?

– 2007 (Qatar, Malaysia, Brunei) – 2009 (UAE, maybe Australia) – Asian imports exceed LTC 2003–2007 and 2011–2013 – High oil prices 2007–present

I Shifts up in 2009–2011 I Fukushima—only Indonesian revisions

Mark Agerton 13 / 31

Other countries: LNG vs oil prices

• ●
• ●
• ●
• ●
• ●
• ● ●
• ●
• ●
• ● ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●
• ● ●
• ●
• ●
• ●
• ●
• ● ●
• ●●
• ●
• ●
• ● ●
• ●
• ●
• ●
• ●
• ●
• ●
• ●● ●
• ● ●
• ●
• ● ●
• ●●
• ● ●
• ●
• ●
• ●
• ●
• ● ●
• ●
• ●
• ●● ●
• ●
• ●
• ● ●
• ● ●
• ●● ●
• ●
• ●
• ●
• ● ●
• ●
• ●
• ●●
• ●
• ●
• ● ●

Korea—Qatar Korea—Oman Taiwan—Indonesia Spain—Qatar Spain—Algeria Spain—Nigeria 5 10 15 20 5 10 15 20 50 100 50 100 50 100

Oil LNG

Regime

• 2

1

Mark Agerton 14 / 31

• S. Korea, Taiwan, Spain

I Fewer and smaller revisions I No S-curves I Slopes

– South Korea and Taiwan: approximately 0.17 ≈ 1/5.8 – Spain: similar to S-curve tail (0.06–0.10)

Mark Agerton 15 / 31

Findings

I Japan

– S-curves to reduce risk – Abandoned: rising oil prices, tight market – Fukushima: LTCs as insurance? – Poorer fit in recent years

I South Korea and Taiwan

– No S-curves – Smaller countries, fewer firms and revisions

I Spain

– More substitutes and weaker indexation – Fewer revisions, no S-curves

Mark Agerton 16 / 31

Conclusion

I Pricing with oil-indexation, not JKM I Heterogeneity and revisions in pricing, not rules-of-thumb I Must account for structural breaks

Mark Agerton 17 / 31

Japan Estimates I

Japan–Malaysia Japan–Brunei Japan–USA Start 1988(05) 2000(03) 2007(04) 2009(10) 1988(08) 2000(05) 2007(11) 2010(05) 1988(04) 1991(04) 1999(12) 2006(01) 2009(04) End 2000(02) 2007(03) 2009(09) 2014(02) 2000(04) 2007(10) 2010(04) 2014(02) 1991(03) 1999(11) 2005(12) 2009(03) 2011(03) c 1.039∗∗∗ 3.058∗∗∗ 2.438∗∗∗ −1.280∗∗∗ 1.091∗∗∗ 3.271∗∗∗ 1.281 1.588∗ 1.454∗∗∗ 0.974∗∗∗ 2.469∗∗∗ 3.451∗∗∗ 0.829∗∗ (0.118) (0.041) (0.596) (0.256) (0.127) (0.093) (0.895) (0.686) (0.130) (0.173) (0.059) (0.059) (0.238) δ 0.131∗∗∗ 0.055∗∗∗ 0.101∗∗∗ 0.168∗∗∗ 0.125∗∗∗ 0.042∗∗∗ 0.126∗∗∗ 0.139∗∗∗ 0.096∗∗∗ 0.134∗∗∗ 0.070∗∗∗ 0.041∗∗∗ 0.146∗∗∗ (0.006) (0.001) (0.007) (0.002) (0.007) (0.002) (0.011) (0.006) (0.007) (0.010) (0.002) (0.001) (0.003) θ0 0.167 0.074 −0.464 0.069 0.134 0.300 −0.444 0.114 0.074 0.181 0.075 0.043 −0.009 (0.137) (0.118) (0.260) (0.055) (0.142) (0.343) (0.266) (0.145) (0.212) (0.225) (0.119) (0.062) (0.064) θ1 −0.246 −0.116 0.326 −0.126 −0.154 −0.204 0.386 0.004 0.292 −0.053 −0.059 −0.503∗∗∗ −0.052 (0.273) (0.186) (0.449) (0.085) (0.282) (0.532) (0.470) (0.240) (0.472) (0.385) (0.188) (0.116) (0.087) θ2 0.149 0.203 0.115 0.182∗ −0.050 0.252 0.479 0.122 −1.033∗ −0.740∗ 0.082 0.589∗∗∗ 0.085 (0.306) (0.196) (0.440) (0.085) (0.318) (0.559) (0.443) (0.253) (0.483) (0.368) (0.190) (0.120) (0.088) θ3 0.039 0.333 0.558 0.513∗∗∗ 0.369 0.136 −0.207 0.229 1.667∗∗∗ 1.612∗∗∗ 0.902∗∗∗ 0.870∗∗∗ 0.976∗∗∗ (0.275) (0.199) (0.497) (0.079) (0.325) (0.570) (0.507) (0.251) (0.222) (0.204) (0.122) (0.058) (0.065) θ4 0.891∗∗∗ 0.507∗∗∗ 0.466 0.362∗∗∗ 0.417 0.278 0.499 0.357 (0.133) (0.124) (0.256) (0.051) (0.325) (0.570) (0.502) (0.251) θ5 0.220 0.280 −0.464 0.263 (0.322) (0.570) (0.421) (0.255) θ6 0.168 0.418 0.008 −0.073 (0.292) (0.567) (0.454) (0.243) θ7 −0.105 −0.460 0.744∗∗ −0.016 (0.151) (0.374) (0.232) (0.141) T 142 85 30 53 141 90 30 46 36 104 73 39 24 ˆ σi 0.129 0.104 0.532 0.312 0.120 0.148 0.754 0.488 0.125 0.164 0.084 0.103 0.234 ˆ σLR,i 0.220 0.136 0.744 0.286 0.216 0.318 0.862 0.569 0.156 0.247 0.133 0.083 0.180 ˆ ρi 0.662 0.360 0.429 −0.082 0.760 0.709 0.444 0.466 0.405 0.491 0.467 −0.262 −0.236 ˆ σ 0.225 0.300 0.141 ˆ σLR 0.315 0.380 0.180 ˆ ρ 0.279 0.491 0.279 D-Fuller

• 13.160
• 10.198
• 12.410

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05

Mark Agerton 19 / 31

Japan Estimates II

Japan–Qatar Japan–Australia Japan–UAE Japan–Indonesia Start 1997(01) 2004(02) 2007(04) 2009(05) 2012(01) 1989(08) 2005(10) 2009(01) 1993(06) 2005(01) 2009(03) 1988(04) 2005(01) 2011(04) End 2004(01) 2007(03) 2009(04) 2011(12) 2014(02) 2005(09) 2008(12) 2014(02) 2004(12) 2009(02) 2014(02) 2004(12) 2011(03) 2014(02) c 1.310∗∗∗ 1.818∗∗∗ −0.504 2.650∗∗∗ −0.090 2.019∗∗∗ −0.535 1.225 1.335∗∗∗ 0.489 −0.397 0.951∗∗∗ 2.337∗∗∗ −3.729 (0.129) (0.258) (0.340) (0.397) (2.127) (0.132) (0.521) (0.683) (0.252) (0.321) (0.325) (0.118) (0.176) (2.519) δ 0.122∗∗∗ 0.091∗∗∗ 0.142∗∗∗ 0.126∗∗∗ 0.153∗∗∗ 0.088∗∗∗ 0.115∗∗∗ 0.129∗∗∗ 0.113∗∗∗ 0.106∗∗∗ 0.156∗∗∗ 0.141∗∗∗ 0.089∗∗∗ 0.191∗∗∗ (0.006) (0.005) (0.004) (0.005) (0.019) (0.006) (0.007) (0.007) (0.011) (0.004) (0.003) (0.005) (0.002) (0.022) θ0 0.235 0.367 −0.155 0.014 −0.036 −0.334 0.152 0.046 0.073 −0.121 0.174 0.992∗∗∗ 0.932∗∗∗ 0.340∗∗ (0.174) (0.229) (0.085) (0.144) (0.099) (0.306) (0.207) (0.247) (0.442) (0.179) (0.097) (0.159) (0.094) (0.105) θ1 −0.128 0.038 0.058 −0.389 0.104 0.056 −0.348 0.090 −0.120 −0.097 −0.237 0.077 0.146 0.041 (0.298) (0.346) (0.162) (0.208) (0.145) (0.546) (0.412) (0.387) (0.719) (0.344) (0.148) (0.288) (0.176) (0.169) θ2 0.319 −0.074 0.508∗∗ 0.766∗∗ 0.121 0.191 0.290 0.089 0.204 0.297 0.211 −0.087 −0.090 −0.133 (0.307) (0.373) (0.160) (0.213) (0.158) (0.573) (0.483) (0.382) (0.756) (0.348) (0.148) (0.293) (0.175) (0.170) θ3 0.241 0.118 −0.048 0.606∗ 0.420∗ 0.951 0.802 0.631 −0.222 0.428 0.637∗∗∗ 0.018 0.012 0.752∗∗∗ (0.306) (0.395) (0.172) (0.217) (0.155) (0.555) (0.510) (0.360) (0.754) (0.350) (0.146) (0.167) (0.099) (0.109) θ4 0.133 0.410 0.075 −0.123 0.356∗ 0.135 0.104 0.144 1.065∗ 0.494∗ 0.215∗ (0.311) (0.403) (0.163) (0.198) (0.149) (0.319) (0.309) (0.195) (0.460) (0.187) (0.083) θ5 0.121 −0.138 0.037 0.074 0.066 (0.304) (0.390) (0.163) (0.203) (0.140) θ6 0.079 0.279 0.524∗∗∗ 0.052 −0.031 (0.178) (0.255) (0.086) (0.117) (0.099) T 85 38 25 32 26 194 39 62 139 50 60 201 75 35 ˆ σi 0.128 0.297 0.335 0.491 0.337 0.189 0.553 0.686 0.279 0.430 0.397 0.152 0.287 0.480 ˆ σLR,i 0.274 0.354 0.299 0.412 0.279 0.565 0.757 1.123 0.788 0.619 0.511 0.440 0.376 0.499 ˆ ρi 0.717 0.306 0.143 −0.089 0.017 0.816 0.407 0.466 0.791 0.376 0.305 0.812 0.363 0.193 ˆ σ 0.287 0.395 0.342 0.241 ˆ σLR 0.279 0.720 0.631 0.372 ˆ ρ 0.103 0.500 0.509 0.381 D-Fuller

• 12.788
• 9.806
• 8.857
• 11.652

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05

Mark Agerton 20 / 31

Estimates for others

Korea–Qatar Korea–Oman Taiwan–Indonesia Spain–Qatar Spain–Algeria Spain–Nigeria Start 1999(09) 2000(04) 2000(01) 2008(11) 2003(04) 1995(01) 2010(10) 2000(02) 2008(01) End 2013(04) 2013(04) 2008(10) 2013(03) 2013(04) 2010(09) 2013(04) 2007(12) 2013(04) c 0.795∗∗∗ 0.733∗∗∗ 1.035∗∗∗ 0.409 1.312∗∗∗ 0.764∗∗∗ 2.172 1.165∗∗∗ 2.935∗∗∗ (0.106) (0.147) (0.119) (0.278) (0.253) (0.130) (1.421) (0.173) (0.302) δ 0.155∗∗∗ 0.155∗∗∗ 0.151∗∗∗ 0.164∗∗∗ 0.081∗∗∗ 0.098∗∗∗ 0.077∗∗∗ 0.080∗∗∗ 0.061∗∗∗ (0.002) (0.002) (0.002) (0.003) (0.003) (0.003) (0.013) (0.004) (0.003) θ0 −0.072 −0.080 0.210∗∗ 0.021 0.126 0.357∗ −0.320 0.110 0.136 (0.079) (0.104) (0.076) (0.074) (0.192) (0.168) (0.364) (0.231) (0.162) θ1 0.076 0.025 0.702∗∗∗ 0.745∗∗∗ −0.018 −0.056 0.048 −0.159 −0.009 (0.146) (0.190) (0.136) (0.117) (0.328) (0.281) (0.497) (0.319) (0.288) θ2 0.115 0.050 0.088 0.234∗∗∗ −0.002 −0.194 1.440∗ −0.170 0.136 (0.150) (0.194) (0.088) (0.064) (0.335) (0.283) (0.561) (0.329) (0.292) θ3 0.042 0.610∗∗ 0.354 0.092 −1.029 0.142 0.171 (0.150) (0.192) (0.337) (0.284) (0.572) (0.328) (0.294) θ4 0.313∗ 0.395∗∗∗ −0.223 0.245 0.806 0.430 −0.527 (0.148) (0.104) (0.331) (0.283) (0.552) (0.326) (0.287) θ5 0.527∗∗∗ 0.763∗∗∗ 0.048 −0.305 0.648∗∗ 1.092∗∗∗ (0.080) (0.192) (0.282) (0.498) (0.244) (0.158) θ6 0.508∗∗ 0.360 (0.170) (0.283) T 164 157 106 53 121 189 31 95 64 ˆ σi 0.446 0.576 0.556 0.516 0.653 0.528 0.906 0.318 0.548 ˆ σLR,i 0.639 0.841 0.588 0.496 0.987 0.959 0.664 0.656 0.502 ˆ ρi 0.527 0.411 0.103 0.004 0.438 0.546 −0.259 0.679 −0.082 ˆ σ 0.543 0.583 0.423 ˆ σLR 0.554 0.820 0.475 ˆ ρ 0.061 0.327 0.169 D-Fuller −7.031 −7.955

• 11.721

−6.815

• 10.475
• 10.587

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05

Mark Agerton 21 / 31

Related literature

1 ) Energy market integration

– Process of price convergence... not discrete breaks – No a-priori DGP – De Vany and Walls (1993) – Neumann and Cullmann (2012); Neumann, Siliverstovs, and Hirschhausen (2006)

2 ) Spot–LTC interactions

– Theoretical – Predict increasing flexibility – Brito and Hartley (2007) – Ikonnikova et al. (2009) – Hartley (2013)

Mark Agerton 22 / 31

References I

Andrews, Donald WK (1991). “Heteroskedasticity and autocorrelation consistent covariance matrix estimation”. In: Econometrica: Journal of the Econometric Society 59.3, pp. 817–858. Bai, Jushan and Pierre Perron (1998). “Estimating and testing linear models with multiple structural changes”. In: Econometrica 66.1, pp. 47–78. – (2003). “Computation and analysis of multiple structural change models”. In: Journal of Applied Econometrics 18.1, pp. 1–22. Brito, Dagobert L and Peter R Hartley (2007). “Expectations and the evolving world gas market”. In: The Energy Journal, pp. 1–24. De Vany, Arthur and W. David Walls (1993). “Pipeline Access and Market Integration in the Natural Gas Industry: Evidence from Cointegration Tests”. In: The Energy Journal 14.4. Hartley, Peter R (2013). The Future of long-term LNG Contracts. Tech. rep. The James A. Baker III Institute for Public Policy, Rice University. Ikonnikova, Svetlana et al. (2009). “Strategic model of LNG arbitrage: analysis

• f LNG trade in Atlantic Basin”. In: 32nd International Association for Energy

Economics Conference Proceedings. Citeseer.

Mark Agerton 23 / 31

References II

Kejriwal, Mohitosh and Pierre Perron (2008a). “Data dependent rules for selection of the number of leads and lags in the dynamic OLS cointegrating regression”. In: Econometric Theory 24.5, p. 1425. – (2008b). “The limit distribution of the estimates in cointegrated regression models with multiple structural changes”. In: Journal of Econometrics 146.1,

• pp. 59–73.

– (2010). “Testing for multiple structural changes in cointegrated regression models”. In: Journal of Business & Economic Statistics 28.4, pp. 503–522. Neumann, A and A Cullmann (2012). “What’s the story with natural gas markets in Europe? Empirical evidence from spot trade data”. In: European Energy Market (EEM), 2012 9th International Conference on the. IEEE, pp. 1–6. Neumann, Anne, Boriss Siliverstovs, and Christian von Hirschhausen (2006). “Convergence of European spot market prices for natural gas? A real-time analysis of market integration using the Kalman Filter”. In: Applied Economics Letters 13.11, pp. 727–732. Weems, Philip R (2006). “Evolution of long-term LNG sales contracts: Trends and issues”. In: Oil, Gas & Energy Law Intelligence 4.1.

Mark Agerton 24 / 31

Contracts analyzed

Japan

25 50 75 2000 2005 2010 Australia Brunei Indonesia Malaysia Qatar UAE USA Others

Korea

10 20 30 40 2000 2005 2010 Oman Qatar Others Source: GIIGNL

Taiwan

5 10 2000 2005 2010 Indonesia Others

Spain

5 10 15 20 2000 2005 2010 Algeria Nigeria Qatar Others

Mark Agerton 25 / 31

Japan: mostly LTCs

Australia Brunei Indonesia Malaysia Qatar UAE USA

5 10 15 20 2 4 6 5 10 15 5 10 15 5 10 15 2 4 0.0 0.5 1.0 2000 2005 2010

million metric tons

Actual LT Actual Spot + ST LTC

Source: GIIGNL and author's calculations

Mark Agerton 26 / 31

Others: mostly LTCs

Korea−Oman Korea−Qatar Taiwan−Indonesia Spain−Algeria Spain−Nigeria Spain−Qatar

2 4 5 10 1 2 3 2 4 6 2 4 6 8 1 2 3 4 2000 2005 2010 2000 2005 2010 2000 2005 2010

million metric tons

Actual LT Actual Spot + ST LTC

Source: GIIGNL and author's calculations

Mark Agerton 27 / 31

Benchmark selection

I Either Brent crude or Japanese Crude Cocktail (JCC) I Most statistically significant set of lags in

PLNG,t = ↵ +

12

X

k=0

Brent,kPBrent,t−k +

12

X

k=0

JCC,kPJCC,t−k + ut

I Japan: JCC I Others: Brent

Mark Agerton 28 / 31

Lag selection

I DOLS model

yt = c + zt +

p

X

s=0

⇡s∆zt−s + ut

I Use BIC to detmine p (Kejriwal and Perron, 2008a)

Mark Agerton 29 / 31

The sup Wald test

I Test for k breaks (Kejriwal and Perron, 2008b, 2010)

FT(, k) = T − (p + 3) × (k + 1) k × SSR0 − SSRk SSRk

I Break fractions ∈ Λ must be at least ✏ > 0 apart

i+1 − i = (Ti+1 − Ti)/T > ✏

I sup Wald just minimizes the SSR:

arg max FT(, k) = arg min SSR

Mark Agerton 30 / 31

Testing procedure

I Recommended by Bai and Perron (2003):

1 ) UD max = maxk=1,...,M sup FT(, k) 2 ) SEQ(k|k + 1) for null of k vs k + 1 breaks 3 ) sup Wald = supλ∈Λ FT(, k)

I Robust version is just scaling by long-run variance

– Scale by (ˆ 2

u/ˆ

2) – ˆ 2 is a modified Andrews (1991) estimator (Kejriwal and Perron, 2010)

I Delta-method for variance of implied ✓s

Mark Agerton 31 / 31