Valuing the Risks and Returns to the Spot LNG Trading

Transcription

1 Valuing the Risks and Returns to the Spot LNG Trading Prepared for the 27th USAEE/IAEE North American Conference, Houston, September 16-19, 2007 Hiroaki Suenaga School of Economics and Finance Curtin University of Technology Perth, Western Australia

2 Motivation Two common observations in world gas industry in the last decade: 1. Increase in level and volatility of regional gas price 2. Reduction in cost to liquefy gas and transport LNG These observations have followed extensive discussions on potential gains from short-term LNG trading and, as a consequence, integration of world gas market. Only a little attempt has been made to analytically elucidate how these changes affect LNG trading practices.

3 Objectives 1. Examine the stochastic properties of the US natural gas and crude oil prices, and 2. Consider their implications for the risks and returns from the short-term LNG trading in Asia-pacific region.

4 Literature Hayes (2006) examined the values of flexible LNG supply in Atlantic region. A trading strategy considered: LT commitment (full cover) to supply to the US divert shipment to Europe when price favors replace the reduced LNG supply to the US by pipeline gas purchased at the Henry Hub spot market. The strategy is practicable only in the presence of welldeveloped spot gas market. It is not possible in Asia due to the absence of active spot gas market.

5 Values of flexible supply vs. Gains from ST trade (as opposed to LT contract) This paper evaluates the gains from ST trading as opposed to LT contracting. An LNG producer needs to choose between 1. Supply to Japan under LT contract (forward LNG price is stochastic as it is linked to spot crude oil price), 2. Supply to either Japan or US through ST trading. Value differs from Hayes (2006) due to uncertainty in base gas price. This paper considers, for each of ST trading and LT contracting: (1) w/o cross-hedging with futures contracts for the related energy, (2) w/ cross-hedging.

6 Gains from ST trade vs. LT contract (1) No cross-hedging Sells forward: Sells spot: where TR i TR F = F LNG Q TR S = P max Q = revenue (i = F if forward, i = S if spot), F LNG = forward price (linked to the spot price of crude oil, P CO ), P max = max{p LNG,j }, P LNG,j = spot gas prices in market j, net of transportation cost (j = 1 represents the market to which LNG is supplied under LT contract), Q = production capacity

7 Gains from ST trade vs. LT contract (2) With cross-hedging Sells forward: Sells spot: where TR FC = F LNG Q (f CO,1 + f CO,0 ) x TR SC = P max Q (f CO,1 + f CO,0 ) x f CO,i = futures price of crude oil at period i (i = 0 for time of forward contracting, i = 1 for time of physical delivery of LNG) x = futures position

8 Gains from ST trade vs. LT contract (2) With cross-hedging (cont.) Note 1: Consider LNG pricing formula, F LNG = a + b P CO, Taking x = bq units in futures position, Sells forward: TR FC = b(p CO f CO,1 )Q + (a + bf CO,0 )Q Sells spot: TR SC = (P max a bf CO,1 )Q + (a + bf CO,0 )Q

9 Gains from ST trade vs. LT contract (2) With cross-hedging (cont.) Note 1: Consider LNG pricing formula, F LNG = a + b P CO, Taking x = bq units in futures position, Sells forward: TR FC = b(p CO f CO,1 )Q + (a + bf CO,0 )Q =0 if futures market is efficient (P CO = f CO,1 ) Sells spot: TR SC = (P max a bf CO,1 )Q + (a + bf CO,0 )Q = (P max a bp CO )Q 0

10 Expected return and variance of return Expected return Without futures With futures Forward (a + be[pco])q (a + be[pco])q Spot E[Pmax]Q E[PLNG,1]Q E[Pmax]Q E[PLNG,1]Q Variance of return Without futures With futures Forward (b Q) 2 V[P CO ] (b Q) 2 V[P CO f CO,1 ] = 0 Spot V[P max ]Q 2 b 2 V[P CO ]Q 2? V[P max b f CO,1 ]Q 2 >0

11 Gains from ST trade vs. LT contract (2) With cross-hedging (cont.) Note 2: Revenue from short-term trading can be decomposed into, (1) TR S = (P max P LNG,1 )Q + P LNG,1 Q (2) TR SC = (P max P LNG,1 )Q + (P LNG,1 a b f CO,1 )Q + (a + b f CO,0 )Q

12 Gains from ST trade vs. LT contract (2) With cross-hedging (cont.) Note 2: Revenue from short-term trading can be decomposed into, (1) TR S = (P max P LNG,1 )Q + P LNG,1 Q (2) TR SC = (P max P LNG,1 )Q + (P LNG,1 a b f CO,1 )Q + (a + b f CO,0 )Q Gains from spatial arbitrage Gains from supplying into market 1 through ST trading as opposed to forward contracting Revenue from futures trading

13 Gains from ST trade vs. LT contract (2) With cross-hedging (cont.) Note 2: Revenue from short-term trading can be decomposed into, (1) TR S = (P max P LNG,1 )Q + P LNG,1 Q (2) TR SC = (P max P LNG,1 )Q + (P LNG,1 a b f CO,1 )Q + (a + b f CO,0 )Q Gains from spatial arbitrage Negatively correlated = Risk reduction from multiple destination Gains from supplying into market 1 through ST trading as opposed to forward contracting Revenue from futures trading

14 Cost and benefit from ST trade Main benefit: Gains from spatial arbitrage: (P max P LNG,1 )Q 0. Main cost: A greater price risk in ST trading than under LT forward contracting. Cost or benefit? The option to choose its supply destination from multiple regional markets can either increase or decrease the spot price risk.

15 Valuing risks and returns to ST trading Approach 2 steps: 1. Estimate models of commodity price dynamics for US NG and Brent CO. 2. Monte Carlo simulation: Generate simulated price series from the estimated model and obtain the distribution of revenue to an LNG producer under (1) LT contracting, (2) ST trading.

16 Model of commodity price dynamics For each of HH natural gas (i =1) and Brent crude oil (i = 2), the spot price, P i,t, is modeled by, P i,t = f i (t) + g i (t) e i,t e i,t = ρ i e i,t 1 + u i,t The innovation, u i,t, is specified by GARCH(1,1) process with CCC model of Bollerslev (1990). For u t = (u 1,t, u 2,t ), E[u t u t I t 1 ] = H t = D t R D t D t = diag( h h ii,t = (1 ρ i 2 )(1 γ i,1 γ i,2 )+ γ i,1 h ii,t 1 + γ i,2 (u i,t 1 ) 2 where R = (r ij ) is a symmetric PD matrix with r ii = 1 and r ij = ρ ji for i j., h 1 / 2 1 / 2 11, t 22, t )

17 Model of commodity price dynamics (cont.) Two functions representing seasonality in mean price and variance of deviation from mean price are specified by non-parametric functions, f ( t) i = a i,0 + a i,1 t + 5 j= 1 a i,2 j sin 2jπτ t a 2jπτ 365 t i,2 j+ 1 cos g ( t) i = b i,0 + 5 j= 1 b i,2 j sin 2jπτ t + bi 365,2 j 2jπτ t + 1 cos 365 The model is estimated by ML with daily spot prices of HH natural gas and Brent crude oil price from Nov 1, 1993 to March 23, 2007.

18 Estimation results The estimated function, f i (t), accounts for 69.4 % (71.7%) of daily price variations for natural gas (crude oil). AR(1) coefficient, ρ, is (0.993) for natural gas (crude oil), deviation from seasonal mean is highly persistent. The sum of the two GARCH coefficients is (0.952) for natural gas (crude oil) conditional variance is also highly persistent. Correlation of two prices is before controlling for seasonality in mean price. It drops to once seasonality is controlled for, and further down to if autoregressive process is controlled for in addition to seasonality in mean and variance.

19 Estimate of seasonal mean price, f i (t) Crude Oil Natural gas Crude oil Natural gas Day in year 1.8

20 Estimate of seasonality in variance, g i (t) Crude Oil Natural gas Crude oil Natural gas Day in year 0.0

21 Simulation Design 1 DGP of three prices The simulated data on the HH natural gas and the Brent CO spot price are generated from the estimated model. The simulated price series in a hypothetical spot LNG market in Japan is generated from the following process, 1 ln P LNG, t = ln FLNG, t ln(1+ σ LNG ) e LNG, t e + u LNG, t = ρ LNG e LNG, t 1 where F LNG,t =.15 P JCC,t. t u 2 2 N(0, ln(1 +σ )(1 ρ Values of two parameters are set as σ LNG =.20 and ρ LNG =.90 in the base case. Sensitivity is examined for σ LNG = [0,.50] and ρ LNG = [0,.90]. CO futures price is obtained as conditional expectation of spot CO price. t ~ LNG LNG ))

22 Simulation Design 2 Revenue calculation Annual revenue is calculated for a hypothetical LNG producer located in the North West Shelf (NWS), Australia. The producer has finalized its decision to invest on an LNG production project. On the sales side, the producer can choose: 1. LT supply/purchase arrangement deliver to Japan 2. ST trading supply to either the US spot market or a hypothetical spot market in Japan. For each of the two trading strategies, the firm either do or do not cross-hedge with crude oil futures. Arbitraging decision as well as processing and shipping of LNG is assumed instantaneous. One year of daily data in the three prices and the annual revenue to the producer are obtained.

23 Simulation Results (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Symbol F LNG P MAX P LNG P MAX F LNG Description Forward LNG price Maximum spot natural gas prices of US and Japanese Market LNG price in hypotherical Japanese spot market Annual gain from spot trading with spatial arbitrage as opposed to forward trading flng ρ LNG = ρ LNG = ρ LNG = Mean σ LNG = SD σ LNG =

24 Simulation Results (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Symbol F LNG P MAX P LNG P MAX F LNG Description Forward LNG price Maximum spot natural gas prices of US and Japanese Market LNG price in hypotherical Japanese spot market Annual gain from spot trading with spatial arbitrage as opposed to forward trading flng ρ LNG = ρ LNG = ρ LNG = Mean σ LNG = SD + 5.9% % σ LNG =

25 Simulation Results (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Symbol F LNG P MAX P LNG P MAX F LNG Description Forward LNG price Maximum spot natural gas prices of US and Japanese Market LNG price in hypotherical Japanese spot market Annual gain from spot trading with spatial arbitrage as opposed to forward trading flng ρ LNG = ρ LNG = ρ LNG = Mean All roughly identical to F LNG σ LNG = SD σ LNG =

26 Simulation Results (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Symbol F LNG P MAX P LNG P MAX F LNG Description Forward LNG price Maximum spot natural gas prices of US and Japanese Market LNG price in hypotherical Japanese spot market Annual gain from spot trading with spatial arbitrage as opposed to forward trading flng ρ LNG = ρ LNG = ρ LNG = Mean σ LNG = SD σ LNG = SD[P LNG ] < SD[P max ] when σ LNG and ρ LNG are high

27 Simulation Results (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Symbol F LNG P MAX P LNG P MAX F LNG Description Forward Maximum spot natural gas prices of LNG price in hypotherical Japanese Annual gain from spot trading with Mean LNG price forward trading flng ρ LNG = ρ LNG = ρ LNG = σ LNG = SD US and Japanese Market spot market σ LNG = spatial arbitrage as opposed to SD[P max P LNG ] < SD[P max ] substantially (roughly half in base case).

28 Monte Carlo distribution of firm s revenue Forward trading without cross-hedge (a) Forward trading without cross hedge (F LNG ) Mean = 3244 SD = 375 Skewness = Kurtosis = Revenue (US$/MBTu capacity year)

29 Monte Carlo distribution of firm s revenue Spot trading without cross-hedge (b) Spot trading with spatial arbitrage, no cross hedge More positively skewed (low risk of low revenue) Mean = 3436 SD = 403 Skewness = Kurtosis = Revenue (US$/MBTu capacity year)

30 Monte Carlo distribution of firm s revenue Spot trading with cross-hedge (c) Spot trading with spatial arbitrage, with cross hedge Even more positively skewed Mean = 3436 SD = 212 Skewness = Kurtosis = Revenue (US$/MBTu capacity year)

31 Frequency of trade to the US Sensitivity to the volatility of JP spot price (a) Sensitivity to the volatility of spot forward LNG price deviation in Japan % 50% % of time exported to US 40% 30% 20% 10% 0% Month

32 Conclusions Expected gain from spatial arbitrage through ST trading is about 6% of the expected revenue from forward contracting. ST trading associates with high revenue volatility (12%), yet only slightly higher than already volatile revenue from forward contracting. Cross-hedging with CO futures not only mitigates the forward price risk but also reduces the spot price risk volatility of revenue from ST trade is less than 7% of mean revenue. Revenue from ST trading is highly positively skewed it exceeds the revenue from forward sales 90% of time. Historical prices of the US NG and CO imply a positive return from the ST LNG trading with reasonably low risk. Whether LNG producers actually shift to ST trade depends on other factors, such as quantity risk and cost associated with finding trading partners they will remain high in the absence of well-developed gas markets in Asia.

33 Comparison with Hayes (2006) Values of flexible LNG supply (Hayes, 2006) V flex = (P max P LNG,1 )Q Gains from ST trading as opposed to LT contracting w/o cross-hedge: TR S TR F = (P max F LNG ) Q w/ cross-hedge: TR SC TR FC = {P max (a + b P CO )} Q Difference from Hayes = cost to LNG producers to receive arbitrage profit through ST trading, in the absence of well developed spot and futures natural gas markets in Asia. (1) TR S V flex = P LNG,1 Q (2) TR SC V flex = {P LNG,1 (a + b f CO,1 ) + (a + b f CO,0 )}Q

34 Simulated price series One realization (a) Crude oil Crude Oil $/bbl Day in year (b) Natural gas and LNG Henry Hub spot NG LNG Forward (FOB to Japan) LNG Spot (FOB basis to Japan, base case) $/MBTu Day in year

35 Simulated price series Descriptive stat. (1) (2) (3) (4) (5) (6) (7) (8) (9) Symbol P CO P HH F LNG P LNG P MAX Description Spot crude oil price Spot HH gas price Forward LNG price LNG price in hypotherical Japanese spot market Maximum spot natural gas prices of US and Japanese Market ρ LNG = ρ LNG = Mean σ LNG = SD σ LNG =

36 Frequency of trade to the US Sensitivity to the persistence of JP spot price (b) Sensitivity to the prsistence of spot forward LNG price deviation in Japan % 45% 40% % of time exported to US 35% 30% 25% 20% 15% 10% 5% 0% Month