Can the Shanghai LNG Price Index indicate Chinese market? An econometric investigation using price discovery theory

Abstract

China became the world’s second largest liquefied natural gas (LNG) importer in 2018 but has faced extremely high import costs due to a lack of bargaining power. Assessments of the Shanghai LNG Price Index, first released in 2015, are vital for improving the understanding of these cost dynamics. This paper, using the LNG price index data from the Shanghai Petroleum and Gas Exchange (SHPGX) coupled with domestic and international LNG prices from July 1, 2015 to December 31, 2018, estimates several econometric models to evaluate the long-term and short-term equilibriums of the Shanghai LNG Price Index, the responses to market information shocks and the leading or lagging relationships with LNG and alternative energy prices from other agencies. The results show that the LNG price index of the SHPGX has already exhibited a long-term equilibrium and short-term adjustment mechanisms to reflect the average price level and market movements, but the market information transparency and price discovery efficiency of the index are still inadequate. China’s LNG market is still relatively independent of other natural gas markets, and marketization reforms are under way in China. The influence of the SHPGX LNG price index on the trading decisions of market participants is expected to improve with further development of China’s LNG reforms, the formation of a natural gas entry-exit system, and the increasing liquidity of the hub.

Keywords

liquefied natural gas / price index / Shanghai Petroleum and Gas Exchange / price discovery / market reforms

Introduction

Within the Paris Agreement adopted at the United Nations Conference on Climate Change in 2015, China’s commitments mainly include peaking its carbon emissions by approximately 2030 and reducing its CO₂ emissions per unit of GDP by 60%–65% from the 2005 level [¹]. Natural gas, as the cleanest fossil fuel, has gained increasing recognition for its importance in achieving carbon emission targets and improving domestic environmental quality [²–⁴]. China increased its natural gas consumption from 186.9 billion m³ in 2014 to 280.3 billion m³ in 2018, while its domestic gas production increased from 130.2 to 160.3 billion m³ during the same period [⁶]. The rapidly expanding gap between supply and demand implies there is a great increase in natural gas imports in China. Due to restrictions on pipeline construction, China has built a significant number of LNG terminals after 2014. From 2015 to 2018, China’s monthly liquefied natural gas (LNG) import volume increased from 1.6 to 4.5 million tons per month; its share of global LNG imports rose from approximately 8% to 17% [⁷], and China became the second largest LNG importer in the world, with an external dependence of 44.7% for natural gas in 2018 [⁸]. In 2018, China also suffered an 82% year-on-year increase in LNG import costs, while the import volume increased by only 41% [⁹].

To ensure the security of its natural gas supply, the Chinese government urgently needs to change the market structure and the pricing mechanism of LNG. In March 2015, the first national transaction center for natural gas, called the Shanghai Petroleum and Gas Exchange (SHPGX), was incorporated to guarantee the security of China’s natural gas supply, improve its bargaining power in the natural gas market, and gradually establish a regional benchmark price [¹⁰]. This move is also considered as a strategic maneuver by the Chinese government to liberalize its monopolistic gas market [¹¹]. Under the guidance of the going-out strategy, Chinese oil companies have signed long-term contracts mainly with suppliers from Australia, South-east Asia, and the Middle East [¹²]. The source countries for Chinese LNG imports are highly concentrated. Before the 2010s, more than 80% of China’s LNG imports originated from Australia. In 2018, as imports expanded, the number of China’s main LNG suppliers reached 18, among which Australia and Qatar accounted for 43.7% and 17.3% of total imports by volume, respectively [⁷].

Before the establishment of the SHPGX, the daily LNG price and benchmarks could be provided only by western price reporting agencies (PRAs) and some domestic reporters [¹³]. It is difficult to obtain detailed price information on long-term contracts because of confidentiality. Only a small share of LNG spot prices or short-term contract prices are based on the data from PRAs, which are also relatively opaque. Published information on the quantity and price of LNG imported into China can be acquired only via the official website of China’s General Administration of Customs. However, monthly reports on imports have a one-month lag. The Chinese daily domestic wholesale LNG prices are reported by domestic PRAs or commercial public platforms. The reliability of the information is mainly based on the goodwill of PRAs. The PRA price information is widely accepted by the market and taken as an important trading reference by market participants. Moreover, PRAs provide trading platforms, negotiation services, and financing services for their paid members, but their information is not available to nonmembers and the public.

After the foundation of the SHPGX, daily LNG prices became publicly available in China. The SHPGX platform provides two means of exchange: quote trading and auction trading. The principal LNG price indices released by the SHPGX include the LNG import terminal price, the Chinese national LNG ex-factory price index, the South China LNG transaction price index, and the Chinese subprovincial ex-factory price index, all of which are derived from quote trading. State-owned oil companies are encouraged to report the LNG transaction information from both their receiving stations and liquefaction plants to the SHPGX. Meanwhile, market participants are also encouraged to negotiate and trade on the SHPGX platform. Thus, based on the information from the platform, the SHPGX can formulate day-ahead price indices for its reports. Another important type of trading, auction trading, is infrequently held by the SHPGX with major state oil companies for the buying and selling of both pipeline natural gas (PNG) and LNG. Nevertheless, the details of auction transactions are not released publicly.

As the first LNG hub price index in China, the SHPGX price has attracted growing public attention. This paper intends to assess the LNG price index, and describe and analyze the LNG price since it began to be released upon the establishment of the SHPGX in China. Moreover, it compares the SHPGX hub LNG price index with international LNG market prices and the price data from PRAs. Furthermore, it conducts an empirical analysis of delivery orders and the responsiveness to market information between the SHPGX LNG price index and other price indices to assess the market features of the LNG price index, thereby clarifying the market role of the SHPGX.

Literature review

The price discovery theory is an important tool for studying the degree of interaction and uniformity of prices in multiple markets. Its applications lie in differentiating the dominant market from satellite markets to verify that the dominant market plays a leading role in the discovery of potentially permanent prices of commodities, while the satellite market reflects the basic value of assets or commodities, as originally proposed by Garbade and Silber [¹⁴]. This theory is also applied in adopting econometric methods to test the transmission of market information in all market segments and using system models to analyze the long-term and short-term effects of information transmission. The main methodologies underlying this theory are methods based on simple linear regression [¹⁵–¹⁷], and the systematic cointegration analysis methods (based on linear and nonlinear cointegration; and combined with causality testing of long-term stable correlations of the time series and lead-lag relationships between variables [¹⁸–²¹], and three kinds of information sharing models, including the basic information sharing model [²²], its subsequent evolution into the generalized information sharing model [²³], and the permanent-temporary sharing model [²⁴]). The price discovery theory has been gradually expanded to various fields from its initial application to bulk agricultural products, and tends to use monthly, daily, or even higher-frequency trading data for analysis to reflect more detailed market fluctuations.

The price discovery theory was first used in the international energy market for oil and gas and extensively applied in empirical research on these sectors, such that for international commodities such as crude oil, fuel oil, and gasoline, the spot price and the price of four futures contracts can be generally given at the same time. Price discovery research always analyzes these 5 prices and judges their lead or lag relationships with market information transmission to ensure that the oil futures market and spot market are dominant [¹⁷,²⁵–²⁷]. With the development of the natural gas market, the price discovery theory can also be extended to the empirical study of price discovery in this market. The literature mainly focuses on the study of the “early or late” information transmission relationship among different natural gas spot prices [²⁸], between spot prices and futures prices [²⁹,³⁰], and between prices of natural gas and alternative energy sources such as oil or shale gas [³¹,³²].

The current literature mostly centers on the empirical analysis of information transmission prior to the existence of futures and spot prices in a complete futures pricing system and multimarket price transmission based on market liberalization. For emerging hubs, the literature mainly focuses on qualitative analysis. The early construction of hubs and hub-based gas pricing in Europe have been analyzed in depth by Stern and Rogers [³³,³⁴], and the development of European hubs is constantly tracked and indexed [³⁵–⁴⁰]. The available emerging hub literature tends to focus mainly on the analysis of the European and American gas markets. Compared with studies on European hubs [⁴¹], only a few empirical studies deal with Asian LNG price discovery [¹¹,⁴²,⁴³] because both gas market liberalization and gas hubs are in their nascent stages and there still remain many obstacles to the building of an Asian gas benchmark price [⁴⁴]. In some studies, Shanghai hub prices has be taken into account only as a future scenario [¹¹].

This paper contributes to filling this gap in regards to the Chinese LNG market. Although the LNG price indices of the SHPGX are not perfect, the data volume is sufficient enough to conduct an empirical analysis of the SHPGX using the price discovery theory to test the price responsiveness in the Chinese market. By applying the existing econometric methods of the price discovery theory, it is possible to at least analyze whether the SHPGX LNG price index is related to other market prices and whether there is any adjustment mechanism to test its market integration level. Causality testing is used to determine the lead-lag of LNG price indices relative to similar price indices. In addition, the response of the indices to market information shocks is also studied. The above work can provide an overall description of the price discovery ability of the indices of the SHPGX in the LNG market, lay a foundation for future research, and provide references for further research on the development of both SHPGX and other Asian gas hubs.

Methodology

The traditional econometric analysis methods of the price discovery theory are mainly applied in this paper. Since there is no LNG futures benchmark price in the Asian market, the LNG index of the SHPGX can be studied only with other daily market data available. If the LNG price index of the SHPGX has a long-term cointegration relationship with domestic LNG prices, international LNG prices and alternative energy prices, and a short-term adjustment mechanism exsits, the price index is basically successful. If the index performs well in the lead-lag and response-to-stimulus tests using other data, the index displays market efficiency. To this end, this paper presents three hypotheses to be tested.

Hypothesis 1: If the hub price reflects market information in a gas spot market where no benchmark price is given by the futures market, there should be at least a long-term equilibrium relationship and a short-term adjustment mechanism in the data. If there is no response at all, it indicates that the trading center does not fulfill an independent market pricing function.

Hypothesis 2: If an efficient hub is developed in a gas spot market without benchmarks from the futures market and there is no significant difference in pricing methodology, the hub and PRA price indices at the same prices can be issued at the same time, with neither price leading or lagging, or the PRA price index will lead.

Hypothesis 3: Alternative energy prices will respond faster than natural gas prices to the same market changes if alternative energy markets are more liberalized.

Econometric models

Cointegration and vector error correction model

The cointegration and VECM framework is used in this paper, which is suitable for evaluating unstable time series without too much loss of data information, to test the long-term equilibrium and short-term adjustment mechanism and assess Hypothesis 1. The cointegration model is appropriate for nonstationary time series and can test whether there is a long-term equilibrium relationship between variables [⁴⁵]. The premise of the cointegration test is that each data series is of the same order of integration [²⁹,³⁰].

If the data have a long-term cointegration relationship, the error correction term of the cointegration model can be used as an explanatory variable and combined with other variables to establish a VECM for the analysis of short-term mediation mechanisms.

The procedures of the cointegration test are as follows:

In the first step, regression of one variable on another is performed for the series X_t and Y_t of the same order.

(1)Y_t=\alpha +\beta X_t+{\epsilon }_t,

where X_t and Y_t are the values of the original series of X and Y in the time of t, and a and b are regression coefficients.

The residual terms of the model for X_t and Y_t are expressed as

(2){\epsilon }_t=Y_t-\alpha -\beta X_t,

where e_t is the estimated value of the model residual.

The second step is to perform a series stability test, such as an ADF test, a Phillips-Perron (PP) test, or a Kwiatkowski-Phillips-Schmidt-Shin (KPSS) test, for e_t in Eq. (1). If the results show that e_t is a stationary series, it can be concluded that X_t and Y_t have a cointegration relationship, and Eq. (1) is the cointegration regression equation.

Vector autoregression model

A vector autoregression (VAR) model is constructed, and impulse response analysis and variance decomposition are conducted to test Hypothesis 2. As an unstructured model, a VAR model can construct a dynamic equilibrium system using the actual economic data instead of the economic theory to judge the dynamic structure of the economic system, but it cannot establish a direct causal relationship between variables. Predictions based on VAR models may be successful in studies without structural constraints (i.e., government intervention) but are unlikely to be reliable when structural constraints are present. The impulse response analysis and variance decomposition of the VAR model can reveal the contribution of each impact to the change in the endogenous variables in the system [⁴⁶].

The p-order vector autoregression model, which is also the VAR(p) model, can be written as

Y_t=c+A_{\mathrm{1}}{Y}_{t-\mathrm{1}}+A_{\mathrm{2}}{Y}_{t-\mathrm{2}}+\cdots +A_p{Y}_{t-p}+e_t,where Y_t is the value of the original series of Y in the time of t, Y_t–i are the values in the lag period i of the original series of Y, c is an n×1 constant vector, A_i is an n×n matrix (i=1,2,…,p), n is the number of endogenous variables, and e_t is an error vector that satisfies the following requirements: the mean value of the error term is 0, the covariance matrix of the error term is an n×n positive definite matrix, and there is no autocorrelation of the error term.

Granger causality test

The Granger causality test is used to test Hypothesis 3 by analyzing the comparative lead-lag response to market information of different variables. In 1969, Granger proposed a two-variable VAR model test for stationary series (also known as the pairwise Granger causality test) to investigate the time lead-lag relationship between the two variables [⁴⁶]. If a multivariate VAR model is established, one can use the output for each endogenous variable from the VAR model equations to assess if one lags behind and determines all the endogenous lag variables of significance and joint significance (also known as VAR Granger causality/block exogeneity Wald tests).

Under the condition that the same lag order is chosen, the two Granger test methods are comparable [⁴⁵]. Although the Granger causality test will make mistakes if important variables are missing from the VAR model [⁴⁷], it is still widely used in the absence of other methods to determine sequencing.

The Granger causality test requires the estimation of the regression model expressed in Eqs. (4) and (5).

(4)Y_t={\displaystyle \sum _{i=\mathrm{1}}^m{\alpha }_i{X}_{t-i}}+{\displaystyle \sum _{i=\mathrm{1}}^m{\beta }_i{Y}_{t-i}}+{\mu }_{\mathrm{1}},

(5)X_t={\displaystyle \sum _{i=\mathrm{1}}^m{\lambda }_i{Y}_{t-i}}+{\displaystyle \sum _{i=\mathrm{1}}^m{\delta }_i{X}_{t-i}}+{\mu }_{\mathrm{2}},

where X_t and Y_t are the values of the original series of X and Y in the time of t; X_t–i and Y_t–i are the values in the lag period i of the original series of X and Y; {\alpha }_i,{\beta }_i,{\lambda }_i,{\delta }_i are regression coefficients; and μ₁ and μ₂ are the error terms.

The Granger test is completed by constructing F-statistics and using the F-test. If for the hypothesis that X is not the Granger cause of Y, that is, for the hypothesis that the parameters before the X lag term in Eq. (4) are all zero, regressions with and without the X lag term are performed. The residual sum of squares of the former is written as , and the residual sum of squares of the latter is written as. Then the F-statistic is calculated as

(6)F={\displaystyle \frac{({\mathrm{RSS}}_l-{\mathrm{RSS}}_f)/n}{{\mathrm{RSS}}_f/(N-\mathrm{2}n-\mathrm{1})}},

where n is the number of lagged terms of X, and N is the sample size.

If the calculated value of F is greater than the critical value F_{\alpha }(n,N-\mathrm{2}n-\mathrm{1}) of the F response distribution at a given significance level α, the null hypothesis is rejected, that is, X is recognized as the Granger cause of Y. If X and Y are single integrals of the first order and there is no cointegration, causality can be determined by a standard F-test of the first-order difference model.

Data

The LNG price variables and data used in this paper include: SHLNGT: LNG terminal quotations released by the SHPGX; OGLNGT: LNG terminal quotations reported by Chinese PRAs; PLJKMSD: Platts JKM SPOT DES (JKM SPOT); HHNGS: The spot price of natural gas in the Henry Hub (HH Spot) released by the US EIA; SHLNGWS: Chinese national ex-factory price index for LNG released by the SHPGX⁷; SHHNSE: LNG alternative energy price index released by the SHPGX⁷; OGLNGWS: Chinese national ex-factory price index for LNG released by Chinese PRAs⁸; and OGCBMWS: LNG alternative energy price (mainly ex-factory price of coalbed methane) reported by Chinese PRAs⁾⁸.

The high-frequency trading data in the database are converted into a daily price index using arithmetic means. All data units are also transformed from USD/MBtu to yuan/t by using the midpoint between the CNY and USD exchange rate published by China’s Foreign Exchange Management Center while one ton is equal to 52 MBtu. Table 1 and Fig. 1 show the basic characteristics of the data.

In the whole sample period, the average imported LNG price from Chinese terminals is close to the domestic ex-factory price, but the domestic price has a high volatility. In fact, due to the domestic gas shortage at the end of 2017, the difference between these two expanded rapidly, as has been mentioned in other studies [¹¹,⁴²,⁴³]. Compared with the gas prices in the American market and the DES LNG price in Japan or South Korea, the prices in the Chinese market are significantly higher. In the subsequent correlation analysis, an interesting finding is that the correlation of LNG prices is proportional to the distance to market (Fig. 2).

Results

Stationarity tests

The data stationarity test is the basis for the cointegration test, the VAR model, and the Granger test. Table 2 presents the main results of the stationarity tests using the augmented Dickey-Fuller (ADF), PP, and KPSS tests. All the results show that the original data are not stable, but the data are integrated of order one, which satisfies the basic condition for the cointegration test. The returns of the data are stationary and provide the basis to establish the VAR model and perform a Granger casualty test.

Results of cointegration test and error correction model

Equation (3) is used with the original data for a Johansen cointegration test, and the trace and max eigenvalue statistics are used to determine the existence of cointegration relationships. Table 3 shows that there exist at least 4 cointegration relationships at the 5% significance level. Table 4 shows the three cointegrations, taking the three main LNG indices of the SHPGX (SHLNGT, SHLNGWS and SHHNSE) as dependent variables.

The relative error correction models are established from the cointegration relationships. Table 5 lists the results. It is shown that the VECM equations pass the significance tests based on both their F-statistics and t-statistics. The unit roots are all in the unit circle, indicating the stability of the model.

The cointegration test results indicate that the LNG price given by the SHPGX has a stable proportional relationship or equilibrium relationship with other market prices in the long run. Furthermore, the error correction model also suggests that there is a short-term adjustment mechanism among various price variables in the model that reflects short-term market fluctuations. This shows that the SHPGX price responds to basic market situations and has a certain market usefulness.

Impulse response and variance decomposition results based on VAR model

Data stationarity was first tested before the establishment of the VAR model. This was done by calculating the rate of return after adjusting all data to realize a zero-order sheet (see Table 2). This paper uses R and the variable name on behalf of the rate of return, with the equation being R = logx– logx(-1), where x(-1) means a first-order lag term of x.

The optimal lag of the VAR model is 2 according to the AIC and LR tests, allowing for establishment of an unconstrained VAR (2) model. All unit roots are in the unit circle, and the model is stable. Only the SHLNGWS, the SHLNGT, and the SHHNSE data were analyzed in the VAR (2) for impulse responses to one standard deviation (S.D.) innovations and variance decomposition using Cholesky degrees of freedom (d.f.) adjusted factors since this study mainly analyzes LNG data released by the SHPGX. The results are presented in Fig. 3.

The results show that China’s LNG import price is insignificantly affected by other price factors, and other price variables have a negligible impact on its rate of change. This situation may arise due to the LNG import price itself being relatively stable or influenced by out-of-sample factors in this analysis. Given the fact that China’s LNG imports are mainly subject to long-term negotiated prices, the former is more likely to occur. Moreover, the market acuity of the domestic LNG wholesale prices is relatively higher, according to the results of the impulse response tests, and these prices reflect other market price variables in a more positive feedback loop. Meanwhile, it can be seen from the variance decomposition results that the South China LNG Trading Price Index contributes greatly to the rate of change. This indicates that the price is also subject to the long-term value of energy itself. Furthermore, the South China LNG Trading Price Index captures the actual value of LNG, which is relatively stable and in line with the original intention of setting the price.

The three major LNG-related indices of the SHPGX (SHLNGWS, SHLGNT, and SHHNSE) are taken as dependent variables for the Granger causality test, with the other variables excluded (Table 6). The results show that the tests for SHLGNT and SHHNSE can exclude other variables overall, which indicates that these indices are generally not correlated with other market data. The test for SHLNGWS cannot exclude other variables, indicating that it is highly correlated with other market data. Besides, in the inspection results, which are not listed in Table 6, the performance of the same type of prices from the SHPGX and PRAs, respectively, in the Granger test display large differences. Although the statistical analysis of the data from different agencies is not very diversified, the Granger test results for the PRA data are strongly linked with other market data, which means that they have a better ability to reflect market information than the data from the SHPGX. In addition, the Henry Hub spot gas price data and Platts LNG JKM SPOT DES index do not show Granger causalities with China’s LNG price in the test results, which indicates that China’s current LNG price is not significantly correlated with the US market and/or the Japan/South Korea spot market.

Discussion

Obstacles to the trade liquidity and data quality of the SHPGX

The SHPGX was created to secure China’s gas supply, increase China’s bargaining power in the gas market, and gradually establish regional benchmark prices [¹⁰]. Despite the fairly small proportion of LNG traded in the SHPGX, China had an advantage over Singapore as the most likely candidate to become an LNG trading hub in Asia due to its enormous volume of natural gas imports [⁴⁴]. Unfortunately, the empirical results show that the SHPGX LNG Price Index has achieved a certain marketability but there is still a lack of efficiency. This is mainly due to the lack of liquidity and data transparency in the LNG trade of the SHPGX.

China’s natural gas market is a state monopoly relying on an administrative licensing system. Government regulations entitled “Special provisions for spot market transactions” implemented in 2014 clearly stated that “no standardized contract transaction shall be conducted by means of a centralized transaction.” This clause effectively limited the churn rate of all Chinese gas spot platforms, including both the SHPGX and the CQPGX. Compared with the general churn rate (above 10) of successful European gas trading hubs such as the British National Balance Point (NBP) and the Dutch Title Transfer Facility (TTF) [⁴⁸], the SHPGX churn ratio (approximately equal to 1) is too low to attract enough traders. According to the data collected in this paper, the SHPGX’s unilateral LNG and PNG trade volume accounted for approximately 4.45% and 45%, respectively, of China’s total gas trade in 2018 (Fig. 4). This is an enormous barrier to attracting trading participants, especially third parties, to trade on the platform.

In addition to the liquidity problem, the SHPGX displays issues of data opacity. From the data and results obtained in this paper, several pieces of evidence supporting this can be found. First, SHPGX public release information is missing for unknown reasons in several cases. For example, the prices of the SHLNGT and OGLNGT are daily listing prices in LNG terminals. The data on these listings hosted by PRAs are more complete than those of the SHPGX, with 1014 observations for the former versus only 784 for the latter in the same sample period. Second, the SHLNGWS and OGLNGWS average ex-factory wholesale prices from LNG liquefaction plants in China are also broadcasted daily. For the same study period, the sample size of PRA observations is 996 broadcasts, which is far above the 513 broadcasts published by the SHPGX.

Difference between SHPGX prices and PRA prices

There are differences in sample size, mean value, and standard deviation when the SHPGX and PRAs broadcast their index of the unified target price. The two broadcast prices in the current stage of market development are essentially different. The SHPGX collects the quotation information reported by member units, organizes and edits it, and then publishes it daily. This is in fact a kind of market-history information release, reflecting the overall level and trend of the market.

The price data presented by PRAs serve as a settlement reference for the members of the agencies. Therefore, PRAs must ensure that the information is reliable and reflects the latest market information. Hence, PRAs collect the actual transaction data from their trading platform; verify the actual transaction information from buyers, sellers, shippers, and other market parties through their own market channels; and finally publish the data after editing it to ensure the reliability of the information. Although the information collected by PRAs may not be as extensive and comprehensive as that of the SHPGX, the information they collect is more reflective of market movements.

As an illustration, the severe supply shortage that occurred in China’s domestic natural gas market from November 2017 to April 2018 could be examined, which provides a good opportunity to test the degree of responsiveness of the domestic LNG price of the SHPGX. Except for SHHNSE, the other three prices all responded to the sudden shortage to different degrees (Fig. 5). As an LNG alternative with a small market scale and a flexible market response, the CBM energy price had the fastest and most obvious response, while the domestic LNG ex-factory price of the SHPGX had the weakest response, and of the SHPGX, its domestic LNG ex-factory price was higher than its own LNG factory price index.

It is true that the current Chinese PRAs are more responsive to the market, but at the moment, they rely mainly on the commercial prestige in China. State-owned oil and gas companies do not trade on PRA platforms. Therefore, their turnover and liquidity are major constraints on the growth of the companies. Nonetheless, in any energy market, PRAs are a market information force to be reckoned with Ref. [⁴¹].

Reasons for high LNG import prices

From the collected and unified daily listing prices of LNG terminals reported by the SHPGX and the PRAs in this paper, it is clearly observed that China’s LNG import price is the highest in the world, even higher than that of Japan and South Korea (Fig. 6). There are two possible reasons for this. First, China’s import sources are too concentrated (60% of LNG imports in 2018 came from Australia and Qatar) and are dominated by long-term associations. Second, China’s LNG terminal prices are listed for LNG imported for the domestic market. Therefore, taxes are included. The terminal price indices reported by the SHPGX and PRAs incorporate China’s zero-tariff policies and 9% value-added tax on LNG imports. This to some extent also raises the price for the receiving station.

Conclusions

This paper conducted an empirical study using an econometric model on the market price discovery ability of LNG price information released by the SHPGX from July 1, 2015 to December 31, 2018, and verified the three hypotheses proposed. Based on the findings, the following conclusions can be reached.

Stable cointegration relationships are established and VECM equations reveal that there exist systematic long-term and short-term adjustments between the SHPGX LNG indices and other prices.

The impulse response and variance decomposition analyses conducted in this paper indicate that the response speed of similar prices from PRAs is faster than that from SHPGX and that the overall market efficiency of PRAs is higher for various reasons.

Granger causality tests based on a VAR model show that prices from a more liberalized market respond faster than other prices. The LNG ex-factory price has a higher sensitivity than the LNG import price to impacts from the market. The SHHNSE indicator, calculated on the calorific value of alternative energy in South China, reflects the long-term market value of LNG. It does not display strong feedback of short-term fluctuations but has guiding and reference functions for assessing the long-term value of LNG.

The LNG indices of the SHPGX have begun to play a role in the market. As an emerging hub, the SHPGX is nascent but has started to release public one-day-ahead LNG price references for China and the world. It is still too early to consider its function to be the provision of a benchmark price; however, this paper serves as a starting point for empirical research regarding the price indices of the SHPGX. With the progress in the construction of the SHPGX, the research on the possibility of building an LNG benchmark in Asia, the evaluation of energy hub performance inside China, the building of an expanded price matrix, and even the liberalization of China’s natural gas market can be explored in the future.

References

[1] Dong C, Dong X, Jiang Q, Dong K, Liu G. What is the probability of achieving the carbon dioxide emission targets of the Paris Agreement? Evidence from the top ten emitters. Science of the Total Environment, 2018, 622–623: 1294–1303

[2] Toichi T. Development of the natural gas market in the Asia-Pacific region. ASEAN Economic Bulletin, 1989, 6(2): 149–156

[3] Dong C, Dong X, Gehman J, Lefsrud L. Using BP neural networks to prioritize risk management approaches for China’s unconventional shale gas industry. Sustainability, 2017, 9(6): 979

[4] Dong K, Sun R, Dong C, Li H, Zeng X, Ni G. Environmental Kuznets curve for PM_2.5 emissions in Beijing, China: what role can natural gas consumption play? Ecological Indicators, 2018, 93: 591–601

[5] Samah Zaki NAJI, Ammar Ali ABD. Sensitivity analysis of using diethanolamine instead of methyldiethanolamine solution for GASCO’s Habshan acid gases removal plant. Frontiers in Energy, 2019, 13(2): 317–324

[6] China’s National Development and Reform Commission. 2019–11–29, available at website of ndrc.gov.cn

[7] BP. BP statistical review of world energy. 2019–6-20, available at website of bp.com

[8] Wang Z, Jiang Q, Dong X, China’s Oil and Gas Industry Development Analysis and Outlook Report Blue Book (2018–2019). Beijing: China Petrochemical Press, 2019

[9] China’s General Administration of Customs. Statistical bulletin (6). 2019–1-14, available at website of customs.gov.cn

[10] Tong X, Zheng J, Fang B. Strategic analysis on establishing a natural gas trading hub in China. Natural Gas Industry B, 2014, 1(2): 210–220

[11] Shi X, Padinjare Variam H M. Gas and LNG trading hubs, hub indexation and destination flexibility in East Asia. Energy Policy, 2016, 96: 587–596

[12] Vivoda V. LNG import diversification and energy security in Asia. Energy Policy, 2019, 129: 967–974

[13] Johnson O. The Price Reporters: a Guide to PRAS and Commodity Benchmarks. New York: Routledge, 2017

[14] Garbade K D, Silber W L. Price movements and price discovery in futures and cash markets. Review of Economics and Statistics, 1983, 65(2): 289–297

[15] Oellermann C M, Brorsen B W, Farris P L. Price discovery for feeder cattle. Journal of Futures Markets, 1989, 9(2): 113–121

[16] Bopp A E, Lady G M. A comparison of petroleum futures versus spot prices as predictors of prices in the future. Energy Economics, 1991, 13(4): 274–282

[17] Schultz E, Swieringa J. Price discovery in European natural gas markets. Energy Policy, 2013, 61: 628–634

[18] Bekiros S D, Diks C G. The relationship between crude oil spot and futures prices: cointegration, linear and nonlinear causality. Energy Economics, 2008, 30(5): 2673–2685

[19] Yousefi S, Weinreich I, Reinarz D. Wavelet-based prediction of oil prices. Chaos, Solitons, and Fractals, 2005, 25(2): 265–275

[20] Chang C, Lee C. Do oil spot and futures prices move together? Energy Economics, 2015, 50: 379–390

[21] Schultz E, Swieringa J. Catalysts for price discovery in the European Union emissions trading system. Journal of Banking & Finance, 2014, 42(5): 112–122

[22] Tse Y, Booth G G. Information shares in international oil futures markets. International Review of Economics & Finance, 1997, 6(1): 49–56

[23] Lien D, Shrestha K. Hedging effectiveness comparisons: a note. International Review of Economics & Finance, 2008, 17(3): 391–396

[24] Zhao L T, Yan J L, Cheng L, Wang Y. Empirical study of the functional changes in price discovery in the Brent crude oil market. Energy Procedia, 2017, 142: 2917–2922

[25] Moosa I A, Silvapulle P. The price-volume relationship in the crude oil futures market some results based on linear and nonlinear causality testing. International Review of Economics & Finance, 2000, 9(1): 11–30

[26] Kaufmann R K, Ullman B. Oil prices, speculation, and fundamentals: interpreting causal relations among spot and futures prices. Energy Economics, 2009, 31(4): 550–558

[27] Alzahrani M, Masih M, Altiti O. Linear and nonlinear Granger causality between oil spot and futures prices: a wavelet based test. Journal of International Money and Finance, 2014, 48: 175–201

[28] Park H, Mjelde J W, Bessler D A. Price interactions and discovery among natural gas spot markets in North America. Energy Policy, 2008, 36(1): 290–302

[29] Asche F, Misund B, Sikveland M. The relationship between spot and contract gas prices in Europe. Energy Economics, 2013, 38: 212–217

[30] Ghoddusi H. Integration of physical and futures prices in the US natural gas market. Energy Economics, 2016, 56: 229–238

[31] Shrestha K. Price discovery in energy markets. Energy Economics, 2014, 45: 229–233

[32] Caporin M, Fontini F. The long-run oil-natural gas price relationship and the shale gas revolution. Energy Economics, 2017, 64: 511–519

[33] Stern J, Rogers H V. The Dynamics of a Liberalised European Gas Market. Oxford: Oxford Institute for Energy Studies, 2014

[34] Stern J, Rogers H V. The Transition to Hub-Based Gas Pricing in Continental Europe. Oxford: Oxford Institute for Energy Studies, 2011

[35] Heather P. European Traded Gas Hubs: a Decade of Change. Oxford: Oxford Institute for Energy Studies, 2019

[36] Heather P. “A Hub for Europe” The Iberian promise? Oxford: Oxford Institute for Energy Studies, 2019

[37] Heather P. The Evolution of European Traded Gas Hubs. Oxford: Oxford Institute for Energy Studies, 2015

[38] Heather P. Continental European Gas Hubs: Are they fit for purpose? Oxford: Oxford Institute for Energy Studies, 2012

[39] Heather P, Petrovich B. European Traded Gas Hubs: an Updated Analysis on Liquidity, Maturity and Barriers to Market Integration. Oxford: Oxford Institute for Energy Studies, 2017

[40] Petrovich B. European Gas Hubs Price Correlation: Barriers to Convergence? Oxford: Oxford Institute for Energy Studies, 2014

[41] Stern J. International gas pricing in Europe and Asia: a crisis of fundamentals. Energy Policy, 2014, 64: 43–48

[42] Zhang D Y, Shi M, Shi X P. Oil indexation, market fundamentals, and natural gas prices: an investigation of the Asian premium in natural gas trade. Energy Economics, 2018, 69: 33–41

[43] Wang T, Zhang D, Ji Q, Shi X. Market reforms and determinants of import natural gas prices in China. Energy, 2020, 196: 117105

[44] Kim Y. Obstacles to the creation of gas trading hubs and a price index in Northeast Asia. Geosystem Engineering, 2019, 22(2): 59–71

[45] Gao T. Econometric Analysis Methods and Modeling. Beijing: Tsinghua University Press, 2006

[46] Li Z, Pan W. Econometrics. 4th ed. Beijing: Higher Education Press, 2015

[47] Michael P M. Econometrics: A Modern Introduction. Beijing: Machinery Industry Press, 2009

[48] Miriello C, Polo M. The development of gas hubs in Europe. Energy Policy, 2015, 84: 177–190

RIGHTS & PERMISSIONS

Higher Education Press