How Far is Gas from becoming a Global Commodity?

• Author: Aguiar-Conraria, Luis

1. INTRODUCTION

This paper analyzes the North American, European, and Asian natural gas markets integration and its relation to oil markets. Although the literature on these topics is extensive, as far as we know, nobody has looked at this problem from a time-frequency perspective. To do that, we rely on multivariate wavelet analysis, which gives us a very natural framework to estimate how these relationships behave at different frequencies and how they evolve. As we will see, energy markets integration has changed, not only over time, but also across frequencies. We reach some results that would be difficult to disentangle with more traditional time-series methods.

Until 2000, most authors would consider 'crude oil as a world market while coal and natural gas belonged to geographically segmented markets'; see Bachmeier and Griffin (2006) and references therein. From that point on, the literature has evolved from a yes or no answer to a more nuanced view of market integration. In particular, some authors started to assess the degree of market integration.

Because we work with prices, we must rely on prices' behavior to operationalize the concept of integrated markets. If the markets are truly integrated, prices must be the same in the different regions, by the law of one price. In a global market, shocks are felt globally. An increase in the demand for gas in Russia should lead to a price increase everywhere. On the other hand, if markets are entirely segmented, only the Russian prices would react. In segmented markets, we expect prices to be independent. In that sense, we can use price synchronization in different regions as a proxy for market integration. The stronger the synchronization, the more integrated the markets are. In the context of the wavelet analysis that we employ, as we explain later, that amounts to say that coherency between prices is very high. We will take the oil market as the benchmark for a global market1 and show that the coherency between West Texas Intermediate (WTI) and Dubai oil prices is consistently close to one.

Several authors investigated the impact of decreasing transport costs of Liquefied Natural Gas (LNG). Neumann (2009) concluded that the increase in LNG trade had accelerated the integration of previously segmented markets in North America, Europe, and Asia. At about the same time, Aune et al. (2009) argued that gas markets integration would keep increasing, thanks to the LNG effect. This prediction was later confirmed by some authors, like Barnes and Bosworth (2015). However, Chiappini et al. (2019) concluded that, despite the increasing interdependence between European and American prices, gas markets were still not global. Oglend et al. (2020) provided one possible explanation. They argued that time commitments associated with inter-continental Liquefied Natural Gas trade increase other types of costs, weakening the ties between global natural gas markets.

Although not as many, some authors also investigated the impact of the shale gas revolution. It seems that the main contribution of shale gas was to separate the markets, not to integrate them. For example, Wakmatsu and Aruga (2013) concluded that the U.S. market had a one-side influence on the Japanese market before 2005, but, thanks to the shale gas revolution, that influence disappeared afterward. Aruga (2016) also concluded that the U.S. gas market became independent after the shale gas revolution. The price linkage between the U.S. and international gas markets became weaker than before.

Overall, there is ample evidence that gas markets are regionally very integrated. We can, therefore, treat them as a single market; e.g., Renou-Maissant (2012), Asche et al. (2013), Yorucu and Bahramian (2015), Bastianin et al. (2019), and Garaffa et al. (2019), regarding European markets, and Park et al. (2008) and Avalos et al. (2016), for North American markets. Nevertheless, they are not globally integrated. It is worth noting that most studies identified some degree of integration between the European and the Asian markets. It is the North American market that is mostly independent. E.g., Li et al. (2014) concluded that there is some convergence between European and Asian markets. However, they also pointed out that this is probably the result of the contract structure that links gas prices to the oil price. Siliverstovs et al. (2005) had reached a similar conclusion earlier: high natural gas market integration between the European and Japanese markets, but not with North America. Chai et al. (2019) even concluded that 'the price linkage relationship between the United States and European natural gas markets had gradually declined in recent years.'

Another strand of literature, which explains some of the described stylized facts, explores the relationship between oil and natural gas markets. Because oil and gas are substitutes, and the oil market is global, oil prices tie the gas markets together (Brown and Yucel, 2008) or, at least, coordinate gas prices across different regions (Brown and Yiicel, 2009). With very few exceptions, notably Batten et al. (2017), the consensus is that causality runs from oil to gas markets; see, e.g., Erdos (2012) or Geng et al. (2017). Most studies also concluded that this relation is more robust for Europe and Asia than North America (Geng et al., 2016; Zhang et al., 2018; Zhang and Ji, 2018). Lin and Li (2015) concluded that European and Japanese gas prices are co-integrated with Brent oil prices, but the U.S. gas price is decoupled from oil due to natural gas market liberalization and shale gas expansion. Additionally, they confirmed the results of other authors when they claimed that their results support the presence of price spillover from crude oil markets to natural gas markets, but not the reverse.

One controversial issue in the oil gas relationship is its stability. For example, while Ji et al. (2018) found a stable contemporaneous causal flow from crude oil to natural gas, Brigida (2014) only found a co-integration relationship once he allowed for shifts in the co-integrating vector. Ramberg and Parsons (2012) also concluded that the co-integrating relationship is not stable through time.

The paper proceeds as follows. Section 2 starts with a discussion showing why wavelet analysis is particularly well-suited to study market synchronization and energy markets, followed by a very brief description of the Continuous Wavelet Transform tools used in this study. We leave the technical details about these tools to an appendix, and in, this section, instead, we apply them to the oil markets, which will serve as a global market benchmark. In Section 3, we present our data, and Section 4 delivers our first results regarding gas market synchronization between North America, Asia, and Europe. In Section 5, we explore the gas-oil relationship and describe how it helps to explain the results of Section 4. Surprisingly, we uncover a long-run relationship between the regional gas markets. It started in the early 2000s and had not been revealed before, to the best of our knowledge. Section 6 concludes

(2.) METHODOLOGY AND THE DEFINITION OF THE OIL MARKET BENCHMARK

We use wavelet analysis to study market prices synchronization. We do not claim that wavelet analysis is better than other more traditional methods. We only argue that this technique is particularly appropriate to investigate this issue. It performs the estimation of the spectral characteristics of a time-series as a function of time, revealing how its different periodic components evolve, which is crucial, because market equilibria are a combination of features operating on different frequencies. Moreover, relations may be different for different frequencies, since different economic agents are concerned with different time-horizons. Some agents focus on short-run (high frequencies) movements and co-movements, while other agents are concerned with longer horizons (lower frequencies). It is entirely conceivable that, at high frequencies, markets may be independent, but they move together at low frequencies. For example, Nick and Thoenes (2014) found that, in the short-run, the German natural gas market is affected by local conditions, like temperature, storage, and supply shortfalls. In the long-run, oil and coal are key determinants. With a global oil market, and assuming that it impacts all gas markets, it is likely that regional gas markets synchronization looks very different in the short- and long-run. Additionally, causality relations need not be the same at different frequencies. It is possible that, at high frequencies, shocks in the gas markets have impacts in the oil markets but that, in the long-run (at lower frequencies), causality runs from oil to natural gas prices.

It is also a fact that energy price dynamics are firmly non-stationary with unit roots, volatility clustering, structural breaks, etc. Therefore, it is essential to use methods that do not require stationarity. Moreover, Kyrtsou et al. (2009) showed that several energy markets display consistent nonlinear dependencies. Thanks to its localized nature, wavelet analysis is particularly well-suited to study data with all these characteristics.

There are mainly two ways to apply wavelets to data. One uses the discrete wavelet transform (DWT) and the other the continuous wavelet transform (CWT), which is the technique used in this paper. With DWT, one decomposes a time-series into a sum of time series of different frequencies. It is similar to applying several bandpass filters to isolate the behavior of a variable for each band. Yogo (2008) showed that multi-resolution with wavelet analysis (very quickly performed with DWT) allows for the decomposition of a variable into a trend, cycles of different periodicities, and noise, in a way very similar to bandpass filtering. Yang (2019) combined this technique with the connectedness measure proposed by...