However, the situation is completely different in the second time sub-period. The value of is quite low in both regressions (.226 and .177) that is an evidence of the low quality of the model. To be more precise, this means that only 26% of TTF prices variance is shared by Brent and BAFA prices while other 74% of variance is explained by some other factors (omitted values in the model in this case). The same situation is observed with NBP spot prices: only 17.7% of NBP prices variance is shared by Brent and BAFA prices and the majority of the variance (82.3%) is explained by other variables that were not included in the model. While examining significance levels for Brent and BAFA prices presented in the Table 11, it could be concluded that only b-coefficients of BAFA prices are different from 0 in both regressions of the second sub-period (Sig. = 0.01 for TTF prices as the dependent variable, Sig = 0.04 for NBP prices as the dependent variable that is < .05 significance level). B-coefficients for Brent prices in both regressions of this sub-period is highly probable equal to 0 and thus the change in NBP and TTF prices as dependent variables are almost not associated with changes in Brent prices as the predictor. According to the correlation matrix for the time period 01.01.2012 - 31.12.2014, the values of the correlation coefficient r for TTF, NBP and Brent prices are approaching to 0 (Appendixes C and D) that is a signal that Brent prices had a very slight influence on TTF and NBP hub-based prices. At the same time the values of r for NBP, TTF and BAFA prices fluctuate around the value of .45 that indicates a moderate influence of BAFA gas long term contract prices on TTF and NBP spot prices. Thus, it may be concluded that in the second time sub-period Brent prices had very slight influence on NBP and TTF prices, while the strength of relationship between BAFA prices and NBP and TTF hub-based prices was on the moderate level (although it has still decreased in comparison with the first sub-period). The correlation strength in the second sub-period between hub-based prices and BAFA long-term prices is considerably higher than the relationship strength between hub-based prices and oil prices.

After the proficient analysis carried out it is supposed that a reason for such prices decoupling and a decrease in influence of Brent oil prices on NBP and TTF gas spot prices is skyrocketed prices for oil at the beginning of 2011 (Figure 16).

Figure 16 The dynamics of Brent oil prices over the period 2008 - 2018

Source: Bloomberg Finance, 2018

This trend of quite high oil prices was not short-term one; it was rather long-term period of skyrocketed prices for oil, especially for Brent oil that lasted from the beginning of 2011 till the middle of 2014. As it was mentioned before while reviewing scientific literature, normally spot gas prices followed oil prices and their dynamics particularly in 2011-2013 when spot markets were not developed enough and had low liquidity levels. However, it is logically to make an assumption that if oil prices are too high and this trend is long-term one, it is inefficient and roughly speaking expensive for gas spot prices to follow this “mad” oil prices dynamics. Thus, it is supposed that high prices for Brent oil might be a reason for prices decoupling.

Since the correlation strength between hub-based prices and BAFA prices has also dropped (but not so much as the correlation strength between hub-based prices and Brent oil prices did) it could be concluded that skyrocketed oil prices with long-term trend influence the relationship between hub-based prices and BAFA prices in the same manner.

For the further analysis it was decided to neglect the second sub-period since the model quality is quite low. Furthermore, as it has been mentioned earlier in the Chapter more than a half of the variance of both dependent variables are explained by other variables that are not included in the analysis in accordance with the research gap and the research questions. Thus, it is logical to neglect this sub-period when analyzing the changes in correlation strength over the sub-periods.

According to the analysis results there is quite strong correlation between spot prices at TTF and NBP, Brent oil prices and BAFA long-term contract gas prices for the time period 31.12.2008 - 31.03.2018. Thus, the hypotheses H1, H2, H5 and H6 are not rejected.

As it was mentioned earlier in the Chapter, in order to assess the trend regarding the correlation strength between the dependent and the independent variables it has been decided to neglect the second-sub-period for the analysis since according to the regression results and especially value and significance values of predictors described earlier it was concluded that other variables excluded from the research (since they do not correspond to the research gap and the research questions formulated) affected hub-based prices for gas at the second sub-period. Thus, this sub-period is excluded from the analysis since it is aimed to analyze the correlation between hub-based prices, oil prices and BAFA prices and how the correlation strength have changed over time. Other variables that affect hub-based prices are not of the interest of this research paper.

Thus, after analyzing two regression results in the first sub-periods and two regression results in the third sub-period, it could be concluded that the correlation strength between both NBP and TTF hub-based prices and Brent oil prices has weakened over time (Table 13).

As it is observed from the Table 13, TTF prices are affected nowadays to the less extant in comparison with the first sub-period. The same conclusion could be made for NBP prices. Thus, the hypotheses H3 and H4 are not rejected.

Table 13 R-coefficients for the first and the second sub-periods

Table 14 Hypotheses results summary

Analyzing r-coefficients between hub-based prices and BAFA prices, it could be concluded that the correlation strength between TTF hub-based prices and BAFA prices has increased since 2008. However, there is no increase of the correlation strength in case of NBP prices and BAFA prices: the r-coefficient has slightly decreased since 2008 (however, the decrease is not significant). Thus, it could be concluded that the hypothesis H7 is not rejected, while the hypothesis H8 is rejected.

Thus, after the regression analysis it is concluded that there is still a correlation between hub-based prices, Brent prices and BAFA prices and the strength of the correlation is quite high even nowadays. Finally, it was discovered that the correlation strength has weakened since 2008 that indicates graduate price decoupling and European hub market development. One of the reasons of price decoupling is a sharp increase in oil prices that has a long-term trend that was observed in the second sub-period. In such situations hub-based gas prices do not correlate with Brent oil prices, but they still follow BAFA prices (correlation coefficients r in correlation matrix for the second sub-periods for TTF, NBP hub-based prices and BAFA prices are around .4 that indicates moderate correlation strength).

Despite the fact that the correlation between hub-based prices at two main European hubs - TTF and NBP - is weakening over time, nowadays hub-based prices are still affected by Brent oil prices and BAFA prices.

4. DISCUSSION&IMPLICATIONS

4.1 Results discussion

The regression models for the first time period is discussed firstly. According to the results, hypotheses H1, H2, H5 and H6 are not rejected. This gives the ground to claim that, firstly, hub-based prices at two highly liquid European hubs - TTF and NBP - correlates with Brent oil prices. The correlation is positive and the strength of the relationship between spot prices at TTF and NBP and Brent oil prices is a little bit high of the middle level magnitude (see Appendix A). Thus, the higher the Brent prices, the higher prices at TTF and NBP since hub-based prices follow the oil price dynamics

The correlation of hub-based prices with oil prices were previously discovered in researches carried out by Gazprom Export and particularly by S. Komlev (2016), by Brown and Yucel (2008), by Asche et. al. (2013), Bunn et al. (2017) etc. Particularly, the analysis carried out by S. Komlev (2016) has shown quite similar results with those got from the current analysis in terms of correlation coefficient measure that shows the strength of relationship between two variables. According to the results of correlation matrix for the time period 31.12.2008 - 31.03.2018 and hub-based prices as the dependent variables and Brent prices as the independent variable for both regression models, the correlation coefficient equals .758 for TTF prices and Brent prices and .724 for NBP prices and Brent prices (Appendix A). The coefficient got by S. Komlev after the analysis equals .67 and .64 correspondingly. Thus, the results received from the current analysis prove conclusions of the previous research about the quite strong correlation between hub-based prices and oil prices.

Secondly, according to the regression analysis results, hub-based prices at TTF and NBP also correlate with BAFA long-term contract prices for gas delivery. The correlation has positive direction and the strength of the relationship between the variables is quite high (Appendix A). Thus, the higher BAFA prices for gas delivery are, the higher spot prices at TTF and NBP are.

In this regard it should be emphasized that the findings in the current research proves the argument reported by S. Komlev (2016) in his research: it was stated that hub-based prices in the European gas market play “a secondary” role in relation to long-term contract prices for gas delivery that guide hub-based prices and create “a price ceiling” for them (BAFA prices in the current research has been chosen as a benchmark for LTC prices for gas). The figures of correlation coefficient shown in the research carried out by S. Komlev (2016) and of the current research paper are around the same level that also highlights the fact of the “secondary” nature of spot prices that follow LTC gas prices by their essence. Furthermore, according to the reports provided by ACER/CEER (2016), hub-based pricing formation depends on a number of factors, but long-term contract gas prices are the main parameter that should be considered and that have considerably high impact on spot prices dynamics.

Thus, the statements in previous research on the existence of correlation between spot prices at TTF and NBP gas hub, Brent oil prices and long-term contract gas prices (BAFA prices in this particular case) have been proven in the current research and the results are quite the same.

In regard of the sub-period regression analysis the hypotheses H3, H4, H7 are not rejected. There was logical rational that the correlation between hub-based prices and oil prices should weaken over a particular time period due to increasing liquidity of European hubs (Hulshof, 2015). As it was mentioned earlier, the markets with high liquidity level (in terms of high number of participant and large gas volume traded) are able to produce prices that are genuine signals of market changes, particularly demand and supply fluctuations (Stern & Rogers, 2014). In this regard it was supposed that the increase in the liquidity level of the two main European gas hubs - TTF and NBP - the correlation should weaken, so hub-based prices at highly liquid gas hubs should become an independent pricing mechanism and reflect real market changes. The assumption was proved by the results of the regression analysis.

As it was mentioned in the previous Chapter when the analysis results have been discussed, the correlation between hub-based prices at TTF and NBP, Brent oil prices and BAFA long-term contract prices was quite strong in the first sub-period 31.12.2008 - 31.03.2011. It is quite logical since in that time gas hubs in Europe were not developed enough interms of number of participant, traded volumes and the liquidity levels so that gas prices formed at gas hubs could act as independent pricing approach and have no impact from the side of oil and LTC prices. In that time the churn ratio that acts as a measure of the liquidity level in a market for both hubs were below 10 points that is a minimum figure for a market to be named as liquid one. Thus, there should be some some factors on the basis of which hub-based prices could be formed. These facts were oil prices and LTC prices for gas delivery. As it was mentioned in the Chapter 1, oil prices determine the gas prices formation from the very beginning of the gas market creation since prices for contracts for gas delivery (Groningen pricing mechanism with oil indexation) use pricing formula where oil indexation is used (Konoplyanik, 2010). For a new pricing paradigm - hub-based pricing - oil prices also play similar role: oil prices at the beginning of hub market development in Europe strongly determined hub-based prices. The majority of trade and gas delivery were under long-term contract conditions that has been without any doubt the dominant pricing approach that also affected hub-based prices (Konoplyanik, 2010; Komlev, 2016). Thus, in the first sub-period 31.12.2008 - 31.12.2011 the strong positive correlation between hub-based prices at TTF and NBP, Brent oil prices and BAFA long-term contract gas prices may be logically explained by low liquidity levels at European hubs that made European gas hub-based prices dependent on oil prices and LTC prices. In that sub-period hub-based prices could not act as independent pricing mechanism.

This conclusion is proved by the scientific literature overviewed in the previous Chapter. Brown and Yucel (2008) in the research highlight that spot prices strongly correlate with oil prices. Two years before Asche et. al. (2006) reported quite the same results: there was strong positive correlation observed between spot prices and oil prices. One of the reasons for such situation was as was explained earlier: low liquidity levels at European gas hubs did not allow spot price act as independent pricing paradigm since the spot trading in that time only started to evolve and the majority of gas trading was carried out under long-term contracts. The lack of market participants (in this particular case at spot trading platforms) and little traded volumes led to low market liquidity at hubs that has been denoted as a reason for such a strong correlation between prices (Heather, 2015).

According to the results received from the regression analysis, there was very slight correlation between hub-based prices at TTF and NBP gas hubs and oil prices in the second sub-period 01.01.2012 - 31.12.2014. As it was discussed earlier in the Chapter 3, it has been decided to neglect the second sub-period in the analysis. The quality of the model was quite low that indicated that there are other variables that affect hub-based prices in the second sub-period. However, these omitted variables are not included in the model since they do not correspond to the formulated research goal and research questions. In the current research paper different factors that have an influence on hub-based prices are not analyzed. The aim of the research paper is to investigate the existence of the correlation between hub-based prices, oil prices and LTC prices for gas delivery. Thus, only these variables have been chosen for the analysis. Thus, since in the second sub-periods hub-based prices dynamics could be explained by some other variables, this sub-period is excluded from the analysis of the trend regarding correlation strength between hub-based prices, oil and BAFA prices.

After the proficient analysis carried out it is supposed that a reason for such prices decoupling and a decrease in influence of Brent oil prices on NBP and TTF gas spot prices is skyrocketed prices for oil at the beginning of 2011.This trend of quite high oil prices was not short-term one; it was rather long-term period of skyrocketed prices for oil, especially for Brent oil that lasted from the beginning of 2011 till the middle of 2014. As it was mentioned before while reviewing scientific literature, normally spot gas prices followed oil prices and their dynamics particularly in 2011-2013 when spot markets were not developed enough and had low liquidity levels. However, it is logically to make an assumption that if oil prices are too high and this trend is long-term one, it is inefficient and roughly speaking expensive for gas spot prices to follow this “mad” oil prices dynamics. Thus, it is supposed that high prices for Brent oil might be a reason for prices decoupling.

Hulshof et al. (2015) reckons that the correlation between hub-based prices at TTF and oil prices is very slight and insignificant. In contrast, the scientist claims that supply and demand define spot prices. It is notably that the research was carried out by Hulshof et. al. using the regression analysis of data for the years 2011-2014 that covered the data analyzed in the second sub-period of the current research. Thus, the insignificant correlation between hub-based prices and oil prices in this time period was observed not only in the current research, but also by other scientists that might be a prove for results received for the second sub-period.

It should be mentioned that the correlation between hub-based prices at TTF and NBP and BAFA long-term contract prices for gas delivery still continued to exist in the second sub-period when oil prices were at the very high level in long-term perspective. This observations could also be considered as a prove for the S. Komlev's argument that hub-based prices play a role of “followers” of LTC prices and are secondary pricing signals in relation to LTC prices.

In the third sub-period the results of the regression analysis show that the correlation between hub-based prices, oil and LTC prices is high. There is a strong positive correlation between hub-based prices at TTF and NBP, oil prices and BAFA long-term contract prices. It should be highlighted that both TTF and NBP prices have stronger positive correlation with BAFA prices rather than with Brent prices. However, TTF prices are affected by both Brent oil and BAFA prices to higher degree in comparison with NBP prices. The general conclusion regarding correlation matrix of the third sub-period is that there is strong positive relationship between TTF and NBP prices and Brent and BAFA prices.

Thus, in the third sub-period the correlation between hub-based prices, Brent oil and BAFA prices still exists despite the fact of high liquidity levels at two main European hubs - TTF and NBP - in terms of traded volumes, market participants and churn ratios. However, for NBP prices the correlation with Brent prices is <.5 (to be precise r=.417) that indicates quite moderate but not strong correlation.

After comparison of the r-coefficients for the first and the third sub-period it could be concluded that the correlation strength between both NBP and TTF hub-based prices and Brent oil prices has weakened over time (Table 13). This could serve as an indicator of the graduate European hub-based gas market development since prices formed at hubs are becoming more and more independent in relation to oil prices. Thus, the European hub gas market is becoming mature and high liquidity levels at hubs in terms of market (hubs) participants, traded volumes (at hubs) and churn ratios facilitate this process.

However, currently NBP prices are affected by Brent oil prices to considerably less extent in comparison with TTF prices. The results in this case could be quite unexpected since nowadays TTF gas hub is more liquid in terms of churn ratio and number of hub (market) participant that according to the theoretical literature (Hulshof et al., 2015) indicate the high maturity degree of a market and high degree of price independence formed in this market. Following this logic, the correlation strength for TTF prices with Brent oil prices should be smaller in comparison with NBP prices which liquidity levels are nowadays lower than TTF ones. However, the results of the regression analysis are opposite: TTF prices are positively affected by Brent oil prices to higher degree than NBP prices are. One explanation could be that NBP gas hub has a longer process of historical development since it is the first gas hub in Europe that gave a start for hub-based gas trading in the continental Europe after the construction of Interconnector (Dickx et al., 2014). Nevertheless, the correlation strength between hub-based prices at TTF and NBP and Brent oil prices has decreased over the last ten years that is a signal of the European hub gas market and trading development. Thus, the hypotheses H3 and H4 are not rejected.

On the basis of the r-coefficients analysis for hub-based prices at both hubs and BAFA prices in the first and the third sub-periods the following conclusion could be made: the correlation between TTF prices and BAFA prices has increased over the last ten years. As it was mentioned earlier in the Chapter 1, BAFA pricing formula has a special “spot” component in line with “oil” component. Since every long-term contract for gas delivery has revision sessions when pricing formula is changed in order to be representative reflection of current market changes, this “spot” component in BAFA pricing formula is increasing after each revising sessions since hub gas market in Europe is gradually developing (Konoplyanik, 2010). Thus, the increase in correlation between TTF and BAFA prices is logically explained and the hypothesis H7 is not rejected. However, there is no increase in correlation between NBP and BAFA prices. Furthermore, there is very slight decrease in the correlation but overall could be considered as insignificant one (Table 13).One explanation could be that NBP has a longer development historical process since this is the first gas hub in Europe where hub-based trading take the beginning in the European market (Dickx et al., 2014). Thus, NBP prices could be considered as more independent even from BAFA prices, where “spot” component is presented. The hypothesis H8 is rejected.

It could be concluded that nowadays there is a strong positive correlation between TTF and Brent oil prices and a positive moderate correlation between NBP and Brent oil prices. Hub-based prices are still affected by oil prices and follow their dynamics. This fact could be an obstacle to make a statement that currently hub-based prices represent independent pricing mechanism in the European gas market. Furthermore, prices at both TTF and NBP gas hubs are positively affected by BAFA prices that supports the argument made by S. Komlev (2016) that hub-based prices play “a secondary” role in relation to LTC prices for gas delivery. Since hub-based prices follow the dynamics of Brent oil and BAFA long-term contract prices, they cannot purely represent independent pricing mechanism.

However, the correlation of hub-based prices with oil prices is weakening over time in line with the European hub trading development in terms of liquidity levels that nowadays are quite high at both TTF and NBP gas hubs. Thus, this decreasing trend is a signal that exporters should adapt their export strategy regarding LTCs and price formation in accordance with the European gas market development. Finally, the development of BAFA pricing formula and increased correlation between hub-based prices and BAFA prices proves the aforementioned argument that hub-based trading in Europe is constantly developing and influences price formation for LTCs also.

4.2 Scenario analysis

On the basis of the empirical findings made in the Chapter 3, it is possible to carry out scenario analysis in order to get deep insights of risks for Gazprom in the current market situation and potential strategy paths how to adjust the company's export strategy to volatile new market reality.

Below there is a 2x2 scenario matrix that describes four possible scenarios for Gazprom at the European gas market. On the y-axis Brent prices were put, while for the x-axis gas market supply was chosen. Such a selection allows to track the behavior of spot prices since there is a correlation between spot prices and Brent prices that was proven during the empirical analysis. Furthermore, the market supply reflects the amount of gas available at the market that also influences gas prices. BAFA prices cannot be chosen for the scenario analysis since there is oil indexation in every long-term contract for gas delivery that means that Brent and BAFA prices correlate that is a deterrent for carrying out a scenario analysis.

The scenario “No worries” implies low Brent prices and low market supply (e.g. there is scarcity of gas in the market, so demand exceeds gas supply). As it was discovered during the analysis in the Chapter 3, hub-based prices do not correlate with Brent prices in a situation when Brent prices are extremely high and this trend is long-term. In other situations Brent prices influence hub-based prices at TTF and NBP quite strongly. Moreover, as it was mentioned earlier, BAFA contracts and all long-term contracts for gas delivery contain oil indexation in their pricing formula. Thus, Brent prices and BAFA prices also correlate to some degree. In this regard, BAFA prices will be low following oil prices dynamics in this scenario. In the situation of low market supply one could think that hub-based prices may increase since the demand for gas will be high while there is a scarcity in gas supplied in the market. However, according to the scientific literature (Komlev, 2016), long-term contract prices for gas create “a ceiling” for spot prices. That means that hub-based prices cannot not go higher than LTC prices because of arbitrage opportunities. Thus, hub-based prices will also be at low level following oil and LTC prices dynamics. In such a scenario gas buyers will not stick only to one type of prices for gas (either hub-based prices at TTF and NBP or LTC prices): if they start to buy gas only at hub, this will lead to increase in spot prices, while LTC prices will stay the same. At some point it will be more costly to buy gas at hub-based prices and more profitable to buy gas at LTC prices. Thus, buyers in this scenario will balance their buying decisions between spot and LTC prices. Following this rational, for Gazprom it is not vitally necessary to adapt the strategy to market changes since buyers will buy gas at both spot and LTC prices.

In the Scenario “Lucky one” the Brent oil prices are high while gas supply in the market is low. Hub-based prices will decouple from oil prices since the analysis has shown that if oil prices are high spot prices do not follow their dynamics. BAFA prices will adapt to high oil prices since there is an oil indexation in the LTC pricing formula and will be also relatively high. One may think that in this situation it is more profitable to buy gas at hub-based prices since they do not follow oil prices level and are under BAFA prices level. However, there is low gas supply at the market. This factor drives hub-based prices up making them approaching BAFA prices. Moreover, there is quite strong correlation between spot and BAFA prices. Thus, spot and BAFA prices will be at the same level. This scenario is a better situation for Gazprom since as in the Scenario “No worries” buyers will not stick to only one type of gas prices for the same reasons explained above. Moreover, as LTC prices will be higher for the same gas amount delivery, it is a more profitable scenario for Gazprom; there is no dire need for the company to adapt its export strategy.

In the Scenario “Time for changes” high oil prices are accompanied by high gas supply (e.g. there is an abundant amount of gas supplied in the market). In such a situation hub-based prices do not correlate with high oil prices. In contrast, there is positive relationship between Brent oil prices and BAFA prices that makes BAFA prices approach to oil prices. Thus, BAFA prices with their oil indexation in their pricing formula are also relatively high in this scenario. It may be logical to suppose that spot prices will approach to the BAFA prices level. However, due to abundant amount of gas supplied to the market spot prices will be dumped in accordance with the economic law of supply and demand. In this situation it will be more profitable for buyers to acquire gas at hub-based prices since they are supposed to be lower than high expensive BAFA prices that have to follow oil dynamics and are not adapt to the market supply conditions quickly. Thus, in this scenario it is highly important and vital for Gazprom to adapt its export strategy to new market reality where it is more profitable to trade at spot markets. For instance, the company may increase the “spot component” in the LTC pricing formula so that LTC prices can be more flexible and responsive to market supply changes. Without such an adaptation there is a high risk for the company to lose a part of its market share in the European gas market and as a consequence to lose a considerable part of its export profit.

In the Scenario “Bad fortune” oil prices are at the low level while the gas supply in the market is high. In this situation spot prices will follow oil prices dynamics and will be at lower level in comparison with BAFA prices. Thus, all prices - hub-based prices, Brent and BAFA prices - will be at the low level. Taking in the account the abundant amount of gas supplied in the market, one may claim that spot prices will to same degree react to such high gas supply and will be dump even lower. Thus, hub-based prices in this scenario will become the lowest and the cheapest prices for gas. As a logical consequence buyers will in most cases prefer to buy gas at hub-based prices rather than at BAFA or other LTC prices since this will be the most profitable and the cheapest decision for them. In this scenario Gazprom will lose a crucial part of its market share since LTC prices are not compatible with spot prices even if there is currently “a spot component” in the pricing formula of Gazprom's LTCs. Thus, Gazprom should in this case adapt its export strategy by, for instance, further adaptation of its LTC pricing formula.

4.3 ManagerialImplications

In the research it has been proven that there is still a positive correlation between hub-based prices at TTF and NBP and Brent oil and BAFA prices. The existenc...