Journal of Energy Markets
ISSN:
1756-3607 (print)
1756-3615 (online)
Editor-in-chief: Derek W. Bunn
Need to know
- We model natural gas returns as a linear function of gas storage and weather variables, where the coefficients of this function may vary continuously over time.
- Our method allows for market participants to continuously try to improve their forecasts of market prices.
- Our model also allows us to calculate a time series of the proportion of volatility attributable to each factor.
- We show applications for hedging and derivatives trading in natural gas markets.
Abstract
The parameters linking the demand for and supply of natural gas to natural gas prices are likely to be dynamic throughout the year and over time. However, estimating a constant coefficient for a deseasoned gas storage or weather variable implicitly assumes that market participants react identically throughout the year (and over each year) to that variable. In this analysis, we model natural gas returns as a linear function of gas storage and weather variables, and we allow the coefficients of this function to vary continuously over time. This formulation takes into account that market participants continuously try to improve their forecasts of market prices, and this likely means that they continuously change the scale of their reaction to changes in underlying variables. We also use this model to calculate conditional natural gas volatility and the proportion of volatility attributable to each factor. We find that natural gas return volatility is higher in the winter, and we show that this increase is due to weather and natural gas storage. We provide time series estimates of the changing proportion of volatility attributable to each factor, which is useful for hedging and derivatives trading in natural gas markets.
Introduction
1 Introduction
The amount of natural gas in storage reflects the ready supply of this commodity, and it is therefore an important factor in pricing natural gas. Often, when estimating a statistical model of natural gas prices, an analyst will estimate a static coefficient for a deseasoned gas storage or weather variable. Doing so implicitly assumes that market participants react identically throughout the year (and over each year) to that variable. Further, it assumes that market participants find the variable no more meaningful in winter, and that they do not adapt to what has occurred in the market. These are unrealistic assumptions of economic behavior. Market participants continuously try to improve their forecasts of market prices, and this likely means they continuously change the scale of their reaction to changes in underlying variables. Moreover, it is likely that natural gas uncertainty is not simply attributable to the regression error term, but also due to changes in how market participants link prices to storage and weather (the coefficients) and the uncertainty in these parameter estimates.
In this analysis, we will model natural gas price returns as a linear function of gas storage and weather variables, and we will allow the coefficients of this function to vary continuously over time. This may be referred to as a time-varying-parameter (TVP) model, and it can be estimated using the Kalman filter. The TVP model allows market participants to adapt their reactions to information contained in variables and affords us an estimate of conditional heteroscedasticity due to both parameter uncertainty and a standard error term.11 1 Note that an autoregressive conditional heteroscedasticity (ARCH) framework does not specify the source of the conditional heteroscedasticity. We will use this to estimate a time series of the proportion of the total volatility that is attributable to each independent variable, using the parameter uncertainty estimates made available by the Kalman filter. Our results are useful for hedging and derivatives trading in natural gas markets.
Our procedure differs from previous work on volatility spillovers (Diebold and Yilmaz 2009, 2012) in several important ways. First, they consider how total volatility in one variable spills over into another, whereas we measure the effect of volatility induced by uncertainty in the parameters of the function linking the variables. Second, to estimate a time series of volatility spillovers, they must use a rolling window approach. Because our method is based on the Kalman filter, time-varying estimates of parameter uncertainty are built into the model.
Our paper builds on previous analyses of fundamental drivers of volatility in natural gas markets. In particular, Mu (2007) found evidence that weather significantly affects the conditional volatility of natural gas futures returns. The analysis was conducted in a multivariate generalized autoregressive conditional heteroscedasticity (GARCH) framework, which, unlike our Kalman filter approach, does not allow the estimated parameters in the variance equation to vary over time. Geman and Ohana (2009) found a negative correlation between natural gas inventory and price volatility. This negative correlation was particularly pronounced in winter and when inventories were in a period of scarcity. Chiou-Wei et al (2014) showed that deviations in the weekly natural gas storage report from market expectations have a significant effect on natural gas prices. Chan et al (2009) found that both weather and natural gas inventories affect the intensity and size of natural gas futures price jumps. Herbert (1995) showed that trade volume affects the variability of futures volatility, which is evidence of the influence of market microstructure. In an analysis of futures versus spot natural gas prices, Herbert (1993) found that futures prices inaccurately incorporate information into prices. Further, Movassagh and Modjtahedi (2005) found that natural gas futures are biased predictors of future spot prices consistent with normal backwardation. Their results were inconsistent with the bias being explained by the effect of systemic risk on risk premiums in natural gas futures prices.
This paper is organized as follows. In Section 2, we describe the data and provide preliminary evidence for our TVP model. In Section 3, we test for parameter instability and the structural form of time-varying parameters. In Section 4, we introduce our main model and the calculation of conditional variances. In Section 5, we summarize our results and perform diagnostic tests. In Section 6, we highlight the applications of our method. Section 7 concludes.
2 Data and preliminary evidence
Our natural gas prices and storage data are from the US Energy Information Administration (EIA) and were gathered from the Administration’s application programming interface using the EIAdata R package (Brigida 2015). Natural gas prices are NYMEX continuous front-month futures prices in USD/MMBtu (EIA API Series ID: NG.RNGC1.W). These contracts are for delivery at the Henry Hub in Louisiana.22 2 There are many different delivery points throughout the United States, such as Transco Zone 6 (New York City, New York) and Algonquin Citygates (Boston, Massachusetts). Spot prices at these different delivery points are regularly the reference rate for forward and, to a lesser extent, futures contracts. However, the Henry Hub delivery is the most liquid in both the spot and futures markets, and the futures market for delivery at the Henry Hub is where price discovery occurs. We convert these prices into log returns. We convert the storage series into percent deviations from the past five-year norm. The “norm” is defined as the average of the storage value over the same week in each of the previous five years.
Heating degree day () data is from the US National Oceanic and Atmospheric Administration (NOAA). We use HDD deviations from the norm to measure temperature fluctuations. Consistent with the literature, this deviation is in HDD values and not percent, given the effect of a small denominator in many weeks. This measures both warm () and cold () temperatures.33 3 We do not include deviations in cooling degree days because, in general, a large positive deviaton from the norm in tends to coincide with a large negative deviation from the norm in (the variables have a less than 0.50 correlation coefficient). All data is sampled at a weekly frequency and ranges from January 1, 1999 to June 20, 2014. Summary statistics for our data set are in Table 1. Time series plots of natural gas futures prices and returns are in Figure 1.
Statistic | ng | Stor | HDD |
---|---|---|---|
Mean | 0.1080% | 6.9241% | 4.6009 |
Median | 0.1947% | 6.7040% | 2.0000 |
Maximum | 64.2103% | 69.4137% | 52.0000 |
Minimum | 43.3280% | 55.1525% | 82.0000 |
Skewness | 0.47727 | 0.1270 | 0.4808 |
Kurtosis | 11.7062 | 4.8712 | 4.8471 |
Jarque–Bera | 145.82 | 119.91 | 2579.3 |
ADF statistic | 9.7970 | 4.7701 | 7.3830 |
ADF -value | 0.01 | 0.01 | 0.01 |
We first use a series of ordinary least squares (OLS) regressions to investigate whether the sensitivity of natural gas prices to changes in storage and the weather () is time varying, and whether the proportion of natural gas volatility attributable to these variables is time varying. To do so, we first calculate the average natural gas log returns (), storage deviation () from a five-year norm, and deviation for each week of the year. We reject a unit root in each of these variables. We then run fifty-two (one for each week) OLS regressions,
where denotes a particular week of the year, and ranges over the fourteen years in our sample. From these regressions, we extract a set of coefficients for each week as well as the proportion of total natural gas volatility attributable to each independent variable.
A plot of the regression coefficients over each week of the year is in Figure 2. Figure 3 shows the proportion of volatility from storage and weather for each week. This offers some evidence that the sensitivity of natural gas returns to storage and the weather varies throughout the year. Further, the proportion of volatility from storage and the weather is generally higher after the fortieth week of the year. These initial results are consistent with our hypothesis, but since the sample is small, they serve to motivate a more thorough examination of these time-varying relationships.
3 Parameter stability tests
Prior to implementing the time-varying parameter model, we first test for parameter instability. If we can reject stability in the parameters, this is evidence in favor of time-varying parameters. We then investigate the functional form of the state equation.
In this section and onward, we treat our sample as one three-dimensional time series . That is, we no longer divide the sample into fifty-two series, each of which contains the same week over multiple years. Thus, in these parameter stability tests, we are testing if the coefficients on our time series variables are stable from one week to the next.
In order to test for parameter instability, similar to Kim and Nelson (1989), we use a test proposed by Brown et al (1975) to detect departures from constancy in time series regression relationships. The specific test is referred to by the authors as the “homogeneity test” (see Section 2.5. Moving Regressions). The null hypothesis of the test is that the regression parameters are equal at each time point (stable regression coefficients).
The sample period is split into nonoverlapping intervals of arbitrary length , and the “between group over within groups” ratio of mean sum of squares is calculated as the test statistic. Under , the test statistic is distributed as , where is the number of regressors, is the number of intervals and is the number of observations.
Applying this test to , for values of ranging from 20 to 50, we are able to reject our null hypothesis at the 5% level of significance for all (most -values are below 0.1%). We therefore reject the stability of the regression coefficients.
3.1 Structural form of the parameters
Also of interest is the structural form of the time-varying regression coefficients. Engle and Watson (1985) suggest a random walk in cases where market participants adjust their estimates of a state only upon the arrival of new information. In natural gas markets, however, participants will also likely adjust parameter estimates based on season.
To test whether the coefficients are random walks, we will estimate using OLS. We also estimate separate equations for each individual independent variable. We then use the fact that one-half times the regression sum of squares of
where is the vector of independent variables, is distributed under the null hypothesis of stable coefficients. The alternative hypothesis is that the OLS regression exhibits heteroscedasticity consistent with random walk coefficients. This result is from Breusch and Pagan (1979), and this test is also employed in Kim and Nelson (1989).
Using this test, we are able to reject the null hypothesis for the coefficient at the 5% level. We do not reject the null hypothesis for the storage coefficient (). Considering this with the results of the previous section implies the coefficient is time varying, although not in a random walk fashion. This is consistent with our intuition that market participants are adjusting the coefficient based on season. Jointly testing all regression parameters, we do not reject the null hypothesis either.
Given the mixed results of the above tests, we modeled the coefficients both as autoregressive and unit root processes. The models with unit root coefficient processes were a better fit to the data. Moreover, Dangl and Halling (2012) found evidence that random walk coefficients can more quickly learn changes in the relationship between variables. Given these findings, we will report results from a time-varying parameter model with random walk coefficients. This allows for adjustment based on season (greater HDD means we are moving toward winter); however, it does not impose seasonality on the process a priori. This is particularly important given how differently natural gas prices may behave from one winter to the next.
4 Time-varying parameter model and conditional variance
The preliminary regression results and parameter stability tests are evidence in favor of parameters that vary with time. However, it is likely that market participants (as rational economic agents) use past and present information when deciding on the appropriate present and future coefficients. This motivates a model where the coefficients are allowed to be updated in a Bayesian fashion when new information arrives, much like the views of market participants. An appropriate specification in this case is a time-varying-parameter model, where the parameters are updated using the Kalman filter.
For the main model of our analysis, the measurement equation is
(4.1) |
and for coefficient we let the transition equation take the form of a random walk:
(4.2) |
Note denotes log returns in natural gas futures prices, while and represent deviation from normal storage and heating degree days, respectively.
Estimation of the model is done using the Kalman filter and prediction error decomposition. The likelihood function was maximized using the optim function in the R programming language (R Core Team 2014).
This time-varying-parameter model will estimate varying regression coefficients, but it also affords us an estimate of conditional volatility through the conditional variance of forecast errors from the Kalman filter (see Kim and Nelson 1989). The present analysis further decomposes this conditional volatility into the contribution from each factor.
From the Kalman filter, we estimate the conditional variance as
where is the vector of explanatory variables, is the variance–covariance matrix of the inference on conditional on information available at time (), and is the variance of the disturbance term.
To calculate the proportion of natural gas volatility attributable to a particular variable (say ), we set that variable to zero in and any row or column in that involves that variable. Denote these as and , respectively. We then recalculate
which affords the conditional variance of natural gas prices without that variable. The difference between the full conditional variance and the conditional variance without the variable () affords us the conditional variance attributable to that variable.
5 Results
5.1 Diagnostic tests
To begin with, we test the heteroscedasticity-adjusted one-period-ahead forecast errors, , for serial dependence using the Box–Pierce and Ljung–Box tests. We then run tests for lags from one to fifty-two weeks. Neither test rejects the null hypothesis of no serial dependence for all lags.
To test for ARCH effects, we use both the Lagrange multiplier test of Engle (1982) and the Ljung–Box test on the squares of the heteroscedasticity-adjusted one-period-ahead forecast errors . The tests disagree, however, with the Ljung–Box test failing to reject and the Lagrange multiplier rejecting the null hypothesis of no ARCH effects. We therefore conclude that there is evidence of ARCH effects.44 4 This may be a suitable avenue for further research. This analysis will focus on the components of volatility from TVP and leave the addition of ARCH effects, if any, to further analyses.
5.2 Results summary
The model was estimated using varying initial parameters, and the maximum loglikelihood over the many estimations was 1650. Figure 4 provides a chart of the Kalman filtered estimates of the time-varying regression coefficients. The standard deviation of the error terms in the measurement equation () is 5.49%. The standard deviations of the error terms in the intercept, and transition equations () are 0.0015, 0.4811 and 0.0001, respectively (these reflect the units of independent variables).
While each coefficient shows substantial variation, the storage coefficient potentially exhibits seasonal variation. The storage coefficient appears to be a stationary series and has a mean of 0.08. The range of variation through the seasons is from 2.40 to 1.99. The weather coefficient shows some seasonal variation; however, the mean of this coefficient seems to vary with time. For the period 1999–2007, the mean was 0.0010. For the period 2008–14, the mean dropped to 0.0002. This 80% drop is evidence that market participants vary their reaction to underlying variables over multiyear periods as well as throughout the year. Further research may be able to determine why market participants became less sensitive to weather after 2007. Perhaps it was a shift in overall winter weather severity? It may also be due to increased overall natural gas supplies due to shale gas. The supply abundance all but ensures ample supply regardless of the severity of winter weather.55 5 We thank an anonymous referee for pointing this out.
To confirm the above, we tested for a unit root in each of the coefficient series. Both the intercept and the weather coefficient contained a unit root. However, the storage coefficient rejected a unit root at a 1% level of significance. The ADF test for a unit root was employed.
Figure 5 shows a time series of weekly natural gas volatility using the TVP and GARCH(1,1) models specified by and for , a sequence of independent and identically distributed (iid) standard normal random variables, and where , and are estimated parameters. The mean forecast uncertainty (that is, the mean of the one-period-ahead forecast errors ) from the TVP model is 10.60%, whereas the mean absolute value of natural gas returns is 5.31%. The unconditional standard deviation of returns from the GARCH(1,1) model is 8.45%. This shows, on average, that there is more forecast uncertainty in natural gas returns than would be implied by the error term alone. That is, parameter uncertainty plays an important role in natural gas return volatility.
In Figure 6, we see the time series of the proportion of natural gas forecast uncertainty due to each underlying factor (, and the intercept). In part (b), we generally see that, in the summer, storage accounts for 50% of the forecast uncertainty, with approximately 25% of the uncertainty coming from weather and the intercept term.
However, in the winter, the proportion of forecast uncertainty due to weather often becomes the prime component of volatility. It often accounts for 40% of total volatility, and the portion attributable to storage drops to around 30%. This is evidence of the seasonal effect of weather and storage on natural gas price behavior.
This shows the marked effect of winter on the drivers of natural gas volatility. Moreover, it is consistent with common accounts of traders focusing on storage amounts during the summer injection season, as this is an indicator of whether there will be enough working gas in storage to meet winter demand.
5.3 Including lagged natural gas returns
We can also include lagged natural gas returns in the TVP model. This specification allows market participants to adapt the parameters of how they link past with present natural gas price behavior over time. In this case, the measurement equation is
(5.1) |
for coefficient , we let the transition equation take the form of a random walk:
(5.2) |
where denotes log returns in natural gas futures prices, denotes log returns lagged one week, and and represent deviation from normal storage and heating degree days, respectively.
The loglikehood of the model specification including the lagged natural gas return is 1676.978, whereas the loglikehood value of the model not including lagged returns is 1566.076. We can consider the model without lagged natural gas returns in the measurement equation to be a restricted model, and in that case the likelihood ratio test statistic is . Comparing this to the chi-square distribution with one degree of freedom, we are able to reject the null hypothesis with a -value near . Since the null hypothesis is that the coefficient on lagged natural gas returns is , this is evidence in favor of including the lagged natural gas term in the model.
Turning to the coefficient on lagged natural gas returns, it is generally positive, reflecting that prices tend to trend. However, the coefficient did become negative around the 2003–5 period.
Including the lagged natural gas term has little effect on the estimate of the time-varying proportion of volatility from storage and weather. Volatility attributable to lags in natural gas prices mainly reduces the contribution of the intercept (which reflects an unknown source). In fact, the correlation between the proportions of volatility attributable to lagged returns and the intercept is 0.73.
6 Applications
Below we will highlight two applications of the TVP model of natural gas returns and its accompanying decomposition of natural gas conditional volatility.
6.1 Application to hedge ratios
It is common for power producers to hedge input prices by buying natural gas futures, and to hedge demand risk by buying/selling weather derivatives (heating and cooling degree days). Say a company buys gas and sells HDD futures. Let and denote the hedge ratio for the gas and HDD futures contracts, respectively. Hedge ratios adjust for differing correlations and variances between spot and futures prices. Let and denote the change in futures and spot prices for natural gas, respectively. Similarly, and denote the futures and spot HDD prices. To calculate the optimal hedge ratios the company will seek to minimize the variance of the combined position:
(6.1) |
The optimal hedge ratios that solve this minimization problem are (proof in the online appendix)
(6.2) | ||||
and | ||||
(6.3) |
We can see the optimal hedge ratios are functions of the variances and covariances of the changes in spot and futures prices.
Note that the coefficient of HDD in our TVP model is a Kalman filtered estimate of ; so, from the above equations, time variation in the parameter estimate will also cause time-varying hedge ratios. Importantly, this time-varying effect on the optimal hedge ratio cannot be captured simply by using seasonal covariances. This is because the proportion of uncertainty from the weather can fluctuate greatly from one year’s winter to another.
6.2 Natural gas trading
The results of this analysis are useful to natural gas traders and other market participants. In fact, this analysis models how, in aggregate, market participants’ parameters linking storage and weather evolve over time. This can help market participants to understand their own adaptation processes.
Importantly, the filtering algorithm affords forecasts of next week’s coefficients, and coefficient uncertainty, as well as next week’s natural gas return, given this week’s data. This allows the creation of trading strategies based on the estimate of the natural gas return in the following period.
To investigate this, we calculated the profits on a simple trading strategy. We first estimate the out-of-sample return in the following week. If it is positive we buy one natural gas contract, and if the return is expected to be negative we sell one contract. Each week, we update our position based on the latest estimate and are long or short one contract at all times.
Over the fifteen-year period studied, the strategy earned a profit of 994%, which is an annualized profit of 16.54%. On average, we predicted the correct direction of natural gas returns (positive or negative) 58% of the time (significantly larger than 50% at the 1% level). However, if returns were larger than average in absolute value, we predicted the correct direction 62% of the time.
7 Conclusion
In this analysis, we have modeled natural gas returns explicitly, allowing for market participants to learn over time and to react differently to present changes in economic variables. This learning and adaptation, and the attendant parameter uncertainty, constitutes another source of time-varying conditional volatility.
In so doing, we have found evidence of significant variation in the coefficients linking natural gas returns and its underlying fundamental factors. Further, we found that the time series of the Kalman filtered estimates of the coefficient did not contain a unit root. This implies that we can make inferences about future coefficient values. The modeling and out-of-sample prediction of future coefficient values would be useful to include in future research. We also found evidence that the weather () coefficient does contain a unit root and that weather became a less important determinant of natural gas returns in 2007.
In an original application of the TVP model, we decomposed conditional volatility into a time series of each factor contributing to that volatility. This showed that storage is the dominant component of natural gas volatility throughout the year, with weather being the largest contributing factor only during winter periods. Finally, we showed that the results of this analysis have particular applications to hedging and trading in natural gas markets.
Declaration of interest
The author reports no conflicts of interest. The author alone is responsible for the content and writing of the paper.
References
- Breusch, T. S., and Pagan, A. R. (1979). A simple test for heteroscedasticity and random coefficient variation. Econometrica 47(5), 1287–1294 (https://doi.org/10.2307/1911963).
- Brigida, M. (2015). EIAdata: R wrapper for the Energy Information Administration (EIA) API. R package, Version 0.0.3. URL: http://CRAN.R-project.org/package=EIAdata.
- Brown, R. L., Durbin, J., and Evans, J. M. (1975). Techniques for testing constancy of regression relationships over time. Journal of the Royal Statistical Society B 37, 149–192 (https://doi.org/10.1111/j.2517-6161.1975.tb01532.x).
- Chan, W. H., Wang, G. H. K., and Yang, L. (2009). Weather, inventory, and common jump dynamics in natural gas futures and spot markets. Preprint, Social Science Research Network (https://doi.org/10.2139/ssrn.1537762).
- Chiou-Wei, S. Z., Linn, S. C., and Zhu, Z. (2014). The response of US natural gas futures and spot prices to storage change surprises: fundamental information and the effect of escalating physical gas production. Journal of International Money and Finance 42, 156–173 (https://doi.org/10.1016/j.jimonfin.2013.08.009).
- Dangl, T., and Halling, M. (2012). Predictive regressions with time-varying coefficients. Journal of Financial Economics 106, 157–181 (https://doi.org/10.1016/j.jfineco.2012.04.003).
- Diebold, F. X., and Yilmaz, K. (2009). Measuring financial asset return and volatility spillovers, with application to global equity markets. Economic Journal 119(534), 158–171 (https://doi.org/10.1111/j.1468-0297.2008.02208.x).
- Diebold, F. X., and Yilmaz, K. (2012). Better to give than to receive: predictive directional measurement of volatility spillovers. International Journal of Forecasting 28(1), 57–66 (https://doi.org/10.1016/j.ijforecast.2011.02.006).
- Engle, R. F. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica 50(4), 987–1007 (https://doi.org/10.2307/1912773).
- Engle, R. F., and Watson, M. W. (1985). Application of Kalman filtering in econometrics. Unpublished Report, World Congress of the Econometric Society, Cambridge, MA.
- Geman, H. I., and Ohana, S. (2009). Forward curves, scarcity and price volatility in oil and natural gas markets. Energy Economics 31, 576–585 (https://doi.org/10.1016/j.eneco.2009.01.014).
- Herbert, J. H. (1993). The relation of monthly spot to futures prices for natural gas. Energy 18, 1119–1124 (https://doi.org/10.1016/0360-5442(93)90084-Q).
- Herbert, J. H. (1995). Trading volume, maturity and natural gas futures price volatility. Energy Economics 17, 293–299 (https://doi.org/10.1016/0140-9883(95)00033-Q).
- Kim, C. J., and Nelson, C. R. (1989). The time-varying-parameter model for modeling changing conditional variance: the case of the Lucas hypothesis. Journal of Business and Economic Statistics 7, 433–440.
- Movassagh, N., and Modjtahedi, B. (2005). Bias and backwardation in natural gas futures prices. Journal of Futures Markets 25, 281–308 (https://doi.org/10.1002/fut.20151).
- Mu, X. (2007). Weather, storage, and natural gas price dynamics: fundamentals and volatility. Energy Economics 29, 46–63 (https://doi.org/10.1016/j.eneco.2006.04.003).
- R Core Team (2014). R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL: http://www.R-project.org/.
Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.
To access these options, along with all other subscription benefits, please contact info@risk.net or view our subscription options here: http://subscriptions.risk.net/subscribe
You are currently unable to print this content. Please contact info@risk.net to find out more.
You are currently unable to copy this content. Please contact info@risk.net to find out more.
Copyright Infopro Digital Limited. All rights reserved.
As outlined in our terms and conditions, https://www.infopro-digital.com/terms-and-conditions/subscriptions/ (point 2.4), printing is limited to a single copy.
If you would like to purchase additional rights please email info@risk.net
Copyright Infopro Digital Limited. All rights reserved.
You may share this content using our article tools. As outlined in our terms and conditions, https://www.infopro-digital.com/terms-and-conditions/subscriptions/ (clause 2.4), an Authorised User may only make one copy of the materials for their own personal use. You must also comply with the restrictions in clause 2.5.
If you would like to purchase additional rights please email info@risk.net