Using rational expectations storage model to explain natural gas price

Natural gas is a key energy source for residential, commercial, electric power and industrial use. Residential, commercial and electric power sector consumption is primarily driven by weather conditions and displays obvious seasonal patterns while production is relatively stable throughout the year. As weather condition is uncertain and both consumption and production are inelastic in the short term, natural gas price is quite volatile especially in the heating season. Due to an imbalance between production and consumption, storage plays an important role in ensuring availability and smoothing price between low and peak consumption seasons. Storage is also a key driver of price volatility. The existing literature confirms the importance of inventory and weather conditions in determining price and its variance. Most studies to date use time series models and focus on the historical price realizations while providing little insight into how price patterns are determined by market participants’ behavior. In addition, the impact of inventory and weather variables on price volatility has not been analyzed in detail.This thesis aims to construct a model that can mimic the major market participants’ behavior and reproduce the natural gas price with mean and standard deviation patterns consistent with historical observations: higher average level and standard deviation in peak consumption season. We construct a monthly rational-expectations competitive storage model to better reflect monthly variations in price. Natural gas consumption and production are specified in a way that the current period volume is highly correlated with previous period volume so as to capture the stickiness and gradual change in natural gas markets. Imposing non-arbitrage condition, price is inter-temporally correlated. Net storage cost consists of both physical storage cost and convenience yield obtained from holding stock at hand. Normal storage level for each month is introduced to reflect the yearly cycling of natural gas inventory and is used in the convenience yield calibration. It denotes the normal storage level each month that is needed to balance seasonal demand-supply relationship. Convenience yield is high if the inventory falls below normal storage level and high convenience yield pushes up the price and decreases current consumption to accumulate more natural gas for future use.The model is solved using numerical methods because analytical solutions are not feasible. In order to validate the result, accuracy tests are conducted and the major assumptions are tested as well. The model’s approximation errors are reasonable. The model is further validated by comparing simulated price series with historical observations by using historical weather variables in the solved model. The simulated model generates prices that largely replicate the key features of historical data, including the price level, price variance, price sensitivity under unusual weather conditions and price autocorrelation.Weather conditions and total natural gas availability are the main drivers for price and price standard deviation. The model finds that in winter high heating degree days (HDD) or low inventory drives price and price volatility higher while price and its variance decrease with low HDD and high inventory. The case is similar in summer with cooling degree days (CDD) instead of HDD as the weather variable. When inventory is low, weather shocks have a larger impact on price than when inventory is high. The effect is more pronounced in winter than in summer because the supply is tighter in heating season.Using the validated competitive storage model, this thesis further assesses the potential impact of LNG export on the U.S. domestic natural gas market. Given the large pricing spread between the United States and the rest of world, along with policy promotion and the completion of LNG facilities, U.S. LNG exports are poised to expand dramatically. This study covers two major types of LNG export scenarios: exogenous fixed volume and endogenous export volume depending on the price spread between US and world prices. Four export scenarios are analyzed and compared with the benchmark scenario of no LNG export. The first two scenarios are fixed export volume with 6 bcf/day and 12 bcf/day respectively, to be consistent and comparable with an EIA 2014 report and the existing literature. One of the endogenous scenario scenarios assumes no consumption and supply growth for importing countries and the other one assume 100% increase of demand and 50% increase of supply in LNG importing countries by 2036.Because of high shipping cost and inelastic natural gas demand in importing countries, the U.S. LNG is not competitive under current market condition, if no growth is expected. The U.S. LNG export volume is very small and decreases over time. Due to small export volumes, the domestic price impact is minimal. For all scenarios analyzed in this study, the long-term price impact is less than 8%, or around $0.33 per thousand cubic feet. In the long-term, the endogenous export with growth assumption scenario shows the largest price increase compared to the no export benchmark scenario. The export level is around 12 bcf per day.The U.S. domestic price variance becomes smaller if an endogenous export sector is added while the price variance becomes higher under fixed export volume scenarios. If the LNG export is endogenously determined, when domestic price increased, LNG export decreases. This provides an additional buffer to the U.S. domestic market if there is shock to push up natural gas consumption and price. In contrast, fixed volume export makes the total natural gas consumption less responsive to price change and thus increases price variance.Most of the LNG export volumes will be satisfied by production increases instead of domestic consumption reductions in the long term. In all four scenarios analyzed in this study, production catches up gradually in response to price increase due to LNG export. In the beginning period when production is constrained by production capacity, most of the export is covered by domestic consumption reduction. In the long term, as production increase, domestic consumption recovers to similar level as in the no export scenario.