عوامل و اقتصاد کلان و بازار آینده نفت : یک مدل غنی از داده ها
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|15893||2010||9 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Energy Economics, Volume 32, Issue 2, March 2010, Pages 409–417
I study the dynamics of oil futures prices in the NYMEX using a large panel dataset that includes global macroeconomic indicators, financial market indices, quantities and prices of energy products. I extract common factors from the panel data series and estimate a Factor-Augmented Vector Autoregression for the maturity structure of oil futures prices. I find that latent factors generate information that, once combined with that of the yields, improves the forecasting performance for oil prices. Furthermore, I show that a factor correlated to purely financial developments contributes to the model performance, in addition to factors related to energy quantities and prices.
During the past year, oil prices have made the headlines of the financial press almost every day. Since the beginning 2008, the spot price of crude oil traded in the New York Mercantile Exchange (NYMEX) has almost doubled at peak. This has raised serious concerns among market participants and policymakers worldwide. Comments released to the press have often denoted a deep disagreement on the causes of the price spikes and, in general, on the mechanics of oil market. Bernanke (2008) has represented the central bankers' view in a timely manner, stating that1 “…the price of oil has risen significantly in terms of all major currencies, suggesting that factors other than the dollar, notably shifts in the underlying global demand for and supply of oil, have been the principal drivers of the increase in prices. (…) Another concern that has been raised is that financial speculation has added markedly to upward pressures on oil prices. (…) However, if financial speculation were pushing oil prices above the levels consistent with the fundamentals of supply and demand, we would expect inventories of crude oil and petroleum products to increase as supply rose and demand fell. But in fact, available data on oil inventories show notable declines over the past year.” Since oil commodities are traded through futures and derivatives contracts, market views shape the pricing of oil commodities. In this sense, the financial press has pushed the hypothesis that purely ‘financial’ considerations, unrelated to ‘real’ market developments, have been behind the recent spikes (see Chung, 2008 and Mackintosh, 2008). The distinction between financial and real determinants of oil prices in the long run is also present in the academic literature. A large number of papers suggest that oil prices are mainly driven by demand and supply considerations. For instance, Kilian (2008b) suggests that a proper measurement of the business cycle effects of energy prices requires disentangling the role of demand supply shocks in energy markets. Kilian (2008a) decomposes the real price of crude oil into supply shocks, shocks to the global demand for industrial commodities, and demand shocks that are idiosyncratic to the oil market. The role of energy quantity factors is stressed also in Alquist and Kilian (2008), who show that spread between oil futures prices of different maturities are related to uncertainty about supply shortfalls. The literature on the financial determinants of oil prices has produced various contributions on the role of market uncertainty and volatility for oil pricing. Askari and Krichene (2008) model the jump intensity of daily crude oil prices between 2002 and 2006. They find that measures of market expectations extracted from call and put option prices have incorporated no change in underlying fundamentals in the short term. Chong and Miffre (2006) document the presence of a significant pattern of risk premia earned by investors on a number of commodities futures since 1979, including crude oil. Gorton et al. (2007) show that, although commercial positions on oil futures are correlate with inventory signals, they do not determine risk premia. The presence of two opposing views on price formation in the oil market over the long run implies that a number of key questions are not dealt with in the literature. The issue of causality between spot and futures prices across the maturity structure is largely unsettled. Suppose that oil futures contain information about spot prices. Omitting futures prices would bias the results in favour of a strong role for demand–supply factors to drive the spot price. Moreover, the role of macroeconomic factors for the dynamics of oil prices is typically studied in isolation from the conditions prevailing in financial markets.2 In this paper I study whether the interplay between real and financial factors can play a systematic role for explaining oil prices changes over a long time period. I exploit the information from a large panel to investigate the sources of changes in the term structure of futures prices for WTI crude oil. Like Bernanke et al. (2008), I extract common factors form the large panel dataset, and I model the joint dynamics of the factors and the oil prices in a ‘Factor-Augmented’ Vector Autoregression (FAVAR). The factors mimic the drivers of oil prices that are ‘latent,’ in the sense that they are not directly observed by the econometrician from the information set. In standard Vector Autoregressive (VAR) models, the econometrician is required to choose what observable variables best represent theoretical concepts, such as supply and demand. The supply of oil can be measured with data on oil production. However, these data series are affected by measurement errors of different types, for instance arising from aggregation. As argued by Bernanke et al. (2008), the use of sparse information in the form of factors extracted from a large dataset mitigates this problem. This modelling strategy has already been applied by Ludvigson and Ng (forthcoming) and Mönch (2008) for the construction of pricing models for the yield curve of government bonds, and it presents several advantages. The model can capture the interdependence between oil price changes and the factors of different nature. The FAVAR also allows to model jointly the relevant maturities of oil futures prices in a flexible way. It should be stressed that the literature features a long list of contributions on the role of unobservable factors for oil price dynamics.3 These contributions differ from the present paper in two dimensions. The factors are typically meant to drive the time-varying volatility of observed at a daily frequency. Instead here I use monthly data, and I abstract from the role of high-frequency price movements. The panel dataset from which I extract common components include over 200 data series with detailed information on energy demand and supply, energy prices, macroeconomic and financial variables. I show that a latent factor correlated with the open interest on oil futures prices contributes significantly to the joint model of the oil price returns. This appears to corroborate the conjecture of Trichet (2008) on the financial determinants of oil prices. The other factors are strongly correlated with data on energy quantity and prices, as typically suggested by the macroeconomics literature. I find that augmenting the information from the term structure of oil futures prices with latent factors improves the forecasting performance of the model. This paper is organized as follows. In Section 2, I outline the structure of the FAVAR model. Section 3 presents the dataset. Section 4 describes the results. Section 5 concludes.
نتیجه گیری انگلیسی
This paper models the dynamics of the term structure of oil futures prices by using information from a panel dataset including over 230 series with global macroeconomic indicators, financial market indices, quantities and prices of energy products. I estimate a Factor-Augmented Vector Autoregression with latent factors extracted from the panel. I show that latent factors generate information which, once combined with that of the returns, improves the forecasting performance for oil prices. Furthermore, I find that a factor correlated to purely financial developments contributes to the model performance, in addition to factors related to energy quantities and prices. The results presented here can be extended in a number of directions. I amplanning to use Bayesian model averaging to study the performance of the best-performing subset of factors for forecasting the term structure of oil prices. Moreover, the factors could be used to identify the impact of oil demand and supply shocks. In this sense, it would be important to understand what role purely financial market variables can play for the persistence and magnitude of the estimated shocks.