Mobile Edition
Print
Get the Mag
Weekly UpdatesIntroduction
Crude oil markets have been heavily influenced by OPEC's behavior for almost 30 years. In the 1980s, OPEC changed its pricing policies to allow the marketplace to have a larger impact on price. Prices became more transparent with the accompanying evolution of spot and futures markets. The availability of international supply and demand data has also improved. This has afforded analysts more opportunity to analyze and forecast crude oil prices.
During much of the 1990s, crude oil prices fluctuated around $20 per barrel. Towards the end of the 1990s, the world supply-demand balance shifted significantly. A combination of supply increases from OPEC, including the return of Iraq's crude oil volumes to the marketplace following the first Gulf War, and diminished demand as the Asian financial crisis occurred, produced excess supply in petroleum markets that resulted in crude oil prices, as represented by West Texas Intermediate (WTI) crude oil, plunging to almost $10 per barrel. Because of this price collapse, in March 1999, OPEC changed its market behavior and reduced oil production. With less supply and the subsequent recovery of demand, petroleum markets tightened. Prices rose and settled at around $30 per barrel, well above the prior average of $20 per barrel. Then in 2004, prices began their second ascent, and by the middle of 2008, were well over $130 per barrel.
From the early 1990's until early in the current decade, crude oil prices could be explained and forecasted using only OECD inventory data (see, for example, Ye et al. 2005). During that time, OPEC had excess production capacity that could be used to meet unexpected demand increases. However, as world demand grew, this excess capacity diminished, reducing the perceived ability of producers to meet demand increases in the short term. Improvements in forecasting capabilities were seen when an excess production capacity variable was added as an additional predictor in the forecast model (see, for example, Ye et al. 2006). However, since early 2004, estimated inventories and excess production capacity significantly under-predict WTI prices, even with shift factors added to the model to reflect possible structural changes.
The objective of this study is to develop a short-run forecast model that can provide improved monthly forecasts for crude oil prices. An additional variable, the cumulative excess capacity, is derived and incorporated into the forecast model to capture the so-called Duesenberry Ratchet Effect observed in the crude oil market in recent years, reflecting the changing behaviors on both the demand and supply sides. This new model provides significantly improved forecasts for the entire post-Gulf War 1 time period over previous models.
The three predictor variables, inventory, excess production capacity, and cumulative excess capacity are supported by the economic literature both theoretically and empirically. First, the relationship between commodity inventory levels and short-run price has been studied for nearly a century. (1) Inventory balances supply (2) and demand; it captures expected seasonality and general trends in production and demand, as well as unexpected supply or demand shifts. It is the immediate "supply" when needed and can also become a demand to cushion fears of shortages.
[FIGURE 1 OMITTED]
Second, the economic literature has recognized the impact of excess production capacity, or capacity utilization rate, on commodity markets in general (e.g., Dixit and Pindyck 1998) (3), and on petroleum markets in particular (e.g., Dahl and Yucel 1991, and Powell 1990). Recently Kaufman et al. (2004) tested quarterly models of OPEC behavior and found capacity utilization to be an important explanatory variable. Thus, both theoretical and empirical studies indicate that excess production capacity must be considered when analyzing the petroleum industry, which has nearly costless short-run production quantity adjustments, but which exhibits a large capital investment cost to develop new reserves.
Finally, a ratchet effect, also termed as a Duesenberry Effect, was developed in the 1970s and has since been applied in various economic models. (4) Theoretically, the ratchet effect is justified in dynamic optimization with asymmetric behavior (e.g., Dybvig 1995) and empirically, it enables modelers to capture persistent inflation and increasing money demand (Enzler et al. 1976). In energy markets, it has been applied as a tool for forecasting electricity prices (Stevens and Adams 1986).
In the next section, the data and their sources are described. The "Regime Changes" section identifies three regimes reflecting structural changes in the crude oil market. The "A Forecast Model" section shows a monthly crude oil price forecast model with a cumulative excess production capacity term. The last section offers some conclusions and suggestions for future work.
Data
Data on crude oil prices, petroleum inventories, and excess production capacity were collected for the period from January 1992 to December 2007. The nominal WTI crude oil spot price was used as a representative price for the world crude oil price. Daily WTI spot prices, obtained flown Reuters, were converted to monthly averages to correspond with the inventory data which is only available at a monthly frequency.
[FIGURE 2 OMITTED]
A consistent monthly data series for industrial inventories of crude oil and petroleum products for OECD countries is available beginning in December 1990 from the International Energy Agency (IEA). This study analyzed markets beginning in January 1992 to avoid the impact that the first Gulf War had on petroleum markets.
Estimates of monthly OPEC excess production capacity are created by the U.S. Energy Information Administration (EIA). Prior to January 2004, excess production capacity was estimated using internal EIA information on annual capacity and OPEC production. The current estimation procedure assumes that only OPEC has excess capacity and calculates monthly changes in OPEC production. These monthly estimates are then benchmarked with the more accurate annual capacity estimates, with any known changes in production capacity explicitly included. Values of excess production capacity for January 2004 onwards can be found in the Short Term Energy Outlook (5)
Regime Changes
Crude oil prices for the time period from the early 1990's to the present can be divided into three distinct regimes, as shown in Fig. 1. Regime 1 (R1), which represents a stable market with WTI prices averaging about $20/bbl, encompasses the January 1992 to June 1999 time period. Regime 2 (R2), which reflects OPEC's attempt to reestablish control of the crude oil market, extends from July 1999 to May 2004 with WTI prices averaging about $30/bbl (except for post-9/11/2001 period). Finally Regime 3 (R3), wherein visually no excess production capacity exists for crude oil, runs from June 2004 to December 2007, with crude oil prices increasing.
[FIGURE 3 OMITTED]
Figure 2 shows a 3-D graph of the three key variables, WTI prices, OECD inventories, and excess production capacity, color coded for the three regimes with dark blue representing Regime 1, light blue for Regime 2, and green and red (6) for Regime 3. Figure 3 shows the surface plot of WTI price being a function of OECD inventories and excess production capacity, created from observed historical data with the colors roughly corresponding to the regimes shown in Fig. 2.
The three regimes are seen in Fig. 4, where the WTI prices from Fig. 3 are projected onto the (2-dimensional) plane of OECD inventories and excess production capacity. The figure readily shows that Regime 1 (roughly the dark blue area) and Regime 2 (light blue) are associated with low WTI prices and relatively high excess capacity; and Regime 3 (the green, yellow and red areas) with low excess capacity (and with inventories that are often relatively high).
The changing relationships between WTI prices and OECD inventories as well as between WTI prices and excess production capacity in the three regimes can be readily seen in Figs. 5 and 6, which are quasi 2-D projections derived from Fig. 2. (7) The changing relationship between WTI price and these two key market variables reflects OPEC policy changes, significant reductions in excess crude oil production capacity, and changing market behavior in both demand and supply, including fear of little additional crude oil production capacity despite a rapidly growing global demand.
[FIGURE 4 OMITTED]
[FIGURE 5 OMITTED]
[FIGURE 6 OMITTED]
As can be seen from Fig. 5, the relationship between WTI prices and OECD inventories has changed from the early 1990's to December 2007. The dark blue points represent Regime 1 (January 1992 to June 1999); the lighter blue points represent Regime 2; (July 1999 to May 2004); while the green and red points reflect the relationship in Regime 3 (the period from June 2004 to December 2007); this regime may be an (upward) shifting of the usual downward sloping price/inventory relationship.
While a $10 per barrel shift in price for the same inventory level between Regimes 1 and 2 is clearly indicated, the expected negative relationship between price and inventories remains unchanged. This indicates that the normal negative price/inventory relationship did not change, only that the market accepted a ratcheting upward of the equilibrium price level by about $10/Bbl. However, in Regime 3, an unusual positive relationship between crude oil price and OECD inventories is observed, which cannot be explained under equilibrium conditions and had not been observed previously. This unusual market behavior of higher crude oil prices being associated with greater inventory (or inventory demand) in Regime 3 implies a rapidly shifting transitional relationship with little indication of a stable market as exhibited in Regimes 1 and 2.
FEED