Buyer Beware: The Asymmetric Impact of the Strategic Petroleum Reserve on Crude Oil Prices.

AuthorStevens, Reid B.
  1. INTRODUCTION

    The Strategic Petroleum Reserve (SPR) was created in response to the 1973-1974 oil embargo by the Organization of the Petroleum Exporting Countries, and was intended to "store petroleum to reduce the adverse economic impact of a major petroleum supply disruption to the United States" (U.S. Congress, 1975). The oil supply disruptions from the 1950s to the 1970s led economists to propose a government controlled crude oil reserve to protect the economy from insecure oil suppliers (Cabinet Task Force on Oil Import Control, 1970; National Petroleum Council, 1973; Nordhaus, 1974). These calls for the U.S. government to create a publicly managed crude oil reserve intensified during the Arab Oil Embargo (Tolley and Wilman, 1977). The theory behind these policy proposals is simple: the government could offset supply disruptions by increasing oil supply with the SPR rather than by restricting oil demand with unpopular quotas, tariffs, or taxes. Since research was beginning to connect oil price increases in the 1970s with the subsequent economic downturns (Fried et al., 1975; Haberler, 1976), a policy mechanism that controlled oil prices had broad political appeal (Weimer, 1982).

    Since the SPR was constructed in 1977, over 150 million barrels of crude oil have been released from the Reserve. During these releases, crude oil from the SPR made up 5 to 12 percent of weekly domestic oil supply. The government has purchased over 850 million barrels of crude oil for the Reserve in this period, which accounted for between 2 to 4 percent of U.S. crude oil consumption during purchase weeks. Despite this long history of purchases and releases, little is known about the effect of the SPR on crude oil prices. Much of the early work on the SPR focuses on optimal purchase and release strategies, and does not directly estimate the effect of the SPR on crude oil prices (Tolley and Wilman, 1977; Nichols and Zeckhauser, 1977; Teisberg, 1981; Balas, 1981; The views expressed in this article do not necessarily represent the views of the Federal Reserve or the United States government.

    Chao and Manne, 1983; Samouilidis and Berahas, 1982; Zhang et al., 2009). These papers generally assume the SPR has an asymmetric effect on crude oil prices, one that is favorable to releases: crude oil released during a supply disruption will lower oil prices, but purchases when the oil market is calm will not affect prices. Unsurprisingly, these models show that under optimal management, the benefits of the SPR far outweigh the costs of construction, oil acquisition, and storage.

    The estimated benefits of the SPR are even higher when the negative macroeconomic effects of oil price increases are taken into account (Weimer, 1982; Bohi and Toman, 1996; Leiby and Bowman, 2000). In 2006, when Congress approved plans to expand the Reserve capacity from 727 million barrels to 1 billion barrels, SPR proponents echoed the conclusions of these papers, praising the Reserve's ability to lower oil prices and stabilize the economy during oil supply disruptions (Government Accountability Office, 2006; Congressional Research Service, 2006).

    The few empirical estimates of the SPR price effect vary widely. In early empirical research, SPR releases are estimated to lower the price of crude oil by 3 to 32 percent, while SPR purchases are estimated to increase the price of crude oil by 0.4 to 32 percent (U.S. Senate, 2003; Verleger, 2003; Considine, 2006). More recently, Newell and Prest (2017) estimate that past SPR releases have reduced spot prices of crude oil by as much as 15-20 percent. The lack of consistency among these estimates is evidence that identifying the effect of SPR policy on oil prices is difficult. Some SPR crude oil releases occurred in response to severe oil market supply disruptions (e.g., 1991 Desert Storm sale) while others occurred when oil prices were relatively low and stable (e.g., 1996 sales to reduce the deficit). Similarly, the largest SPR purchases were a response to high oil prices following the Arab Oil Embargo. Other large purchases for the SPR were also made in the late 1980s, when there was excess capacity in oil markets and prices were relatively low.

    The key difficulty in the empirical literature is to isolate the effect of SPR policy from the state of the oil market to which the policy responds. SPR purchase and release decisions are made by the President and carried out by the Department of Energy (DOE). Isolating the effect of SPR policy is challenging because the President's policy decisions depend, in part, on the state of the oil market. Changes in the price of oil following an SPR policy action reflect the policy, the market conditions to which the policy is responding, and the market conditions to which the policy is not responding. The endogeniety of the price of oil and SPR policy has likely led to the widely varying estimates of the SPR's effect on the price of oil.

    We contribute to the literature by using a structural vector autoregression (VAR) model of the U.S. oil market to disentangle these effects. Beginning with Kilian's (2009) structural VAR model of the global oil markets, VAR models have been used to study country-level oil markets (Baumeister and Peersman, 2013a; Kilian and Murphy, 2014), volatility in oil markets (Alquist and Kilian, 2010; Kellogg, 2014), and storage in oil markets (Kilian and Murphy, 2014). We extend this literature by postulating a model of the U.S. oil market that explicitly includes SPR purchase and release variables.

    We model the U.S. oil market, rather than the global market, because SPR oil sales are limited to U.S. buyers who process the oil at U.S. refineries. Though the U.S. market is connected to the global market, significant frictions exist in the oil market, including transportation costs and infrastructure constraints. The significance of these frictions can be seen in the divergence between the WTI price and Brent prices from 2012 through 2016. Frictions in the global oil market allow regional supply and demand shocks to disproportionately impact the local oil market.

    As a starting point of our analysis, we estimate a recursively identified model of the U.S. oil market, where the effect of SPR policy is estimated by imposing a temporal ordering on all structural shocks to the oil market. For example, the identifying assumptions in this model imply that oil production and economic activity respond only with a lag to SPR policy shocks. To make these recursive ordering assumptions more plausible, we use weekly data to estimate the model, rather than the monthly or quarterly frequency typically used in the oil market literature. In this model, an unanticipated SPR purchase raises the price of the price of oil about 1 percent over 20 weeks following purchase, but an unanticipated SPR release does not have a statistically significant effect on the price of oil.

    Though recursively identified VAR models have been used in the oil market literature Kilian, 2009), we recognize this approach is not robust to violations of the recursive ordering of responses to exogenous shocks, which may occur in data measured at a monthly, or even weekly, frequency. Because of this potential limitation, we estimate two additional VAR models with different identifying assumptions. These distinct identification strategies yield results consistent with the baseline result. (1)

    In particular, instead of assuming a recursive ordering of the residuals, we identify policy shocks using information that is external to the VAR model, but correlated with unanticipated policy actions (Stock andWatson, 2008; Montiel Olea et al., 2015). This approach has been used many times to identify oil supply shocks in VAR models (Hamilton, 2003; Kilian, 2008; Arezki et al., 2015). We construct two separate instruments to estimate the effects of SPR purchase and release shocks in a partially identified VAR model.

    First, we use the SPR purchase schedule as an instrument to identify unanticipated SPR purchases. From 1999 to 2010, the DOE hired private contractors to purchase oil for the SPR. Over 225 millions barrels of oil were delivered to SPR during this period. These contractors were required to deliver a specified amount of oil over a fixed time period. For example, a contractor would agree to deliver 9.4 million barrels of oil from April to July 2005. The timing of these purchases was set in advance by the DOE, the purchase schedules were not immediately announced, and exemptions from the schedule were rarely granted, which makes the purchase schedule correlated with unanticipated SPR purchases. The schedule was also set well ahead of the delivery window, so the purchase schedule is uncorrelated with other structural shocks at the time of purchase. Using the purchase schedule as an instrument to partially identify the VAR model, we estimate that SPR purchase shocks cause crude oil prices to rise about 2 percent over 20 weeks following purchase.

    Next, we use crude oil futures data to identify the effect of SPR releases on crude oil prices. This model is inspired by the work of Cochrane and Piazzesi (2002), who use changes in the Federal Funds futures price following policy announcements to estimate monetary policy shocks. In our model, we use tick (minute by minute) data for the benchmark U.S. crude oil futures contract, West Texas Intermediate (WTI), before and after after SPR release announcements to measure the extent to which the policy was unanticipated by market participants. Some SPR releases, like the April 1996 budget reduction sale, were anticipated by the market and so the crude oil futures price was unchanged on the day of the announcement. Other SPR release releases, like the June 2011 sale during the Arab Spring, were, to some extent, unanticipated by the market and were immediately followed by a drop in the crude oil futures price. Using these estimated policy shocks, we find that...

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT