"Do Foreign Gifts Buy Corporate Political Action? Evidence from the Saudi Crude Discount Program" (2014)

"Do Foreign Gifts Buy Corporate Political Action? Evidence from the Saudi Crude Discount Program" (2014)
Do Foreign Gifts Buy Corporate Political Action?
Evidence from the Saudi Crude Discount Program∗
Jennifer R. Peck†
September 2014
Abstract
Between 1991 and 2003, Saudi Aramco sold crude to U.S. refineries at a substantial discount,
with a total cost of approximately 8.5 billion dollars. This paper assesses the incidence of these
discount rents in the U.S. market and yields the first empirical evidence for the use of oil as a
tool of political leverage through transfers to American firms. Using panel variation in discount
receipts, I find that rents were captured by refiners as profits and did not pass through into retail
gasoline prices. Discounts were also associated with increases in refiners’ political donations and
strategic reallocations of these contributions.
∗
I am extremely grateful to Michael Greenstone and Esther Duflo for their extensive feedback on this project. I
also thank Christopher Knittel, James Poterba, Jessica Leight, Laura Ralston, Horacio Larreguy, Joseph Shapiro and
Eliza Forsythe as well as seminar participants at MIT for their insightful questions and comments. The George and
Obie Shultz fund provided financial support to acquire the data for this project.
†
Swarthmore College Department of Economics, 500 College Avenue, Swarthmore, PA 19081.
(email: jpeck1@swarthmore.edu)
1
Introduction
Political and economic concerns are fundamentally linked in the world petroleum market. Access
to a dependable oil supply is critical to economic and political stability, and these supplies are very
dependent on local political climates. Oil is used as a political tool by suppliers, and is a frequent
topic of diplomatic intervention. At the center of this is Saudi Aramco, the world’s largest producer
of crude oil and holder of 25 percent of global reserves. Because of its position as a global swing
producer, the economic and political drivers of its output decisions attract a great deal of attention.
In addition to deciding how much to produce, however, Saudi Arabia also decides where to
sell its crude, a decision which receives much less attention but which is perhaps no less strategic.
From 1991 to 2003, Saudi Arabia maintained a position as the top supplier of foreign crude to
U.S. refiners in order to support its political alliance with the United States. Although oil is often
thought of as having one world price, Saudi Aramco supported this export strategy by selling the
same crude at different prices in different geographic markets. Maintaining the pricing differentials
required by its export targets appears to have been both politically strategic and quite expensive:
between 1991 and 2003, Saudi Arabia spent approximately 8.5 billion dollars selling discounted
crude to the United States. The per-barrel discount relative to the Asian price reached a high of
6.30 dollars, 30 percent of the U.S. crude price in 2001, and was worth 1.9 billion dollars that year
alone. In achieving its export target, Saudi Arabia therefore transferred substantial rents to the
U.S. oil industry in the form of discounted crude supplies. In addition to determining the total
value of this transfer through its export quotas and pricing policies, Saudi Arabia also controlled
how these rents were distributed within the United States; discounted crude was targeted at specific
refineries using highly restrictive sales contracts.
Despite the magnitude of this transfer, the Saudi crude discount program has received almost
no attention in the academic literature or in the popular press.1 The first task of this paper is to
document the size of these rents and identify the refiners that received this gift. This paper then
evaluates the success of this allocation mechanism as a political tool by identifying where rents
accrued as a result of the policy and how these rents affected political action by discount recipients.
To do this, I first estimate the effect of discounts on refinery-owner profits and local gasoline prices. I
1
Despite considerable attention to Saudi Arabia’s petroleum production and exports, The Quest – Daniel Yergin’s
history of the world energy market post 1991 – never mentions this program.
2
then examine the effect of discount receipts on one particular type of measurable corporate political
action: contributions to congressional campaigns. The analysis follows the money from the original
oil shipment to the campaign funds of federal politicians.
The empirical analysis relies on several key features of the market for Saudi crude in the United
States. All U.S. refiners pay the same per-barrel price for Saudi crude each month, but quantities
differ both across refineries and within a single refinery from month to month. These quantities
are based on long-term contracts, but are subject to unilateral changes by Saudi Aramco each
month. This yields variation in the total value of the discount to each refinery, which in turn
creates variation in discount receipts across refining companies and in the geographical distribution
of discounted crude. It is therefore possible to estimate the effect of the discount on company-level
outcomes (profits and political contributions) and market-level prices for refined products.
There are several key results. First, there was a great deal of heterogeneity in the value of the
discount received by different companies and significant geographical dispersion in the destination
of Saudi crude. Despite their similar refining capacities, for example, Marathon and Tosco received
very different amounts of Saudi crude: a total of 680 million barrels were delivered to Marathon’s
refineries, and just 20 million to Tosco refineries over the period. The amount received also varied
considerably from year to year; in just two years, Marathon’s receipts increased from 41 million
barrels in 1997 to 75 million barrels in 1998 and then fell to 62 million barrels in 1999. There is also
variation across refineries and within a single refinery over time; Chevron’s Richmond refinery near
Oakland received 118 million barrels of Saudi crude over the period, with annual receipts varying
by as much as 24 million barrels from year to year. The nearby (and similarly-sized) refinery in
Martinez, which was owned by Shell for most of the period, processed no Saudi crude.
This variation allows for an estimation of the impact of discount receipts both on refiner profits
and gasoline prices in local markets. Correspondingly, the second result is that most of the discount
rents were captured by refinery owners as profits rather than passed through to consumers as lower
gasoline prices. This is to be expected given that the discount was targeted only at certain refineries,
so the effect on market-level costs tended to be infra-marginal. The capture of rents appears to have
been almost complete, supporting the idea that the discount was purposefully targeted at specific
refiners. Finally, I find that discount receipts affected refining company political contributions, with
recipients targeting more of their financial support to politicians on committees that considered
3
bills of interest to Saudi Arabia. I also find that funds tended to be diverted away from congressmen
who received donations from pro-Israel interest groups.
The results described above tie together several different areas of study. First, although the
Asian price premium (or, U.S. discount) for Saudi crude has attracted occasional comment in the
trade press on petroleum markets, it has received very little attention in the academic energy
literature. Exceptions are several papers that have attempted to explain the premium in terms of
models of price discrimination2 or regulatory distortions3 . Other papers have discussed strategies
for Asian consuming nations to reduce or eliminate the premium through regulatory reform [Ogawa
2003] or by improving pipeline infrastructure [Jaffe & Soligo 2004].
This paper also provides evidence on the incidence of non-marginal cost changes in the oil
refining industry. Borenstein & Kellogg [2012] examine the effect of a similar change in average
input costs caused by the oil glut in the U.S. Midwest beginning in early 2011. They also find that
the relative decrease in local crude oil prices did not pass through into wholesale gasoline and diesel
prices, and conclude that refiners must have received the rents generated by the crude price shock.
This paper complements their analysis by examining a setting with richer variation in cost changes
and showing that not only was the cost decrease not passed through into prices, but that it was
instead captured by refiners as profits rather than by retailers or other market intermediaries.
In addition to documenting the existence and incidence of the discount, this paper also adds
a quantitative dimension to the literature on the political economy of global energy markets. As
far as I know, this paper is the first to provide empirical evidence for the use of Saudi oil not
only directly as a tool of political leverage, but through transfers to American companies. The
political motivations behind Saudi supply patterns have been previously examined from a historical
perspective, notably by Moran [1981] and more recently in a comprehensive history by Jaffe & Elass
[2007]. Moran [1981] argues that OPEC behavior is best understood by looking at the political
motivations of Saudi Arabia between 1973 and 1980. In an unpublished working paper, Jaffe &
Elass [2007] similarly argue that Saudi Aramco’s behavior has often been guided by the kingdom’s
2
Soligo & Jaffe [2000] model Saudi Aramco’s pricing decision as that of a dominant firm operating in two fully
separated markets and assert that Asian/European price ratio is consistent with reasonable values of supply and
demand elasticities and Saudi market shares. In a response, Parsons & Brown [2003] propose an alternative model
of international oil markets as a Cournot duopoly, with Gulf OPEC producers competing in both markets with local
suppliers.
3
Horsnell [1997] argues that government involvement in Asian procurement has been at least partially responsible
for the higher prices Asian firms pay for oil.
4
foreign policy goals; they outline the kingdom’s policy of maintaining a position as the top global
supplier of crude to the United States beginning in 1990, and provide a historical survey that follows
the eroding relationship between the two countries and subsequent policy reversal in 2003.
More generally, this paper also adds to the broader literature on the relationship between
business and politics and the determinants of political giving by corporations. There is a long
literature suggesting that campaign contributions may be a key way for companies to engage in
rent-seeking, including Pittman [1977] and Krueger [1974]. Empirical work by Fisman [2001],
Johnson & Mitton [2003], Sapienza [2004], Faccio [2006], Jayachandran [2006] and Cooper, Gulen
& Ovtchinnikov [2010] indicates that these political connections can be valuable to firms. There
is also some recent evidence on what determines political giving and how firms decide how much
to contribute and how to allocate donations. Notably, recent work by Holland, Hughes, Knittel, &
Parker [2014] finds evidence that firms allocate campaign contributions to support their regulatory
interests. There is also evidence to suggest that political donations are a consumption good for
executives, with contributions increasing in executive pay [Ansolabehere, de Figueiredo & Snyder
2003] and correlated with agency problems within the firm [Aggarwal, Meschke & Wang 2012].
This paper adds to this literature by providing new evidence on how these preferences are formed.
In particular, the Saudi crude discount program is a non-standard lobbying channel that offers
a window into how firm preferences over the amount and allocation of their political giving are
affected by their other business relationships.
The rest of the paper proceeds as follows. Section 2 gives some background on U.S.-Saudi
relations, and Section 3 describes the history of the discount policy and how Saudi Aramco targeted
the discount rents. Section 4 provides evidence on who captured the discount rents, and Section 5
discusses how discount receipts affected political action by U.S. oil refiners. Section 6 concludes.
2
Background: U.S.-Saudi Relations
2.1
Aramco Before Nationalization
Although the discount policy did not begin until 1990, the defense of Saudi political interests
by American oil companies began in the 1930s with the simultaneous foundation of Saudi Arabia
and its oil concession to an American oil company. What followed over the next sixty years was the
5
evolution of a complex web of political and economic connections between the Saudi government,
a consortium of American oil companies, and the U.S. government. Throughout this period, the
Saudi government used its oil to try to influence U.S. foreign policy, with U.S. oil companies acting
as both its willing conduits and active partners.
The Kingdom of Saudi Arabia was founded by King Abdul-Aziz Ibn Saud in September of 1932,
four months after Standard Oil of California (SoCal, now Chevron) discovered oil in Bahrain. The
United States recognized the new country in 1933, the same year that the King granted SoCal an
exclusive sixty-six year concession for oil exploration and production in the al-Ahsa region (FTC
1952). The King hoped that this concession to the geographically distant United States would
protect the kingdom from the more immediate interference of the British [Jaffe & Elass 2007]. In
1936, SoCal joined with Texaco to form the California-Arabian Standard Oil Company (Casoc),
which first struck oil at Dammam Well No. 7 in 1938. Impressed by the discovery, the king extended
Casoc’s original concession to cover a 440,000 square mile area of Saudi Arabia [Sampson 1975,
109]. This discovery also prompted President Roosevelt to charge the Secretary of State with
protecting U.S. interests in the Saudi oil concession [Jaffe & Elass 2007]. Following the subsequent
discovery of three more major oil fields, the company was renamed Aramco (the Arabian-American
Oil Company) in 1944.
The consortium of American partners expanded following the Second World War, with what
later became Mobil and Exxon buying in to the partnership. The kingdom also began to take a
more active role in the company, with Saudi representatives joining the newly-formed Executive
Committee in 1950. To bolster its alliance with Saudi Arabia and to protect the company against
nationalization pressures, the U.S. government arranged a deal in December 1950 to give 50 percent
of Aramco’s profits to the kingdom as a “tax”, which was deducted from the taxes owed by the
consortium companies to the U.S. government [Yergin 1991, 447]. This was the first in a series of
moves by the U.S. government to use Aramco as a channel of influence between the two countries.
By the Arab-Israeli wars of the 1960s and 1970s, the company had become both a source of
leverage for the United States as well as a representative of Saudi foreign policy toward the U.S. In
1967, oil minister Sheikh Yamani warned the U.S. government of the “consequences” of giving aid
or support to Israel, sending his message through Aramco for emphasis [Brown 1999, 268]. Another
account quotes Yamani as saying that “if the United States directly supports Israel, Aramco can
6
anticipate being nationalized ‘if not today, then tomorrow.’ If the U.S. does not stay out of this
conflict, the U.S. is finished in the Middle East” (Foreign Relations of the United States, 1967).
The company managed to resist full nationalization over the next few decades, with the country instead opting to slowly increase its profit share and replace American managers with Saudi
personnel. As Saudi influence over the company grew and the political situation in the Middle
East deteriorated, the American consortium began to feel increasing tension between the interests
of the Saudi and U.S. governments. Following the outbreak of the Yom Kippur War, Aramco’s
American partners advocated for the Saudi political agenda. In May of 1973, the directors of
Exxon, Mobil, Texaco and SoCal called on their contacts at the State Department, the Pentagon
and the White House to urge the government to support the Saudi position in the Arab-Israeli hostilities [Sampson 1975, 292-293]. On June 21, 1973, Mobil published an “advertorial” in the New
York Times calling on the U.S. government to join with the Soviet Union in insisting on a peace
agreement in the Middle East. The article argued that continued American prosperity depended
on U.S. support of Saudi interests in the Middle East, and warned that “political considerations
may become the critical element in Saudi Arabia’s decisions, because we need the oil more than
Saudi Arabia will need the money” (New York Times, 1973). Oil minister Yamani later wrote to
Mobil’s president praising the ad and calling it a “positive step”.4 That July, SoCal’s chairman
sent out his own letter to the company’s employees and shareholders asking them to pressure their
representatives to “acknowledge the legitimate interests of all the peoples of the Middle East” and
encourage Washington to improve relations with Arab governments [Sampson 1975, 294]. On October 12, six days after the start of the Yom Kippur War, the chairmen of Exxon, Texaco, Mobil and
SoCal sent a joint memo to President Nixon’s chief of staff warning that U.S. aid to Israeli forces
would result in “a critical and adverse effect on our relations with the moderate Arab producing
countries” and the loss of American influence in the region “to the detriment of both our economy
and our security” [Sampson 1975, 300]. Indeed, the oil embargo began five days later on October
4
In an extremely bizarre incident, Mobil Oil took out another full-page ad in the New York Times on May 8,
1980 regarding a documentary called “Death of a Princess” that was set to air on PBS two days later. The film
told the story of a young Saudi princess who had been publicly executed for committing adultery. It was considered
extremely unflattering by the Saudi regime, and Mobil (which was one of the major supporters of PBS) called on the
network to cancel the broadcast. The last line of the article read:
“We hope that the management of the Public Broadcasting Service will review its decision to run this film and
exercise responsible judgment in the light of what is in the best interest of the United States.” (New York Times,
1980).
7
17th, and Aramco was compelled to enforce the Saudi boycott of oil sales to the United States and
other Israeli supporters to maintain its concession.
In 1981, American oil companies again inserted themselves in the middle of Saudi-U.S. diplomacy concerns preceding the controversial sale of the AWACS (Airborne Warning and Control
System) surveillance planes, which was proposed by the Reagan administration.5 Once again, the
oil industry launched an extensive political campaign in support of Saudi Arabia. Mobil in particular spent more than half a million dollars on full-page advertorials in the New York Times, again
emphasizing the importance of the economic partnership between Saudi Arabia and the United
States. Mobil’s president also personally called Arkansas senator David Pryor to lobby for the
sale [Bard 2010]. Congress eventually approved the landmark sale, despite strong opposition from
American voters, the State of Israel, and the Israel lobby.
In the meantime, Aramco’s ownership had been slowly transferred to the Saudi government, with
Saudi Arabia completing its purchase of Aramco’s assets by the end of 1980. American interests
continued to manage the company, however, and the final paperwork for full nationalization was
not signed until 1990 [Yergin 1991, 652]. During this time, a precursor to the discount scheme
resulted in Yamani’s termination as oil minister. In 1986, Yamani had been charged by the king
to both increase Saudi production and increase the worldwide oil price. Yamani thought this
was impossible, and instead offered some customers a secret 50 cent-per-barrel discount on their
contract price to try to meet the production quotas. The discount caused overall oil prices to
fall and spurred tensions with other OPEC members, and the King relieved the previously highly
successful oil minister of his post. This action announced the King’s intention to control oil policy
more directly rather than working through OPEC (New York Times 1986, PIW 1986, Financial
Times 1986, Adelman 1995, Jaffe & Elass 2007).
After nearly sixty years of Saudi-U.S. partnership, control of Aramco passed fully to the Saudi
government in 1990. Even without its American partners, the newly-independent Saudi Aramco
would continue in its role as a critical conduit for diplomatic relations between Saudi Arabia and
the United States.
5
James Atkins, former Ambassador to Saudi Arabia, predicted that Saudi Aramco would keep the crude price at
32 dollars if the AWACS sale was allowed to proceed [Adelman 1993].
8
2.2
Diplomacy After Nationalization: “Oil for Security”
Just as the American-owned Aramco served as an intermediary for American and Saudi interests
through the 1980s, Saudi Aramco has continued to work to achieve the kingdom’s foreign policy
goals since nationalization. While the company’s core mission is to maximize oil profits, this goal is
often superseded by the government’s foreign policy agenda. The most significant examples of this
are the use of Saudi Arabia’s considerable excess production capacity to stabilize U.S. oil supplies
and the U.S. export quota policy.
Saudi Aramco has used its excess capacity at key moments to serve U.S. economic interests
during oil supply emergencies. The first such incident occurred just after full nationalization in
1990, when Iraqi forces invaded Kuwait at the start of the first Gulf War. As promised, American
forces were immediately deployed to Saudi Arabia, and President Bush sent a letter to King Fahd
requesting that the kingdom immediately increase production in order to replace lost U.S. crude
supplies from Iraq and Kuwait. The King agreed, and within three months the company had raised
production by 2 million barrels per day (bpd) to 7.3 million bpd. The oil price, which had surged
to 40 dollars per barrel on the news of the invasion and UN boycott, fell immediately back to
pre-invasion levels [Jaffe & Elass 2007]. Another notable instance of this strategic export expansion
occurred in 2002, when Saudi Aramco pledged to replace U.S. supplies lost during the Venezuelan
oil strike. In return for this type of strategic assistance, the United States provided the Saudis with
domestic intelligence, protection against external threats, and advanced weaponry. The American
commitment to protect the Gulf states was first explicitly stated in 19806 and later reiterated by
President Reagan, who specifically articulated a commitment to intervene to protect Saudi Arabia
in particular. This commitment was borne out most clearly at the beginning of the Gulf War, when
American forces arrived in Saudi Arabia less than two weeks after the invasion of Kuwait.
6
7
“Let our position be absolutely clear: An attempt by any outside force to gain control of the Persian Gulf region
will be regarded as an assault on the vital interests of the United States of America, and such an assault will be
repelled by any means necessary, including military force.”
President Jimmy Carter, State of the Union Address, 1980.
7
The United States is also by far the main supplier of military equipment to Saudi Arabia, with U.S. supplies
accounting for 66.4 percent of Saudi arms imports between 1990 and 2010 (SIPRI, 2011). In 1990, President Bush
waived a number of congressional bans to proceed with a multi-billion dollar arms sale to Saudi Arabia that included
F-15s, Stinger and Patriot missiles and launchers, M60A3 and Abrams tanks, Apache and Blackhawk helicopters and
other equipment (SIPRI, 2011). The size of these arms deals continued to increase throughout the 1990s under both
the Bush and Clinton administrations (Figure A.1).
9
2.3
Saudi Interests in the Middle East Peace Process
In addition to security and weapons, Saudi Arabia’s other main diplomatic concern has been
American engagement in the Israeli-Palestinian peace process. Saudi Arabia has long supported
Palestinian claims to sovereignty, and has used its economic and diplomatic relationship with the
United States to pursue this agenda. Indeed, the most overt use of Saudi oil as a political weapon
occurred before 1990, with the threatened expropriation of Aramco in 1967 and the oil embargo
of 1973, both of which occurred in response to American support of Israeli military actions. Since
then, Saudi advocacy has focused mostly on U.S. diplomacy toward Israel.
There have been several notable instances where the Saudi regime has put diplomatic pressure
on the United States to intervene in the Arab-Israeli conflict. Following the surge in violence that
followed the failure of the 2000 Camp David summit, Crown Prince Abdullah reportedly refused
an invitation to visit Washington in June 2001 to indicate Saudi displeasure over insufficient U.S.
efforts to prevent Israeli military action against Palestinians (New York Times, 2001a). Later
that year (and less than two months after the September 11th attacks), the Saudi Foreign Minister
declared that the Saudi government was “angrily frustrated” with the Bush administration’s failure
to engage with the Palestinian leadership in the peace process (New York Times, 2001b). Crown
Prince Abdullah subsequently proposed a new Saudi-backed peace plan, which was adopted by
the Arab League in 2002. Despite repeated diplomatic pressure from Saudi Arabia, however, the
United States failed to pressure Israel to accept the plan and U.S.-Saudi relations continued to
deteriorate. As described in the next section, this deterioration in diplomatic relations coincided
with the abandonment of the discount policy.
3
The Discount Policy
3.1
History
The discount policy began in 1990 as a politically-motivated export quota policy. Following
Aramco’s full nationalization and transition to Saudi control, King Fahd directed the company to
maintain a position as the top supplier of foreign crude to the United States [Jaffe & Elass 2007].
The stated purpose of this policy was to cement Saudi Arabia’s strategic alliance with the United
10
States by making itself critical to U.S. energy security. Achieving the quota on a month-to-month
basis was initially quite straightforward; Aramco supplied as much as a third of total U.S. imports
in the early 1990s while keeping the U.S. price for Saudi crude roughly the same as the world price.
By the end of the decade, however, increasing competition from Mexico, Venezuela and Canada
forced Aramco to cut its U.S. prices to defend its market share (Figure A.2). The discount reached
a high of $6.30 per barrel in 2001, 30 percent of the U.S. selling price at the time (Figure I). The
total value of the discount over the policy period from 1991-2003 was 8.5 billion dollars.
As the discount became larger, the Saudi government faced increasing pressure from Aramco
to abandon the expensive quota policy.8 Faced with a string of political disappointments in the
early 2000s, the Saudi government also began to question the benefits of a close alliance with the
United States.9 Saudi Arabia also began to feel the political strain of its relationship with the
United States. Although Saudi Arabia declined to participate in Operation Iraqi Freedom in 2003
(and voiced its disapproval of American intervention in Iraq), it became a target of a string of Al
Qaeda attacks in the aftermath of the war. Tensions again worsened after the 9/11 report described
alleged Saudi links with the hijackers and funding sources from within the Saudi government. At
the same time, Saudi pressure for the United States to intervene in the escalating Arab-Israeli
conflict went unheeded, and President Bush declined to intervene with Israel in support of King
Abdullah’s proposed peace plan. In light of the lackluster geopolitical response of the United States
to Saudi interests, the commercial costs of the export policy seemed too great.
The policy was suspended in 2003, with Aramco allowing the average discount to fall to zero
and ceding the top exporter spot to Canada.10 Saudi crude exports, which had averaged 15.9
percent of total gross imports in the first half of 2003, fell to 13.1 percent in the second half of
the year and then to 11 percent in the beginning of 2004. Industry observers noted the apparent
policy change11 , and Saudi Aramco President Abdullah Jumah publicly acknowledged the shift in
8
For more details see Jaffe & Elass [2007].
Saudi arms purchases also fell dramatically after 2000, dropping from 1.7 billion dollars a year from 1990-1999
to .5 billion from 2000-2011.
10
At the same time, all American troops were withdrawn from Saudi Arabia in September 2003.
11
“Has Saudi Arabia abandoned its pivotal policy of being the largest crude supplier to the U.S.? The world’s
top oil exporter has long had the biggest share of the world’s largest oil market and has favored volume over price
to retain the top spot. Intriguingly, since the spring of last year, in the wake of the U.S.-led invasion of Iraq, the
Saudis have ceded their prime position to others.”(“Saudis Drop Volume, Lose U.S. Top Spot” Petroleum Intelligence
Weekly, May 12, 2004.)
9
11
February 200512 .
3.2
Discount Distribution
The prices that refiners pay for Saudi crude deliveries are determined by Aramco’s Official
Selling Prices, or OSPs. Aramco’s OSP announcement is made about a month in advance of
delivery, and consists of a differential relative to the spot price of a different benchmark crude
for each of the four markets. Over the policy period, these benchmark crudes were West Texas
Intermediate (WTI) for the United States, dated Brent for Europe and the Mediterranean, and
the Oman-Dubai average (ODA) for Asia. These differentials reflect factors such as the quality
differences between the marker crude and the Saudi crudes, transportation costs,13 and product
prices in different markets. The price that refiners pay is then calculated using the previously
announced differential and the average price of the marker crude over the month.14
For example, the January differentials for Arab Light to the United States and Asia were -4.05
and +0.25, respectively. During January, the average WTI price was 19.48 dollars per barrel, and
the average spot prices of the Oman and Dubai crude markers was 18.34 dollars per barrel. The
price paid by U.S. refiners for January deliveries of Arab Light was 19.48 − 4.05 = 15.43 dollars per
barrel, and the price for Asian refiners was 18.34 + 0.25 = 18.59 dollars per barrel. The value of
the “discount” that month was therefore the difference between the realized Asian and U.S. prices,
i.e. 18.59 − 15.43 = 3.16.
Discount = Asian Price − U.S. Price
= (ODA + Asia Differential) − (WTI + U.S. Differential)
= (18.34 + 0.25) − (19.48 − 4.05)
= 18.59 − 15.43
= 3.16.
12
Jad Mouawad and Simon Romero, “Saudis in Strategy to Export More Oil to India and China,” The New York
Times. February 18, 2005.
13
Although Aramco considers differences in transportation costs in determining the price differentials, all OSPs
reflect fob prices and do not include transportation.
14
Figure A.3 shows a sample OSP release from Platts from December 2001 announcing the pricing differentials on
January deliveries.
12
In addition to dictating the prices that refiners pay, Saudi Aramco also directly controls how
much crude each refinery receives each month. Saudi Aramco only sells crude through long-term
contracts with specific refineries, and crude resale is not permitted. Although many petroleum
producers trade their crude on a spot basis and/or allow their crude to be traded by buyers ex
post, Saudi Aramco does neither and keeps tight control over where its crude is processed. Any
resale by refiners results in permanent blacklisting from future contracts, and even companies
that refine Saudi crude in multiple locations must specify the particular refinery where they will
process each shipment. Refineries must therefore be prepared to either process or store whatever
quantity they receive in a particular month. Contract quantities are specific to refinery capacity to
process Saudi crude, and potential buyers submit detailed contract applications to the Crude Oil
Sales Department in Dhahran that provide refinery specifications and operating figures as well as
audited balance sheets and detailed corporate profiles. Once approved, these contracts may then
be renewed on an annual basis at the discretion of the refinery and Saudi Aramco.
Despite the rigidity of the contract rules for refiners, the actual delivered quantities vary significantly from month to month. While annual contracts specify the quantity that refiners are
willing to buy in terms of barrels per day, deliveries are subject to unilateral cuts by Aramco.
Quantities may be cut, for example, following a decrease in OPEC quotas or a mandate from the
Saudi government.15 Indeed, there is significant monthly variation in the quantities that buyers
receive each month, both at the refinery and company level. Although Aramco frequently changes
quantity deliveries, it is very unusual for refiners to cut their orders even when prices are high.16
Refinery inputs are constrained by physical capital in the short run, and most refineries require the
composition of their input blend to be fairly consistent. Refiners can replace a shortage of Saudi
crude by buying a fairly close substitute or higher-grade crude on the spot market, but these prices
are usually higher than the Saudi term contract prices. Aramco also tries to keep its contract terms
15
In practice, Saudi Aramco can at its discretion give some flexibility to buyers to switch between the several
crude grades it produces or to ask for a change in the total quantity. In the month prior to delivery, Saudi Aramco
announces its price differentials. Customers then indicate their preferences, and Saudi Aramco takes their requests
into account when issuing the final allocation.
16
“Despite their grumbling, customers - sensitive to the strategic importance of Saudi contracts - will not dare
threaten to reject their March supplies. ‘We can only hope that we’re compensated next month with a deep discount
on the April formula,’ says an industry source. . . . One US contract holder, none too pleased with the March price
adjustment, carefully described the change as ‘very inconsistent with the market.’ . . . . Again, however, loyalty to
Aramco seems certain to win out over momentary irritation. ‘We are painfully aware of the new prices, but we’re
dealing with it,’ the customer added.” Petroleum Intelligence Weekly, “Buyers Bemoan Saudi Price Hikes, Eyes Now
on Iraq” (Feb. 8, 1999).
13
fairly attractive on average so that customers want to renew their contracts over the long term and
so that they can continue to place their desired quantities in the U.S. market.
4
Where Did the Discount Go?
The first task of this paper is to track down where discount rents accrued when Saudi Aramco
sold discounted crude to the United States, whether captured by refiners as excess profits or passed
on to consumers in the form of lower product prices. I next provide an overview of the relevant
features of the U.S. refining industry and a basic model to give some intuition on the incidence of
the discount. I then discuss the data required for this part of the analysis, the empirical framework,
and the results from the estimation.
4.1
Market Details
At their most basic, refineries blend input crude oil and then distill it into its constituent hydrocarbons, isolating the molecules and then blending them into end products like propane, gasoline,
jet fuel and diesel. Although all refineries perform this same basic process, they vary a great deal
in their complexity and in what sorts of crude oil they can process and what finished products
they output. Topping refineries, the most basic type, include only distillation units and produce
mainly unfinished oils. Hydroskimming refineries add a hydrotreating and reforming unit to the
basic topping refinery configuration, allowing the refinery to remove sulfur from more sour crudes
so that outputs conform to environmental standards. The most versatile (and most expensive)
refinery type are catalytic cracking or coking refineries, which also feature gas-oil conversion plants,
olefin conversion plants, and coking units to reduce or eliminate the production of residual fuels.
These refineries are able to break larger (and less valuable) molecules and reform them into lighter,
more valuable products like gasoline and jet fuel. The product mix is determined both by the blend
of molecules in the input crude mix as well as the sophistication of the refinery.
Most refineries blend different crude oils before distillation begins. This allows them to maintain
consistent processing conditions and mitigate the corrosive effects of cheaper sour crudes. Refiners
periodically run linear programming models to determine the optimal quantity and quality of their
inputs and outputs and to make (small) adjustments to their refinery operating parameters. Most
14
secure a certain amount of “baseload” crude under term contracts, which are less flexible but offer
more attractive prices, and then balance their remaining crude slate in the spot markets. If a
refinery gets a cut in its monthly crude order from Saudi Aramco, then, they will buy other similar
crudes on the spot market to fill the shortfall rather than curtailing their run. Mexican Isthmus-34,
for example, is a good substitute for a refinery using Arabian Light in its crude mix and is available
on the spot market at most locations where Saudi crude can be delivered. Many refineries can also
substitute to higher-quality crudes if necessary. Spot market prices for these inputs are usually
higher than Saudi term contract prices, however, so replacing a shortfall of Saudi crude on the spot
market tends to increase input costs. Because Saudi crude is delivered to the U.S. by ship, refiners
have several months to secure a replacement crude once they receive notification of a cut in their
delivery.
Once the crude oil has been fully processed, refined products are transported to wholesale racks
by either pipeline, barge, truck or rarely by railroad [Association of Oil Pipelines 2009]. Pipelines
are the least expensive way to move products,17 and mainly connect areas of high refining output
with those of high demand. Trucks usually make only local trips, and most trips by petroleum tank
trucks are no longer than 50 miles [Untiet 1984]. Refiners sell products to retailers out of these
local wholesale racks, who then mark up the prices to sell to consumers. These retail margins make
up only a small proportion (about ten percent) of consumer prices.18
The two primary possible destinations for Saudi discount rents are therefore refiner profits and
consumer prices.
4.2
Framework
Although changes in crude oil costs certainly influence the consumer price of gasoline,19 there
is reason to think that discounts targeted at specific refineries will be captured as profits rather
17
Transportation costs by pipeline, barge and truck are estimated at 2, 4.5 and 35 cents per gallon per thousand
miles of transportation [Jacobs 2002].
18
See http://www.eia.gov/petroleum/gasdiesel/ for the components of the price of a gallon of gasoline.
19
The literature on the nature of the dynamics of crude price pass-through into gasoline prices began with Bacon
[1991], though the link between crude and gasoline prices was already well-established. Further detailed work on
the magnitude and speed of gasoline price responses to crude prices can be found in Borenstein & Shepard [1996],
Borenstein, Cameron & Gilbert [1997], Borenstein & Shepard [2002], and Bachmeier & Griffin [2003]. Even though
several of these papers find that the pass-through of crude price changes to gasoline prices is not immediate, full
pass-through of all common shocks occurs within a month or two of the initial shock. Because I examine pass-through
at the annual level, I avoid these dynamic pass-through concerns and assume that all discount effects pass through
to gasoline prices in the year that they occur.
15
than passed on to consumers, even in a competitive market.20 Over the policy period, Saudi oil
accounted for about ten percent of crude refined in the United States, and crude receipt varied
substantially across refineries in the same local area. This section therefore models the receipt of
Saudi crude as a heterogeneous cost shock across refineries in a given market.
The gasoline wholesale market is characterized by a relatively small number of refineries with
fixed capacities. At most, cities have 12 refineries that supply refined gasoline to wholesale racks
for retail distribution. Most refineries operate at full capacity for most of the year, and even with
regular shutdowns for maintenance overall capacity utilization is around 90 percent. Though fixed
costs are large, the majority of marginal costs are the price of crude inputs. These vary across
time as the price of crude fluctuates, and across refineries according to refinery sophistication.
While common shocks to costs do pass through into gasoline prices, cost shocks targeted to specific
refineries will tend to affect profits but not prices. To make this clear, I consider a simple model
of the wholesale refined product market characterized by a finite number of refineries with fixed
short-run capacities.
Consider a city (or wholesale market) with n possible refineries with marginal costs ci such that
c1 ≤ c2 ≤ · · · ≤ cn and capacities k1 , . . . , kn . Demand is given by D(p), where p is the gasoline (or
product) price. In this market, equilibrium will consist of an m and p∗ such that
P
1. No additional firms want to begin producing: cm+1 ≥ D−1 ( m ki )
2. No producing firms want to exit the market: D−1
P
m−1
ki > cm
3. The marginal refinery (m) is maximizes profits:
h
i
P
p∗ = argmaxp:Dr (p)≤km (p − cm ) D(p) − m−1 ki
The equilibrium price p∗ therefore comes from the marginal firm’s profit maximization facing the
residual demand from all lower-cost firms producing at capacity.
Therefore (for non-corner solutions):21 p∗ = 1+cm1 where rD is the elasticity of residual demand:
r
D
Pm−1
r
∗
D (p) = D(p) −
ki . Note that p is only a function of the marginal firm’s marginal costs,
20
Borenstein & Kellogg [2012] find that refiners, not consumers, captured the rents generated by a temporary
depression in crude prices in the Midwest. As in this section, their model generates this lack of pass-through into
product prices as a consequence of the fact that most refiners are capacity-constrained, and are therefore operating
at the vertical part of their supply curves.
21
For solutions where the marginal firm produces at capacity, decreases in costs will not affect prices even for
marginal firms. Increases will sometimes increase prices if they are sufficiently large.
16
i.e. p∗ = p∗ (cm ). The profit of the marginal firm is therefore: πm (cm ) = (p∗ (cm ) − cm )Dr (p∗ (cm )).
For inframarginal firms (i < m): πi (ci , cm ) = (p∗ (cm ) − ci )ki . Therefore non-marginal shocks to
marginal cost will only affect refiner profits (πi ) and will not pass through into prices. Decreasing
the costs of the marginal firm (cm ) will lower prices, increase profits of the marginal firm (πm ) and
lower profits for other firms (πi ).22 Consequently, this simple model shows that we would expect
most receipts of discounted crude to affect refiner profits and not to pass through into product
prices. Since most refiners are inframarginal, most of the discounted crude is also inframarginal;
only changes in the amount of Saudi crude sold to the one marginal producer in each market would
even partially be passed through into the price.
4.3
Data
In this part of the analysis, I follow the discount through each stage of the refining process from
the delivery of the discounted crude to the refinery to local gasoline prices.
4.3.1
Total Discount Value
The key independent variable for the analysis in this paper is the refinery-level value of the
discount received from Saudi Aramco, i.e. the per-barrel discount multiplied by the quantity of
crude delivered to each refinery. I calculate the per-barrel discount that U.S. refiners receive relative
to Asian refiners using Saudi Aramco’s official selling prices. As described above, Saudi Aramco
takes a variety of measures to prevent the emergence of a secondary market in its crude, so these
prices should accurately reflect the prices that refiners pay to purchase Saudi crude oil. The OSP
releases are published each month by Platts for each of the major crude grades in each market,
and the time series of the realized selling prices (these differentials plus the relevant benchmark)
are available from Bloomberg. The series for Arabian Light, Arabian Medium, and Arabian Heavy
are available for Europe, the United States, and Asia beginning in January 1991, and I use these
series to construct a single average discount for Saudi crude.23 I use the difference between the U.S.
22
See Figure II for a simple example of how a common shock to costs (such as a change in the world price of
crude) would affect both profits and prices, but a targeted decrease in costs to inframarginal refineries (such as the
discount) would affet profits but not product prices.
23
Since only overall sales volumes for all the crude types are available, I calculate the discount as a weighted
average of the discounts on the three different crude grades. The weights are left fixed over time and are based on
the volumes published in the IEA Oil Market Report over the 1998-2003 period. The average discount is calculated
17
price and the Asian price to approximate the discount for several reasons. First, East Asia was the
primary alternative destination for Saudi crude exports. Asian markets received about half of total
Saudi crude exports by 2003 and 65 percent by 2010. At the same time, the share that went to the
North American market fell from what had been a steady 25 percent from 1991 to 2003 down to
18 percent in 2010.24 Additionally, the differential against the Asian price returned to zero once
the discount policy was abandoned in 2004, supporting the idea that the Asian and U.S. prices
for Saudi oil would likely have been the same in the absence of the policy. For these reasons, the
U.S.-Asian price differential represents a meaningful approximation of both the cost to Aramco of
pursuing the policy as well as the benefits to U.S. refiners.25
In order to estimate the effects of the discount on refinery profits and local retail prices, it is
critical to identify which refineries received crude shipments from Saudi Arabia in each month.
All of this information comes from the Energy Information Administration (EIA), which monitors
domestic petroleum refining operation and which collects data on all foreign crude imports through
the EIA-814 Monthly Imports Report. This data is publicly available at the refinery-level from the
EIA starting in 1986. I match this data to a list of all operating refineries in each year constructed
using the EIA Refinery Capacity Report (from EIA-820), which lists all operating U.S. refineries
in each year for 1994, 1995, 1997, and 1999-2010 and reports capacity, owner, and state for each
refinery. There are 178 refineries listed in the report. When matched with the importer data, this
lists all refineries operating in a given year and how much crude each refinery imported from Saudi
Arabia and elsewhere.
Receipts vary a great deal across refineries and within each refinery from month to month, which
leads to considerable variation in discount values among refinery owners and across states. Of the
178 operating refineries, 58 used crude imported from Saudi Arabia between 1991 and 2003. Most
deliveries of Saudi crude go to refineries along the Gulf Coast, but a surprising amount also goes
to inland refineries and to refineries in California and along the East Coast (Figure III). Coastal
refineries are mostly supplied by tanker and offshore terminal, and inland refineries receive deliveries
as
AvgDiscountt = 0.35 · LightDiscountt + 0.46 · M ediumDiscountt + 0.19 · HeavyDiscountt .
24
Export figures are from United Nations Commodity Trade Statistics Database.
The main results are also robust to using the European price rather than the Asian price as a benchmark to
approximate the discount value.
25
18
by pipeline within a couple of weeks of delivery by tanker.26
4.3.2
Refiner Profits
To estimate the impact of the Saudi crude discount on refiner profits, I use data on annual
profits from refining operations for the 40 publicly-traded companies that owned at least one U.S.
refinery during the sample period. These data come from Standard and Poor’s Compustat North
America dataset. Because many U.S. refineries are owned by large corporations with multiple
business lines, I use profits data from the Business Segments Dataset, which includes companies’
self-reported balance sheet data by business type. The relevant segments were identified using the
segment-specific NAICS classification for petroleum refining and the segment name as reported
by the company. The analysis focuses on operating profits, which represents sales of the refining
business segment less its allocated share of operating costs and expenses. For comparison, I also use
overall net income as a measure of total company profits. This is also available from Compustat,
and represents quarterly income (or loss) after subtracting all expenses and losses from all revenues
and gains. The reporting for this item is more consistent across companies, but is a much noisier
indicator of the variable of interest.
In order to link company profits to refinery imports, it is also necessary to determine annual corporate ownership of each refinery. Ownership of individual refineries was established using the EIA
Refinery Capacity Reports and supplemented using corporate profiles from the Moody’s/Mergent
Industrial Manuals. Refineries owned through a join enterprise were assigned to either the U.S.listed corporation or to the majority stakeholder.27 Refineries that changed ownership mid-year
were assigned to the company that owned the refinery for the majority of the year.28 Discount
receipts for these companies also vary tremendously, and are not necessarily related to company
size. Chevron and Texaco, for example, received over 4 billion dollars worth of crude discounts,
an amount equal to 10% of their overall refining profits. The similarly-sized Shell, however, re26
There is also a great deal of variation in the fraction of inputs that come from Saudi Arabia. The Texaco
refinery in Delaware, for example, got more than half of its inputs from Saudi Arabia over the period. Mississippi
and Arkansas also got over half of their crude inputs from Saudi Arabia (Figure A.4).
27
Texaco is an exception to this, and is excluded from the regressions due to its Motiva joint venture with Saudi
Aramco. In general, including Texaco in the regressions either has no effect or slightly decreases point estimates and
increases their statistical significance.
28
Table A.1 provides some summary statistics on the publicly-traded refinery owner companies that appear in the
profits analysis.
19
ceived only 0.08 billion in discounted crude, 0.2% of total refining profits. Smaller companies like
Marathon and Valero received about a billion dollars in discounts over the period as well. On a
yearly basis the value of discount receipts relative to profits could also be quite large: in 2001,
Valero received 0.25 billion in discounted crude, an amount equivalent to a quarter of its total
refining profits that year. Although a substantial share of the discounted crude went to the former
Aramco partners (ExxonMobil, Chevron and Texaco), other companies like BP, ConocoPhillips,
Marathon and Valero were also top recipients. There is also great deal of variation in the fraction
of inputs that come from Saudi Arabia. Texaco, for example, received Saudi imports at seven of
its ten refineries and devoted approximately 45 percent of its processing capacity to refining Saudi
crude.
4.3.3
Retail Gasoline Prices
In addition to examining the effect of the discount on refining profits, I also look at the impact
on consumers though changes in the retail price of gasoline in local refinery markets. For this part
of the analysis, refineries were matched to the largest city within an hour travel time by road.
Monthly retail price averages for the 75 cities with local refineries was provided by the Oil Price
Information Service (OPIS). OPIS uses data from credit card receipts to capture daily stationspecific retail gasoline prices for up to 120,000 stations throughout the United States, and their
data include prices for most major retailers regardless of ownership. This daily station data for
regular unleaded gasoline was aggregated up to the city-month level for this paper.
I link this price data for each city to refining capacity and Saudi crude quantities in its wholesale
market. I define a city’s market in two different ways. In the first, local markets are defined using
the assumption that refineries serve only cities within an hour travel time by truck. I alternatively
define a refinery’s market as any city that is “down-pipe” of the refinery using directional product
pipeline information from 2004 (Figure A.6).29 Under the first definition, for example, retail prices
in St. Paul and Chicago are assumed to be affected only by the refineries operating in their local
area. The second market definition takes into account that there is a product pipeline going from
Bismarck to St. Paul, and from St. Paul to Chicago. Prices in Bismarck are still only affected by
29
This is similar in spirit to Muehlegger [2006], who also uses product pipelines (as well as truck and barge access)
to calculate the approximate transportation costs for a refinery to serve markets in each state.
20
refineries in the local area, but St. Paul prices are now also affected by refineries in Bismarck as
well as in St. Paul, and prices in Chicago by refineries in Bismarck and St. Paul.
4.4
Empirical Framework and Results
In this section, I investigate what refiner characteristics determine receipt of Saudi crude, and
where discount rents went. To determine whether the discount was captured as profits or passed
through to consumers, I estimate the effect of discount receipts on refinery owner profits and local
retail gasoline prices.
4.4.1
Who received the discount?
To get a sense of which features made companies most likely to receive Saudi crude, I first
construct a dummy variable (D(Recipient)) equal to 1 if the company received Saudi crude at
any point over the policy period. I estimate a simple descriptive linear probability model for this
dummy variable on a set of refiner characteristics. These include total refining capacity over the
period as well as a dummy variable indicating whether the company owned any refineries that had
processed Saudi crude in the 1989/1990 period, indicating that the company had the technical
capacity (without any further investment) to process crude from Saudi Arabia. I also construct a
set of dummy variables that indicate whether the company owned a refinery in a particular state
to get at geographical effects. Anticipating the results in section 5, I also include a set of variables
indicating political contributions by these companies in the 1989/1990 period to check whether
recipients differed from non-recipients in their political leanings.30
The results from the LPM for discount receipt on refiner characteristics (Table I) reveal that the
most important requirements to have received Saudi crude are technical. Companies that owned at
least one refinery with demonstrated capacity to process Saudi crude were 65 percent more likely
to have received Saudi crude over the policy period. Of course, the availability of substitute crude
inputs means that these firms do not require Saudi crude, and that refineries that did not process
Saudi crude in the pre-period may still have had the capacity to do so. Firms may also have
updated refinery specifications during the period to enable Saudi crude capacity. Nonetheless, this
variable is significant in all seven specifications and its magnitude is quite consistent. There is no
30
The political contributions data are discussed in detail in section 5.
21
indication that firms with larger overall refining capacity were more likely to receive Saudi crude,
and these coefficients are close to zero and statistically insignificant. Former Aramco partners were
surprisingly no more likely to get Saudi crude than other refiners controlling for technical refining
capacity.
The coefficients for the state dummy variables are not reported in the table, and are jointly
significant but mostly individually not statistically significant. These estimates tend to be positive
for states that are easily accessible by barge (e.g. Alabama, Louisiana, California, New Jersey) and
negative for those that are difficult to reach (e.g. Alaska, Arizona, Illinois, Nevada). (Including
these dummy variables in the other LPM regressions has very little effect on the point estimates,
but makes the standard errors on the other coefficients larger.)
The estimates on the aggregates for overall contributions, contributions to politicians and contributions to committees are all small and statistically insignificant. The aggregates by politician
characteristics (party, pro-Israel donation recipients) are also very small and close to zero, indicating that the overall contribution patterns of recipients and non-recipients were very similar in the
pre-period. Recipients and non-recipients do appear to have varied in how their contributions were
allocated across members of various committees, however; companies with more contributions to
members of the House Armed Services and Senate Foreign Relations committees were more likely
to receive Saudi crude. Companies with more donations to members of the House Energy, House
Foreign Affairs, and Senate Armed Services committees were less likely to get Saudi crude.
4.4.2
Refining Profits
To determine the extent to which the discount was captured as profits, I estimate the relationship
between company-level refining profits and total annual discount value as well as the relationship
between log profits and log quantity:
πjt = β · DiscV aluejt + αj + γt + jt
saudi
log(πjt ) = β̃ · log(qjt
) + α̃j + γ̃t + ˜jt
(1)
(2)
The motivation for these two specifications comes from the expression for the refiner’s profits. The
value of the discount to a refiner can be expressed in terms of the percentage discount on the crude
22
price (dt ), the per-barrel price of crude (ct ), and the number of barrels the refiner receives from
saudi . We can then write the refiner’s profit function:
Saudi Arabia in year t, qjt
πjt = (pt − ct )qjcap + β · DiscV aluejt + jt
saudi
= µt qjcap + β · dt · ct · qjt
+ jt
where pt is the average product price and qjcap is refinery capacity. µt is the difference between
the average product price and the cost of crude, i.e. the refining profits per barrel or the “crack
spread”.
In this case, β can be consistently estimated using firm fixed-effects as long as µt is fixed over
time and does not vary with the crude price, i.e. µt = µ. This leaves us with:
πjt = µqjcap +β · DiscV aluejt + jt
| {z }
(3)
αj
This was approximately true from 1983-2003, with the crack spread, which estimates the value
added by the refining operation, remaining fairly constant in real terms at around 12 dollars per
barrel (Figure A.5).31
A more serious empirical problem is created by integrated refiners. Since these firms also sell
crude, the crude price now enters the calculation for the discount value as well as the profitability
of their exploration and production business lines:
saudi
πjt = µ · qjcap + β · dt · ct · qjt
+ ct · qjcrude + jt
{z
} |
{z
}
|
DiscV aluejt
ηjt
Now Cov (DiscV aluejt , ηjt ) > 0, and β̂ overestimates the effect of the discount on refiner profits.
Intuitively, the discount per barrel can only be large when the price of crude is also large – you
cannot have a six-dollar discount when the crude price is only five dollars. We would also expect
refiners that produce crude to have higher profits when the oil price is higher. Even though I use
31
The crack spread is calculated as the difference in cost between a barrel of crude and a representative mix of
typical outputs. For example, a simple version (the 2-1-1) is calculated as the difference in cost of 2 barrels of crude
and a barrel of gasoline and a barrel of heating oil/diesel. Here I use the 6-3-2-1 crack spread, which is the difference
in the total cost of 6 barrels of crude and outputs of 3 barrels of gasoline, 2 barrels of heating oil/diesel, and 1 barrel
of residual fuel oil. Trends are the same for each of the four standard spreads.
23
refining profits in this analysis to try to mitigate this effect, these self-reported segment profits are
likely to co-move with overall company profits.
One solution to this problem is to use a log-log specification and add time fixed effects to remove
time-varying factors (ct , dt ) from the discount. The coefficient β̃ now estimates the percent change
in profits associated with an increase in the quantity of Saudi crude that a refiner receives.
saudi
log πjt = β̃ · log(qjt
) + α̃j + γ̃t + ˜jt
(4)
Since the discount per barrel is mostly positive over this period, a positive estimate indicates that
an increase in the total discount value to a refiner is associated with an increase in firm profits.
Results from the profit regressions (Table II) indicate that most of the discount appears to be
captured by refiners as profits. On average, the discount value is equal to about two percent of
refining profits, and the average discount is one dollar per barrel over the period. Full capture
would therefore be consistent with a coefficient of 0.02 in column 1. The actual point estimate is
0.016, i.e. a 1.6 percent increase in profits associated with doubling the amount of Saudi crude
delivered. When the sample is restricted only to observations with non-zero amounts of Saudi crude
and positive profits, the point estimate increases to 0.17 relative to the full pass-through benchmark
of 0.09 for this subsample.
The point estimate from the level regression in column 3 is large, implying a three dollar increase
in profits for every dollar of discount, and may be biased upward for the reasons discussed in the
previous section.32 Nonetheless, the test of this coefficient against the full pass-through benchmark
of 1 cannot be rejected.33
To support the interpretation that the receipt of discounted crude increases profits through
reducing costs rather than by increasing output, I also estimate the effect of annual district-level
Saudi crude receipts on total refinery output.34 Because events like refinery shutdowns and capacity
32
Consistent with the hypothesis that the upward bias of the estimate in the level regression is due to measurement
error due to other business lines, the point estimate in column 3 decreases to 2.48 when I exclude the supermajors
(BP, ExxonMobil, Chevron and Shell), which had the largest incomes from crude production. The estimates in
columns 1 and 2 decrease slightly but remain statistically significant.
33
Controlling for refiner size (as captured by refining capacity) does not affect the main results in these regressions
or in the gasoline price regressions in the next section.
34
Although refining capacity is available at the refiner level, capacity utilization rates are only released at the
district level. The EIA defines twelve such refining districts: East Coast, Appalachian No. 1, Indiana-IllinoisKentucky, Minnesota-Wisconsin-North and South Dakota, Oklahoma-Kansas-Missouri, Texas Inland, Texas Gulf
24
expansions are likely to affect both production levels and inputs, I control for total district refining
capacity. I use time fixed effects to capture trends in capacity utilization rates and district fixed
effects to control for district-level heterogeneity in refinery efficiency.
Table III reports the results from this regression of the log of production on Saudi crude receipts
controlling for refining capacity and year and district fixed effects. These estimates are all very
small; a doubling in the Saudi crude delivery is associated with around a 0.3 percent increase in
production even though about ten percent of crude inputs came from Saudi Arabia. This is in
contrast to the estimate in column (1) of Table III, which indicates a 1.6 percent increase in profits
for the same increase in Saudi crude quantity. The primary channel for the impact of the receipt of
discounted crude on profits is therefore the discount on costs rather than an increase in production.
4.4.3
Retail Gasoline Price
As for profits, I examine the extent to which the discount was passed through to consumers by
estimating the effect of the city-level discount per gallon of refined gasoline and quantity share in
production on local gasoline prices:
RetailP ricemt = ρ · DiscGalmt + λm + ηst + mt
saudi q
+ λ̃m + η̃st + ˜mt
log (RetailP ricemt ) = ρ̃ · log mt
cap
qmt
The retail gasoline price can be expressed as a per gallon markup on the crude costs per barrel
(pgt ) plus some state-level tax τst . The observed price is this counterfactual price (ηst ) plus the
effect of the discount, which is expressed here as the discount value per gallon of gasoline refining
capacity multiplied by ρ, the share of the discount which is passed on to consumers. This yields:
RetailP ricemt = pgt + τst +ρ · DiscGalmt + mt
| {z }
(5)
ηst
Using the constant crack spread µ, pgt can be expressed as a markup µ over the price of crude ct
Coast, Louisiana Gulf Coast, North Louisiana-Arkansas, New Mexico, Rocky Mountain, and the West Coast.
25
divided by the number of gallons of gasoline per barrel:
pgt =
µ + ct
19.5
where 19.5 is the average number of gallons of gasoline refined from a 42 gallon barrel of crude oil.
The discount per gallon is calculated as the total discount to refineries in market m divided by the
total number of gallons of gasoline refined in that market:
DiscGalmt =
saudi
dt ct qmt
cap
qm
As discussed earlier, one of the primary challenges here is defining the local market in order to
link refineries to the appropriate retail gasoline prices. I do this two ways: first by assuming that
refineries only serve cities within an hour travel by truck, and second by linking each refinery to all
saudi and q cap are calculated.
cities “down-pipe” of the refinery. These two definitions affect how qmt
m
Under the first definition, they refer to Saudi imports and refinery capacity only in refineries within
an hour’s travel from the city. Under the second, they refer to imports and capacity of all refineries
up-pipe.
Again, taking logs and using state by month fixed effects removes any bias caused by the crude
price appearing in both the counterfactual gasoline price and the value of the discount.
saudi
cap
log (DiscGalmt ) = log(dt ct ) + log(qmt
) − log(qm
)
| {z }
| {z }
γt
λm
The primary specification is therefore
log (RetailP ricemt ) = ρ̃ log
saudi
qmt
cap
qm
+ λ̃m + η̃st + ˜mt
(6)
The pass-through coefficient ρ̃ is therefore identified off of variation in the changes in Saudi crude
shares across cities within the same state.
The results from these regressions show no evidence that discounts were passed on consumers in
the form of lower refined product prices (Table IV). Estimates in column 1 give fairly precise zeros
(with standard errors clustered at the city as well as the state by year and month level), even when
26
compared with a full pass-through benchmark of -0.0435 . The full pass-through benchmark for cities
that receive positive amounts of Saudi crude (column 2) is -0.15, so these estimates are also very
close to zero. Large standard errors on the estimates in column 3 make their interpretation more
difficult, but the point estimates are nonetheless quite small. Panel A of column 3, for example,
indicates that a one-cent increase in the per-gallon discount value decreases gasoline prices by 0.03
cents. These estimates are consistent with the large pass-through into refining profits in Table II.
Taken together, the profits and retail price results indicate that, as expected, the discount
rents were captured by refining firms as profits rather than passed on to consumers. Since refining
companies appear to have been the primary beneficiaries of the discount, the next section examines
how discount receipts affected political action by these firms.
5
Political Action by Discount Recipients
While it is impossible to directly measure the effect that the discount actually had on U.S.
foreign policy toward Saudi Arabia, we can observe how the companies that received the discount
behaved as a result of the policy. As described earlier in the paper, there are plenty of cases where
American oil companies took (sometimes rather extraordinary) political action to support Saudi
interests. The most easily measurable type of corporate political action is financial contributions
to political campaigns. In this section, I examine the relationship between discount receipts and
political contributions by American refiners during the policy period. In particular, I focus on refiner
contributions to members of Congress, particularly those who serve on key committees and those
who appear to be most sympathetic to Israeli political interests. It is important to note that direct
contributions to politicians from individual donors represent only a fraction of financial support
from companies to politicians, and a smaller fraction still of total political action by corporations.
Overall, individual donors account for only about half of the money that goes to House candidates
and two thirds of the money that goes to Senate candidates [Center for Responsive Politics 2012].
The rest comes from PACs and candidates’ personal resources. The advantage of focusing on
individual donations is that it shows the direct link from refiners to politicians so that patterns of
giving can be examined in addition to overall levels of giving. Patterns that are seen in this small,
35
The average unleaded gasoline price is 1.34 dollars, and Saudi inputs make up about 5.5 percent of total inputs,
so a one percent increase in the Saudi input share would decrease prices by 0.04 percent.
27
transparent part of political action by discount recipients may be suggestive of overall patterns of
behavior as it relates to the receipt of the crude discount.
5.1
Saudi Arabia and the U.S. Congress
There is a large amount of work in the public choice literature that suggests that campaign contributions may be a key way for companies to engage in rent-seeking.36 This is particularly likely
to happen in industries where the potential benefits are concentrated among a small number of
players, as they are in the refining industry. However, the empirical results in the literature on political contributions and policy influence are mixed. In particular, despite the conventional wisdom
in political economy, it has proven very difficult to show direct causal links between contributions
and voting behavior, though there is some evidence that regulatory outcomes are influenced by political contributions [de Figueiredo & Edwards 2007].37 In their study of the incidence of the costs
and benefits of several transportation-sector environmental regulations, Holland et al. [2014] find
patterns in the political donations by organizations and politician voting behavior that suggest that
districts used campaign contributions to influence the House vote on the Waxman-Markey cap and
trade bill. In particular, they find evidence that organizations that opposed the Waxman-Markey
bill were more likely to donate money to House members from districts that would be negatively
impacted by the bill, with a similar pattern in the contributions of supporters to candidates from
districts that would benefit. Further, they find that political contributions from organizations that
opposed the bill were associated with a large reduction in the likelihood of voting for the bill. As
discussed by Ansolabehere et al. [2003], campaign contributions likely affect policies in other ways
besides through roll call votes, including securing access to legislators at other stages of the policymaking process. Indeed, politicians likely find it desirable to promote client interests in ways that
are observable to clients but not to voters. Because of these difficulties in observability, I assume
that corporate donors allocate their donations strategically, whether with the intent of affecting
policy or to demonstrate loyalty to Saudi interests.
In particular, contributions to federal legislators are an important channel corporations can use
36
See for example [Pittman 1977] and Krueger [1974].
In particular, it is still an open debate over whether politicians are influenced by contributors, or whether
contributors simply support the candidates who are known to favor their interests. See for example Bronars & Lott
[1997] for some evidence of the latter explanation.
37
28
to advocate for Saudi interests. Many bills pass through Congress that affect Saudi Arabia, including bills regarding arms sales, trade, aid38 , and immigration39 . Most of these are referred to the
House Committee on Foreign Affairs or the Senate Committee on Foreign Relations. Committees
exercise a great deal of power over the legislation they review, and can both block legislation by
tabling it or by revising the bill before they send it to the floor for a vote. In 1992, for example, none
of the bills opposing the sale of F-15XP fighter planes to Saudi Arabia made it out of committee.
In another example, the House Foreign Affairs Committee tabled the Persian Gulf Security Cost
Sharing Act (2001), which would have required Saudi Arabia to defray the cost of U.S. military
deployments in the region.
Committees can also amend popular bills to turn them from aggressive foreign policy changes
into purely symbolic gestures. The House Committee on Foreign Affairs had added Presidential
waiver authority to several anti-Saudi bills, including the Anti Economic Discrimination Act of 1995,
which would have stopped the sale of military equipment to countries participating in the boycott
of Israel. Another notable example of this was an amendment to the 2005 Foreign Appropriations
bill, which stated that “[n]one of the funds appropriated or otherwise made available pursuant to
this act shall be obligated or expended to finance any assistance to Saudi Arabia.” Changes to this
amendment made in committee allowed the President to waive this rule provided that he certified
to the Congressional Appropriations committees that Saudi Arabia was cooperating in the war
against terrorism. Although the amendment passed,40 it was immediately waived by Presidential
Determination. A subsequent bill (the Prohibit Aid to Saudi Arabia Act of 2005) that attempted
to impose the ban on U.S. aid to Saudi Arabia without waiver authority was not passed on by the
House Foreign Affairs Committee.41
In the rest of the paper, I look at how the level and composition of direct corporate donations
to Members of Congress vary with discount receipts.
38
Examples of aid-related bills include H.R.3137.IH (2003) which prohibited assistance or reparations to Cuba,
Libya, North Korea, Iran, Syria and Saudi Arabia, as well as Amendment 708 to H.R. 4818 (the Foreign Operations
Appropriations bill for FY2005).
39
See for example, H.R. 604 and 3934. These bills would have halted the issuance of visas to Saudi citizens until
the President certified that the Saudi government did not discriminate in its visa policies on the basis of religious
affiliation or cultural heritage.
40
Vote 217-191, see roll call http://clerk.house.gov/evs/2004/roll389.xml
41
In practice the United States provides almost no aid to Saudi Arabia, and the ban targeted a small 25,000 dollar
International Military Education and Training grant for Saudi military training.
29
5.2
Data
In this part of the analysis, I match company-level discount receipts to corporate political donations. Although it is illegal for corporations to contribute to political campaigns directly, in practice
they do contribute through personal donations by their managers and employees. These individual
donors account for approximately two-thirds of campaign money to Senate candidates and half of
the financing for House candidates. By federal law, all contributions to federal candidates, political
action committees (PACs), or parties of over 200 dollars must be reported to the Federal Election Commission (FEC). These reports include the name and address of the donor, as well as the
donor’s employer and occupation.42 These data are published by the FEC and aggregated by the
Center for Responsive Politics (among others). I use these data to construct total annual campaign
contributions by employees of all U.S. refining companies to each member of the House and Senate
for the 1991-2003 period. I merge these with data on congressional committee assignments from
Stewart & Woon [2012] and Nelson [2012] to construct total contributions by committee.43 The
positive relationship between Saudi imports and total political contributions over the period can be
seen graphically in Figure IV, which plots the log of total company campaign contributions against
the log of total Saudi imports over the discount policy period controlling for refining profits.
For the politician-level contributions analysis, committee assignments and political contributions are matched to the set of all Members of Congress collected in Stewart & Woon [2012], which
begins in 1993. The merged politician by refiner-level dataset contains information on politician
characteristics including party affiliation, committee assignments, chamber seniority, and contributions from other types of interest groups. In particular, the analysis classifies politicians according
to the degree to which they receive contributions from donors affiliated with pro-Israel groups.
One categorization simply assigns a dummy variable equal to one for politicians who received any
contributions from pro-Israel donors. Another categorizes politicians according to their approximate quintile of average pro-Israel donations. The first group is the approximately 27 percent of
politicians who never received any contributions from pro-Israel organizations during their time in
42
All observations where the industry is coded as oil and gas report the name of the organization related to
contributor (employee, owner, spouse of owner, etc.). In the full sample 95% of contributions list an affiliated
industry, and 60% list the company name.
43
Table A.2 shows some summary statistics on political contributions for the 26 companies that received crude
from Saudi Aramco, both public and private. Though companies that received no Saudi crude are excluded from this
table, all 126 refining companies are used in the analysis.
30
Congress. The second group collects the 27th through 40th percentile of recipients, the third the
40th through 60th percentile, the fourth the 60th through 80th percentile, and the fifth the 80th
percentile and above.44 The median contribution to a single politician is only around one thousand
dollars per year, though the maximum contribution is very high; in 1996, Koch Industries affiliates and employees contributed almost 45,000 dollars to Representative Sam Brownback’s (R-KS)
campaign. The median total giving by a company in a single year is 7,404 dollars, but again the
maximum is quite high; ExxonMobil donated over 374,000 dollars in direct contributions in 2000.
5.3
5.3.1
Results
Contribution Aggregates
As in the profits analysis, I estimate the relationship between discounts and political contributions using the regression
Contribjt = θ · DiscV aluejt + αj + γt + jt
(7)
where j is the contributing refining company and t is the year. One possible concern here is the
effect of refiner profits on contributions. If corporations increase overall contributions or change
their contribution patterns when their profits are high, θ may capture the spurious common effect
of overall crude price on contributions (via profits) and the discount value or the indirect effect of
the discount through its effect on profits. I therefore also run this regression on a the subsample of
public companies (for which I have refining profits) both with and without profit controls.
Results for the aggregate contribution regressions are reported in Table V. The top panels
show the results for the whole sample of refiners (both public and private). Column 1 shows the
relationship between discount receipts and total donations, and Columns 2 and 3 break this out
into aggregates for contributions to politicians and to PACs. The estimate in Panel A of Column 1
indicates a small but positive relationship between the discount and overall political contribution
by refiners – a one-million dollar increase in the discount value is associated with an approximately
385 dollar increase in contributions. This is consistent with the fact that contribution magnitudes
44
Table A.3 gives some summary statistics for the total annual contributions at the company by politician level.
Since most politician-refinery-year observations are zeros, the left hand panel shows statistics conditional on a donation
occurring.
31
tend to be much smaller than the discount value. Among discount recipients, the average discount
amount is 45 million dollars per year and often varied substantially from year to year; the average
annual fluctuation in the value of the discount was 25 million dollars. The average discount, then, is
associated with a contributions increase of 17,325 dollars, which is quite substantial compared with
the annual company average of 18,578 dollars from Table A.3. Positive and significant estimates of
the elasticity of contributions with respect to discount receipts in Panel B confirm this relationship.
The lower panels display the same results for the subsample of corporations that report profits data.
The results are consistent with the estimates for the full sample, though the point estimates tend
to be larger. Controlling for profits has almost no effect on the estimates for this subsample, which
is reassuring for the full sample results.45
Interestingly, these estimates are close to previous estimates of the relationship between corporate profits and political contributions. Jayachandran [2006] finds that for every dollar companies
had donated to the Republican party, firms lost 2,313 dollars of their market value following the
shift of control in the U.S. Senate when Senator Jim Jeffords switched his party affiliation from
Republican to Democrat in May 2001. This estimate implies that a one million dollar decrease in
market value would have been associated with a contribution of 432 dollars to the Republican party.
My estimates of the effect of discount receipts on overall political giving indicate an increase of 385
dollars in response to the receipt of one million dollars in discounted crude. Ansolabehere et al.
[2003] find that top corporate executives increased their personal giving to congressional campaigns
by 510 dollars for every additional million dollars of income per year.
I also address the direct effect of profits on contributions by comparing the effect of the discount with that of non-discount profit shocks. To capture non discount-related variation in refiner
profits, I use hurricane landfall events as an instrument for annual profits. Hurricanes affect oil
and petroleum product production in several ways. As a hurricane approaches the coast, offshore
platforms are evacuated and wellheads are plugged to prevent leakage. These offshore platforms
are sometimes destroyed by hurricanes, which imposes direct costs on producers and disrupts oil
supplies. In addition to shutting down upstream crude, hurricanes also affect downstream infrastructure. Underwater pipelines can be damaged, and ports may stop accepting shipments. Onshore
45
When the same specification is run for the post-policy period (2004-2011), the point estimate on the relationship
between discount receipts and political contributions is negative and not statistically significant.
32
refineries can also be heavily damaged; Hurricane Rita, for example, caused refining capacity losses
of more than 4.9 million bpd due to direct damage or power interruptions along the Texas coast
[EIA 2006].46
To construct the hurricane instrument, I create a dummy variable equal to 1 if a company had
a refinery in a state that was hit by a hurricane in that year. This is further split into dummy
variables for each hurricane category – i.e. dummy variables indicating the most severe hurricane
level experienced by a refinery owned by company j in year t. I focus on the most damaging
hurricanes, i.e. those that were classified as category 2, 3, or 4 on the Saffir-Simpson scale [Simpson
& Saffir 1974], with sustained surface wind speeds of 82-95 knots, 96-112 knots, and 113-136 knots,
respectively.47
I then use these hurricane indicators to instrument for profits in a regression of political contributions on refining profits and discount receipts.48
(F S)
P rof itsjt = δ · DiscV aluejt + ζ1 · 1(Category 2)jt + ζ2 · 1(Category 3+ )jt + αj + γt + jt
hurr
P\
rof itsjt
(SS)
= ζb1 · 1(Category 2)jt + ζb2 · 1(Category 3+ )jt
hurr
Contribjt = β1 · DiscV aluejt + β2 · P\
rof itsjt
+ αj + γt + νjt
(8)
I then test β1 > β2 to determine whether discount receipts had a larger effect than other types
of profits. Table VI reports the results for the regression of total contributions on discount value
and profits orthogonal to the discount. Panel A reports results from the first stage regression,
which confirm that discounts increase profits and hurricanes decrease them, and that the hurricane
instruments and the discount have comparable explanatory power for profits. The results in panel
B show that only discount receipts seem to have an effect on total contributions, with a one million
dollar increase in the discount associated with a 317 dollar increase in contributions (column 1).
46
Other papers have examined the effect of hurricanes and tropical storms on U.S. refined product markets,
including Fink, Fink & Russell [2010], Kaiser, Dismukes & Yu [2009] and Lewis [2009].
47
There are five relevant hurricane landfall events in the sample. Two of these were category two hurricanes: Bob
(New York 1991) and Georges (Mississippi 1998). The two category three hurricanes were Andrew (Louisiana 1992)
and Bret (Texas 1999), and there was one category four hurricane, Iniki (Hawaii, 1992). Hurricane landfall and
severity data comes from NOAA [2006].
48
Allowing for more flexibility in the effects of hurricane landfall events on profits has little impact on the results.
In particular, indicators for refineries owned in adjacent states were statistically insignificant in the first stage and
had no effect on the two-stage least squares results.
33
Column 2 shows the corresponding elasticity estimate, and a ten percent increase in the discount
is associated with a 3 percent increase in political contributions. Profits not associated with the
discount have no statistically significant effect on contributions, and the point estimates are close to
zero as well. In both regressions we can reject the hypothesis that discount receipts and orthogonal
profit shocks have the same effect on contributions.
The second set of results (Table VII) shows the relationship between the discount value and
the way that total contributions were allocated among various committees. For the foreign policy
committees, for example, the regression is of the share of congressional contributions that go to
members of the House or Senate foreign policy committees.
ShareFjtP =
ContribFjtP
Contribtotal
jt
= θ · DiscV aluejt + αj + γt + jt
(9)
Though the standard errors on these estimates are large, the pattern is suggestive. The coefficient on contributions to the foreign policy committees is not statistically significant, but there is
a substantial increase in contributions to members of the Appropriations committees, which also
review Saudi-relevant legislation. The coefficient in Column 2 indicates that a one billion dollar
increase in the discount is associated with a 39 percentage point increase in the share of contributions that go to Appropriations committee members. The average discount of 45 million dollars
would therefore correspond to a 1.8 percentage point increase in contributions to members of the
Appropriations committees. This is about seven percent of the average share of 27 percent. I also
include some results for “placebo” committees, that would be less likely to be concerned with Saudi
affairs. Most of these coefficients are negative and significant with the exception of the House Post
Office and Senate Labor committees, which appear to be unaffected.49
It is not possible to draw conclusions about the direction of causality from this analysis, and
both contributions and discount are certainly the product of a complex relationship between Saudi
Aramco and U.S. refiners. Nonetheless, we can examine the time structure of the relationship
between contributions and discounts by running regression (7) for lags and leads of contributions
on company-level discount value in year t. Figure V summarizes the results from these regressions,
with the timing of contributions relative to discounts on the x-axis and the separate point estimates
49
As in Table IV, most of these results remain the same when I control for profits, though the further loss of
observations increases the standard errors.
34
of θ from these regressions on the y-axis. The largest statistical relationship is between discounts
and contributions in the previous year, with a 951 USD increase in contributions in year t − 1
associated with a one million dollar increase in discounts in the following year.
5.3.2
Politician-Level Contributions
To further explore the effect of the discount on political giving, I also examine the effect of the
discount on corporate contributions at the politician level. In this part of the analysis, companyyear level discount receipts are interacted with a set of politician characteristics (gi ) to estimate
how discount receipts affected contributions to different types of politicians. These politician characteristics include a dummy variable for receiving contributions from pro-Israel interest groups,
quintiles of the level of pro-Israel contributions, party affiliation and committee membership. This
yields:
Contribijt = θ · gi · DiscV aluejt + Xit β + αij + γt + jt
(10)
where the subscript i indicates the politician, and j and t the company and year. Controls Xit are
dummies for the time-varying politician characteristics (chamber seniority and committee membership).
In Table VIII, we see that the discount tended to decrease giving to pro-Israel politicians, with a
one billion dollar increase in discount receipts decreasing annual contributions by an additional 180
dollars to politicians who received pro-Israel funding (column 1). Column 2 shows this same result
split up by quintile, and the point estimates (though not statistically significant) indicate larger
penalties for politicians in the highest pro-Israel funding quintiles associated with the discount.
The pattern is similar in the regression with the full set of interactions and the coefficients are
individually significant. Republican politicians benefit more from the discount than Democrats
and independents. Republican candidates get a 166 dollar boost from companies that received an
additional billion dollar discount relative to Democrats and independents (column 3). Again, the
pattern is the same in column 5, with Republicans taking a 141 dollar increase. The evidence on
the committee membership patterns is much more mixed, and contrasts with the patterns in the
aggregate contribution regressions. Point estimates on the committee membership interactions are
very sensitive to controls, and are large and negative for the House Foreign Affairs committee in
35
both columns 4 and 5. Estimates for members of other committees are noisy, and although point
estimates are quite large none are statistically significant.50
In general, these results indicate significant politician-level heterogeneity in the impact of the
discount on funding by refining companies. Top beneficiaries were Republicans and candidates
who did not receive funding from pro-Israel groups. If Israeli interest groups donated more to
candidates that were perceived as being pro-Israel, this could indicate that the discount caused
refining companies to divert funds away from pro-Israel politicians, a move consistent with the
Saudi political agenda toward the Arab-Israeli conflict.
It is important to note here that all of these estimates are quite small. The 253 dollar penalty to
pro-Israel donation recipients, for example, implies that the average discount of 45 million dollars
would be associated with just an eleven-dollar decrease in contributions to these candidates. From
Table A.3, the average contribution to these candidates, however, is only 38 dollars, so the elevendollar increase is a reasonable proportion of the total, though difficult to think of as practically
significant. Nonetheless, the pattern of the change in this type of contribution is likely suggestive
of other political responses, both financial and otherwise, by discount recipients.
In addition to categorizing politicians by how much money they received from other interest
groups, we can also examine how recipients voted on anti-Saudi legislation. As discussed earlier,
most of the Saudi-relevant legislation introduced in congress over the period was stopped at the
committee-level. A notable exception occurred at the end of the policy period, with Amendment 708
to H.R. 4818, the amendment to the FY2005 Foreign Appropriations bill prohibiting any spending
on aid to Saudi Arabia. The controversial amendment was approved by the House on July 15,
2004 by a vote of 217-191, with 25 Members abstaining. Table IX shows the results from an LPM
regression of a dummy variable for whether a member of the House voted “No” on the amendment
on contribution totals from different groups over the 1991-2003 period. Here “contributions by
recipients” indicates the total amount of contributions received from refiners in years in which they
received Saudi crude. “Contributions by non-recipients” are all contributions from refiners in years
50
When the sample is restricted to years when the discount was large (1998-2003, Table A.4), the patterns become
more clear for the pro-Israel and committee assignment results. In particular, the estimate of the discount penalty
to pro-Israel donation recipients increases to 253 dollars, and a statistically significant pattern of larger penalties to
politicians who received more pro-Israel money emerges (column 2). The large positive effect on members of the
Senate Appropriations committee is statistically significant in columns 4 and 5 at around 2,400 dollars. The negative
estimate for members of the House Foreign Affairs committee decreases to 621 dollars (column 4), and the magnitude
of the (still statistically insignificant) point estimate for the Senate Foreign Affairs members decreases as well.
36
in which they did not receive Saudi crude.51 There is evidence that contributions from discount
recipients were associated with an increased likelihood of voting “No”, an increase of two percentage
points for every million dollars of total contributions. Contributions from non-recipient refiners and
from pro-Israel interest groups were associated with small but statistically significant decreases in
the likelihood of voting against the amendment. The point estimates on the effect of contributions
by recipients decrease to one percentage point per million dollars in contributions when controls for
committee membership and political ideology are included. The impact of including the political
ideology controls in these regressions and the importance of party affiliation in the allocation of
funds indicates that contributions may be used both as a way to support known allies as well as to
influence voting preferences.
6
Conclusion
This paper examines the effect of the Saudi crude discount program on the political behavior
of U.S. firms. This policy is an empirically appealing example of Saudi strategic political behavior
both because it had a clearly measurable cost and because it was effectively directed at specific
recipients. There are three main sets of results. First, I calculate that the program transferred
substantial rents to the U.S. market, with a total discount value of 8.5 billion dollars over the policy
period. Second, I find that the discounted crude was almost entirely a gift to refinery owners, with
full pass-through of discount receipts into refiner profits. Finally, the results suggest that this
gift induced some amount of pro-Saudi political action on the part of recipients. In particular, I
find that discount receipts were associated with higher levels of overall political giving and with
an increase in the share of funds directed toward members of key congressional committees. On
the politician level, there is also some evidence that recipients tended to divert funds away from
congressmen who received donations from pro-Israel interest groups.
There are two possible explanations for this observed pattern in the data. First, the pro-Saudi
shift in corporate political contributions could reflect a direct and intentional effort by Saudi Arabia
to influence the U.S. political process. However, the pattern is also consistent with rent-seeking by
firms trying to protect their future access to discounted crude. This behavior could reflect either a
51
The results are robust to changing these definitions so that “contributions by recipients” includes all contributions
(in every year) from refiners that ever received discounted Saudi crude.
37
desire to be seen as friendly by Saudi Arabia, or simply efforts to protect the interests of a country
with which they had developed close economic ties. In any case, the results indicate that U.S.
refiners responded to the discount, and that the policy successfully influenced their political action.
References
Adelman, M. A. [1993], ’The Cartel in Retreat’, Working paper, MIT Center for Energy and
Environmental Policy Research. MIT-CEEPR 93-005WP.
Adelman, M. A. [1995], The Genie out of the Bottle: World Oil since 1970, MIT Press.
Aggarwal, R. K., Meschke, F. & Wang, T. Y. [2012], ‘Corporate political donations: Investment
or agency?’, Business and Politics 14(1).
Ansolabehere, S., de Figueiredo, J. M. & Snyder, James M., J. [2003], ‘Why is there so little
money in U.S. politics?’, Journal of Economic Perspectives 17(1), 105 – 130.
Association of Oil Pipelines [2009], Annual Report on Shifts in Petroleum Transportation: Statistics. Available at http://www.aopl.org/pdf/Shift_Report_Posted_September_2_20091.pdf.
Bachmeier, L. J. & Griffin, J. M. [2003], ‘New evidence on asymmetric gasoline price responses’,
Review of Economics and Statistics 85(3), 772–776.
Bacon, R. W. [1991], ‘Rockets and feathers: the asymmetric speed of adjustment of UK retail
gasoline prices to cost changes’, Energy Economics 13(3), 211–218.
Bard, M. [2010], The Arab Lobby : The invisible alliance that undermines America’s interests in
the Middle East, Harper, New York.
Borenstein, S., Cameron, A. C. & Gilbert, R. [1997], ‘Do gasoline prices respond asymmetrically
to crude oil price changes?’, The Quarterly Journal of Economics 112(1), 305–39.
Borenstein, S. & Kellogg, R. [2012], ‘The incidence of an oil glut: Who benefits from cheap crude
oil in the Midwest?’, NBER Working Paper No. 18127.
Borenstein, S. & Shepard, A. [1996], ‘Dynamic pricing in retail gasoline markets’, RAND Journal
of Economics 27(3), 429–451.
Borenstein, S. & Shepard, A. [2002], ‘Sticky prices, inventories, and market power in wholesale
gasoline markets’, RAND Journal of Economics 33(1), 116–139.
Bronars, S. G. & Lott, J. R. [1997], ‘Do campaign donations alter how a politician votes? or, do
donors support candidates who value the same things that they do?’, Journal of Law and Economics
40(2), 317–50.
Brown, A. C. [1999], Oil, God and Gold: The Story of Aramco and the Saudi Kings, Houghton
Mifflin, Boston.
Cameron, A. C., Gelbach, J. B. & Miller, D. L. [2011], ‘Robust inference with multiway clustering’,
Journal of Business & Economic Statistics 29(2), 238–249.
38
Carroll, R., Lewis, J., Lo, J., McCarty, N., Poole, K. & Rosenthal, H. [2011], ‘Dw-nominate scores
with bootstrapped standard errors’, Available at: voteview. com/dwnomin. htm .
Carter, J. [1980], Remarks by the President in the State of the Union Address, Washington, DC,
The White House, Office of the Press Secretary.
Center for Responsive Politics [2012], ‘Federal Campaign Finance’. Data retrieved May 1, 2012
from http://data.influenceexplorer.com/docs/contributions/.
Cooper, M. J., Gulen, H. & Ovtchinnikov, A. V. [2010], ‘Corporate Political Contributions and
Stock Returns’, Journal of Finance 65(2), 687–724.
Daniels, L. A. [1986], ‘Yamani Ousted as Oil Minister of Saudi Arabia’, New York Times . October
30, 1986.
de Figueiredo, Rui J. P., J. & Edwards, G. [2007], ‘Does private money buy public policy? campaign contributions and regulatory outcomes in telecommunications.’, Journal of Economics and
Management Strategy 16(3), 547 – 576.
EIA [2006], The Impact of Tropical Cyclones on Gulf of Mexico Crude Oil and Natural Gas
Production, U.S. Energy Information Administration.
EIA [2011a], EIA-814 Monthly Imports Report, U.S. Energy Information Administration.
EIA [2011b], EIA-820 Annual Refinery Report, U.S. Energy Information Administration.
Faccio, M. [2006], ‘Politically Connected Firms’, American Economic Review 96(1), 369–386.
URL: http://ideas.repec.org/a/aea/aecrev/v96y2006i1p369-386.html
Federal Trade Commission [1952], ‘The international petroleum cartel’, Staff report; released
through the Subcommittee on Monopoly of Select Committee on Small Business, U.S. Senate, 83rd
Congress, 2nd Session.
Fink, J. D., Fink, K. E. & Russell, A. [2010], ‘When and how do tropical storms affect markets?
the case of refined petroleum’, Energy Economics 32(6), 1283–1290.
Fisman, R. [2001], ‘Estimating the Value of Political Connections’, American Economic Review
91(4), 1095–1102.
Holland, S. P., Hughes, J. E., Knittel, C. R., & Parker, N. C. [2014], ‘‘Some inconvenient truths
about climate change policy: The distributional impacts of transportation policies’’, The Review
of Economics and Statistics .
Horsnell, P. [1997], Oil in Asia: Markets, Trading, Refining, and Deregulation, Oxford University
Press, Oxford.
International Energy Agency [1998-2003], Oil Market Report. http://omrpublic.iea.org/.
Jablonski, W. [1986], ‘The ‘Inside Story’ of Why Yamani Was Actually Dismissed’, Petroleum
Intelligence Weekly . November 24, 1986.
Jacobs, S. [2002], ‘Pipeline factors affecting gasoline prices’, Colonial Pipeline Presentation to the
FTC, May 8, 2002. Available at http://www.ftc.gov/bc/gasconf/comments2/jacobsstevee.
pdf.
39
Jaffe, A. M. & Elass, J. [2007], ‘Saudi Aramco: National flagship with global responsibilities’,
Working paper, Rice University, James A. Baker III Institute for Public Policy.
Jaffe, A. M. & Soligo, R. [2004], ‘The future of Saudi Price Discrimination’, Working paper, Rice
University, James A. Baker III Institute for Public Policy.
Jayachandran, S. [2006], ‘The Jeffords effect’, Journal of Law and Economics 49(2), 397–425.
Johnson, S. & Mitton, T. [2003], ‘Cronyism and capital controls: evidence from Malaysia’, Journal
of Financial Economics 67(2), 351–382.
Jones, R. [1986], ‘OPEC After Yamani, Discount in the Kingdom Forces Sacking of Stylish Servant’, The Financial Times . October 31, 1986.
Kaiser, M. J., Dismukes, D. E. & Yu, Y. [2009], ‘The value of lost production from the 20042005 hurricane seasons in the Gulf of Mexico’, Journal of Business Valuation and Economic Loss
Analysis 4(2), 4.
Krueger, A. O. [1974], ‘The political economy of the rent-seeking society’, American Economic
Review 64(3), 291–303.
Lewis, M. S. [2009], ‘Temporary wholesale gasoline price spikes have long-lasting retail effects:
The aftermath of hurricane Rita’, Journal of Law and Economics 52(3), 581–605.
Memorandum for the Record [May 24, 1967], Foreign Relations of the United States, 1964-1968,
Volume XIX, Arab-Israeli Crisis and War, 1967, Document 56 . Accessed January 15, 2012 at
http://history.state.gov/historicaldocuments/frus1964-68v19/d56.
Mergent [2011], ‘The North American retailing sectors’. Retrieved August 15, 2011, from Mergent
WebReports database.
Mobil Oil Corporation [1973], ‘The U.S. stake in Middle East peace: I’, New York Times . June
21, 1973.
Mobil Oil Corporation [1980], ‘A New Fairy Tale’, New York Times . May 8, 1980.
Moran, T. H. [1981], ‘Modeling OPEC behavior: economic and political alternatives’, International
Organization 35(2), 241–272.
Mouawad, J. & Romero, S. [2005], ‘Saudis in strategy to export more oil to India and China’,
New York Times . February 18, 2005.
Muehlegger, E. [2006], Gasoline price spikes and regional gasoline content regulation: A structural approach, Working Paper Series rwp06-015, Harvard University, John F. Kennedy School of
Government.
Nelson, G. [2012], ‘Committees in the U.S. Congress, 1947-1992 Senate and House’. Data retrieved
Feb 7, 2012 from http://web.mit.edu/17.251/www/data_page.html#0.
NOAA [2006], Chronological List of All Hurricanes which Affected the Continental United States:
1851-2005, National Oceanic and Atmospheric Administration: Atlantic Oceanographic and Meterological Laboratory.
40
Ogawa, Y. [2003], Proposals on measures for reducing Asian premium of crude oil: Changes in
pricing, utility among consuming countries and preparation of oil market, Working paper, Institute
of Energy Economics, Japan.
Parsons, C. & Brown, J. [2003], The ‘Asian premium’ and dependency on Gulf oil, Working Paper
Series CITS WP 2003-02, Center for International Trade Studies.
Perlez, J. [2001], ‘U.S. Gingerly Discusses Taking More Active Role’, New York Times . May 17,
2001.
Pittman, R. [1977], ‘Market structure and campaign contributions’, Public Choice 31, 37–52.
Sampson, A. [1975], The Seven Sisters: The Great Oil Companies and the World They Made,
Viking Press, New York.
Sapienza, P. [2004], ‘The effects of government ownership on bank lending’, Journal of Financial
Economics 72(2), 357–384.
Sciolino, E. & Tyler, P. E. [2001], ‘Saudi Charges Bush with Failure to Broker Mideast Peace’,
New York Times . November 9, 2001.
Simpson, R. H. & Saffir, H. [1974], ‘The hurricane disaster-potential scale’, Weatherwise, American
Meteorological Society 27(4), 169.
Soligo, R. & Jaffe, A. M. [2000], ‘A note on Saudi Arabian price discrimination’, The Energy
Journal 21(1), 121–134.
Standard & Poor’s [2011], ‘Complete Financial Statements 2004-2009’. Retrieved August 22, 2011
from Compustat database.
Stewart, C. & Woon, J. [2012], ‘Congressional Committee Assignments, 103rd to 112th Congresses,
1993-2011: Senate and House’. Data retrieved Feb 7, 2012 from http://web.mit.edu/17.251/
www/data_page.html#0.
Stockholm International Peace Research Institute [2011], ‘SIPRI military expenditures database’.
Accessed October 15, 2011 at http://www.sipri.org/contents/milap/milex/mex_database1.
html.
Untiet, C. J. [1984], Competition in the Oil Pipeline Industry: A Preliminary Report, Antitrust
Division, U.S. Department of Justice.
Yergin, D. [1991], The Prize: The Epic Quest for Oil, Money, and Power, Simon and Schuster,
New York.
41
Figures
Figure I: Saudi Crude Quantity, Discount, and Discount Value
20
Quantity (mil barrels per month)
30
40
50
60
70
(a) Saudi Crude Deliveries to U.S. Refineries, mil bbl per month. Source: EIA-814
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
Year
ï8
ï6
ï4
Discount (USD)
ï2
0
2
4
6
8
(b) Saudi Crude Discount to U.S. Refineries, (pASIA − pU S ). Source: Platts
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
Year
ï400
Discount Value (mil USD per month)
ï200
0
200
400
(c) Total Value of the Saudi Crude Discount, mil 2000 USD per month
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
Year
42
Figure II: Common cost shocks affect both price and profits, but inframarginal shocks affect only
profits.
(b) Discount to inframarginal producers
(a) Common crude price shock
p
p
p*
p! *
c!m
!m
!! m
!ˆ1
!! 1
c1
c1
q! m
!
!
k1
qm
!
D( p)
Q
km
D( p)
!
k1
Q
km
Figure III: Saudi crude quantity by refinery, mil bbl: 1991-2003
!
!
!
!
!
!!
!
!
!
!!
!!
!
!
!
!
!
! !!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!!!!
!
!
! !
!
!
!
!
!
! !
!
!!
!
!
!
!
!
!
!
!
!"!#
!!!!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
! !!
! !
!
!
!
!
!
!
!
#
#!
!
#$!!!
43
!
ï4
Log Congressional Campaign Contributions
ï2
0
2
Figure IV: Correlation of Total Saudi Imports and Total Campaign Contributions by Company
ï4
ï2
0
Log Saudi imports
2
4
Note: This figure plots the log of total congressional campaign contributions (mil USD) over the 1991-2003
period against the log of total Saudi imports (thousands of barrels) controlling for refining profits.
Regression line is ln(amount) = 0.488 * ln(Saudi imports) + .381 ln(Refining Profits) - 0.543; F(2,12) =
17.30.
Figure V: Time Structure of Correlations of Campaign Contributions with Discount Receipts
'#$%
"#&%
"#$%
$#&%
$#$%
!$#&%
!"#$%
(!)%
(!*%
(!'%
(!"%
(%
(+"%
(+'%
,-.(/0123-.%4056%76/0-8%
Note: This figure plots the coefficient on discount value from regressions of lags and leads of company-level
campaign contributions on discount value in year t. Dashed lines represent the 90% confidence interval. All
regressions include company and year fixed effects. Standard errors are clustered at the company level.
44
Tables
Table I: Refiner Characteristics and Saudi Crude Receipts
Table 2: Refiner Characteristics and Saudi Crude Receipts
Saudi Crude Refinery
(89/90)
Total Refining Capacity
Former Aramco Partner
Overall Political Contrib
(89/90)
Politicians
(1)
0.647***
(0.086)
0.025
(0.030)
-0.012
(0.113)
(2)
0.639***
(0.125)
0.062
(0.058)
0.036
(0.247)
(3)
0.637***
(0.090)
0.007
(0.032)
-0.080
(0.130)
0.001
(0.001)
(4)
0.642***
(0.090)
-0.005
(0.039)
-0.090
(0.130)
(5)
0.635***
(0.089)
0.008
(0.035)
-0.079
(0.104)
(6)
0.637***
(0.087)
0.002
(0.035)
-0.070
(0.132)
(7)
0.626***
(0.103)
0.062
(0.082)
-0.554!
(0.676)!
0.001
(0.001)
-0.001
(0.002)
Committees
Pro-Israel Recipients
0.001
(0.001)
Republican
0.001
(0.001)
0.001
(0.001)
Democratic
House Appropriations
0.072
(0.053)
0.116**
(0.056)
-0.058**
(0.026)
-0.024
(0.073)
-0.213**
(0.087)
-0.199
(0.135)
0.011
(0.015)
-0.158*
(0.083)
-0.026
(0.024)
0.095*
(0.054)
0.145
(0.102)
0.088
(0.068)
House Armed Services
House Energy
House Educ & Labor
House Foreign Affairs
House Post Office
Senate Appropriations
Senate Armed Serv.
Senate Energy
Senate For. Rel.
Senate Govt Affairs
Senate Labor
State Dummies
x
R2
0.373
0.620
0.379
0.381
0.376
0.380
0.434
N
115
115
115
115
115
115
115
Notes: Results for LPM regression of a dummy for whether a refining company received Saudi crude
during the policy period on a set of refiner characteristics. These include whether the company ever
owned a refinery that had processed Saudi crude in the pre-policy period (1989-1990), total refining
capacity in billions, an indicator for the former Aramco partners, and pre-period political
contributions to Members of Congress by party affiliation and committee membership. All political
contributions are in thousands of 2000 USD. State dummies indicate whether a company owned any
refineries in a particular state.
*** p"0.01, ** p"0.05, * p"0.1
45
Table II: Impact of the Discount on Refining Profits
Table 3: Impact of the Discount on Refining Profits
Discount Value
H0: full pass-through (p-value)
Companies
Quantity: log
(1)
Quantity: log
(2)
Value: level
(3)
0.016**
0.169***
3.38**
(0.007)
(0.045)
(1.56)
0.620
0.101
0.135
39
16
39
N
317
81
317
Notes: Columns 1 and 2 report results from the regressions of log annual company-level refining
profits on log quantity of Saudi crude delivered to refineries owned by the same company in that
year for the 1991-2003 period. In the regression in Column 1, 1 is added to profits and 1 to
quantity to avoid losing observations (both in billions) and column 2 drops observations with
negative profit observations or zero Saudi quantities. Column 3 shows the results from a
regression of the level of refining profits on total discount value. The second row reports the pvalue of an F-test of the coefficient against the full pass-through benchmark of 0.02 for column 1,
0.09 for column 2 and 1 for column 3. All regressions include company and year fixed effects.
Standard errors are clustered at the company level.
*** p!0.01, ** p!0.05, * p!0.1
Table III: Saudi Crude Receipts and Refinery Output
Table 4: Saudi Crude Receipts and Refinery Output
(1)
ln(Saudi Crude Receipts)
ln(Refining Capacity)
District FE
(2)
(3)
0.001
0.004*
0.003*
(0.002)
(0.002)
(0.001)
1.103***
1.142***
0.661***
(0.081)
(0.065)
(0.174)
x
x
x
District FE * t
x
Year FE
x
x
N
156
156
156
Notes: This table shows the results from the regression of log annual districtlevel production on the log of Saudi crude receipts to all refineries in the district
for the 1991-2003 period. Refinery output is calculated at the district level for
twelve refining districts using refining capacity from the EIA Refining Capacity
Report and capacity utilization rates from the EIA-810 Monthly Refinery
Report. One is added to all three variables to avoid dropping observations.
Standard errors are clustered at the district level.
*** p!0.01, ** p!0.05, * p!0.1
"
"
"
46
Table IV: Impact of the Discount on City-Level Retail Gasoline Prices
Table 5: Impact of the Discount on City-Level Retail Gasoline Prices
Quantity: log
(1)
Quantity: log
(2)
Panel A: Local Markets
Discount per Gallon
-0.0002
-0.001
(0.0008)
(0.003)
H0: full pass-through (p-value)
$0.001
$0.001
Panel B: Pipeline-Connected Markets
Discount per Gallon
0.0001
-0.001
(0.0010)
(0.003)
$0.001
$0.001
H0: full pass-through (p-value)
Cities
75
31
Value: level
(3)
-0.031
(0.120)
$0.001
-0.002
(0.154)
$0.001
75
N
5326
1375
5326
Notes: Columns 1 and 2 report results from the regressions of the log of monthly
city-level gasoline prices on log of Saudi crude deliveries as a share of total local
refining capacity for the 1998-2003 period. In the regression in Column 1, 0.01 is
added to share to avoid losing observations and column 2 drops observations
with zero Saudi quantities. Column 3 shows the results from a regression of
gasoline price levels on monthly discount per gallon of local refining capacity.
Panel B reports the same results accounting for all imports and refining capacity
“up-pipe” of the city. The second row in both panels reports the p-value of an
F-test of the coefficient against the full pass-through benchmark of -0.04 for
column 1, -0.15 for column 2 and -1 for column 3. All regressions include city
and state x year x month fixed effects. Standard errors are clustered at the city
and state by year by month level using the two-way clustering technique from
Cameron, Gelbach, and Miller (2011).
*** p$0.01, ** p$0.05, * p$0.1
47
Table V: Discount Correlation with Refiner Political Contributions
Table 8a/b combined
Total
(1)
Politicians
(2)
PACs
(3)
FULL SAMPLE
Panel A: Levels
Discount Value
ln(Discount Value)
Companies
N
0.385**
0.107
0.271***
(0.170)
(0.092)
(0.086)
Panel B: Elasticities
0.440***
0.432***
(0.149)
(0.114)
115
115
1033
1033
0.418**
(0.186)
115
1033
PUBLIC FIRMS
Discount Value
Discount Value
ln(Discount Value)
ln(Discount Value)
Companies
Panel C1: Levels (Controlling for Profits)
0.929***
0.423***
(0.211)
(0.060)
Panel C2: Levels (No Profit Controls)
0.914***
0.370***
(0.235)
(0.060)
Panel D1: Elasticities (Controlling for Profits)
0.466***
0.363**
(0.153)
(0.150)
Panel D2: Elasticities (No Profit Controls)
0.462***
0.352**
(0.150)
(0.147)
38
38
0.507***
(0.167)
0.542***
(0.198)
0.523**
(0.162)
0.533***
(0.177)
38
N
316
316
316
Notes: Panel A shows the results from regressions of total annual political contributions
on total discount value (in thousands) in the same year over the 1991-2003 period, and
Panel B reports the elasticity of giving with respect to changes in the value of the
discount. Panels C and D report the same results for regressions that include only
publicly-listed firms, with panels C1 and D1 controlling for profits and C2 and D2
showing the same regressions without profit controls. Column 1 shows the correlation
with total overall political contributions, and columns 2 and 3 break this overall effect
into the effects on contributions to individual politicians and to PACs. All regressions
include company and year fixed effects. Standard errors are clustered at the company
level.
*** p!0.01, ** p!0.05, * p!0.1
48
Table VI: Profit Shocks and Total Political Contributions
Table 9: Profit Shocks and Total Political Contributions
Levels
(1)
Elasticity
(2)
Panel A: First Stage (Profits on Discount and Hurricane Indicators)
Discount Value
3.89**
1.90***
(1.67)
(0.57)
I(Category 2)
-2.57***
-0.93***
(0.24)
(0.05)
I(Category 3+)
-0.21**
-0.11*
(0.09)
(0.06)
2
Partial R (Discount Value)
0.004
0.006
2
Partial R (Hurricane Indicators)
0.010
0.010
Panel B: Second Stage (Contributions on Profits)
Discount Value
0.317***
0.302**
(0.089)
(0.092)
Profits from Hurricane Shocks
-0.005
-0.002
(0.013)
(0.030)
F-test p-value
0.001
0.007
Companies
38
38
N
316
316
Notes: Results from a two-stage least squares procedure showing the effects of
discount receipts and profits shocks associated with hurricanes on total refiner political
contributions over the 1991-2003 period. Profits are in thousands of 2000 USD and
contributions in 2000 USD. Regressions include company and year fixed effects. One is
added to contributions, discount, and profits (all in billions of 2000 USD) to avoid
losing observations in the regressions in column 2.
*** p!0.01, ** p!0.05, * p!0.1
"
Table
VII: Discount Value and Share of Political Contributions by Committee
Table 10: Discount Value and Share of Political Contributions by Committee
Discount Value
"
"
"
FP
(Both)
(1)
Approp
(Both)
(2)
Energy
(Both)
(3)
Educ
(House)
(4)
Post
Office
(House)
(5)
Labor
(Senate)
(6)
Indian
Affairs
(Senate)
(7)
0.113
0.385**
-0.508**
-0.677**
-0.035
0.021
-0.358**
(0.084)
(0.156)
(0.254)
(0.280)
(0.067)
(0.068)
(0.148)
Sample Avg
0.104
0.271
0.223
0.040
0.010
0.030
0.065
Companies
77
77
77
77
77
77
77
N
532
532
532
532
532
532
532
Notes: This table reports the relationship between the total Saudi discount value and share
of total refiner political contributions by committee membership over the 1991-2003 period.
Discount value in billions of 2000 USD. All regressions include company and year fixed
effects. Standard errors are clustered at the company level.
*** p!0.01, ** p!0.05, * p!0.1
"
"
"
"
"
49
"
!
Table11a:
VIII:
Discount
Receipts
and Politician-Level
Contributions
Table
Discount
Receipts
and Politician-Level
Political Political
Contributions:
1993-2003
!
(1)
(ISR=1)*DiscValue
(2)
(3)
(4)
(5)
-118.21*
(66.70)
(ISR27-40)*DiscValue
-65.65
-20.51
(92.88)
(90.36)
(ISR40-60)*DiscValue
-72.93
-47.54
(74.46)
(75.18)
(ISR60-80)*DiscValue
-140.73
-152.78
(89.07)
(94.70)
(ISR80+)*DiscValue
-176.70
-162.53
(138.87)
(135.90)
Republican*DiscValue
166.23**
140.54*
(73.18)
(House Approp)*DiscValue
(Senate Approp)*DiscValue
(House For. Aff.)*DiscValue
(Senate For. Rel.)*DiscValue
(77.91)
92.99
57.16
(402.97)
(403.16)
1685.37
1705.26
(1079.05)
(1080.74)
-1177.70***
-1212.62***
(398.24)
(402.23)
-1570.44
-1543.64
(1278.91)
(1262.46)
N
464,954
464,954
464,954
464,954
464,954
Notes: This table shows the results from the regression of annual politician by company level
political contributions on the interaction of company discount values with politician
characteristics. All regressions include politician by company and year fixed effects and controls
for committee membership and discount value. Discount value in billions of 2000 USD,
contributions in 2000 USD. Standard errors are clustered at the politician by company level.
*** p"0.01, ** p"0.05, * p"0.1
Table
Total
Politician-Level Contributions
Contributions and
Votes
on Amendment
708 708
Table
IX: 12:
Total
Politician-Level
and
Votes
on Amendment
!
Contributions by Recipients
Contributions by Non-Recipients
Contributions by Pro-Israel Groups
(1)
(2)
(3)
(4)
0.021***
0.018***
0.010*
0.010*
(0.006)
(0.006)
(0.006)
(0.006)
-0.008**
-0.007*
-0.004
-0.004
(0.004)
(0.004)
(0.004)
(0.004)
-0.001***
-0.001***
-0.0004
-0.001*
(0.0003)
(0.0003)
(0.0003)
(0.0003)
x
x
DW-NOMINATE Score Controls
Committee Membership Controls
R
!
2
x
0.109
0.208
x
0.255
0.307
N
433
433
433
433
Notes: Results for LPM regression of a dummy for whether House member voted “No” on an
amendment barring U.S. aid to Saudi Arabia on contribution totals from the 1991-2003 period.
There were 217 “Aye” votes, 191 “No” votes, and 25 Members not voting. The House approved the
amendment on July 15, 2004. Contribution totals include political donations from refiners in
years in which they received Saudi crude, contributions from non-recipient refiners, and proIsrael interest groups. Columns 2 and 4 include controls for committee membership. Columns
3 and 4 control for political preferences captured by two-dimensional DW-NOMINATE scores
(Carroll et al. 2011). All contributions are in millions of 2000 USD.
*** p<0.01, ** p<0.05, * p<0.1
50
A
Appendix: Additional Figures and Tables
0
Arms Sales Value, mil 1990 USD
500
1000
1500
2000
Figure A.1: Annual Value of U.S. Weapons Sales to Saudi Arabia, mil 1990 USD. Source: SIPRI
1980
1985
1990
1995
year
2000
2005
2010
Figure A.2: U.S. Crude Imports by Country, mil bbl per month (12mma). Source: EIA-814.
70 Canada 60 Saudi Arabia Venezuela 50 40 Mexico 30 20 10 0 1988 1990 1992 1994 1996 1998 2000 51
2002 2004 2006 2008 2010 Figure A.3: Sample OSP Releases for December 2001. Source: Platts
Singapore (Platts) - 4 Dec 2001/9.58 pm EST/2.58 GMT
Benchmark
Extra Light WTI
Arab Light WTI
Arab Medium WTI
Arab Heavy WTI
Aug
-2.75
-4.30
-5.35
-7.00
Sep
-3.35
-4.80
-6.05
-7.40
Oct
-3.25
-5.10
-6.40
-6.80
Nov
-3.05
-4.75
-5.95
-6.30
Dec
-3.05
-4.75
-5.75
-6.65
Jan
-2.75
-4.05
-4.85
-5.65
All prices are FOB Ras Tanura.
--Platts Global Alert--
!
!
Singapore (Platts) - 4 Dec 2001/7.20 pm EST/0.20 GMT
Benchmark
(Oman+Dubai)/2
(Oman+Dubai)/2
(Oman+Dubai)/2
(Oman+Dubai)/2
(Oman+Dubai)/2
Super Light
Extra Light
Arab Light
Arab Medium
Arab Heavy
Aug
+2.85
+1.65
+0.50
-0.30
-1.10
Sep
+1.95
+1.15
+0.30
-0.20
-0.80
Oct
+1.55
+1.15
+0.30
-0.20
-0.80
Nov
+1.15
+0.75
+0.20
-0.10
-0.55
Dec
+1.05
+0.55
+0.20
-0.20
-0.65
Jan
+1.05
+0.55
+0.25
-0.10
-0.50
Sales of Saudi crude into Asia loading FOB Ras Tanura are priced
versus the average of Oman/Dubai, plus or minus a differential.
--Platts Global Alert--
!
Figure A.4: State-level Saudi inputs as a share of total refining capacity
!
!
!
!
!
!!
!
!
!
!!
!!
!
!
!
!
!
! !!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!!!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!!!!!
!
!
! !
!
!
!
!
!
! !
!
!!
!
!
!
!
!
!
!
!
!!!!
!
!
!
!!
!
!
!
!
!
!
!
!
!
!
! !!
! !
!
!
!
!
!
!
!
!"
!#$!"
$!#%!"
&!#'!"
'!#(!"
!
)*(!"
52
!
Figure A.5: Crack Spread and Average Refiner Crude Cost
110 100 90 Average Refiner Crude Cost 6-­‐3-­‐2-­‐1 Crack Spread 80 70 60 50 40 30 20 10 0 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 Note: The crack spread is calculated as the difference in cost between a barrel of crude and a representative
mix of typical outputs. For example, a simple version (the 2-1-1) is calculated as the difference in cost of
2 barrels of crude and a barrel of gasoline and a barrel of heating oil/diesel. Here I use the 6-3-2-1 crack
spread, which is the difference in the total cost of 6 barrels of crude and outputs of 3 barrels of gasoline,
2 barrels of heating oil/diesel, and 1 barrel of residual fuel oil. Trends are the same for each of the four
standard spreads. Source: EIA.
Figure A.6: Product Pipelines, 2004.
Note: The pipelines in this map were used to construct the markets in panel B of table IV.
Source: http://www.theodora.com/pipelines/united states pipelines.html#map
53
Table
A.1:
Refinery owner
owner characteristics
Table
1: Refinery
characteristics and
and receipts,
receipts, 1991-2003
1991-2003
Disc
Saudi
Refining
Share
No.
Refining
Value
Imports
Capacity Inputs No. Saudi
Profits
Disc/
Owner
(bil USD)
(bil bbl)
(bil bbl)
Saudi Refs Refs (bil USD) Profits
0.10
Chevron / Texaco
4.05
3.26
10.16
0.32
18
12
39.84
0.09
Marathon
1.01
0.68
3.50
0.19
7
6
10.81
0.01
ExxonMobil
0.97
0.89
9.03
0.10
10
8
73.94
0.18
Valero
0.84
0.39
1.82
0.21
14
6
4.71
0.02
BP / Amoco
0.57
0.45
9.28
0.05
15
6
34.68
0.39
ConocoPhillips
0.46
0.37
4.97
0.07
17
5
1.17
0.002
Royal Dutch Shell
0.09
0.13
4.13
0.03
15
7
40.13
0.07
Premcor
0.08
0.04
1.70
0.02
6
2
1.10
0.02
Tosco
0.06
0.02
2.53
0.01
10
2
3.22
0.12
Murphy
0.05
0.02
0.61
0.03
2
1
0.33
0.03
Sunoco
0.05
0.11
3.45
0.03
6
2
1.52
0.03
Lyondell
0.03
0.04
1.25
0.04
1
1
0.99
0.50
Alon USA
0.02
0.01
0.34
0.02
2
1
0.04
0.20
Fina
0.01
0.02
0.78
0.03
3
1
0.05
Total
0.002
0.002
0.62
0.003
3
1
13.72
!0.001
UDS
0.001
0.001
1.27
0.001
4
2
3.31
!0.001
0
Amerada Hess
0
0
2.46
0
2
0
0.13
0
Citgo
0
0
2.37
0
4
0
1.53
0
Tesoro
0
0
0.94
0
6
0
1.12
0
UnoCal
0
0
0.67
0
3
0
0.57
0
Mapco
0
0
0.56
0
2
0
0.64
0
Farmland Ind
0
0
0.54
0
2
0
0.07
0
Crown Central
0
0
0.47
0
1
0
-0.04
0
PDVSA
0
0
0.39
0
1
0
0.99
0
Frontier
0
0
0.35
0
2
0
0.42
0
Holly
0
0
0.31
0
3
0
0.55
0
United Refining
0
0
0.29
0
1
0
0.09
.
Delek
0
0
0.26
0
1
0
0
0
Pennzoil / Quaker
0
0
0.28
0
3
0
0.72
0
Giant Industries
0
0
0.18
0
3
0
0.42
.
Big West
0
0
0.11
0
1
0
0
.
Calumet
0
0
0.10
0
3
0
0.00
0
Huntway Refining
0
0
0.06
0
2
0
0.06
0
AIPC
0
0
0.03
0
1
0
-0.06
0
Suncor
0
0
0.02
0
1
0
0.34
.
Greka Energy
0
0
0.01
0
1
0
0.00
.
SABA Petrol
0
0
0.01
0
1
0
0
Notes: This table provides sample statistics on the refinery-owners that are publicly-listed. Columns 16 are calculated using refinery-level observations on monthly Saudi imports and crude prices and annual
refinery capacity reports from the EIA. Columns 5 and 6 indicate the number of unique refineries that
were owned by the company during the policy period and the number of these that received any
amount of Saudi crude during the period. Total refining profits is the sum of the annual operating
profits variable for refining operations from the Compustat Business Segments data, and total net
income is the sum of the quarterly net income entries from the Compustat aggregation of the SEC-10K
filings. Column 8 reports the ratio between the value of discount receipts relative to total refining
profits. All dollar values are billions of real 2000 USD.
"
"
"
54
Table
A.2:
Political Contributions
Table
6: Political
Contributions and
and Discount
Discount Value
Value by
by Company,
Company, 1991-2003
1991-2003
Share
Share
Total
Politician
PAC
Disc
Inputs
Saudi
Refining
Disc /
Corporation
Contrib
Contrib
Contrib Value
Saudi
Exports
Profits
Profits
Chevron/Texaco
7.720
2.704
4.928
4,049
0.683
0.475
39,840
0.10
Marathon
3.806
2.053
1.682
1,005
0.441
0.131
10,810
0.09
ExxonMobil
7.093
4.296
2.669
970
0.185
0.129
73.940
0.01
BP/Amoco/ARCO
8.592
2.586
2.681
567
0.130
0.066
34,680
0.02
Valero
1.636
0.745
0.852
835
0.211
0.056
4,710
0.18
ConocoPhillips
3.403
1.370
1.938
460
0.218
0.053
1,170
0.39
Ergon
0.263
0.202
0.049
172
0.345
0.020
.
.
Royal Dutch Shell
1.245
0.810
0.384
85
0.030
0.018
40,130
0.002
Sunoco
1.696
0.561
1.112
45
0.033
0.016
1,520
0.03
Lyondell
0.0002
0
0.0002
34
0.035
0.006
990
0.03
Premcor
0.014
0.009
0.003
81
0.024
0.006
1,100
0.07
Hunt
1.036
0.286
0.700
62
0.232
0.005
.
.
Tosco
0.793
0.306
0.474
64
0.009
0.003
3,220
0.02
Fina
0.113
0.052
0.058
13
0.030
0.003
50
0.26
Murphy
0.688
0.150
0.496
45
0.029
0.003
330
0.14
Coastal
2.762
1.136
1.575
27
0.014
0.002
.
.
Phibro
0.007
0.007
0
2
0.018
0.002
.
.
Chalmette
0
0
0
19
0.027
0.002
.
.
Alon USA
0.010
0.007
0.003
22
0.024
0.001
40
0.55
Koch
6.214
2.471
3.615
1
0.001
0.0003
.
.
!.001
Total
0.037
0.031
0.006
2
0.003
0.0003
13,720
Basis
0.001
0.001
0
0
0.021
0.0003
.
.
Orion
0
0
0
1
0.005
0.0002
.
.
Sinclair
0.046
0.040
0.004
-1
0.002
0.0001
.
.
!.001
UDS
0.186
0.150
0.010
1
0.001
0.0001
3,310
Flint Hills
0.008
0.008
0
1
0.001
0.0001
.
.
Notes: Includes all public and private companies with any Saudi imports over the period. Columns 1-3
are calculated using data from the FEC filings available at
http://data.influenceexplorer.com/docs/contributions. Column 2 includes only direct contributions to
campaigns for House and Senate seats, and Column 3 totals contributions to political committees.
Columns 4-6 are calculated using refinery-level observations on monthly Saudi imports and crude prices
and annual refinery capacity reports from the EIA. Total refining profits is the sum of the annual
operating profits variable for refining operations from the Compustat Business Segments data. Column 8
reports the ratio between the value of discount receipts relative to total refining profits. All dollar values
are millions of real 2000 USD.
55
Table7:A.3:
Summary
Statistics
Politician-LevelPolitical
PoliticalContributions:
Contributions: 1993-2003
1993-2003
Table
Summary
Statistics
forfor
Politician-Level
!
Conditional ("0)
Mean
Median
Max
Unconditional
SD
N
Mean
N
Overall
1,487
997
44,901
1,883
10,680
34
464,954
Received pro-Israel Contribs
1,595
1,026
44,901
2,042
8,218
38
337,583
ISR27-40
1,187
933
38,732
1,852
1,314
26
59,307
ISR40-60
1,182
937
17,618
1,370
2,204
28
92,447
ISR60-80
1,709
1,053
31,402
2,074
2,534
46
93,418
ISR80+
2,130
1,157
44,901
2,496
2,166
50
92,411
Republican
1,572
1,006
44,901
2,014
7,701
52
230,456
Democrat
1,269
958
21,155
1,472
2,977
16
232,655
Independent
721
721
963
342
2
1
1,354
House
1,247
951
44,901
1,584
8,515
28
379,093
Senate
2,431
1,406
31,402
2,554
2,165
61
85,861
House Appropriations
1,244
960
38,732
1,746
1,210
1,224
1,223
Senate Appropriations
2,508
1,877
21,155
2,517
708
2,454
721
House Foreign Affairs
1,136
697
44,901
2,085
591
1,097
606
Senate Foreign Relations
2,478
1,199
31,402
3,103
393
2,439
399
Company-Level Total
34,861
7,404
374,089 57,333
453
18,578
850
Notes: This table reports annual average values for contributions by refining companies to
members of congress. The left panel reports statistics on the non-zero entries in the dataset,
e.g. the size of an annual contribution conditional on a contribution being made by that
company to that politician in a given year. The right panel reports statistics on the entire
annual company by politician sample, including zeros for years in which the politician was a
member of congress and the company existed but no contribution was made. The last row of the
table gives annual averages by company. Contribution values are in 2000 USD.
!
56
Table A.4:
Contributions:
1998-2003
Table
11b: Discount
Discount Receipts
Receiptsand
andPolitician-Level
Politician-LevelPolitical
Political
Contributions:
1998-2003
(1)
(ISR=1)*DiscValue
(2)
(3)
(4)
(5)
-253.11***
(85.52)
(ISR20-40)*DiscValue
(ISR40-60)*DiscValue
(ISR60-80)*DiscValue
(ISR80+)*DiscValue
-196.84*
-178.00
(110.49)
(109.40)
-214.56**
-208.00***
(89.16)
(89.75)
-406.76***
-438.21
(127.28)
(129.67)
-175.11
-243.06
(172.81)
Republican*DiscValue
(House Approp.)*DiscValue
(Senate Approp.)*DiscValue
(House For. Aff.)*DiscValue
(Senate For. Rel.)*DiscValue
(179.81)
30.45
0.19
(93.02)
(102.97)
-605.49
-578.65
(383.80)
(377.23)
2375.64**
2432.25**
(1192.79)
(1185.21)
-621.36*
-616.60
(377.95)
(378.35)
-582.36
-567.95
(1043.25)
(1038.65)
N
223,213
223,213
223,213
223,213
223,213
Notes: This table shows the results from the regression of annual politician by company level
political contributions on the interaction of company discount values with politician
characteristics. All regressions include politician by company and year fixed effects and controls
for committee membership and discount value. Discount value in billions of 2000 USD,
contributions in 2000 USD. Standard errors are clustered at the politician by company level.
*** p!0.01, ** p!0.05, * p!0.1
"
"
"
57
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising