0 
 
The Role of Speculation in Oil Markets:
What Have We Learned So Far?
Bassam Fattouh Lutz Kilian Lavan Mahadeva
Oxford Institute for University of Michigan Oxford Institute for
Energy Studies CEPR Energy Studies
SOAS
July 30, 2012
Abstract: A popular view is that the surge in the real price of oil during 2003-08 cannot be explained by
economic fundamentals, but was caused by the increased financialization of oil futures markets, which in
turn allowed speculation to become a major determinant of the spot price of oil. This interpretation has
been driving policy efforts to tighten the regulation of oil derivatives markets. This survey reviews the
evidence supporting this view. We identify six strands in the literature and discuss to what extent each
sheds light on the role of speculation. We find that the existing evidence is not supportive of an important
role of speculation in driving the spot price of oil after 2003. Instead, there is strong evidence that the comovement
between spot and futures prices reflects common economic fundamentals rather than the
financialization of oil futures markets.
JEL Code: Q43; G15; G28
Key Words: Oil price; spot market; futures market; fundamentals; speculation; financialization.
Acknowledgements: We thank Christopher Allsopp, Ron Alquist, Christiane Baumeister, Jim Hamilton,
Laurence Harris, Daniel Murphy, Michel Robe, and Pasquale Scaramozzino for helpful discussions and
Jim Hamilton and Scott Irwin for providing some of the data.
Bassam Fattouh, Oxford Institute for Energy Studies, 57 Woodstock Road, Oxford OX2 6FA, UK, and
School of Oriental and African Studies (SOAS). Email: bassam.fattouh@oxfordenergy.org.
Lutz Kilian, Department of Economics, 611 Tappan Street, Ann Arbor, MI 48109-1220, USA. Email:
lkilian@umich.edu (corresponding author)
Lavan Mahadeva, Oxford Institute for Energy Studies, 57 Woodstock Road, Oxford OX2 6FA, UK.
Email: lavan.mahadeva@oxfordenergy.org.
1 
 
1. INTRODUCTION
The unprecedented surge in the spot price of crude oil during 2003-08 has sparked a heated public debate
about the determinants of the price of crude oil. A popular view is that this surge cannot be explained by
changes in economic fundamentals, but was caused by the increased financialization of oil futures
markets, which in turn allowed speculation to become a major determinant of the spot price of oil.1
The
financialization of oil futures markets has been held responsible for a variety of phenomena including
changes in price volatility, increased co-movement between oil futures prices and other financial asset
and commodity prices, a breakdown of the statistical relationship between oil inventories and the price of
oil, and an increased influence of the decisions of financial investors such as swap dealers, hedge funds
and commodity index traders on the oil futures price. Most importantly, there is a perception that oil
futures markets no longer adequately perform their function of price discovery and risk transfer. The view
that the entry of financial investors such as index funds has somehow distorted oil prices has been driving
policy efforts to tighten the regulation of oil derivatives markets as part of a larger effort by the G20
governments to regulate international financial markets.
This survey reviews the evidence in the academic literature supporting this interpretation. We
begin by highlighting the role of speculation in oil markets. We discuss the perception that speculation
has been excessive and stress that commonly used accounting definitions of hedgers and speculators are
not informative about the degree to which speculation in oil futures markets has been excessive. Indeed,
the academic literature has instead focused on the problem of finding indirect evidence in support of the
view that financial speculation has affected oil prices. We identify six strands in this literature. One strand
attempts to link the increased participation of financial investors in oil futures markets to evidence of
increased comovement among oil prices, other commodity prices and stock prices. Another strand in this
                                                            
1
 Oil futures contracts are financial instruments that allow traders to lock in today a price at which to buy or sell a
fixed quantity of the commodity on a predetermined date in the future. Futures contracts can be retraded between
inception and maturity on futures exchanges such as the CME Group (formerly the New York Mercantile Exchange)
or the Intercontinental Exchange (ICE). Exchanges offer institutional features that allow traders to transact
anonymously, they reduce individual default risk and they ensure homogeneity of the traded commodity, making the
futures market a low-cost and liquid mechanism for hedging against and for speculating on oil price risks.
2 
 
literature seeks to establish a causal link from the positions taken by index fund traders in particular to
higher oil futures prices. A third strand examines whether higher oil futures prices systematically precede
increases in the spot price of oil. A fourth approach focuses on the breakdown of the historically negative
relationship between oil prices and oil inventories. A fifth approach relies on structural VAR models to
identify the role of speculation, taking account of the mutual endogeneity of all oil market variables
including the spot price and futures price of oil. Finally, yet another approach has been to explore the role
of time-varying risk premia in oil futures markets.
We discuss to what extent each approach sheds light on the role of speculation in the spot market
and in the futures market for crude oil. We find that the existing evidence is not supportive of an
important role of financial speculation in driving the spot price of oil after 2003. Instead, there is strong
evidence that the co-movement between spot and futures prices reflects common economic fundamentals
rather than the financialization of oil futures markets. To what extent this conclusion may have to be
refined in light of recent evidence for time-varying risk premia in the oil futures market is still unclear,
but there is no reason to expect the answer to change fundamentally. In particular, the ability of economic
fundamentals such as unexpectedly strong demand for crude oil from emerging Asia to explain the surge
in the real price of oil between 2003 and mid-2008 does not depend on how the oil futures market is
modeled.
2. WHAT IS SPECULATION?
One of the reasons why discussions about the role of speculation often degenerate into blanket
generalizations is that it is rarely clear how speculation is defined. The most general economic definition
of speculation is provided in Kilian and Murphy (2011) who note that anyone buying crude oil not for
current consumption, but for future use is a speculator from an economic point of view. What is common
to all speculative purchases of oil is that the buyer is anticipating rising oil prices. Speculative buying
may involve buying crude oil for physical storage leading to an accumulation of oil inventories, or it may
involve buying an oil futures contract, provided an oil futures market exists. Either strategy allows one to
3 
 
take a position on the expected change in the price of oil. Standard theoretical models of storage imply
that there is an arbitrage condition ensuring that speculation in one of these markets will be reflected in
speculation in the other market (see, e.g., Alquist and Kilian 2010).
It is immediately clear that speculation defined in this manner need not be morally reprehensible.
In fact, speculation may make perfect economic sense and indeed is an important aspect of a functioning
oil market. For example, it seems entirely reasonable for oil companies to stock up on crude oil in
anticipation of a disruption of oil supplies because these stocks help oil companies smooth the production
of refined products such as gasoline. The resulting oil price response helps alleviate future shortages by
curbing current oil consumption and providing incentives for additional exploration and production.
Hence, it would be ill-advised for policymakers to prevent such oil price increases.
2.1 Excessive Speculation
Notwithstanding these economic arguments, the term speculation in the public debate has a negative
connotation because speculation is viewed as excessive. Excessive speculation may be defined as
speculation that is not required for the oil market to function properly. Excessive speculation might also
be defined as speculation that is beneficial from a private point of view, but would not be beneficial from
a social planner’s point of view. It follows naturally that the public has an interest in preventing this form
of speculation. The broad definition of speculation in Kilian and Murphy (2011) makes no distinction
between socially desirable and undesirable speculation, making it difficult to judge the extent of excessive
speculation unless there is no evidence of any speculation at all, in which case excessive speculation must
be absent as well.
Determining whether speculative trading is excessive is difficult. For example, it does not depend
on who engages in speculation. It is common in the policy debate to define speculation in terms of who is
buying the oil. Traditionally, traders in oil futures markets with a commercial interest in (or an exposure
to this physical commodity) such as oil companies or airlines have been called hedgers, while those
without a physical position to offset have been called speculators. This distinction between hedging and
4 
 
speculation in futures markets is less clear than it may appear, however. First, the oil futures market
cannot function without speculative traders providing liquidity and assisting in the price discovery. The
presence of speculators defined as noncommercial traders tells us nothing about whether speculation is
excessive. Second, in practice, commercial traders may take a stance on the price of a commodity or may
not hedge in the futures market despite having an exposure to the commodity. Both positions could be
considered speculative (see Büyükşahin and Harris 2011). Likewise, there is nothing that prevents oil
companies from acting as speculators in the physical market. Nor is there anything excessive about a
hedge fund, for example, storing crude oil in a tanker in anticipation of rising oil prices, as long as the
hedge fund bears the risk of the price of oil not increasing enough to compensate for the costs of this
holding strategy. These examples illustrate the futility of defining excessive speculation by classifying
different types of traders.
Likewise, efforts to detect speculators on the basis of high ex-post profits are not compelling.
After all, speculators take risky positions and the return on holding oil must reflect that risk. Another
common argument has been based on the relative size of the oil futures market and the physical market
for oil. For example, it is often asserted that the daily trading volume in oil futures markets is at least three
times as high as daily physical oil production, fueling the suspicion that speculators are dominating this
market. This ratio is misleading because it compares a stock in the numerator to a flow in the
denominator. Ripple (2008) shows that this ratio – after taking account of the number of days to delivery
for the oil futures contracts – is only a fraction of about one half of daily U.S. oil usage rather than a
multiple, invalidating this argument.
Yet another approach has been to quantify speculation following Working (1960) as an index
measuring the percentage of speculation in excess of what is minimally necessary to meet short and long
hedging demand (see, e.g., Büyükşahin and Harris 2011). A high Working index number, however, does
not necessarily indicate excessive speculation. One natural benchmark when evaluating this index are
historical values of the same index for other commodity markets for which excessive speculation is not
considered a problem. By that standard, the index numbers for the oil market even at their peak remain in
5 
 
the midrange of historical experience. Moreover, there does not appear to be a simple statistical
relationship between this index of speculation and the evolution of the price of oil. For example, although
the index for near-term contracts for the most part increased from 2001 until 2005, it remained roughly
constant between early 2006 and early 2009, while the average price of oil first rose from $60 to $147 and
then fell to $35. The correlation between the speculative index and daily price changes is near zero (see,
e.g., Büyükşahin and Harris 2011; Alquist and Gervais 2011).
Sometimes excessive speculation is equated with market manipulation. For example, it has been
asserted that financial traders are herding the market into positions from which they can profit, resulting
in excessively high oil prices in the spot market. It is important to stress that market manipulation and
speculation are economically distinct phenomena. For example, the increased financialization of oil
markets does not by itself mean that market manipulation is on the rise. Nor is there widespread evidence
of market manipulation, notwithstanding some isolated cases in recent years.2
We conclude that there is no operational definition of excessive speculation. As a result, no one
has been able to quantify the problem of excessive speculation to date. Rather the public debate has
tended to conflate several recent developments in oil futures markets and in the spot market for crude oil,
taking the existence of excessive speculation as self-evident.
2.2 The Masters Hypothesis
The starting point for the public debate has been the well-documented financialization of oil futures
                                                            
2
 While recognizing the difficulty of defining manipulation in a precise manner, Pirrong (1994) emphasizes the
‘corner’ or ‘squeeze’ as the main form of manipulative activities. A squeeze refers to a situation in which the trader
goes long in the futures market by an amount that exceeds the actual barrels that can be delivered during that month.
The squeezer can then demand the ‘shorts’ to deliver the barrels. The shorts have to secure the barrels often at a
higher cost because they have to bid for these barrels from far away markets. This cost increases with the number of
deliverable barrels and hence the bigger the position, the higher the premium that the squeezer can extract.
Alternatively, the shorts can agree to cash settle at a price above the competitive price. A successful squeeze will
have important effects on prices and the flow of the underlying commodity increasing the spot price both in absolute
value and relative to the price of oil for forward delivery. Also the price of oil in the delivery market will rise
relative to other locations. Thus, manipulation undermines the two main functions of the futures markets: (1) the
price discovery function by distorting absolute and relative prices; and (2) the risk shifting function by reducing the
effectiveness of hedging. There is a general consensus that policymakers must protect the integrity of futures market
against such manipulative strategies. The debate focuses on whether the government should adopt harm-based rules
or preventive measures and whether the existing regulations regarding manipulation are effective.
6 
 
markets in recent years. As shown in Alquist and Kilian (2007), Büyükşahin et al. (2009), and Fattouh
(2010a), starting in about 2003 there was an influx of investors from outside the oil industry (often
collectively and somewhat imprecisely referred to as hedge funds) into oil futures markets and other
commodity futures markets. The presumption is that these new financial investors were attracted by the
prospect of high returns. At about the same time, both spot and futures prices of crude oil began to surge,
soon reaching unprecedented levels. A natural conjecture at the time was that this price surge was caused
by the financialization of oil futures markets. In this interpretation, initially proposed by non-academics
such as Masters and Soros testifying before Congress, hedge funds took speculative positions that resulted
in rising oil futures prices, which in turn caused a similar increase in the spot price of oil.
It is useful to break up the Masters hypothesis into its components, each of which represents a
distinct link in the chain of causation: First, the argument is that financial investors after 2003 entered the
oil futures market in large numbers. Second, it is asserted that these investors’ demand for long positions
drove up the oil futures price. Third, the claim is that higher oil futures prices signaled expectations of
rising spot prices and hence drove up the demand for oil inventories above the ground (and possibly
below the ground). Fourth, higher oil inventory demand is thought to have increased the real spot price of
oil and hence the real price of gasoline.
As we will discuss in more detail below, the veracity of this view is not obvious at all and much
of the academic debate centers on the evidence, if any, supporting Masters’ (2008, 2010) interpretation.
For example, what should matter is not only the number of financial investors, but also what net positions
they take and how their presence changes the risk aversion, horizon, and risk-bearing capacity of traders
overall.3
Moreover, an exogenous shift in the composition of traders is not the only possible explanation
for the surge in the price of oil. An alternative interpretation of the same evidence is that financial
investors merely responded to the same market forces as other market participants and that both spot and
futures prices were driven by the same economic fundamentals. Identifying cause and effect in financial
                                                            
3
  For example, Etula (2010) attributes the evolution of the risk premium in oil futures markets during 2008 to the
arrival of institutional investors with a greater capacity to bear short-term price volatility.
7 
 
markets requires a structural model such as the model we will discuss in section 7. Causality cannot be
inferred from predictive correlations.
One reason that the Masters hypothesis has received disproportionate attention among
policymakers is that it seems to provide an obvious remedy to the problem of rising oil and commodity
prices. To the extent that financial speculation is the cause of the problem of rising oil prices, policies
aimed at restricting trades in oil futures markets can be expected to prevent increases in the spot price of
oil. This interpretation has informed recent policy efforts to tighten regulation in oil derivatives markets
as part of a larger effort by the G20 governments to tighten the regulation of international financial
markets. While these policy measures are perhaps understandable within the broader context of the global
housing and banking crisis, they are not based on solid evidence. The following sections review the
evidence in support of the Masters hypothesis from a variety of angles, mirroring the evolution of the
academic literature on this subject.
3. EVIDENCE FOR THE INCREASED FINANCIALIZATION OF OIL FUTURES MARKETS
While the definition of the term financialization remains vague, it captures the increasing acceptance of
oil derivatives as a financial asset by a wide range of market participants including hedge funds, pension
funds, insurance companies, and retail investors. Financial innovation provided easy and inexpensive
access to financial instruments such as futures, options, index funds, and exchange-traded funds. The
most obvious candidate explanation for this shift toward the use of energy derivatives is a secular change
in investor preferences, involving increased reliance on market-based instruments for consumption
smoothing, insurance and diversification, perhaps spurred by lower transaction costs or lower risk
aversion. Other possible explanations are more temporary in nature and include that risk-free interest rates
were low; that there was excess liquidity in financial markets; that returns in other financial markets were
low or too volatile; or that crude oil was seen as a good hedge against inflation risks and a weak U.S.
dollar.
8 
 
There is clear evidence of increased involvement of financial investors in oil futures markets
starting in 2003 (see Alquist and Kilian 2007; Büyükşahin et al. 2009; Tang and Xiong 2011; Hamilton
and Wu 2011). Much of the initial analysis of the speculation hypothesis focused on documenting
changes in the co-movement of oil prices and financial asset prices following this financialization of oil
futures markets. For example, Büyükşahin, Haigh and Robe (2010) and Silvennoinen and Thorp (2010)
show that the conditional correlation between commodity futures returns and U.S. stock index returns has
increased in recent years, especially in periods of high volatility. Büyükşahin and Robe (2011) report
increased equity-energy price co-movements which appears related to the entry of hedge funds that take
positions in both equity and energy markets. Specifically, holding other things equal, they find that a 1%
increase in the overall share of hedge funds in energy futures is associated with 5% increase in the
dynamic correlation between energy and equity returns. This result, however, does not hold for other
types of traders such as commodity index traders and swap dealers. They also find that the correlation is
weaker in periods of financial market stress. Other studies examine the impact of the entry of index
investors on the price co-movement between energy and other commodities. Tang and Xiong (2010) find
that prices of non-energy commodities became increasingly correlated with oil prices. They also find that
this result is more pronounced for commodities that are included in commodity indices. In a similar vein,
Silvennoinen and Thorp (2010) conclude that the correlation between crude oil future returns and other
commodities such as agricultural commodities has increased. There also is evidence of increased
correlation between returns in foreign exchange and returns on oil (see, e.g., Akram 2009).
Other researchers have studied so-called commodity index funds and shown that the correlation is
higher among returns within the index than among returns outside the index and that the correlation
between energy-based indices and non-energy indices has increased. This has been taken as evidence of
financial market integration driven by the rapid growth of commodity index funds (see Tang and Xiong
2010). Commodity index funds obtain price exposure by entering long positions in futures contracts. In
order to maintain the long exposure, they need to unwind the maturing contracts before they expire and
initiate new long positions in contracts that have later maturity dates. Because the funds have to roll their
9 
 
entire positions forward on a regular basis, the potential pressure on the prices of contracts involved can
be very large, given the size of investment. Mou (2010) shows that the profitability of trading strategies
designed to exploit this market anomaly increases in the share of index fund investment in oil futures
markets and decreases in the amount of arbitrage capital.
The implications of increased co-movement between oil prices, financial asset prices and other
commodity prices on social welfare are not clear. Büyükşahin and Robe (2011) in recognition of this
point stress that “additional work is needed, if one is to ascertain whether the impact of financialization on
cross-correlations represents a welcome improvement in market efficiency or, instead, is a worrisome
development”. Some studies argue that the increased co-movement of prices enhances spillover effects
from other financial markets and from other commodity markets (see Tang and Xiong 2010). Others
suggest that the increased co-movement had the effect of eroding the diversification benefits in
commodity markets, one of the main motives for investors to include commodities in their portfolios in
the first place (Silvennoinen and Thorp 2010). On the other hand, Fattouh (2010b) shows that the
emergence of highly liquid crude oil futures markets has helped physical crude oil markets become more
integrated by reducing transaction costs and facilitating arbitrage across geographically distant markets
and across crude oil of different quality. Similarly, Büyükşahin et al. (2009) find that increased
financialization is associated with more efficient derivatives pricing methods and Pirrong (2011) shows
that greater financial market integration may reduce the market price of risk and increase the level of
inventories by reducing the cost of hedging. While this mechanism induces an increase in the spot price,
the higher level of oil inventories reduces the chances of future price spikes.
Most importantly, evidence of increased comovement between the spot price of oil, oil futures
prices and other asset prices does not imply that the surge in the spot price was caused by financial
speculators. First, economic theory tells us that both spot prices and futures prices are determined
simultaneously and respond to the same economic forces (see Alquist and Kilian 2010). To the extent that
global macroeconomic fundamentals have changed in recent years, for example, that fact could provide
an alternative explanation of the observed comovement in spot and futures prices.
10 
 
Second, there are other pieces of evidence that are inconsistent with the view that financialization
is driving the price of oil. For example, Stoll and Whaley (2010) stress that the price of commodities not
traded on futures exchanges rose as much as or more than the price of exchange-traded commodities.
Indeed, the evolution of the price of crude oil is broadly similar to that of an index of non-exchange
traded commodities (see Figure 1). The same point has been made by Irwin, Sanders, and Merrin (2009).
This evidence is consistent with research quantifying broad-based shifts in the demand for all industrial
commodities during 2003-08 (see, e.g., Kilian 2009a,b; Kilian and Murphy 2011; Kilian and Hicks
2012).4
4. DO INDEX FUND FLOWS PREDICT THE PRICE OF OIL FUTURES?
Subsequent studies have focused on the question of whether the increased importance of index funds
helps predict higher returns on oil futures contracts. On the basis of such predictive correlations,
Singleton (2012), for example, concluded that financial flows affect the return on crude oil. This
prompted Reuters to assert that “Stanford University’s Professor Kenneth Singleton has mounted the most
wide-ranging and influential assault so far on the orthodoxy among policymakers and academics that
speculation does not affect commodity prices” (Kemp 2011). This conclusion is remarkable in several
respects. First, it had been academics that were reluctant to embrace Masters’ hypothesis rather than
policymakers. In fact, the prevailing orthodoxy among policymakers had been precisely the position
Singleton advocates, not the position that speculation does not affect commodity prices. Second, the only
evidence presented by Singleton is that financial flows precede changes in the oil futures price, but
predictability does not necessarily imply causality. Third, the central question not answered by Singleton
is how financial flows explain the corresponding increase in the spot price of oil. Fourth, Singleton’s
evidence was based on a very short estimation period. There is little controversy about the fact that the
                                                            
4
At the same time, persistent deviations between oil futures prices and natural gas futures prices suggest a
potentially important role for supply constraints in individual commodity markets. The development of shale gas in
particular decoupled price developments in the natural gas market from those in oil markets. The view that the
supply side matters is also consistent with evidence in Kilian (2009a) who studies the evolution of oil and non-oil
commodity prices in recent years and notes systematic differences between the price of oil and the price of other
more elastically supplied industrial commodities.
11 
 
price of oil futures rose when index fund investments increased starting in 2003. The question is whether
these index funds moved endogenously in response to economic incentives or exogenously shifted the oil
futures price. In this sense, Singleton’s analysis raises more questions than it answers.
Quite apart from these issues of interpretation there are concerns with the nature of the evidence
provided by Singleton (2012). His empirical analysis – like that of Masters (2008, 2010) – is based on
highly aggregated CFTC data on the positions of index funds (see Figure 2). Büyükşahin in a series of
coauthored papers since 2009 has shown that this heavily aggregated data set is unsuitable for studying
the question of speculation and that very different conclusions about the role of speculation are obtained
when alternative proprietary CTTC data sets are used that include individual traders’ positions. Moreover,
Singleton’s measure of index fund positions in oil futures is in fact inferred from CFTC data on
agricultural futures, which subsequent research has shown to have little relation to the index funds’ actual
positions regarding oil (see, e.g., Irwin and Sanders (2012). Based on an alternative CFTC data set that is
superior to that used by Singleton in terms of coverage and accuracy, Irwin and Sanders show that the
WTI crude oil futures positions computed by Masters and by Singleton are grossly misleading. The
imputed data used by Singleton and several other researchers not only contain large absolute errors, but
have near zero correlation with the best available estimates of actual positions taken by index funds. Irwin
and Sanders conclude that the data used by Singleton at best suffer from severe measurement error and at
worst lead to highly misleading statistical estimates and inference.
The same point has been illustrated by a large number of papers. Brunetti and Büyükşahin (2009),
for example, examine the same question as Singleton with a data set showing individual positions taken
by traders. Unlike Singleton, they find no evidence that financial investor flows predict movements in oil
futures prices or price volatility. Rather such speculative trading was associated with reduced volatility.
Likewise, Büyükşahin and Harris (2011) find no evidence that the positions of hedge funds or other noncommercial
investors predict changes in the futures price; rather futures price changes precede changes in
positions. Brunetti, Büyükşahin, and Harris (2011) study specific categories of traders and test whether
positions taken by speculators such as hedge funds and swap dealers cause changes in oil futures prices or
12 
 
price volatility. Their results are consistent with speculators providing valuable liquidity to the market and
with speculators reacting to market conditions rather than vice versa. Similarly, Irwin and Sanders (2011)
show that there is no statistically significant relationship between growth in the volume of oil futures
contracts on the one hand and oil futures returns, their realized volatility and their implied volatility on the
other hand. When a relationship is found at all, it is negative rather than positive. Similar results are also
reported in Stoll and Whaley (2011) and Sanders and Irwin (2011) based on a different data set. In
addition, Hamilton and Wu (2012) demonstrate that the agricultural index fund positions used by
Singleton paradoxically predict the futures price of oil more accurately than the futures price of
agricultural commodities. Moreover, the same model based on agricultural index fund positions also
predicts the U.S. stock market. The apparent explanation is that the positive predictive correlation found
by Singleton on the basis of a very short sample is driven by the 2008 recession. Indeed, the predictive
correlation breaks down when the sample is extended by two years.
Finally, even if we were to accept the results in Singleton (2012), his evidence is not necessarily
evidence of speculation. It is equally consistent with the existence of frictions in the processing of
information. This point has been stressed in a more general context by Hirshleifer (1988, 1989). Given the
presence of setup costs (possibly related to informational costs) that limit the participation of noncommercial
traders in the futures market, the risk premium depends both on systematic risk and on a
residual risk linked to hedging pressure, with increased hedging pressure increasing the size of the risk
premium and affecting oil futures and spot prices. Chang (1985) and De Roon el al. (2000) provide
evidence that in many markets producers’ hedging positions are correlated with commodity futures price
changes, consistent with the existence of such frictions. In related work, Hong and Yogo (2012) show that
under the assumption of limited risk absorption capacity in futures market, the change in open interest
constitute a more reliable signal about future economic activity and asset prices than the oil futures price.
Consistent with their model, they find that changes in open interest help predict changes in the oil futures
price and in the spot price of oil. In their model, the main driver of open interest is anticipation of higher
economic activity, which induces higher hedging demand.
13 
 
5. DO FUTURES PRICES PREDICT SPOT PRICES?
Another strand of the literature has focused on the link between futures prices and the spot price. It is
sometimes argued that higher futures prices cause increases in the spot price of oil, simply because longterm
oil contracts often use the oil futures price as a reference price. This argument is not compelling
because the oil futures price reflects expectations about future demand and supply conditions in the oil
market. Theoretical models of storage imply that these same expectations must be reflected in the spot
price by construction. For example, Peck (1985) emphasized that “expectations are reflected nearly
equally in current and in futures prices. In this sense cash prices will be nearly as good predictions of
subsequent cash prices as futures prices”, echoing in turn the discussion in Working (1942) who was
critical of the “general opinion among economists that prices of commodity futures are … the market
expression of consciously formed opinions on probable prices in the future” whereas “spot prices are not
generally supposed to reflect anticipation of the future in the same degree as futures prices”. Working
specifically criticized the error of “supposing that the prices of futures … tend to be more strongly
influenced by these anticipations than are spot prices”.
Nevertheless, there has been a presumption in the public debate that the evolution of oil futures
prices at a minimum helps to predict changes in the spot price of oil. This is not the case. Alquist, Kilian
and Vigfusson (2012) examine the out-of-sample accuracy of daily and monthly oil futures prices and
show that there is no compelling evidence that oil futures prices help forecast the spot price of oil.
Although this evidence does not address the central question of what caused the surge in the price of oil
after 2003, it invalidates one of the arguments to which proponents of the Masters hypothesis about the
role of index funds appeal to.
6. CRUDE OIL INVENTORY DATA
Yet another strand of the literature has examined data on crude oil inventories for evidence of speculative
pressures (see, e.g., Hamilton 2009a). Often rising levels of crude oil inventories are interpreted as a sign
of speculative pressures in oil markets. Figure 3 plots the relationship between U.S. commercial stocks of
14 
 
crude oil held above the ground and the nominal West Texas Intermediate (WTI) price of crude oil during
1994-2011. The plot shows that there is little systematic relationship. It is sometimes claimed that the
negative relationship between these variables that used to prevail during 1994-2004 has been overturned
by speculators entering the oil futures market after 2003. Indeed, there is evidence of a positive slope in
post-2004 data. The data in Figure 3 may seem to support this interpretation at first sight, but there are
several problems with this argument.
First, to the extent that there is a systematic relationship driven by speculative demand for
inventories, economic theory tells us that this relationship should be between the speculative component
of the real price of oil (or equivalently the negative of the oil futures spread) and the change in global
crude oil inventories (see Alquist and Kilian 2010). In this sense, both axes in Figure 3 are mismeasured
and the lack of a systematic relationship is not surprising. Second, such a relationship would only be
expected to hold in the absence of other oil demand and oil supply shocks. Third, the conventional
economic argument for a positive relationship between oil prices and inventories after 2003 relies on a
static model in which speculators exogenously drive the futures price above its equilibrium level, which
all else equal encourages new production and discourages consumption. This in turn causes an unplanned
accumulation of inventories, resulting in positive co-movement between the spot price of oil and crude oil
inventories. This model is not realistic because it takes as given the premise that the increase in the
futures price is exogenous with respect to the oil market. It rules out by assumption that these speculators
were responding endogenously to expectations of future shortages. Standard theoretical models of storage
tell us that all relevant model variables will respond endogenously to shifts in expectations about future
oil supply shortfalls, violating the ceteris paribus assumption in the static model. In these models, oil
inventories, the spot price of oil and the futures price are jointly and simultaneously determined rather
than the futures price being exogenously given.
A detailed analysis of such a dynamic rational expectations model can be found in Alquist and
Kilian (2010) who study the relationship between changes in oil inventories, the oil futures spread and
changes in the real price of oil driven by precautionary demand shocks. Their focus is on the effects of
15 
 
increased uncertainty about future oil supply shortfalls, modeled as an exogenous mean-preserving
spread. Similarly, Pirrong (2008) advocates the use of a dynamic rational expectations model for
analyzing the commodity storage problem. In his model, there is a random shock to economic
fundamentals with a stochastic variance that reflects increased uncertainty about future oil supply
shortfalls. As in Alquist and Kilian, forward-looking agents respond to an exogenous increase in this
variance by increasing inventory holdings which requires the spot price of oil to increase. If the variance
is sufficiently high, the reduced-form relationship between inventories and the real spot price becomes
unstable.
In related work, Dvir and Rogoff (2010) study exogenous shocks to the demand for oil in the
form of exogenous shifts in world income. In their model agents increase storage optimally in response to
an income shock, causing the real spot price of oil to increase in equilibrium because expectations of
higher growth in the future amplify the demand for storage. This means that the volatility of the price of
oil increases due to storage contrary to the notion that storage necessarily reduces oil price volatility. Dvir
and Rogoff note that their model is particularly applicable to modern oil markets since 1973.
The absence of a stable relationship between inventories and the real price of oil is also a central
feature of structural oil market models of the type proposed in Kilian and Murphy (2010). Their analysis
focuses on the post-1973 period. Unlike Dvir and Rogoff’s theoretical analysis, the empirical model of
Kilian and Murphy treats all observables as fully endogenous. A popular argument in the literature,
building on the static model with exogenous shifts in the oil futures price that we discussed earlier, has
been that a persistent accumulation of oil inventories implies the presence of speculators. Kilian and
Murphy show that this argument can be misleading once all variables are treated as endogenous. An
observed increase in inventories, for example, is equally consistent with an increase in inventory demand
driven by speculative demand shocks and an unexpected slowing of the global economy or an
unanticipated increase in oil production. Kilian and Murphy also illustrate that historically different
demand and supply shocks in the global oil market often resulted in offsetting changes in oil inventories,
while reinforcing changes in the real price of oil in the same direction, so the absence of changes in oil
16 
 
inventories does not imply the absence of speculation. This evidence reinforces the point that one should
not expect a stable relationship between the real price of oil and inventories. Indeed, the current
consensus in the literature is that “those searching for evidence of speculative excess need look elsewhere
than the price-inventory relation” (Pirrong 2008).
7. STRUCTURAL ECONOMIC MODELS OF OIL MARKETS
The discussion in section 6 highlights that there is no substitute for a structural model of the oil market in
examining the evidence for speculation. Theoretical models of storage such as Alquist and Kilian
(2010) or Pirrong (2008, 2011) make it clear that the notion that the real price of oil is no longer driven by
demand and supply, but by speculation, as is often asserted in the media, is wrong because speculative
demand is simply one component of the demand for oil. Moreover, speculative demand may arise from
expectations of future shifts in economic fundamentals. The challenge is how to model speculation within
an empirical model that nests different explanations of changes in the real price of oil.
7.1 THE KILIAN-MURPHY MODEL
Kilian and Murphy (2011) proposed the first empirical model designed to quantify the effect of
speculative demand shocks on the real price of oil. Their model builds on a growing literature using
structural vector autoregressive (VAR) models to disentangle demand and supply shocks in global oil
markets (see, e.g., Kilian (2009a,b); Kilian and Murphy 2012; Baumeister and Peersman 2012). The set of
variables in Kilian and Murphy (2011) consists of monthly data for the percent change in global oil
production, a measure of global real activity (in deviations from trend), the real price of crude oil, and the
change in above-ground global crude oil inventories. The model is identified based on a combination of
sign restrictions and bounds on the short-run price elasticities of oil demand and oil supply.
The key identifying assumptions for the purpose of this discussion are restrictions on the signs of
the impact responses of the four observables to each structural shock. There are four structural shocks.
First, conditional on past data, an unanticipated disruption in the flow supply of oil causes oil production
to fall, the real price of oil to increase, and global real activity to fall on impact. Second, an unanticipated
17 
 
increase in the flow demand for oil associated with unexpected changes in the global business cycle
causes global oil production, global real activity and the real price of oil to increase on impact.5
Third, a
positive speculative demand shock, defined as an increase in inventory demand driven by expectations
shifts not already captured by flow demand or flow supply shocks, will in equilibrium cause an
accumulation of oil inventories and will raise the real price of oil (see, e.g., Alquist and Kilian 2010). The
accumulation of inventories requires oil production to increase and oil consumption to fall (associated
with a fall in global real activity). The focus on above-ground crude oil inventories is consistent with
conventional accounts of speculation involving the accumulation of inventories in oil-importing
economies. An alternative view is that speculation may also be conducted by oil producers who have the
option of keeping the oil below the ground in anticipation of rising prices (see Hamilton 2009b). Such an
accumulation of below-ground inventories by oil producers in this model would be reflected in a
reduction in flow supply. No attempt is made to distinguish such speculative supply shocks from other
supply shocks. Finally, the model also includes a residual oil demand shock designed to capture
idiosyncratic oil demand shocks driven by a myriad of reasons that cannot be classified as one of the first
three structural shocks. The latter demand shock has little systematic effect on the real price of oil.
The Kilian and Murphy model is designed to remain valid even if there are limits to arbitrage
between oil futures and spot markets. In fact, one of its advantages is that the identification strategy does
not require the existence of an oil futures market, but remains valid in the presence of an oil futures
market. This allows the use of data back to 1973 in estimating the model. Indeed, it can be shown that the
oil futures spread is redundant in this model because all the information about expectations is already
captured by the data on oil inventories. This interpretation is consistent with the fact that the oil futures
spread, when it is available, does not Granger cause the four model variables, which may be viewed as a
test of the model specification (see Giannone and Reichlin 2006).
                                                            
5
The response of above-ground oil inventories to flow demand and flow supply shocks is left unrestricted in the
model because the price responses to these shocks could potentially trigger an accumulation of inventories due to
speculation, while refiners are likely to smooth their production of refined products, resulting in lower inventories.
Which effect dominates is not clear ex ante, but the model estimates indicate a negative response on balance.
18 
 
The estimates of the Kilian and Murphy model show that indeed there were several historical
episodes in which speculation moved the real price of oil including 1979, 1986, 1990 and late 2002. This
evidence is important because it illustrates that the model has the power to detect speculation when it
exists according to anecdotal evidence (see, e.g., Yergin 1992). With regard to the surge in the real price
of oil between 2003 and mid-2008, in contrast, the model estimates unambiguously show that there is no
evidence to support the notion that this surge was caused by speculation. If so, one would have expected
positive speculative demand shocks to have driven up the real price of oil in the spot market. This is not
what the structural VAR model estimates imply. To the extent that there is unlimited arbitrage, as
maintained in the Masters hypothesis, these estimates also imply that there was no speculation in the oil
futures market because by arbitrage such speculation would have caused speculative demand for
inventories to shift. While the latter implication is necessarily more tentative, it is consistent with
evidence based on disaggregate CFTC data that index funds did not drive up the oil futures price.
Likewise there is no evidence for speculation by oil producers or for other negative oil supply
shocks causing the price surge. Negative flow supply shocks during 2003-08 raised the real price of oil
slightly, but the increases were minor compared with the observed surge in the real price of oil. Nor is
there evidence for the peak oil hypothesis having affected the real price of oil. Instead the model indicates
that a sequence of positive flow demand shocks, none of which was particularly large individually, is
perfectly capable of explaining the observed surge in the real price of oil, even during 2007/08. This
empirical result is also consistent with independent evidence in Kilian and Hicks (2012) based on an
analysis of professional real GDP forecasts, which highlights the role of increased demand for crude oil
and other industrial commodities from emerging Asia, and with evidence based on dynamic stochastic
general equilibrium models of the world economy in Bodenstein and Guerrieri (2011). The same type of
flow demand shock in the VAR model also explains the partial recovery of the real price of oil after 2009.
This evidence implies that both oil futures prices and spot prices were driven up by the same economic
fundamentals. In short, the Kilian and Murphy model provides the most decisive and most formal
19 
 
evidence on the question of speculation to date.6
7.2 OTHER STUDIES
Recently, two related structural VAR studies have argued that the Kilian and Murphy model does not
explicitly allow for “financial speculation” and hence may be misleading (see Juvenal and Petrella 2011;
Lombardi and Van Robays 2011).
Juvenal and Petrella (2011) introduce an additional shock into the Kilian and Murphy model
designed to capture speculation by oil producers, while retaining the speculative demand shock in that
model. Their analysis relies on a static economic model in which there is an exogenous upward shift in oil
producers’ expectations of the real price of oil. Oil producers expecting a higher price for future
deliveries, as a result, will withhold oil from the market and accumulate inventories. This accumulation
takes the form of an accumulation of below-ground inventories notably in OPEC countries, resulting in a
reduction in flow supplies, associated with a higher spot price and lower oil consumption and hence lower
real activity. Juvenal and Petrella’s first crucial identifying assumption is that the impact response to a
negative speculative supply shock involves an increase in above-ground oil inventories in oil importing
countries. This assumption is inconsistent with the underlying economic model which predicts an
accumulation of below-ground inventories.7
The only way to defend this assumption is to impose the
hypothesis that speculation by oil producers and speculation in oil-importing countries has been perfectly
correlated since 1973, so we can treat speculative demand and speculative supply shocks as one hybrid
shock. Adding such a hybrid shock to the model renders the original speculative demand shock
redundant, but, instead of dropping the latter demand shock, Juvenal and Petrella impose an additional
sign restriction on the inventory response to flow supply shocks, for which there is no theoretical
                                                            
6
It also has been shown that this model may be used to construct more accurate real-time out-of-sample forecasts of
the real price of oil than alternative models including the random walk model and forecasts based on oil futures
prices (see Baumeister and Kilian 2012). The latter evidence is instructive in that it suggests that the linear VAR
model is a good first-order approximation even in the presence of structural changes such as fluctuations in the risk
premium (see section 8).
7
One reason for this assumption is that the speculative supply shock would be observationally equivalent to a flow
supply shock if we restricted the response of below-ground inventories. This is indeed the reason why Kilian and
Murphy (2011) do not differentiate between flow supply and speculative supply shocks. The other reason is that
time series data for global below-ground oil inventories (more commonly known as “reserves”) are not available.
20 
 
justification. This allows them to retain two “speculative” shocks within their model. Even with these
(and numerous other) changes in the model specification, however, their estimates show that flow demand
shocks are the primary cause of the oil price surge after 2003, confirming the substance of the findings in
Kilian and Murphy (2011). More importantly for the purpose of this survey, even leaving aside all caveats
about their model identification and specification, their analysis does not tell us whether the hybrid
“speculative” shock is actually associated with financial speculation. There is no way of knowing based
on their model. In short, the analysis in Juvenal and Petrella is not informative about the role of financial
speculation.
In related work, Lombardi and Van Robays (2011) propose to augment the Kilian and Murphy
model with data on the oil futures price and to re-estimate the model on data since the late 1980s.8
Their
key innovation is the introduction of an additional ‘nonfundamental’ financial speculation shock. This
financial speculation shock is identified as a shock that raises both the oil futures spread and the oil
futures price on impact. All other impact responses to this shock are left unrestricted. The key identifying
assumption is the positive response of the spread. This is in contrast with the negative response of the
spread to a speculative demand shock in standard models of storage. However, they fail to impose other
restrictions implied by economic theory such as the restriction that the impact responses of oil production
and of global real activity should be negative and that of the real price of oil should be positive. Likewise,
the Lombardi and Van Robays fail to impose the other restrictions on oil inventories implied by the
analysis in Kilian and Murphy and the additional restrictions on the spread implied by the analysis in
Alquist and Kilian (2010). This invalidates the model estimates for the reasons discussed in Kilian and
Murphy (2012). Moreover, their interpretation of the speculative demand shock of Kilian and Murphy
(2011) as an oil-specific demand shock is incorrect, as is the assertion that this shock is designed to reflect
expected changes in fundamentals only and cannot capture other expectations shifts. We conclude that
                                                            
8
 Their model is misspecified in that they model spot and futures prices in differences and in nominal terms,
rendering all estimates inconsistent. A more appropriate specification would have involved the oil futures spread and
the real price of oil in the spot market. Here we abstract from this and other modeling choices in the interest of
focusing on the identification. 
21 
 
Lombardi and Van Robays’ analysis does not shed light on the role of financial speculation shocks.9
A very different approach was taken by Hamilton (2009b) who observes that speculation in oil
markets indeed will have to involve a response of above-ground inventories unless the short-run price
elasticity of gasoline demand is literally zero, allowing increases in the real price of oil to be passed on to
consumers without a reduction in consumption causing an involuntary accumulation of inventories. The
empirical relevance of this limiting case depends on the magnitude of the elasticity in question. Kilian and
Murphy (2011) show that under reasonable assumptions this gasoline demand elasticity is approximately
the same as the short-run price elasticity of oil demand. They also show that traditional estimates of the
short-run price elasticity of oil demand based on reduced-form models are not identified, causing these
estimates to be biased toward zero. Alternative structural estimates that correct for this problem are
consistently far from zero. For example, the preferred estimate of the short-run price elasticity of oil
demand in Kilian and Murphy is about -0.25. Similar results have been obtained using many other
econometric methodologies that address the central issue of identification (see, e.g., Serletis, Timilsina
and Vasetsky 2010; Bodenstein and Guerrieri 2011; Baumeister and Peersman 2012). This conclusion
also is consistent with recent structural estimates of the short-run price elasticity of gasoline demand
based on instrumental variable regressions (see, e.g., Davis and Kilian 2010). An unresolved question is
to what extent the price elasticity of gasoline demand may have declined in magnitude in very recent
years and at what time (see, e.g., Hughes, Knittel, and Sperling 2008). It would take a substantial decline,
however, to make the limiting case of a zero elasticity discussed in Hamilton (2009b) empirically
relevant.
8. MODELS OF TIME-VARYING RISK PREMIA
Buying a futures contract involves risk that is compensated for in equilibrium by a risk premium.
                                                            
9
An additional problem common to both Juvenal and Petrella (2011) and Lombardi and Van Robays (2011) is that
the interpretation of the empirical results is based on median response functions. It has been shown that median
response functions are not a valid summary measure of the central tendency of the impulse response estimates nor
do they have an economic interpretation, making it difficult to interpret the results in these papers (see, e.g., Fry and
Pagan 2011; Inoue and Kilian 2011; Kilian and Murphy 2012).
22 
 
Standard theoretical and empirical models of oil markets treat this risk premium as constant over time.
Indeed, Alquist and Kilian (2010) provide empirical evidence that the risk premium in oil markets is on
average not significantly different from zero. This evidence would also be consistent with time-varying
risk premia which alternate in sign depending on the state of the economy, however. A number of recent
studies have developed models in which the risk premium evolves over time. Of particular concern is the
possibility that the risk premium may have decreased in recent years during the time of the surge in the
real price of oil. A time-varying risk premium could potentially undermine the usefulness of existing
empirical and theoretical models based on a constant risk premium.
Standard models of speculation such as Alquist and Kilian (2010) focus on refiners’ use of oil
inventories to hedge oil price risks, building on the notion of a convenience yield. In other words, oil
storage is valuable because it provides insurance against supply disruptions or against unexpected
increases in demand. In these models, the oil futures market merely provides an additional financial
instrument that serves much the same purpose as accumulating crude oil inventories. Acharya et al.
(2010) propose an alternative model of oil futures markets in which speculators allow oil producers to
hedge. The model highlights the role of the capital constraints faced by speculators for the determination
of futures and spot prices. The key difference is that this model focuses on oil producers’ inventories
rather than refiners’ inventories, as in conventional models.
Acharya et al. (2010) show that tighter capital restrictions faced by commodity investment funds
or increased hedging pressures increase the price of oil futures and raise the cost of hedging for producers
who are naturally short. This induces producers to scale back their above-ground oil inventory holdings,
releasing the oil to the market and thus putting downward pressure on the spot price. They also show that
an increase in hedging pressures and/or a decline in the speculators’ ability to bear risk affect the risk
premium in oil futures markets and helps predict changes in the spot price. Specifically, increases in risk
aversion (proxied by the default risk facing producers) predict an increase in the excess return on shortterm
futures. Moreover, the fraction of the futures risk premium attributable to default risk is higher in
periods when broker-dealer balance sheets are contracting. One reason to be concerned with their model
23 
 
is that the model has direct implications for regulators. In particular, efforts intended to increase the cost
of hedging may reduce the risk-bearing capacity of speculators and hence may raise the spot price of oil.
A key implication of Acharya et al. (2010) is that a decrease in the futures risk premium causes
an increase in the spot price of oil, as producers are willing to hold higher inventory levels when the cost
of hedging is low. In closely related work, Etula (2010) shows that the effective risk aversion of brokers
who are highly leveraged and who tend to play an important role in matching parties in over-the-counter
markets is an important determinant of risk premia in commodities markets. Etula (2010) derives an asset
pricing model in which the risk premium depends both on systematic and non-marketable risk with the
latter increasing in the economy’s effective risk aversion. Etula (2010) shows that risk aversion varies
over time with brokers’ risk bearing capacity. Using aggregate balance sheet data for broker-dealers to
construct an empirical proxy of risk aversion, he finds that changes in risk aversion predict expected
futures returns for crude oil. However, the indicator of effective risk aversion fails to explain oil price
fluctuations during 2008-2009. Etula attributes the break in the relationship to the inadequacy of the
constructed measure of risk aversion to reflect the broader appetite for risk as it does not capture the
various types of financial investors such as hedge funds, index investors and pension funds. He also
shows that changes in risk aversion help predict one-quarter ahead excess returns in crude oil futures out
of sample, although the extent of the reductions in the mean-squared prediction error cannot be inferred
from his results. 
These results on time-varying risk aversion are encouraging, but hardly dispositive. For example
it is difficult to tell from the analysis in Acharya et al. (2010) which of the many regressors is responsible
for how much improvement in predictive accuracy and it is unclear how the out-of-sample forecast
accuracy of these models compares to that of alternative forecasting models based on economic
fundamentals (see, e.g., Alquist et al. 2012; Baumeister and Kilian 2012). Moreover, Acharya et al.
(2010) readily acknowledge that their alternative theory is unlikely to explain the full magnitude of the
rise and fall in oil prices witnessed recently. Indeed, it is not a theory of global oil markets, but of U.S. oil
producers and the available firm data are not rich enough to infer the aggregate hedging position even for
24 
 
U.S. oil producers, not to mention global oil producers. Hence, Acharya et al. view this mechanism as a
likely contributing factor to recent price movements at best. In contrast, existing empirical oil market
models are capable of explaining much of the observed surge in the real price of oil based on unexpected
changes in global real activity without relying on changes in the risk premium (see, e.g., Kilian 2009a,b).
The logical next step would be an evaluation of the empirical content of models of time-varying
risk premia within the context of a more general model that embeds the notion of time-varying risk
premia along with other explanations of oil price fluctuations. It is not clear, however, how to construct
such a model or how to compute global measures of risk-bearing capacity or, for that matter, how to
obtain disaggregate global crude oil inventory holdings for oil producers.
The idea of time-varying risk premia also plays a central role in recent work by Hamilton and Wu
(2011) who provide evidence of large shifts in the risk premium after 2005. Figure 4 shows that,
following a long period of stability, the risk premium on the 8-week oil futures contract became highly
volatile after 2005. An initial surge in the risk premium to an unusually high positive level in 2005/06 is
followed by subsequent equally negative spikes in 2007/08 and 2010. Hamilton and Wu associate these
shifts with index funds and other financial investors having become natural counterparties for commercial
hedgers. They provide a formal theoretical model and improved econometric techniques for the
estimation of time-varying risk premia. The implications of this evidence for the evolution of the spot
price of oil are not clear.
Changes in estimated risk premia are not the only sign of possible structural change in oil futures
markets in recent years. In related work, Büyükşahin et al. (2009) note large changes in the term structure
of oil futures after about 2002 (including a lengthening of the term structure), and Fattouh and
Scaramozzino (2011) provide evidence that the front end and the back end of the term structure of oil
futures has shifted, with the back-end price adjusting to the front-end price. Such a shift could be
explained by a model of Bayesian updating in which the probability distribution about the future price of
oil shifts. Whether this shift is caused by financial investors or more fundamental shocks is a different
question.
25 
 
All this evidence is suggestive of the interpretation that the arrival of financial investors may have
changed the behavior of oil futures markets. There is no consensus, however, that this increased
financialization has been detrimental. For example, Büyükşahin et al. (2009) emphasize that increased
efficiencies in derivatives pricing have caused futures prices at different maturities to become more
closely linked. In related work, Boyd et al. (2010) find evidence of herding in oil futures markets,
resulting in common trading strategies, but conclude that this herding actually stabilized the price.
One of the least understood aspects of financialization is how the entry of new classes of
investors affects the formation of market expectations. Singleton (2012), for example, building on Allen,
Morris, and Shin (2006), stresses that participants in oil futures markets may form expectations not only
in terms of expected fundamentals, but also in anticipation of other market participants’ actions. This
possibility is not likely to be a concern for structural models such Kilian and Murphy (2011), which leave
the formation of expectations unspecified, but it affects the construction of theoretical models of the
formation of oil prices. Likewise, there is a concern that some market participants may overreact to public
information or signals that do not reflect large changes in underlying fundamentals. This concern seems
particularly pertinent when new market participants lack the expertise to understand oil market
developments. Another possibility is that market participants coordinate on a small number of easily
observable signals such as data on oil inventories, the dollar exchange rate, shortages of supply, or
concerns over peak oil. Many of these concerns arise in similar form in other financial markets. In this
sense, we may say that oil futures markets have matured and have become an integrated part of the
financial market system.
9. CONCLUSION
The key question addressed in this survey is whether there is evidence that speculation has been driving
the recent fluctuations in the spot price of oil. The academic literature allows several conclusions. First,
there is clear evidence of the increased financialization of oil futures markets. Whether this
financialization also was responsible for increased co-movement among different asset prices continues to
26 
 
be debated. Although there is some evidence of increased co-movement across asset classes, that cocomovement
is also found in markets in which index funds do not operate and for which there are no
futures exchanges, which is suggestive of an explanation based on common economic fundamentals.
Second, there is no compelling evidence that changes in financial traders’ positions predict
changes in the price of oil futures. Conflicting results in the literature in this regard can be traced to the
use of data sets in some studies that are too aggregated to be informative about these predictive
relationships. To the extent that any evidence of predictive power from index fund holdings to oil futures
prices has been found, that evidence has not been based on rigorous real-time analysis and the extent of
the out-of-sample gains has yet to be quantified. Finally, it cannot be ruled out that this predictive power,
if any, arises from traders’ positions responding to the underlying fundamentals of the oil market.
Third, contrary to widely held beliefs that increases in oil futures prices precede increases in the
spot price of oil, there is no evidence that oil futures prices significantly improve the out-of-sample
accuracy of forecasts of the spot price of oil. This result holds whether one is forecasting the nominal
price or the real price of oil. There is evidence that models based on economic fundamentals help forecast
the spot price of oil out of sample, however.
Fourth, we observed that the simple static model that has been used to explain how an influx of
financial investors may cause an increase in the spot price of oil is inconsistent with dynamic models of
storage. As a result, the oil price-inventory relationship tells us nothing about the quantitative importance
of speculation in oil markets. In particular, the absence or presence of speculative pressures in the oil
market cannot be inferred from studying oil inventory data without a fully specified structural model.
Fifth, structural VAR models that nest alternative explanations of the evolution of the real price
of oil (including speculative demand) provide strong evidence of speculation in 1979, 1986, 1990 and late
2002, but are not supportive of speculation being an important determinant of the real price of oil from
2003 to mid-2008. Instead these models imply that both spot and futures prices were mainly driven by a
common component reflecting economic fundamentals. Alternative structural VAR studies that claim to
have found evidence of financial speculation suffer from identification problems and are uninformative.
27 
 
Nor is there empirical evidence that the short-run price elasticity of gasoline demand is literally zero, as
required for the theoretical model discussed in Hamilton (2009a) to explain increases in the spot price
based on speculation in oil futures markets.
Sixth, recently developed theoretical and empirical models of time-varying risk premia may help
enhance our understanding of fluctuations in oil prices, but it is not clear how representative these models
are for the global market for crude oil, and their ability to explain fluctuations in the price of oil has yet to
be explored in full detail.
Finally, the absence of evidence in favor of speculation does not mean that the financialization of
oil futures markets does not matter. There is evidence of structural instability in the term structure of oil
futures and in the risk premium, for example, that could be related to the financialization of these markets.
We discussed existing research as well as some avenues for future research that may help us understand
these facts. We highlighted in particular the role of heterogeneous expectations and of the limits to
arbitrage in these markets.
One of the problems in this literature and, more importantly, in the public debate about
speculation is that it is rarely clear how speculation is defined and why it is considered harmful to the
economy. For example, the aim of recent regulatory changes in oil futures markets is to reduce price
volatility, when increased oil price volatility was never the problem, but the persistent increases in the
price of oil after 2003. Moreover, the literature has shown that the presence of index funds has, if
anything, been associated with reduced price volatility.10
Nor is there sufficient attention to the possibility
that some forms of speculation may be more harmful than others. One strategy for evaluating the
consequences of speculation would be to quantify the social welfare costs of speculation. This approach
will require the developments of more fully articulated economic models than the current generation of
theoretical and empirical models.
                                                            
10
 This view is also supported by historical analyses on the relationship between futures markets and price volatility
(see, e.g., Jacks 2007). On a related note, Büyükşahin (2011) observed that the unusually high volatility in post-2007
commodity markets is not limited to exchange traded commodities. 
28 
 
REFERENCES
Acharya, Viral V., Lars A. Lochstoer, and Tarun Ramadorai. 2010. “Limits to Arbitrage and Hedging:
Evidence from Commodity Markets.” Working paper. London Business School.
Akram, Q. Farooq. 2009. “Prices, Interest Commodity Rates, and the Dollar.” Energy Economics 31:
838-851.
Allen, Franklin, Stephen Morris, and Hyun Song Shin. 2006. “Beauty Contests and Iterated Expectations
in Asset Markets,” Review of Financial Studies 19: 719-752.
Alquist, Ron, and Olivier Gervais. 2011. “The Role of Financial Speculation in Driving the Price of
Crude Oil.” Working paper. Bank of Canada.
Alquist, Ron, and Lutz Kilian. 2007. “What Do We Learn from the Price of Crude Oil Futures?” CEPR
Working Paper No. 6548.
Alquist, Ron, and Lutz Kilian. 2010. “What Do We Learn from the Price of Crude Oil Futures?” Journal
of Applied Econometrics 25: 539-573.
Alquist, Ron, Lutz Kilian, and Robert J. Vigfusson. 2012. “Forecasting the Price of Oil.” Forthcoming:
G. Elliott and A. Timmermann, eds., Handbook of Economic Forecasting 2. Amsterdam: North-
Holland.
Baumeister, Christiane, and Lutz Kilian. 2012. “Real-Time Forecasts of the Real Price of Oil.”
Forthcoming: Journal of Business and Economic Statistics.
Baumeister, Christiane, and Gert Peersman. 2012. “The Role of Time-Varying Price Elasticities in
Accounting for Volatility Changes in the Crude Oil Market.” Forthcoming: Journal of Applied
Econometrics.
Bodenstein, Martin, and Luca Guerrieri. 2011. “Oil Efficiency, Demand, and Prices: A Tale of Ups and
Downs.” Working paper. Federal Reserve Board.
Boyd, Naomi E., Bahattin Büyükşahin, Jeffrey H. Harris, and Michael S. Haigh. 2010. “The Prevalence,
Sources, and Effects of Herding.” Working paper. Commodity Futures Trading Commission.
Brunetti, Celso, and Bahattin Büyükşahin. 2009. “Is Speculation Destabilizing?” Working paper,
29 
 
Commodity Futures Trading Commission.
Brunetti, Celso, Bahattin Büyükşahin, and Jeffrey H. Harris. 2011. “Speculators, Prices, and Market
Volatility.” Working paper. Johns Hopkins University.
Büyükşahin, Bahattin. 2011. “Volatility: Not Unique to Exchange-Traded Commodities.” IEA Oil Market
Report. March.
Büyükşahin, Bahattin, Michael S. Haigh, Jeffrey H. Harris, James A. Overdahl, and Michel A. Robe.
2009. “Fundamentals, Trader Activity, and Derivative Pricing.” Working paper. Commodity
Futures Trading Commission.
Büyükşahin, Bahattin, Michael S. Haigh, and Michel A. Robe. 2010. “Commodities and Equities: Ever
a ‘Market of One’?” Journal of Alternative Investments 12: 76-95.
Büyükşahin, Bahattin, and Jeffrey H. Harris. 2011. “Do Speculators Drive Crude Oil Futures?” Energy
Journal 32: 167-202.
Büyükşahin, Bahattin, and Michel A. Robe. 2010. “Speculators, Commodities, and Cross-Market
Linkages.” Working paper. Commodity Futures Trading Commission.
Büyükşahin, Bahattin, and Michel A. Robe. 2011. “Does ‘Paper Oil’ Matter? Energy Markets’
Financialization and Equity-Commodity Co-Movements.” Working paper. The American
University.
Chang, Eric C. 1985. “Returns to Speculators and the Theory of Normal Backwardation?” Journal of
Finance 40: 193-208.
Davis, Lucas W., and Lutz Kilian. 2011. “Estimating the Effect of a Gasoline Tax on Carbon
Emissions.” Journal of Applied Econometrics 16: 1187-1214.
De Roon, Frans A., Theo E. Nijman, and Chris Veld. 2000. “Hedging Pressure Effects in Futures
Markets?” Journal of Finance 55: 1437-1456.
Dvir, Eyal, and Kenneth S. Rogoff. 2010. “Three Epochs of Oil.” Working paper. Harvard University.
Etula, Erkko. 2010. “Broker-Dealer Risk Appetite and Commodity Returns.” Working paper, Federal
Reserve Bank of New York.
30 
 
Fattouh, Bassam. 2010a. “Oil Market Dynamics through the Lens of the 2002-2009 Price Cycle.” OIES
Working Paper 39, Oxford Institute for Energy Studies.
Fattouh, Bassam. 2010b. “The Dynamics of Crude Oil Differentials.” Energy Economics 32: 334-342.
Fattouh, Bassam, and Pasquale Scaramozzino. 2011. “Uncertainty, Expectations, and Fundamentals:
Whatever Happened to Long-Term Oil Prices?” Oxford Review of Economic Policy 27: 186-206.
Fry, Renee, and Adrian Pagan. 2011. “Sign Restrictions in Structural Vector Autoregressions: A Critical
Review.” Journal of Economic Literature, 49: 938-960.
Giannone, Domenico, and Lucrezia Reichlin. 2006. “Does Information Help Recover Structural Shocks
from Past Observations?” Journal of the European Economic Association 4: 455-465.
Hamilton, James D. 2009a. “Causes and Consequences of the Oil Shock of 2007-08.” Brookings Papers
on Economic Activity Spring 1: 215-259.
Hamilton, James D. 2009b. “Understanding Crude Oil Prices” Energy Journal, 30: 179-206.
Hamilton, James D., and J. Cynthia Wu. 2011. “Risk Premia in Crude Oil Futures Prices.” Working
paper. University of California at San Diego.
Hamilton, James D., and J. Cynthia Wu. 2012. “Effects of Index-Fund Investing on Commodity Futures
Prices.” Working paper. University of California at San Diego.
Hirshleifer, David. 1988. “Residual Risk, Trading Costs, and Commodity Risk Premia?” Review of
Financial Studies 1: 173-193.
Hirshleifer, David. 1989. “Determinants of Hedging and Risk Premia in Commodity Futures Markets?”
Journal of Financial and Quantitative Analysis 24: 313-331.
Hong, Harrison G., and Motohiro Yogo. 2012. “What Does Futures Market Interest Tell Us about the
Macroeconomy and Asset Prices?” Forthcoming: Journal of Financial Economics.
Hughes, Jonathan E., Christopher R. Knittel, and Daniel Sperling. 2008. “Evidence of a Shift in the
Short-Run Price Elasticity of Gasoline.” Energy Journal 29: 113-134.
Irwin, Scott H., and Dwight R. Sanders. 2012. “Testing the Masters Hypothesis in Commodity Futures
Markets.” Energy Economics 34: 256-269.
31 
 
Inoue, Atsushi, and Lutz Kilian. 2011. “Inference on Impulse Response Functions in Structural VAR
Models.” Working paper. University of Michigan.
Irwin, Scott H., Dwight R. Sanders, and Robert P. Merrin. 2009. “Devil or Angel? The Role of
Speculation in the Recent Commodity Price Boom (and Bust).” Journal of Agricultural and
Applied Economics 41: 377-391.
Jacks, David S. 2007. “Populists versus Theorists: Futures Markets and the Volatility of Prices.”
Explorations in Economic History 44: 342-362.
Juvenal, Luciana, and Ivan Petrella. 2011. “Speculation in the Oil Market.” Working paper. Federal
Reserve Bank of St. Louis.
Kemp, J. 2011. “Stanford Professor Cracks Orthodoxy on Oil Prices.” Reuters, June 20
(http://in.reuters.com/article/2011/06/20/idINIndia-57804520110620).
Kilian, Lutz. 2009a. “Comment on ‘Causes and Consequences of the Oil Shock of 2007-08’ by James D.
Hamilton.” Brookings Papers on Economic Activity 1, Spring 2009: 267-278.
Kilian, Lutz. 2009b. “Not All Oil Price Shocks Are Alike: Disentangling Demand and Supply Shocks in
the Crude Oil Market.” American Economic Review 99: 1053-1069.
Kilian, Lutz, and Bruce Hicks. 2012. “Did Unexpectedly Strong Economic Growth Cause the Oil
Price Shock of 2003-2008?” Forthcoming: Journal of Forecasting.
Kilian, Lutz, and Daniel P. Murphy. 2011. “The Role of Inventories and Speculative Trading in the
Global Market for Crude Oil.” Working paper. University of Michigan.
Kilian, Lutz, and Daniel P. Murphy. 2012. “Why Agnostic Sign Restrictions Are Not Enough:
Understanding the Dynamics of Oil Market VAR Models.” Forthcoming: Journal of the
European Economic Association.
Lombardi, Marco J., and Ine Van Robays. 2011. “Do Financial Investors Destabilize the Oil Price?”
Working paper. European Central Bank.
Masters, Michael W. 2008. Testimony before the U.S. Senate Committee on Homeland Security and
Governmental Affairs, May 20.
32 
 
Masters, Michael W. 2010. Testimony before the Commodity Futures Trading Commission, March 25.
Mou, Yiqun. 2010. “Limits to Arbitrage and Commodity Index Investment: Front-Running the
Goldman Roll.” Working paper. Columbia University.
Peck, Anne E. 1985. “Economic Role of Traditional Commodity Futures Markets.” In A.E. Peck, ed.,
Futures Markets: Their Economic Role. Washington, DC: American Enterprise Institute for
Public Policy Research, pp. 1-81.
Pirrong, S. Craig. 1994. “Squeezes, Corpses, and the Anti-Manipulation Provisions of the Commodity
Exchange Regulation Act.” Regulation 17: 52-63.
Pirrong, S. Craig. 2008. “Stochastic Fundamental Volatility, Speculation, and Commodity Storage.”
Working paper. University of Houston.
Pirrong, S. Craig. 2011. Commodity Price Dynamics: A Structural Approach. Cambridge: Cambridge
University Press.
Ripple, Ronald. 2008. “Futures Trading: What is Excessive?” Oil and Gas Journal, June 9.
Sanders, Dwight R., and Scott H. Irwin. 2011. “The Impact of Index Funds in Commodity Futures
Markets: A Systems Approach.” Journal of Alternative Investments 14: 40-49.
Serletis, Apostolos, Govinda Timilsina, and Olexandr Vasetsky. 2010. “International Evidence on
Sectoral Interfuel Substitution.” Energy Journal 31: 1-29.
Stoll, Hans R., and Robert E. Whaley. 2010. “Commodity Index Investing and Commodity Futures
Prices.” Journal of Applied Finance 20: 7-46.
Stoll, Hans R., and Robert E. Whaley. 2011. “Commodity Index Investing: Speculation or
Diversification.” Journal of Alternative Investments 14: 50-62.
Silvennoinen, Annastiina, and Susan Thorp. 2010. “Financialization, Crisis, and Commodity
Correlation Dynamics.” Working paper. University of Technology, Sydney.
Singleton, Kenneth J. 2012. “Investor Flows and the 2008 Boom/Bust in Oil Prices.” Working paper.
Stanford University.
Tang, Ke, and Wei Xiong. 2011. “Index Investment and Financialization of Commodities.” Working
33 
 
paper. Princeton University.
Working, Holbrook. 1942. “Quotations on Commodity Futures as Price Forecasts.” Econometrica 16:
39-52.
Working, Holbrook. 1960. “Speculation on Hedging Markets.” Stanford University Food Research
Institute Studies 1: 185-220.
Yergin, Daniel. 1992. The Prize. The Epic Quest for Oil, Money, and Power. New York: Simon and
Schuster.
34 
 
0
50
100
150
200
250
300
Crude Oil Non-Exchange-Traded Commodity Index
Figure 1. Price of Non-Exchange Traded Commodities and Price of Crude Oil
Source: Computations of the authors based on data in the IEA Oil Market Report, March 15, 2011.
Figure 2. Comparison of WTI Crude Oil Net Long Positions based on Index Investment Data and
Masters’ Estimates
Source: Irwin and Sanders (2012). Reproduced with permission of the authors.
35 
 
1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
Figure 3. The Crude Oil Inventory-Oil Price Relationship
0
20
40
60
80
100
120
140
160
255 270 285 300 315 330 345 360
US Commercial Crude Stocks vs. WTI
1994 - 2004 2005-2007 2008 2009-2010 2011
(US $/b)
(million barrels)
Source: Computations of the authors based on EIA data.
Figure 4. 8-Week Risk Premium Estimated from Hamilton and Wu (2011) Model with
Sample Split in 2005
Source: Hamilton and Wu (2011).