Content uploaded by Susan Newman
Author content
All content in this area was uploaded by Susan Newman on Dec 21, 2020
Content may be subject to copyright.
Working Paper No 2009/7
M
ARCH
2009
The New Price Makers: An investigation into the impact of financial
investment on coffee price behaviour
Susan Newman
*
A
BSTRACT
The collapse of the international commodity agreements in the 1980s, together with the liberalisation of marketing systems in commodity exporting developing
countries under structural adjustment programs have removed the role of the state in the pricing of commodities. The logic of liberalisation was to remove
impediments to the efficient market discovery of commodities prices w here state intervention was seen as the main culprit. Mainstream economics has also
promoted the role of derivatives exchanges in efficient price discovery on physical markets in international commodity markets with globally fragmented
production. By rapidly exploiting opportunities to profit from price deviations between futures and spot markets, rational and informed speculators are seen to
drive any price deviations in the market quickly to the equilibrium price that reflect international supply and demand realities.
Such a view, however, fails to consider the role of differentiated market participants on international exchanges in driving commodity prices, by assuming that all
trading activities are efficiency enhancing irrespective of the motives behind them. There has been much media interest in the role of speculation on commodity
prices, in particular food commodities, over the recent months. There has, however, been very little systematic evidence gathered on the extent to which
speculative activities affect prices of exchange traded commodities.
Using the example of international prices for coffee, this paper shows that increases in trading activities by non-physical market actors for the purposes of
financial investment have led to a dislocation between prices on international exchanges and supply and demand realities. With increasing participation by various
institutional investors such as hedge and pension funds, prices have become increasingly driven by trading that arises owing to changes in the financial
investment environment more generally, such as the dot-com crash of the late 1990s and the recent credit crunch.
K
EY
W
ORDS
Commodity prices, futures markets
* Susan Newman is a Doctoral student of the NCCR individual project on “Primary Commodities” and a PhD candidate in the Economics Department of the School
of Oriental and African Studies, University of London.
NCCR
TRADE
W
ORKING
P
APERS
are preliminary docum ents posted on the NCCR Trade Regulation website (<www.nccr-trade.org>) and w idely circulated to stimulate
discussion and critical comm ent. These papers have not been formally edited. Citations should refer to a “NCCR Trade Working Paper”, with appropriate
reference made to the author(s).
1. Introduction
This paper explores the extent to which financial investment on futures markets affect price
outcomes such that futures prices deviate from that which is warranted by the supply and demand
conditions in the market for physical commodities.
With the collapse of various multilateral price and income stabilisation schemes, as well as the
dismantling of public marketing systems in commodity producing developing countries, various
private price risk management (PRM) strategies became the only option for market actors engaged in
international commodity systems. Of the available PRM- strategies, futures contracts have been
promoted as the key hedging instrument.
1
Theoretical justifications for the wholesale use of future
contracts rely upon the assertions that: first - given their widespread availability –futures markets
provided effective hedging instruments for cash market actors; second, futures markets are efficient
in the sense that they incorporate all knowable market information in price outcomes and therefore
enable efficient price discovery (Fama 1970); and third, by facilitating speculation, futures markets are
beneficent in that they bring about a better allocation of supplies over time and reduces fluctuations
in price (Goss and Yamey 1978).To the extent that these assertions hold, uninhibited futures markets
therefore have two important functions in income stabilisation for individuals engaged in the trading
of physical commodities, first by providing effective hedging instruments and second by stabilising
prices (McKinnon 1967).
Recent economic literature on futures and/or commodities trading, while emphasising the
importance of futures markets for price and income stabilisation, have not examined the functioning
of these markets in terms of the above assertions together. The financial literature on futures trading
has been preoccupied with applying empirical tests for the efficient market hypothesis (Fama 1998).
On the other hand, recent literature on the behaviour of cash prices of commodities have examined
the relationship between futures and cash prices solely in terms of the co-movement of the two price
series, or the enhancement to the accuracy of cash price forecasts with the incorporation of futures
price information (refs.) with the implicit assumption that efficient price discovery takes place on
futures markets. Neither strands of literature engages in dialogue with the other and the relationship
between futures and cash markets tend to be assumed rather than studied.
This paper argues that once we take into account the nature of trading activities of various market
participants, distinguished according to the motivations behind their engagement in futures markets,
the classical assertions that imply efficient price discovery and price stabilization on futures markets
do not necessarily hold. Commodity futures have become increasingly attractive to financial investors
1
Other private PRM-strategies include the use of forward contracts and income diversification.
who are looking to diversify their portfolios, particularly in times of high inflation and falling equity
prices associated with economic downturns (Gorton and Rouwenhorst 2004). The last 20 years have
seen a large number of new institutional investors entering commodity markets, including hedge and
pension funds (Domanski and Heath 2007).
Section 2 begins by outlining the classical theory of speculation and the implied relationship
between futures and physical markets for commodities, it continues with a critique of the classical
theory from the perspectives of Post-Keynesian economics as well as the theory of performative
economics (McKenzie, date). Section 3 presents two quantitative studies into the extent of financial
investment on the coffee exchange of the New York Board of Trade (NYBOT) and its impact on, firstly
the relationship between prices and supply and demand, and second the impact of financial activities
on price volatility. Evidence is found for a loosening in the relationship between annual prices and
changes in supply and demand conditions associated with a once and for all increase in financial
investment on commodity markets at the time of the dot-com crash in 2000. In addition, we find that,
in some cases, financial activities on coffee markets can lead to increased price volatility.
2. Re-examining the relationship between futures and physicals markets for
commodities
Futures contracts were developed in the context of trade in agricultural commodities as an
exchange traded risk management instrument for physical market traders. As an agreement between
two parties to buy or sell an asset (in this case some quantity of the commodity) at a certain time in
the future at a certain price. This allows the buyer and seller to lock in a price and thus hedge against
risks associated with adverse movements in price that might take place between harvest and delivery.
The trading of futures contracts can be traced back to Japan in the 18
th
Century for rice and silk,
and the Netherlands for tulip bulbs. The modern form of commodities exchanges appeared in the
American Midwest with the establishment of the Chicago Board of Trade (CBOT) in 1848. The
commodities exchanges brought together farmers and merchants. Initially, the main task of the
exchange was to standardise the quantities and qualities of grains that were traded, but within a few
years, futures like contracts were being traded on the exchange. (Hull 2003) Soon after the
introduction of futures contracts, speculators entered the market, trading in futures contracts in order
to profit from price changes without the need to hold the physical commodity.
Futures and many other derivative instruments have subsequently been developed for a wide
range of agricultural and non-agricultural commodities, and financial assets. These are traded on
derivatives exchanges across the world.
2
1.1. 2.1 Stabilising Speculation
Modern textbooks on futures trading tell us that there are three distinct types of actors in futures
market, namely hedgers, speculators and arbitragers (Geman 2006). Hedgers are those that are also
trading in the physical commodity and engage in futures trading in order to cover themselves against
price changes prior to the delivery of the stored, or yet to be produced/harvested, commodity
3
.
Speculators take a position in the market, exposing them to the risk of adverse price movements in
order to gain rewards. Speculators, although not necessarily operating outside of physical markets,
are often assumed to have no desire to trade in the physical commodity rather they are financial
investors that view commodities as an alternative asset class from equities and currencies. The
Arbitrager, unlike the speculator, is said to enter into several transactions in two or more markets in
order to gain a riskless profit. In exploiting arbitrage opportunities, arbitragers quickly eliminate such
price discrepancies across different markets.
4
In the text book case, hedgers do not speculate and
speculators to not hedge.
The classical view of speculation sees speculators as playing an important market making role.
By providing the market with liquidity, speculation is seen to be beneficent in that it both enhances
the allocative efficiency of the economy and reduces fluctuations in price.
5
Marshall reinforces the
2
Currently, the major futures exchanges are the CBOT, the Chicago Mercantile Exchange, the New York Mercantile Exchange,
the New York Board of Trade, the International Petroleum Exchange, the London Metal Exchange and Le Marché à Terme
International de France.
3
In the language of Holbroke Working (1962) hedgers engage in routine hedging and have no expectations how prices will
behave in the future.
4
Even in the mainstream financial literature it has been debated whether the pure arbitrager exists in reality. This together with
the result from the EMH of no arbitrage opportunities, models of derivatives pricing are generally based on the
assumption of no-arbitrage opportunities and thus no arbitragers (Geman, 2006).
5
[Speculative activities] “which enable a deficiency in one place to be supplied from the surplus of another, at a moderate or
even small advance on the ordinary price, render the fluctuations of prices much less extreme than formerly. This effect is
much promoted by the existence of large capitals, belonging to what are called speculative merchants whose business it is
to buy goods in order to resell them at a profit. These dealers naturally buying things when they are cheapest, and storing
them up to be brought again into the market when the price has become unusually high; the tendency of their operations
is to equalize price, or at lease to moderate its inequalities. The prices of things are neither so much depressed at one time,
nor so much raised at another, as they would be if speculative dealers did not exist.
Speculators, therefore, have a highly useful office in the economy of society; and (contrary to common opinion) the most useful
portion of the class are those who speculate in commodities affected by the vicissitudes of seasons. If there were no corn-
dealers, not only would the price of corn be liable to variations much more extreme than at present, but in a deficient
season the necessary supplied might not be forthcoming at all. Unless there were speculators in corn, or unless, in default
dealers, the farmers became speculators, the price in a season of abundance would fall without any limit check, except the
wasteful consumption that would invariably follow. That any part of the surplus in one year remains to supply a
classical view by posing the theory of constructive speculation whereby, the actions of speculators in
aggregate will be tend to stabilise prices even if individual speculators trade in an inefficient
6
way
because it is in the interests of speculators to gather market information and use it in the best way
possible since the wrong trading decisions lead to losses. Inefficient speculators, those that tend to
make the wrong forecasts, lose money and are thus eliminated from the market (Marshall 1932).
The neoclassical approach focuses on the Knightian definition of risk
7
. In this sense, every
decision is risky, and economic actors can assign probabilities to all economic decisions to the limit of
available information. While on average knowable, economics actors are operating in a stochastic
world and thus all decisions to buy and sell are in this sense speculative, but rational (Samuelson
1952). Mainstream theory of derivatives markets would imply that the introduction of futures
markets would reduce cash price volatility owing to the price discovery properties of liquid futures
markets. Movements in prices on futures markets are seen to embody the rapid and unbiased
incorporation of all available information in the supply and demand on cash markets. By rapidly
exploiting arbitrage opportunities that appear on the market, rational and informed speculators are
seen to drive any price deviations in the market quickly to the equilibrium price. If markets are
efficient, futures prices are unbiased predictions of subsequent spot prices. (Mandelbrot
1966;Samuelson 1965) Futures markets can thus perform their second role in terms of cash price
stabilisation and risk reduction through hedging. By enhancing the price discovery mechanism, all
trading activities, regardless of the motivation of the trader (be it to derive income from price
movements, or to hedge against the price risk associated with holding the physical commodity) are
expected to be stabilising. (Working 1962) Speculative activities can act to drive price volatility and
amplify price swings on commodity exchanges, only where rational market actors are performing in
the context of distorted or incomplete information (Grossman and Stiglitz 1980;Stein 1987).
As a consequence of the classical view of stabilising speculation, much of the empirical work
conducted around the operation of futures markets has been in the context of testing the Efficient
Market Hypothesis (EMH). The general definition of an efficient market is one in which price always
“fully reflect” available information (Fama 1970, p383). The liquidity provided by speculators on
futures markets, according to classical theories of speculation greatly enhances the efficiency of these
deficiency of another is owing to farmers who withhold corn from the market, or to dealers who buy it when at the
cheapest and lay it up in store. “ (Mill 1848, bk 4 ch.2 section 4)
6
Such that they trade at prices that deviate wildly from those warranted by market conditions
7
For Knight (1933), risk referred to situations where outcomes are randomly generated by
stochastic processes that are on average knowable. In this way, probability distributions can be
assigned to risky situations. Uncertainty, on the other hand, cannot be anticipated in terms of
mathematical probabilities. Subsequent schools of thought in economics have tended to align
themselves through emphasising one, or the other conceptions of risk and uncertainty.
markets. Rather than a theory proper, the EMH consists of a number of testable hypotheses (Fama
1970;1998). What is implicit is a theory of price determination
8
as well as the assumption that
derivatives markets function to complete markets in the Arrow-Debreu sense (Arrow and Debreu
1954). In this way, the relationship between futures and physical markets are assumed away as if
there is a seamless interface between the two.
2.2 Destabilising speculation
While speculative activities are not conditional on the existence of futures markets
9
, their
development has facilitated an enormous growth in speculative activities on commodity markets by
making it possible for speculation without the speculator having to deal in the physical commodity. It
may be expected, owing to the ease of futures trading, that waves of speculative activity can influence
the price levels and price movements of commodities. Whether or nor speculative trading on futures
markets can influence price such that they deviate from market fundamentals (supply and demand)
for any length of time has been the subject of much debate (refs.).
One of the earliest theoretical challenges against the classical theory of speculation came from
Irwin (1937), who argued that, contrary to the classical proposition, speculators can in fact gain from
trading activities that drive prices away from that which is warranted by physical markets conditions.
With the simplest example of the movement or noise trader, profits need not be made by making the
correct predictions of the market as determined by market fundamentals but “the inertia or
momentum in the price movements which arises mainly from the large number of traders operating
in these markets” (Irwin 1937, p.270 as quoted in Goss and Yamey 1978, p.35). Speculators can,
therefore, profit by riding out price trends, the skill being in getting off before the market turns, rather
than the correct price predictions made on the basis of physical market conditions.
8
Mainstream commodity market theory refers to the “theory of intertemporal price formation on
speculative markets” Ghosh, S, C.L Gilbert, and A.J Hughes Hallet. 1987. Stabilizing Speculative
Commodity Markets: Oxford University Press..
8
The theory focuses on the intertemporal relationships
between quantity, and consequently price. As well as the relationships between market demand (for
consumption) and supply, stock may be demanded for transactions; precautionary or speculative
purposes. Taken together, these relationships define an intertemporal stock equilibrium. Movements
in stockholding imply movements in prices. Prices are however, ultimately determined through the
mechanisms of supply and demand. Market structure is understood as the relationship between the
forces of supply and demand. On these basic tenets, econometric models of commodity markets are
generally constructed to consist of a “demand block”, a “supply block”, together with an equation
relating stocks and prices over time (Labys 1978; Akiyama & Duncan 1982; Adams & Klein 1982)
9
Outside of futures markets, speculation in commodities can take the form of stockholding or forward sales in the face of
future price uncertainty
Post Keynesians have criticised the focus of neoclassical economics on risk and argue that
uncertainty is most relevant when considering economic decision making. (Davidson 1991;Shackle
1979) In the context of a fundamentally uncertain world, Keynes argued that, “the market will be
subject to waves of optimistic and pessimistic sentiment, which are unreasoning and yet in a sense
legitimate” (Keynes 1997, p154). In this way, speculative activities have potentially destabilising
affects on the economy.
As in the case of different types of hedging, there is no reason to assume that all speculators
behave in the same way on futures markets. It has also been argued that particular techniques
employed by the speculator in his trading activities can have destabilising affects. As early as 1937,
Irwin argued that “much of the trading activity of professional speculators on commodity and stock
exchanges did not correspond with the activities subsumed in the classical theory; and that this type
of trading, called movement trading, was destabilizing”(Goss and Yamey 1978, p.34). More recently,
MacKenzie (2005) takes this argument further by stressing that derivatives markets themselves are
social constructs, and as such will embody, and display the vested interest of their participants, who
are far from uniform in terms of their motivations, market power and behaviour. Moreover, the
widespread adoption of particular technical apparatus in the formation of price expectations that
inform trading decisions can in turn determine market outcomes. In illustrating the performativity of
economics in the context of options pricing, MacKenzie shows how socio-historical factors led to the
adoption of Black’s Sheets by options traders, that embodied the predictions of the Black-Scholes-
Merton (BSM) theory of options pricing, resulted in price behaviour on options markets that became
more closely related to the predictions of the BSM model (Black and Scholes 1972;Merton 1973)
1.2. 2.3 Implications of destabilising speculation
The brief discussion above on the potential destabilising affects of speculative trading suggests
that empirical research on the implications of financial investment on international commodity
exchanges on price behaviour should not be limited to whether or not the EMH holds in the context
of futures markets. First, the discussion suggests a need to examine more closely the relationship
between futures prices and, so called fundamental market conditions rather than assume that the
relationship is given by the co-movement of prices on futures and physicals markets. If futures
markets indeed facilitate price discovery, then these efficient prices should reflect changes in the
supply and demand of the physical commodity as implied by a theory of price determination. Second,
given that speculators can profit from different types of trading, there is no reason to assume that all
speculators behave in the same way given that the motivations behind their engagement on futures
markets can differ considerably. There may be periods where the dominant form of speculative
activities act to drive prices away from that warranted by supply and demand conditions over long-
periods of time (such as in the case of the long bull run in coffee prices that has occurred from 2002
until 2007). There may also occur, situations were a large and sudden influx of financial funds act to
destabilise prices to the benefit of short-term financial interests that are trading on price volatility.
3. An investigation into the impact of financial investment on coffee exchange of
the NYBOT and coffee price behavior
The structure of the international coffee system
10
has evolved over the past two decades, together
with the structure of trading activities on international commodity exchanges. Whilst futures markets
were initially developed predominantly to allow the effective management of risks faced by physical
traders, they have evolved to increasingly reflect the needs of financial investors operating entirely
outside of the markets for physical commodities.
The share of derivatives trading taken up by financial investors has increased throughout the
period after the end of the Bretton Woods system. Since 2000, there has been a further explosion in
trading activities on commodities exchanges that was initially triggered by the Dot com crisis which
saw a shift of funds from equities to commodities.
11
The value of outstanding OTC commodity
derivatives in June 2007 was over 7.5 trillion US dollars, compared with 0.77 trillion in 2002 and 0.44
trillion in 1998. (Bank for International Settlements 2007) The growth in commodities has been
sustained as new investors enter the markets, these include various institutional investors such as
hedge funds and to lesser extent pension funds.
12
(Doyle, Hill, and Jack 2007) Unlike hedge funds,
pension funds have relatively long investment periods, 30 years for example, and can help sustain the
level of investment in these markets over time. There has also been a growth in investment on
commodity markets that has resulted from the lowering of entry barriers for smaller hedge funds
with the transition from open outcry to electronic trading platforms that have taken place in many of
the major commodities exchanges. Retail investment is also set to rise with an increase in the number
10
The international coffee system describes the economic, political and institutional structures under which the activities
between production and consumption of coffee take place. For a more detailed description of the structural changes that
have taken place in the coffee system over the past three decades see (Ponte 2001; Daviron & Ponte 2006; Newman 2008).
11
Commodities represent an attractive asset class for investors since commodity prices move in line with inflation,
commodities act as an inflation hedge. More importantly, commodity prices display a negative correlation with the price
of other financial instruments such as bonds and equities and therefore constitute a good asset class in terms of
diversifying portfolios and risk reduction. It is therefore not surprising that commodity markets see increasing investment
during periods of economic downturns.
12
In their 2007 FSA study, Doyle, Hill and Jack, found that75-80% of funds invested in the Goldman Sachs Commodity Index
(GSCI) are from pension funds.
of tradable products, such as Exchange Traded Funds (ETFs), Exchange Traded notes (ETNs) and
Exchange Traded Commodities (ETCs). (Doyle, Hill, and Jack 2007)
The coffee exchange of the NYBOT has not escaped the rise of financial investment. It is possible
to track this increase in futures trading attributed to market participants who are not involved in the
handling of physical coffee since the NYBOT divides all trading members into two categories,
commercial and non-commercial traders. Commercial traders are defined as entities that are trading
on the exchange who are also engaged in physical commodity trade, these will include the large
international commodity trading houses. All other exchange trading members are classified as non-
commercial. In spite of the difficulties in separating speculative and hedging activities based upon the
NYBOT categories, we can use this distinction in calculating a lower limit for speculative activity as a
ratio of total trading activities using the ratio of NPCT open interest to total open interest. Figure 1.
illustrates the evolution of the proportion of total futures trading made up by non-commercial
(financial) actors on the coffee exchange of the New York Board of Trade (NYBOT) between 1986 and
2007. For coffee futures, the proportion of total trading activities made up by non-commercial actors
has been both volatile and increasing on average from between 10 and 30% of total open interest on
the NY exchange in the 1980s to a ratio that varies between 40 and 70% in the 2000s.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
15/01/1986
15/01/1987
15/01/1988
15/01/1989
15/01/1990
15/01/1991
15/01/1992
15/01/1993
15/01/1994
15/01/1995
15/01/1996
15/01/1997
15/01/1998
15/01/1999
15/01/2000
15/01/2001
15/01/2002
15/01/2003
15/01/2004
15/01/2005
15/01/2006
15/01/2007
ratio of non-commercial OI to total OI
Figure 1. Ratio of non-commercial open interest to total open interest in coffee C contracts
on the New York Coffee Exchange.
At the same time as these dramatic increases in the proportion of financial activities on the futures
market, the liberalisation of producer country marketing systems has resulted in New York coffee
futures price gaining considerably in its importance in terms of informing the pricing decision of
market actors engaged in the exchange of Arabica coffee. Of the various market actors in the physical
market system, it is the international coffee traders and roasters that make up the vast majority of
hedgers on the futures market. Results from interviews of stakeholders in coffee chains, conducted
throughout 2007, indicate that for international coffee traders, the New York futures price is the single
most important piece of price information used in pricing decisions for the bulk of coffee trade in
arabicas, whilst the London futures price is relied upon for trade in robustas. The wide use of “price
to be fixed contracts” by international traders and roasters has cemented the relationship between
futures prices and the price at which physical coffee is exchanged at the international trader level.
13
Given the structure of the international coffee system, which is highly concentrated at the
international trader level
14
, prices on physical markets have become increasingly tied to those which
occur on the futures exchange (Newman 2008).
In the remainder of this section, we present the results of two quantitative investigations into the
role of financial investment in driving price movements on the coffee exchange of the NYBOT.
3.1 Supply and Demand vs. Trader Composition
In this section, we examine whether increases in the volume of financial flows onto the New York
coffee exchange has led to a loosening of the relationship between on-exchange prices and actual
physical supply and demand for Arabica coffee. Structural break tests are applied to series that
describe:
i) the evolution of the total annual volume of trade in futures contracts on the New York
Coffee exchange to identify sudden increases in the volume of futures trading;
ii) the fraction of non-commercial futures open interest to total open interest for each year to
see whether jumps in the volume of futures trading can be attributed to market
participants who are not hedging against physical trades; and
iii) the relationship between coffee prices and world supply & demand
13
A price to be fixed contract refers to a contractual arrangement whereby the volume, delivery date and differential price are
specified but the final price at which the commodity is exchanged will depend upon the futures price on the date at which
the price is fixed, up to the delivery date.
14
In 2006, the top 5 trading companies accounted for a market share of over 55%. (Newman 2008).
In this paper we use the Bai-Perron (2003) method to test for multiple structural breaks in a linear
model estimated by least-squares. Bai and Perron (1998b) present a number of alternative tests to
determine the break dates: “a sup-Wald type tests for the null hypothesis of no change versus an
alternative hypothesis containing an arbitrary number of changes and a procedure that allows one to
test the null hypothesis of, say, changes, versus the alternative of changes.
15
The latter is
particularly useful in that it allows a specific to general modelling strategy to consistently determine
the appropriate number of changes present.” (Bai and Perron 1998a) The Bai and Perron method can
be applied to pure and partial structural change models. The data and errors can be allowed to have
different distributions across segments, alternatively a common structure can be imposed across the
data over the entire time period in testing for a partial structural change. In order to estimate the
number of breaks present in the data, Bai and Perron discuss methods based on information criteria
as well as a method based on a sequential testing procedure.
16
Here Bayesian Information Criterion
(BIC) are used to determine the number of breaks contained in the data.
17
Structural break tests on the volume of trading in coffee C contract on the New York Coffee
Exchange (1973-2006).
Before testing for a structural break in the evolution of the annual data on the total volume of
trading for coffee C contracts (ATV), the time series properties of the series must first be analysed in
order to establish how best to model the series. The graph in figure 2 clearly shows that the series is
not stationary. From first inspection, the series appears to be increasing at an accelerating rate over
time, suggesting that the underlying data generation process is an autoregressive process of order one
(eq. 1)
ttt
ATVATV
ρ
β
+
=
−1
(1)
where ρ is a white noise error term.
In reality, there may be both a deterministic and a stochastic time trend. Harvey (1997) argues
that models with deterministic trends are a limiting case with stochastic trends where the hyper-
parameters (which allow for the level and slope of the trend to change) are equal to zero and therefore
suggests that time series are best modelled as autoregressive processes. There is, however, an
15
The test procedures are (i) a sup F type test of no structural break versus the alternative hypothesis that there are a fixed
number of breaks, (ii) a double maximum test, two tests of the null hypothesis of no structural break against an unknown
number of breaks given by some upper bound, iii) A test of versus breaks. This method amounts to the
application of tests of the null hypothesis of no structural change versus the alternative of a single change
( . )
16
The method is based on the sequential application of the .
17
The test procedures have been implemented using the RATS code MULTIPLEBREAKS.SRC written by Tom Doan (2005) that
is available on the Estima website (http://www.estima.com/Stability.shtml). The RATS
procedure does not give the F-test results and BIC are used to estimate the number of structural breaks.
argument for using the model specifications with a deterministic trend since a “non-linear curve can
be seen as a number of straight lines, with varying slopes, if there are stochastic trends, that would be
reflected in a large number of structural breaks in a linear trend” (Rao and Rao 2007, p6). We,
therefore, test for structural breaks on the basis of the AR(1) specification shown in eq 1 as well as a
more general model that includes a time trend (eq. 2)
ttt
tATVATV
δ
γ
α
α
+
+
+
=
−110
(2)
where δ
t
is a white noise error term.
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
Number of contracts (000s)
Figure 2. Total traded volumes of coffee C contracts on NYBOT per year
(Data Source: NYBOT 2007)
Variable Estimated coefficient p-value
Constant 81945.998636 0.40912633
ATV
t-1
1.030368 0.00000000
R-squared = 0.974264
R bar-squared = 0.924633
Table 1. Estimation results for
ttt
ATVATV
ρ
β
+
=
−1
(1973-2006)
Variable Estimated coefficient p-value
Constant -163955.7050
0.24219607
ATV
t-1
0.7049 0.00003242
T 40963.0360 0.02319624
R-squared = 0.978388
R bar-squared = 0.934600
Table 2. Estimation results for
ttt
tATVATV
δ
γ
α
α
+
+
+
=
−110
(1973-2006)
Tables 1 and 2 show the results from Bai Perron tests for multiple structural breaks applied to the
AR(1) and AR(1) with a time trend specifications of ATV. The maximum number of allowed breaks
was set to three to allow for sufficient observations between regime changes. In both specifications,
the Bayesian Information Criteria select a single break, with a break date of 2001 for the AR(1) model
and 1999 for the model including a time trend. These break points coincide with an influx of funds on
commodity markets associated with the dot-com bubble burst at the end of the 1990s when investors
suffered poor returns from both equity and bond markets (Doyle, Hill, and Jack 2007).
18
The break
identified at the date 1994 for both of the model specifications does not coincide with dramatic events
in the physical coffee market but are coincidental with shifts in the volume of financial activities on
commodities exchanges. In 1993-4 hedge funds allocated a small portion of their funds to
commodities which, owing to the large size of hedge funds, amounted to a sizeable influx of money
for the commodity markets (World Bank 1997). From his study of cocoa markets, Gilbert(1994)
concludes that the rise in cocoa prices during 1993-94 was due to fund investment. The recent great
Bull Run on commodities markets began in 2002 and has been sustained into 2008. This has seen
growth in existing commodity index funds as well as the launch of a number of large index funds in
2003, including the Deutsche Bank Liquid Commodity Index.
For the AR(1) model including a time trend, an additional break date, which is absent from the
AR(1) model, at 1980 is found. The year 1980 precedes a period of hight coffee prices that resulted
from a severe frost in Brazil in July 1981.
One of the reasons why the breaks identified in the data differ between the two specifications is
due to the difference in the specifications of the trend term. While the AR(1) specification shows up
breaks in a stochastic trend, a deterministic trend is included in the second specification. It may be
that once and for all changes in the behaviour of physical commodity traders on derivatives markets
are affected by changes in physical market conditions such as a supply shock, while changes that
18
“In part investors were searching for yield, in part they were seeking an inflation hedge, but mainly they were looking to
diversify equity- and bond-heavy portfolios” (Doyle, Hill !& Jack 2007, p16)
affect the behaviour of financial investors will reinforce stochastic trends since future behaviour
depends more on the trader behaviour in past periods than on physical supply and demand.
ttt
ATVATV
ν
+
=
−1
ttt
tATVATV
δ
γ
α
α
+
+
+
=
−110
Best 3 break dates
identified by Bai-Perron
method
2001, 2004, 1994 1980, 1994, 2000
Number if breaks
selected by BIC
1 1
Break Date 2001 1999
Table 3. Results from structural break tests on annual total trading volume on the coffee
futures exchange of the NYBOT (unrestricted – both intercept and trend were
allowed to vary)
Structural break tests on the annual ratio of non-commercial to total open interest for coffee C
contracts (1986-2006).
Since this paper argues that it is a change in the type of trading activities on international
commodity exchanges, as well as the over all volume that has led to changes in the behaviour of
commodity prices it is necessary to examine the nature of the evolution of the composition of trading
activities on these markets. This is done by studying the time series properties of the ratio of non-
commercial to total open interest, aggregated over the period of one year (ANCOIR), data for which is
available in the Commitment of Traders reports published by the U.S Commodity Futures Trading
Commission (CFTC).
19
As in the case for total annual traded volume, the series ANCOIR is modelled by an AR(1) and an
AR(1) model including a time trend (see table 4. and table 5.).
19
All traders identified by the U.S CFTC is classified as either “commercial” or “non-commercial” or "non-commercial." All of a
trader's reported futures positions in a commodity are classified as commercial if the trader uses futures contracts in that
particular commodity for hedging as defined in CFTC Regulation 1.3(z), 17 CFR 1.3(z). A trading entity generally gets
classified as a "commercial" trader by filing a statement with the Commission, on CFTC Form 40: Statement of Reporting
Trader, that it is commercially "...engaged in business activities hedged by the use of the futures or option markets."
(CFTC 2007)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
Ratio of non-commercial open interest to total open interest
Figure 3. Ratio of non-commercial to total open interest for coffee C contracts 1986-2006
(Data Source: CFTC 2007)
Variable Estimated coefficient p-value
Constant 0.0241249612
0.47052084
ANCOIR
t-1
0.9878033977 0.00000000
R-squared = 0.990033
R bar-squared = 0.882110
Table 4. Estimation results for
ttt
ANCTOIRANCTOIR
σ
δ
+
=
−1
Variable Estimated coefficient p-value
Constant 0.0857198814
0.03667971
ANCOIR
t-1
0.4970132484 0.02906118
T 0.0107845222 0.02244070
R-squared = 0.992729
R bar-squared = 0.908944
Table 5. Estimation results for
ttt
tANCTOIRANCTOIR
ω
η
λ
λ
+
+
+
=
−110
ttt
ANCTOIRANCTOIR
σ
δ
+
=
−1
ttt
tANCTOIRANCTOIR
ω
η
λ
λ
+
+
+
=
−110
Best 3
break dates
identified by
Bai-Perron
method
2001, 1993, 2003 1984, 1987, 1990
Number
if breaks
selected by
BIC
1
1
Break
Date
2001 1990
Table 6. Results from structural break tests on the ratio of the annual non-commercial to total
open interest on the coffee futures exchange of the NYBOT (unrestricted – both
intercept and trend were allowed to vary)
Results from the structural break tests applied to the AR(1) and AR(1) with trend specifications
for the series ANCOIR are shown in table 6.
A single break is selected by BIC for the case of the pure AR(1) model at 2001. The best three
break points are in line with those identified by the Bai-Perron method for the AR(1) specification of
ATV above. The break points found for the two models estimated for the series ANCOIR differ
considerably. By contrast, the best break point identified in the AR(1) model with trend is at 1990,
soon after the collapse of the International Coffee Agreement in 1989. Once again, the differences in
the results from the two models can be attributed to the identification of breaks in a stochastic trend
and a deterministic trend respectively. Comparing results from the AR(1) models for ANCOIR and
ATV suggests that shifts in the volume of futures trading on the coffee exchange of the NYBOT can be
attributed to financial investors such as hedge funds wishing to diversify from bonds and equities
and commodity index funds.
Structural break tests on the relationship between coffee prices and world supply and demand
(1978-2006)
Here, the Bai-Perron multiple structural break method is applied a model of coffee prices that
depends upon physical supply and demand. The annual average ICO composite indicator price for
green coffee is used as a measure for the world price for green coffee. Coffee prices (P) are modelled
as a function of the stock level (TST), disappearance in member countries (MC), and consumption in
producing countries (PC).
20
,
21
[ ]
+
=PCMC
TST
fP
(3)
Following the coffee price model of Maizels, Bacon and Mavrotas (1997) the price model is
specified in log linear form. Through successive estimation the functional form shown in eq. 4 was
found to best fit the data.
22
ε
α
α
α
+
+
+
=
−1210 t
LPLSTRTLP
(4)
Where, LP is the log of world coffee price and LSTRT is the log of ratio of producer stocks to
world coffee production.
23
The results from the estimation of the price model in eq (4) over the entire
time period and structural break analyses are shown in table 7 and table 8 respectively.
20
Member countries refer to coffee producing and consuming countries that are members of the International Coffee
Organisation (ICO). In 2006, the top 10 coffee consuming countries made up over 70% of the total market value. Of the top
10, only Russia (ranked 8
th
with 2.61% of total market value in 2006) is not a member of the ICO. Owing to the availability
of data, only ICO member countries within the top 10 consuming countries are included in the model. Consumption from
non-member countries is taken as exogenous.
21
Disappearance is used as a measure of consumption owing to the difficulties of measuring consumption directly.
Disappearance is measured by deducting re-exports from imports and adjusting the resulting figure changes in visible
inventories. (ICO 2003)
22
Only the central price equation of the Maizels, Bacon and Mavrotas model features in this paper. The entire coffee price
model consists of over 50 separately estimated equations for supply in individual producing member countries, demand
equations for individual consuming member countries as well as 3 separate price equations to link i) the retail price to
world price (relating the nominal retail price to the world coffee price, the exchange rate against the dollar, and time
trend); ii)the world price to producer price (relating the nominal retail price to world coffee price, the exchange rate and a
time trend); and the world price to stock level (relating stock ratios on the supply side to stock levels on the demand side
via the world price. Details on the estimation of this system of equations making up the coffee price model between 1982
and 2006 are available on request from the author.
23
Testing for a structural break in this equation is equivalent to testing for a structural break in an equation for price
determination that links the world price to the ratio of disappearance to consumer stocks.
Variable Estimated coefficient p-value
Constant 1.2716433856 0.00154120
LSTRT
t
0.9138159603 0.01662796
LP
t-1
0.5950293368
0.00001777
R-squared = 0.997879
R bar-squared = 0.690437
Table7. Estimation results for the coffee price model:
εααα
+++=
−1210 t
LPLSTRTLP
ε
α
α
α
+
+
+
=
−1210 t
LPLSTRTLP
Best 3 break dates identified
by Bai-Perron method
1994, 2000, 2004
Number if breaks selected
by BIC
1
Break Date 1994
Intercepts 1.4332727529 (p-value 0.00163319)
1.5442321270 (p-value 0.03369707)
Coefficient on LSTRT 1.6469755916 (p-value 0.00248171)
1.3112557084* (p-value 0.10207228)
Coefficient on LP
t-1
0.6761458959 (p-value 0.00002563)
0.5857708346 (p-value 0.01673472)
* not significant at the 5% level
Table 8.
The best three break dates identified by the Bai-Perron method are close to those identified for the
series ATV and ANCOIR, modelled as AR(1) processes, albeit in a different order of significance.
Bayesian Information Criteria select a single break in the relationship between coffee prices, stocks
and prices in the previous period in 1994. Estimates for the coefficients of the price model in the sub-
periods 1978-1994 and 1994-2006 indicate a loosening in the relationship between price and stocks. In
the earlier sub-period, the regression results suggest that, citeris paribus, a 1.65% increase in the ratio
of producer stocks to world coffee production is associated with a 1% increase in the world price
compared with a 1.31% increase in the ratio of producer stocks to world coffee production being
associated with a 1% increase in the world price. The coefficient on LSTRT is less significant in the
period 1994-2006 (p-value = 0.102) compared with the period 1978-1994 (p-value 0.002).
The structural break analysis presented in this section provides evidence for a break in, as well as
a loosening of the relationship between coffee prices and physical supply & demand that coincides
with a break in the data for futures trading volumes on NYBOT. Increased trading volumes for the
coffee C contract can be attributed to increased participation by non-commercial traders such as
hedge funds and commodity index funds.
3.2 Financial investment and short-term price movements
Following the method set out by Laby and Thomas (1975) we examine the simple correlation
between financial investment in coffee futures and a computed instability index, or volatility index
for the futures price. A monthly volatility index is calculated as the normalised standard deviation of
daily closing futures prices within a trading month (equation 5)
( )
−
=∑
n
PP
P
VI
t
2
100
(5)
Where
t
P
is daily closing Coffee C futures price on NYBOT, and
P
is the monthly average
closing Coffee C futures price on NYBOT. Figure 4 shows the monthly volatility index from 1972 to
2006.
0
5
10
15
20
25
Sep-72
Sep-74
Sep-76
Sep-78
Sep-80
Sep-82
Sep-84
Sep-86
Sep-88
Sep-90
Sep-92
Sep-94
Sep-96
Sep-98
Sep-00
Sep-02
Sep-04
Sep-06
Volitility Index (VI)
Figure 4.
As a measure of futures trading that is not for the purposes of hedging we construct a financial
investment index (SPEC) as the natural logarithm of the ratio of the volume equivalent of coffee-C
contracts traded and total exports of green coffee from ICO member countries.
24
The monthly
Financial Investment Index for the period 1972-2006 is shown in figure 5.
0
2
4
6
8
10
12
Sep-72
Sep-74
Sep-76
Sep-78
Sep-80
Sep-82
Sep-84
Sep-86
Sep-88
Sep-90
Sep-92
Sep-94
Sep-96
Sep-98
Sep-00
Sep-02
Sep-04
Sep-06
ln(volume equivalent of coffee C contracts traded to total exports from
ICO member countries)
Figure 5. Natural logarithm of the ratio of monthly coffee volume equivalent of futures
contracts traded on NYBOT to total exports from ICO member countries 1972-2006
(SPEC)
(Data Source: CFTC 2007; ICO 2007)
Equation 6 gives the linear specification for the relationship between the volatility index and the
index of financial investment that is estimated by OLS.
ttt
uSPECVI
+
+
=
21
β
β
(6)
OLS estimation of the relationship between volatility and financial investment result in a positive
and significant slope coefficient on the index of financial investment is significant (see table 9). This is
24
Laby and Thomas (1975) discuss a number of possible quantitative measurements for speculative activities including a
measure based upon the ratio of long and short positions in the market according to the specification by Working (1962)
that the actual nature of market commitment is that hedging is normally net short and that speculation is net long.
Another possible measure of speculative activities on futures markets can be constructed is the ratio of the volume of
physical trade to the volume equivalent of futures contracts traded in a particular period.
indication of a positive statistical relationship between financial investment activities and price
volatility on the New York coffee exchange.
Variable Estimated Coefficient p-value
Constant -6.105975118
0.00035828
SPEC 1.134782785 0.00000001
R-squared = 0.759739
R bar-squared = 0.092990
Table 9. Estimation results for
ttt
uSPECVI
+
+
=
21
β
β
for the period 1980:01 to
2006:12
To test for breaks in the relationship between the volatility and speculation indices, the Bai-Perron
method, described in section 3.1, is applied. Table 10. summarises the results from the Bai-Perron tests.
Bayesian Information Criteria select 1 structural break in October 2000. This coincides with the
structural breaks identified in the data for futures trading activity and financial trading activities,
attributed to the inflow of financial funds into commodity markets as a result of the Dot-com crash of
2000 in section five.
The correlation coefficient (ρ) between the volatility and financial investment indices are
calculated and the simple linear regression model shown in eq. 6 are estimated for four different sub
periods. The four sub periods are based upon the best two structural breaks identified, these are:
1980:01 to 1994:04, 1994:04 to 2000:10, 2000:10 to 2006:12 and 1994:04 to 2006:12. 1994:04 is included as
a break date since it coincides with the entry by hedge funds onto commodity markets. The
correlation coefficients differ for all four sub-periods and is at its highest, 0.46 (2dp), between 1994:04
and 2000:10. The estimated slope coefficients for regression model for each sub-period also differ. In
the period 1980:01 to 1994:04, the slope coefficient is highly significant and suggests that an increase
in the index of financial investment of one unit results in an increase in the volatility index of 0.84
units (2dp). By contrast, the slope coefficient estimated for the period 1994:4 to 2000:10 is 5.66 (2dp),
with a p-value lower than that calculated for the estimated coefficient for the previous period. This
result supports the argument that increasing financial activities on commodity exchanges associated
with the massive inflow of funds from hedge funds in 1993/4 has lead to heightened price volatility
on the coffee exchange of the NYBOT.
The estimated slope coefficient between 2000:10 and 2006:12 was found to be lower than the
previous periods, 0.73 (2dp), and not statistically significant at the 10% level. This is somewhat of a
surprise given that the year 2000 saw a sudden increase in financial investment on commodities
associated with the dot-com crisis, and has continued to increase over the remainder of the period.
This apparent contradiction in the relationship between financial investment on commodity
exchanges and volatility can explained if we consider the type of participants in the market, their
motivations for trading and consequently, the nature of the trading activities.
The structural break in the relationship between price volatility and financial investment on the
NYBOT coffee exchange in 1994 is explained by the inflow of funds associated with the inclusion of
commodities by a number of large hedge funds. While the term ‘hedge fund’ covers a diverse group
of institutional investors, a distinguishing feature of hedge funds operating on commodities markets
is in their ability to sell short that is to bet on falls in the market and profit from them. (Doyle, Hill,
and Jack 2007) Hedge funds can thus profit from price volatility on commodities markets in a way
that other financial investors cannot and provide commodity markets will a large proportion of the
liquidity in them. An estimated 80% of funds in commodity markets are looking for higher-than-
market-following returns (alpha) (Doyle, Hill, and Jack 2007). This alpha money is short-term and
moves quickly into and out of the market. The volatility of fund in- and out-flows has implications on
short-term price movements on commodity exchanges where large flows can rapidly drive prices
away from the price warranted by world supply and demand.
The motivations of financial investors entering commodities markets in 2000 were very different
from the inflows of money due to hedge funds in the 1990s. The burst of the dot-com bubble saw a
dramatic fall in the returns from both equity and bond markets. In addition, commodities had been in
a bear market for around 25 years preceding the dot-com crash, with prices at their all time low.
Falling returns on equity and bond markets, together with the anticipation of inflation led investors to
seek out alternative investment vehicles. The character of commodity markets (with prices that move
in the opposite direction to equities and bonds and in line with inflation) led to an inflow of funds
seeking an inflation hedge and looking to diversify their bond and equity heavy portfolios. The
period since 2000 has thus been accompanied by greater variety in the types of financial investors and
investment strategies in commodity markets (Domanski and Heath 2007). In particular, passively
managed investment and portfolio products have been growing. These investments often pursue a
fully collateralised long futures strategy which is attractive to institutional investors with a longer-
term horizon such as pension funds.(Domanski and Heath 2007) Rather than trading on volatility, this
group of investors profit from long-run price increases. It is therefore, not surprising that the slope
coefficient for the regression equation (6) estimated for the period 2001:10 to 2006:12 is insignificant
and smaller than that estimated for previous periods. Rather than enhanced price volatility, the
period between 2001 and 2006 has seen an upturn in commodity prices that has been largely driven
by financial investors with longer-term horizons as described in the previous section.
ttt
uSPECVI
+
+
=
21
β
β
(1980:01 to 2006:12)
BIC
Best 5 break dates identified by Bai-
Perron method 1981:03, 1981:09, 1987:06, 1994:04,
2000:10 1.33945
Best 4 break dates identified by Bai-
Perron method 1980:05, 1986:11, 1994:04, 2000:10
1.37260
Best 3 break dates identified by Bai-
Perron method 1987:06, 1994:04, 2000:10
1.38572
Best 2 break dates identified by Bai-
Perron method 1994:04, 2000:10
1.44880
Best break date identified by Bai-
Perron method 2000:10 1.53878
Number if breaks selected by BIC 1
Break Date 2000:10
Table 10. Summary of results from Bai-Perron tests of multiple breaks in the relationship between
financial investment and price volatility
Variable Estimated coefficient p-value
1980:01 to 1994:04, R-squared = 0.718231, R bar-squared = 0.048808, ρ = 0.233176
Constant -3.773838265 0.09257627
SPEC 0.836345100
0.00208172
1994:04 to 2000:10, R-squared = 0.839774, R bar-squared = 0.203926, ρ = 0.462745
Constant -45.75715604 0.00009639
SPEC 5.66135965 0.00001756
2000:10 to 2006:12, R-squared =0.861727 ,R bar-squared = 0.019330, ρ = 0.163134
Constant -3.171043700
0.46426916
SPEC 0.731497994 0.12120941
1994:04 to 2006:12, R-squared=0.790368, R bar-squared =0.007724, ρ = 0.180504
Constant -3.407337938 0.51972349
SPEC 0.855760676 0.14160197
Table 11. Estimation results for
ttt
uSPECVI
+
+
=
21
β
β
for the 4 sub-periods between
1980:01 to 2006:12
0
2
4
6
8
10
12
14
0 2 4 6 8 10 12 14
Speculation Index (SPEC)
Volatility Index (V)
Figure 6. Volatility Index Vs. Index of financial investment 1980:01 to 1994:04
0
2
4
6
8
10
12
14
16
18
0 2 4 6 8 10 12
Speculation Index (SPEC)
Volatility Index (VI)
Figure 7. Volatility Index Vs. Index of financial investment 1994:04 to 2000:10
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8 10 12
Speculation Ratio (SPEC)
Volatility Index
Figure 8. Volatility Index Vs. Index of financial investment 2000:10 to 2006:12
0
2
4
6
8
10
12
14
16
0 2 4 6 8 10 12
Speculation Ratio (SPEC)
Volatility Index
Figure 9. Volatility Index Vs. Index of financial investment 1994:04 to 2006:12
4 Conclusion
The results from the quantitative analysis presented in section 3 suggest that the behaviour of
prices for coffee has been affected by the extent, as well as the type, of financial investment that has
taken place on the coffee exchange of the NYBOT. Since the 1990s, financial investors have based
their trading decisions on coffee futures markets in terms of a diversified portfolio of investments that
aim to balance risks against returns, rather than trading on the basis of supply and demand
conditions in the physical market. The flow of funds onto the NYBOT has occurred in relation to the
wider economic environment, such as the down-turn in the price of equities associated with the dot-
com crash in 2000, rather than the market conditions of the commodity in question. A similar inflow
of funds onto commodity markets has taken place since the sub-prime crisis and resulting credit
crunch that began to emerge in the latter half of 2007. Commodity prices have continued to increase
across the board. While in some cases, price increases can be attributed to increasing demand, such as
for metals and minerals, or grains, there are a number of commodities (such as coffee) that are likely
to be carried along with the current commodity boom owing to financial interests on international
exchanges.
Whether or not specific bouts of financial activities on commodity markets can occur to the
benefit of physical market actors will depend firstly, on the affect on price outcomes that fund inflows
have. It has been suggested in this paper that, depending on the type of financial investment, whether
gains are made from longer term price trends or short-run volatility, price outcomes will differ.
Second, depending on the type of price outcome, whether or not physical market actors can benefit
will depend upon the structure of the market, and the extent to which different market actors can deal
with the risks associated with price volatility. Newman (2008) shows how heterogeneous market
actors in Tanzanian and Ugandan coffee chains differ in their ability to harness rewards associated
with price volatility. As hedging instruments, futures contracts are inherently biased against
producers and local exporters since they are unable to benefit from forms of speculative hedging open
to short-hedgers such as international trading companies.
In terms of policy conclusions, this paper suggests that, coffee price stabilization may be desirable
in the context where producers and local exporters do not have easy access to hedging instruments.
Multilateral price stabilization schemes along the lines of the International Commodity Agreements
(ICAs) are neither economically, nor politically feasible today (Gilbert 1996;Newbery and Stiglitz
1981). Moreover, the buffer stock and quota schemes of the ICAs do not address the causes of price
volatility that originate outside of the physical markets. In order to anchor prices that evolve on the
international commodity exchanges to supply and demand conditions, it may be prudent to fix an
upper limit to the proportion of futures trading conducted by non-commercial actors. The suitable
size of such a cap will depend upon the market in question and the types of financial activities that
take place on it.
25
25
For example, it may be expected that cotton prices were less sensitive than coffee to financial investment on the associated
cotton exchange of the NYBOT since the cotton price control of the US-Farm Scheme can act as a price anchor. Having said
this, cotton prices experienced historical highs in March 2008. The causes of this recent price trend need further
investigation.
References
Adams, F. G and S. A Klein eds. 1978. Stabilizing World Commodity Markets: Lexington Books.
Akiyama, T and R. C Duncan. 1982. "Analysis of the World Coffee Market." World Bank Staff
Commodity Working Papers:7.
Arrow, K. J and G Debreu. 1954. "The Existence of an Equilibrium for a Competitive Economy."
Econometrica.
Bai, Jushan and Pierre Perron. 1998a. "Computation and Analysis of Multiple Structural-Change
Models." CIREQ working paper:0998.
Bai, Jushan and Pierre Perron. 1998b. "Estimating and Testing Linear Models with Multiple Structural
Changes." Econometrica, 66, pp. 47-78.
Bai, Jushan and Pierre Perron. 2003. "Critical values for multiple structural change tests." The
Econometrics Journal, 6:1, pp. 72-78.
Bank for International Settlements. 2007. "Semiannual OTC derivatives statistics at end-June 2007."
Black, Fisher and Myron Scholes. 1972. "The Valuation of Options Contracts and a Test of Market
Efficiency." Journal of Finance, 27, pp. 399-417.
CFTC. 2007. "Commitment of Traders Reports: EXplanatory Notes." U.S Commodity Futures Trading
Commision: Washington D.C.
Davidson, P. 1991. "Is Probability Theory Relevant for Uncertainty? A Post Keynesian perspective."
Journal of Post Keynesian Economics, 13:1, pp. 129-43.
Daviron, Benoit and Stefano Ponte. 2006. The Coffee Paradox. London and New York: ZED Books.
Domanski, Dietrich and Alexandra Heath. 2007. "Finanical investors and commodity markets." BIS
Quaterly Review:March 2007.
Doyle, Emmet, Jonathan Hill, and Ian Jack. 2007. "Growth in commodity investment: risks and
challeges for commodity market participants." FSA Markets Infrastructure Department.
Financial Services Authority: London, UK.
Fama, Eugene F. 1970. "Efficient Capital Markets: A Review of Theory and Empirical Work." The
Journal of Finance, 25:2, pp. 383-417.
Fama, Eugene F. 1998. "Market efficiency, long-term returns, and behavioural finance." Journal of
Finacial Economics, 49, pp. 283-306.
Geman, Helyette. 2006. Commodities and Commodity Derivatives: Modeling and Pricing for Agriculturals,
Metals and Energy: John Wiley & Sons Ltd.
Ghosh, S, C.L Gilbert, and A.J Hughes Hallet. 1987. Stabilizing Speculative Commodity Markets: Oxford
University Press.
Gilbert, C.L. 1996. "International Commodity Agreements: An Obituary Notice." World Development,
24:1, pp. 591-616.
Gilbert, Christopher L. 1994. "Commodity Fund Activity and the World Cocoa Market." London
Commodity Exchange: London.
Gorton, Gary and K. Geert Rouwenhorst. 2004. "Facts and Fantasies about Commodity Futures." Tale
ICF Working Paper:04-20.
Goss, B. A and B. S Yamey. 1978. The Economics of Futures Trading. London and Basingstoke: The
Macmillan Press Ltd.
Grossman, S and J Stiglitz. 1980. "On the Impossibility if Informational Efficient Markets." The
American Economic REview, 70:3, pp. 393-408.
Harvey, A. 1997. "Trends, Cycles and Autoregression." The Economic Journal, 107, pp. 197-201.
Hull, John C. 2003. Options, Futures, and Other Derivatives: Prentice Hall.
ICO. 2003. "Data concepts and variables used in the statistics of the Organisation." The International
Coffee Organisation.
Irwin, H. S. 1937. "The Nature of Risk Assumption in the Trading on Organized Exchanges." American
Economic Review, 43.
Keynes, John Maynard. 1997. The General Theory of Employment, Interest and Money. Amherst, New
York: Prometheus Books.
Labys, W. C. 1978. "Commodity Markets and Models: The Range of Experience," in Stabilizing World
Commodity Markets. F. G Adams and S. A Klein eds. Toronto: Lexington Books, pp. 3-34.
Labys, Walter C. and armon C. Thomas. 1975. "Speculation, hedging and commodity price behaviour:
an international comparison." Applied Economics, 7:4, pp. 287.
MacKenzie, Donald. 2005. "Is Economics Performative? Option Theory and the Construction of
Derivatives Markets." The annual meeting of the History of Economics Society: Tacoma, WA.
Maizels, A, R Bacon, and G Mavrotas. 1997. Commodity Supply Management by Producing Countries: A
Case-Study of the Tropical Beverage Crops. Oxford: Clarendon Press.
Mandelbrot, Benoit. 1966. "Forecasts of Future Prices, Unbiased Markets and Martingale Models."
Journal of Business
Marshall, Alfred. 1932. Industry and Trade. London: Macmillan.
McKinnon, R. I. 1967. "Futures Markets, Buffer Stocks, and Income Stability for Primary Producers."
Journal of Political Economy, 75.
Merton, Robert C. 1973. "Theory of Rational Option Pricing." The Bell Journal of Economics and
Management Science, 4:1, pp. 141-83.
Newbery, David M. G and Joseph E Stiglitz. 1981. The Theory of Commodity Price Stabilization: A Study
in the Economics of Risk. Oxford: Clarendon Press.
Newman, Susan A. 2008. "The role of international commodity exchanges in price formation and
transmission of prices and price risk along international coffee chains." NCCR Trade Regulation
Working Paper Series:No 2008/12.
Ponte, Stefano. 2001. "Coffee Markets in East Africa: Local Responses to Global Challenges or Global
Responses to Local Challenges " Centre for Development Research Working Paper.
Rao, B. Bhaskara and Gyaneshwar Rao. 2007. "Structural breaks and energy efficiency in Figi." MPRA
Paper:3258.
Samuelson, P.A. 1952. "Probability, Utility and the Independence Axiom." Econometrica, 20, pp. 670-8.
Samuelson, P.A. 1965. "Proof that Properly Anticipated Prices Fluctuate Randomly." Industrial
Management Review.
Shackle, G. L. S. 1979. Imagination and the nature of choice. Edinburgh: Edinburgh University Press.
Stein, J. 1987. "Informational Externalities and Welfare Reducing Speculation." The Journal of Political
Economy, 95:6, pp. 1123-45.
Working, Holbrook. 1962. "New Concepts Concerning Futures Markets and Prices." American
Economic Review.
World Bank. 1997. "Do hedge funds and commodity funds affect commodity prices." DEC notes:
Research findings: no. 29. From the Development Economics Vice Presidency of the World
Bank.
