What drives agricultural commodity prices?

(1)

What drives agricultural commodity prices?

An investigation into the long run and short run determinants

Prepared by Jacob Dankert July 2011

Abstract

This paper looks at the behavior of agricultural commodity prices in both the long run (1980-2010) and the short run (2000-(1980-2010). Unlike many other papers trying to explain commodity prices, it takes into account a multitude of factors that could be of influence and analyzes it in one framework. Inventory levels, demand growth, biofuel production, speculation and monetary conditions are considered. Evidence is found for the effect of inventory levels, demand growth and biofuel production. Speculation and monetary conditions, as far as working through the interest rate, are not found to be of influence in both the short and long run.

Keywords: agricultural commodities, commodity prices, time-series analysis

Master thesis for MSc International Economics and Business Studentnumber: 1530968

Email: j.j.dankert@student.rug.nl

Supervisor: dr. Bart Los, department of Global Economics and Management Rijksuniversiteit Groningen

(2)

1

1. Introduction

In 2007 we were in the midst of a food crisis felt mostly by the world’s poor as food prices nearly doubled in a year’s time. The effects were felt worldwide. In Mexico’s protests against the government for the exploding food prices erupted into the infamous ‘Tortilla crisis’. The streets of Cairo were filled with people demanding cheaper bread, an omen of discontent with the system in Egypt. Meanwhile Russia, one of the big wheat exporters, put an export ban into place to ensure its food security at home. The sudden rise in agricultural commodity prices which drive the food price indices was unexpected for most observers (The Economist, Apr 17th 2008). As The Economist notes, the price rise could not solely be explained by supply side factors such as a bad harvest in a large producing country. Weather-induced supply shocks are common to agricultural products, but rarely cause such a price hike. Other possible causes mentioned were numerous.

One of the reasons is the advent of biofuel use on a large scale. Although Brazil has been using ethanol on a large scale since the 1970’s and the practice existed in the USA since the 1980’s, only recently has usage amassed to a large scale. Sugar cane and corn are used as inputs and this certainly has an impact on the sugar and grains market. The EU mostly promotes the use of biodiesel, which is made from vegetable oils such as rapeseed oil or soybean oil.

A second reason on the demand side is the growth of the developing economies, mainly in Asia. Increasing numbers of people in emerging economies have more money to spend and upgrade their food budget. This means they are increasingly adding meat and dairy products to their diets in lieu of grains. More meat and dairy products means more poultry, hogs and cattle need to be raised. More grains and soybean meal therefore is needed to feed animals, with obvious consequences for grains and oilseeds markets.

Another widespread belief advocates excessive speculation as the cause for the food price rises. Though speculation is a normal part of economic activity, there is evidence that a speculative bubble has been created in some commodity markets (e.g. Tang & Xiong, 2010; US Senate Permanent Subcommittee on Investigations, 2009; Timmer, 2009).

(3)

2

But are these really the causes of the recent boom in agricultural commodities after decades of relative stability? And what is the contribution of the different factors? Following the commodity boom has been a boom in economists’ attention for commodities. Since 2006 numerous papers have been published trying to answer these questions. These papers rarely take into account multiple determinants of commodity prices but only look at one possible influence.

What has not been done before for agricultural commodities is including market data for the agricultural commodities, such as consumption growth and inventory levels, together with macroeconomic factors and some measure for speculation or monetary influences in one framework. The value added of my research then is twofold. First, I take all three views into account in the long run by using annual data world market prices for the period 1980-2010. Second, I collect monthly data of world market prices for the period 2000-2010 to determine whether macroeconomic conditions and speculation can explain the boom. This has not been done before for agricultural commodities. Supply, demand and inventory data are not available on a monthly basis for agricultural commodities to do different harvest seasons around the world, so this has to be excluded from the short-run analysis. This leads me to my research questions:

What determines agricultural commodity prices in the long run? And can global demand conditions and speculation explain the recent boom?

(4)

3

2. Literature review

Theory on the role of speculation

The starting point of this analysis is the work of Nicholas Kaldor (Kaldor, 1939). He reviews the stabilizing or potentially destabilizing influence of speculation on prices of standardized traded assets such as stocks, bonds and commodity contracts. He defines speculation as the purchase (sale) of a certain stock or commodity with the sole purpose of reselling (rebuying) it at a later date, motivated by an expectation of a relative price change between the moments of purchase (sale) and resell (rebuy). So normal commercial transactions are excluded from this definition, as is arbitrage between markets. Speculators take into account interest costs, carrying costs and a risk premium when deciding what positions to take, based on their expectations of the future price. The traditional theory of speculation then posits that speculation narrows the range of fluctuations of the spot price relative to the future price and therefore has a stabilizing effect on prices. However, the potential stabilizing effect of speculators has a number of assumptions. Most importantly, as Kaldor notes, the supply and demand behavior of speculators should only amount to a small part of the total market, so that only the magnitude of the price change and not the direction is influenced by speculators behavior. If speculators were a large part of the market, speculative activity could exert more influence than fundamental supply and demand factors and it would suffice for a successful speculator to predict the positions of other speculators. A speculative bubble could be formed. This leads me to my first hypothesis.

H1: Speculation contributed to the price rise of agricultural commodities.

Theory on the effect of supply and demand on inventories

(5)

4

predicted values show a similar pattern as real world price behaviour. Speculators let the price fall and rise based on their expectations about the future price. Again, his analysis is highly mathematical and he does not provide explanation how speculators form expectations, based on what kind of information.

Building on this is the seminal work of Deaton and Laroque (Deaton and Laroque, 1992), who introduce the impossibility for commodity markets as a whole to carry negative inventories (e.g. eat the grain today that is harvested next year). In their view, this can account for a high degree of price autocorrelation in normal times and sometimes sudden boom episodes when inventories are low or are predicted to be low, for whatever reason. This price autocorrelation means that prices do not change much in normal times, when inventories are plentiful. The next month’s price is likely to be closely correlated to this month’s price. The sudden boom episodes are not unusual for food commodities, for which demand is relatively inelastic. This kind of model however does not say anything about why inventories are low or high, there is no attempt to explain the supply and demand behavior or the cause of shocks. Deaton and Laroque (1992) just model harvests that are random and independent and identically distributed over time. But it does posit that inventory levels and therefore market fundamentals such as supply and demand (that work to raise or lower inventory levels) are the determinants of commodity prices. Further extensions show that the disturbances can also be modeled as periodic disturbances, instead of random shocks which are independently and identically distributed (Chambers and Bailey, 1996). Both papers find a reasonably good fit for the actual price behavior. What they cannot explain however is the high degree of autocorrelation found in actual data, which might be a result of not so independently distributed harvests (e.g. the biblical 7 years of plenty and the 7 years of famine) or some degree of foresight by market participants about the coming harvests, which makes prices adjust in anticipation of the harvest. This literature leads to the second hypothesis.

H2: Agricultural commodity prices are inversely related to inventory levels.

(6)

5

H3: Biofuel production and other consumption growth fueled by emerging economies contributed to price rises of agricultural commodities.

Theory on the role of monetary policy

The work on monetary influences on commodity prices builds on the notion of Hotelling (Hotelling, 1931), that at low interest rates inventory accumulation of storable commodities becomes more profitable. Moreover, as already noted, other investment opportunities became scarce as real estate and stock markets plummeted in 2007. This led to an excess demand for assets, which formed a bubble in commodity markets (Cabellero, Farhi, & Gourinchas, 2008). The opposite is also true. When interest rates are high, the appetite for carrying commodity inventories is lowered. When commodity prices were low in the 1980’s, it coincided with high interest rates. Thus high real commodity prices can be caused by loose monetary policy (Frankel, 2006; Frankel & Rose, 2010). Monetary policymakers already acknowledge the possibility that their decisions are influential. Ben Bernanke, chairman of the Federal Reserve, noted in a June 2008 speech that is of clear importance for policymakers to understand the factors that drive commodity prices and to be able to forecast price movements (Groen and Pesenti, 2010). This leads to the following hypothesis.

H4: Low real interest rates caused higher prices for agricultural commodities.

Empirical evidence on the role of speculation

Is there evidence for these hypotheses? Speculation certainly has received its fair share of research. In a review of events of 2007-2008 (Piesse J. & Thirtle, 2009), the authors conclude that in the wheat, maize and soybean market bubbles have been created by increases in long-positions by financial institutions. This was fueled by a lack of other profitable investment opportunities as stock- and real estate markets crashed in the financial crises. Speculation in agricultural commodities then becomes more attractive.

(7)

6

Trade were in the hands of index speculators in 2008 (Westhoff, 2010, p.133). For the period from January 2006 to October 2009 the average percentage of open interest long positions across agricultural commodities in the hands of index traders was 28,4%. For short positions, this was only 1,6% (Tang & Xiong, 2010). The expectation of the market therefore was of rising prices and a significant part of the market was captured by speculators, making the price rise a self-fulfilling prophecy. There is more evidence. The role of index investment can also be studied by looking at the return correlations of commodities included in popular indices such as the Goldman Sachs Commodity Index and the Dow Jones - UBS Index and comparing them to the return correlations in off-indexed commodities. Since investors tend to move money in- and out of the index as a whole and are not that sensitive to prices of individual commodities, the expectation is that we see comovement of prices of otherwise unrelated commodities. Evidence for this is found from 2004 onwards, when index investment took off (Tang & Xiong, 2010). Indexed commodities returns correlated stronger with each other than commodities not included in the Goldman Sachs and Dow Jones indices. This updates the finding of Pindyck and Rotemberg that for 1960-1985 there was excess comovement in otherwise unrelated commodities, which proves that commodity prices are influenced by speculators and herd behavior on financial markets in general. (Pindyck & Rotemberg, 1990). Hence, the conclusion from these papers is that speculation played a role in the recent boom, exacerbating the price rise. All these papers acknowledge however that speculators act upon a belief that has to be rooted in the fundamental drivers of the market, such as strong demand growth from emerging countries or bad harvests in an important producing region.

The link between agricultural- and energy markets

(8)

7

boom from September 2007-October 2008 it increased to 0.92. From November 2008 - May 2009 the correlation decreased to 0.56, still higher than prior to the boom (Hertel & Beckman, 2011). To account for these changes, it is reasonable to look at the demand side of the market. As mentioned before, the popular claim is that biofuels cause the increases in prices of agricultural commodities. Although not all agricultural commodities are directly used for ethanol or biodiesel production, most commodities are close substitutes. If a poultry farmer finds the corn price elevated due to increased use of corn for ethanol production, he just feeds wheat to his chicken, or soybean meal for that matter (Westhoff, 2010, p.17-22). So there is substitutability in consumption. Off course, substitutability in production is also present, as farmers decide what to sow each crop cycle. These effects make that most agricultural commodity prices are closely related. Demand growth for agricultural commodities due to biofuel production has certainly added to higher prices, but to what extent? Estimates vary, but an overview of different estimates yields predictions in the order of 30% higher prices over 2006-2007 for the US (Perrin, 2008).

The effect of strong economic growth in emerging economies

This alone cannot explain the more than doubling of prices observed over this period. Another prominent cause of demand growth is the increase in income observed in Asian economies, most prominently China and India. This already leads to more grain demand for food, but increasing meat consumption in these adds a multiplier effect as multiple kilograms of feed are needed to produce 1 kilogram of meat. In grains, China has been able to increase production to offset internal demand, although this year it may need to import (The Economist, Feb 10th 2011). For soybeans however, China is the dominant importer in the world. It uses the soybean both for human consumption and to support a growing livestock sector (Westhoff, 2010, p.198). Since grains and soybeans are substitutes, the price rises of grains and soybeans are synchronized to a large extent.

Empirical evidence on the role of monetary policy

(9)

8

interest rate. In another account of the commodity boom of the 1970s, Barsky and Kilian (2001) find evidence that the rise in oil prices and that of other commodities was a response to macroeconomic conditions ultimately driven by monetary conditions. Low interest rates then caused excess liquidity, which fueled the boom. Newer research also looks into the effect of monetary policy by including a measure of global excess liquidity. Excess liquidity was found to be significant in explaining oil and fine wine movements over the past two decades (Saadi-Sedik and Cevik, 2011). However, these researchers conclude that excess liquidity can also be considered an (albeit imperfect) measure of speculation as liquidity is chasing fewer assets and thereby could propagate an asset bubble. However, Saadi-Sedik and Cevik (2011) find stronger evidence for the influence of global demand growth and especially growth in emerging economies. The influence of this growth on commodity prices is stronger than excess liquidity, although both are significant.

Summing up the evidence

Summing up the evidence on the possible determinants of commodity prices, one can thus distinguish three camps. The first camp argues that fundamental supply-demand factors were the cause for the recent boom, with a large role attributed to increasing biofuel demand and the scramble for commodities by the developing countries in Asia. When world economic growth slowed in the financial crisis, commodity prices dropped. This camp has partly macroeconomic explanations, such as strong global GDP growth and partly market fundamentals such as low inventory levels of commodities caused by biofuel production (e.g. Piesse and Thirtle, 2009; Hertel and Beckman, 2011; Perrin, 2008). The second camp argues that increased investment into commodities, for instance in commodity indices such as the Goldman-Sachs Commodity Index, caused the boom. Here, excessive speculation is the culprit (e.g. Tang and Xiong, 2010; US Senate, 2009; Timmer, 2009). The third camp argues that easy monetary policy was of fundamental influence working through low interest rates (e.g. Frankel, 2006; Barsky and Kilian, 2001).

(10)

9

(11)

10

3. Data

Commodity selection

From the hundreds of agricultural products produced, I focus on maize, wheat, soybeans, soybean meal, soybean oil, sugar and rice as my subjects of interest. The selection process of these crops has been as follows. The first selection criterion applied is worldwide production quantity. The production quantities of the most produced crops can be seen in table 1.

Table 1. Most produced crops worldwide Production, metric tonnes (2009)

Sugar cane 1.682.577.768 Maize 817.110.509 Wheat 681.915.838 Rice, paddy 678.688.289 Potatoes 329.556.911 Vegetables fresh 246.349.029 Cassava 240.989.481 Sugar beet 229.490.296 Soybeans 222.268.904 Source: FAO (2011)

(12)

11

consumption there was no reason for this price rise (Piesse and Thirtle, 2009). Rice is included in the research, but with the caveat that prices in the international market are not very reflective of worldwide supply and demand conditions. This leaves the crops sugar cane, maize, wheat, rice, sugar beet and soybeans. Sugar cane and sugar beets are first processed into sugar or ethanol before they are traded. The other commodities are traded in their raw state or in the case of soybeans also as the products soybean oil and soybean meal. These are the subjects of my research.

However, another caveat has to be made. Agriculture remains one of the few bastions of protectionism. This has as a consequence that sometimes world market prices cannot be seen as very informative about market or macroeconomic conditions, but rather as the result of trade barriers. For the commodities under investigation, this is especially so for the sugar market, which is one of the most distorted commodity markets in the world due to policy interventions (Elobeid and Beghin, 2006). The World Bank therefore reports three sugar prices, the world market price, the EU-price and the US-price. As can be seen in figure 1, the US price and the world price show similar movements, although the levels differ.

Source: World Bank (2011)

0 10 20 30 40 50 60 70 80 2 0 0 4 M 0 1 2 0 0 4 M 0 4 2 0 0 4 M 0 7 2 0 0 4 M 1 0 2 0 0 5 M 0 1 2 0 0 5 M 0 4 2 0 0 5 M 0 7 2 0 0 5 M 1 0 2 0 0 6 M 0 1 2 0 0 6 M 0 4 2 0 0 6 M 0 7 2 0 0 6 M 1 0 2 0 0 7 M 0 1 2 0 0 7 M 0 4 2 0 0 7 M 0 7 2 0 0 7 M 1 0 2 0 0 8 M 0 1 2 0 0 8 M 0 4 2 0 0 8 M 0 7 2 0 0 8 M 1 0 2 0 0 9 M 0 1 2 0 0 9 M 0 4 2 0 0 9 M 0 7 2 0 0 9 M 1 0 2 0 1 0 M 0 1 2 0 1 0 M 0 4 2 0 1 0 M 0 7 2 0 1 0 M 1 0

Figure 1. Sugar prices in US dollars 2004-2010

(13)

12

This is because the US only has import tariffs and tariff-rate quota in place and because the US is a net importer of sugar (Elobeid and Beghin, 2006). The EU not only has import tariffs and import quota, but also export subsidies, price support and production quota. This has the effect that the EU is both a large importer (from most-favored nations such as the ACP-countries) and a large exporter, by dumping sugar on the world market with the use of export subsidies (Elobeid and Beghin, 2006). The EU-price therefore has its very own dynamic. In normal times, the US- and EU-price are higher than the world market price. But starting in October 2009, the world market price has been higher than the EU price. This followed the drop in 2008 of EU prices due to the new common market organization of the EU sugar market, where some barriers were removed (European Commission, 2007) but export restrictions are still in place. European farmers are protected from imports, but recently also pay the price as they cannot export for the world market price outside the EU. Since the aim here is to explain commodity prices on world markets, I use the world market price although I am aware of its limitations. This is in line with other researchers that study the sugar price (e.g. Sariannidis, 2010; Frankel, 2006). For the other commodities, there is a consensus what the world market price is, normally the export price of the largest producing region. These prices are reported by the World Bank (2011) as world prices.

Commodity data

(14)

13

Table 2. Data sources Annual data

Commodity Price data Stocks Production/consumption Biofuel

Maize World Bank (2011) USDA PSD (2011) USDA PSD (2011) USDA ERS (2011)

Wheat World Bank (2011) USDA PSD (2011) USDA PSD (2011) n.a.

Soybeans World Bank (2011) USDA PSD (2011) USDA PSD (2011) n.a.

Soymeal World Bank (2011) USDA PSD (2011) USDA PSD (2011) n.a.

Soyoil World Bank (2011) USDA PSD (2011) USDA PSD (2011) USDA PSD (2011)

Sugar World Bank (2011) USDA PSD (2011) USDA PSD (2011) World Bank (2011)

GDP Growth

IMF World Economic Outlook (2011a) Constant prices, advanced and emerging economies

Interest rates

IMF International Financial Statistics (2011b)

3-Month T-Bill, constant maturity

Monthly data

Commodity prices World Bank (2011)

Freight price rates Kilian (2009) Freight price rates for dry cargo

Interest rates Federal Reserve (2011) 3-Month T-bill, secondary market rate

Notes: USDA PSD = United States Department of Agriculture Production, Supply and Distribution dataset. USDA ERS = United States Department of Agriculture Economic Research Service.

Biofuel data

Biofuel production consists of two types, ethanol and biodiesel. In 2009, 81% of all biofuels produced was ethanol, the remaining 19% was biodiesel (US Department of Energy, 2011). Ethanol is mostly produced in the United States and Brazil. Together, they accounted for approximately 89% of world production of ethanol in 2008 (US Department of Energy, Biomass Energy Data Book, 2010, p.34). In the United States, ethanol production uses corn as its primary feedstock, accounting for approximately 98% of ethanol produced in 2009 (US Department of Energy, Biomass Energy Data Book, 2010, p.38). Data on corn inputs in biofuel production in the US were obtained from the US Department of Agriculture Economic Research Service (2011) and are available from 1985 to 2010.

(15)

14

and do react swiftly to the relative prices of sugar and ethanol, making the price movements of both cointegrated (Rapsomanikis and Hallam, 2006). Only limited data on Brazilian ethanol production is available from the Brazilian Sugarcane Industry Association (UNICA, 2011). I therefore opt to use the crude oil price in the regression. The links between sugar-, ethanol- and crude oil prices in Brazil are found to be strong. Rapsomanikis and Hallam (2006) find that oil prices cause both ethanol and sugar prices for the period 2000-2006. It is therefore useful to include it as an explanatory variable in the regression for the sugar price, because the link between the sugar and crude oil market runs through the substitutability in production of sugar and ethanol and the substitutability in consumption of ethanol and oil. The inclusion of the oil price for the study of the sugar price is used in other research as well (e.g. Tokgoz and Elobeid, 2006). The oil price data comes from the World Bank Global Economic Monitor (GEM) Commodities (World Bank, 2011).

Biodiesel is produced from vegetable oils and animal fats and production can use various inputs. In this research, only soyoil is studied. In the United States, soybean oil is the largest feedstock for biodiesel production (US Department of Energy, 2011). Soybean oil normally is used almost entirely for human consumption, only recently has industrial use of soybean oil increased due to biofuel production. The United States Department of Agriculture supplies data on industrial use of soybean oil, which is taken as a proxy for biodiesel input, since virtually no other industrial uses of soybean oil exist. Although the US is not the biggest producer of biodiesel, it is the largest country that produces biodiesel from soyoil (the others are Brazil and Argentina) and the only one from which data are available.

Macroeconomic data

(16)

15

constructs a monthly index of global real economic activity based on dry cargo freight prices. Increased economic activity leads to higher demand for shipping services, which leads to higher freight prices. The dry cargo freight price index includes transport price rates for iron ore, coal, fertilizer, scrap metal and, importantly, grains and oilseeds and is therefore especially fitted to capture demand shifts in global markets for industrial and agricultural commodities. Shipping indices such as The Baltic Dry Index are used more often to proxy for increasing economic activity, Groen and Pesenti (2008) for instance include The Baltic Dry Index in their effort to forecast commodity prices. The Kilian index is available for the period 1968-2010 and is constructed using dry cargo single voyage freight prices available from Drewry Shipping Consultants Ltd. Fixed effects for different routes are eliminated and the series is deflated by US CPI and detrended to control for advances in shipbuilding (see Kilian (2009) for more details).

(17)

16

4. Methodology

In this section, I will describe the models and methods of both the long-run estimation and the short-run estimation. The long-run estimation uses the annual data as described in the previous section, the short-run estimation uses the monthly data.

Model of the long run estimation

For reasons of data availability, the annual regression is estimated for 1980-2010 for all commodities. For the annual data, the equation to be estimated is:

log log log log

log !" #log !$ %log& '

Where log is the natural logarithm of the world price of commodity x at time t and is a constant. log is the natural logarithm of the world’s inventory level. The inventory level of a certain year is calculated by dividing the ending stock of the commodity x at time t-1 (which is the beginning stock at time t) by the consumption of commodity x at time t: ()*+,_+*1(-./0

-0 .

log is the natural logarithm of total consumption use of commodity x at time t.

log is the natural logarithm of production of biofuel from commodity x at time t.

The biofuel term of the regression is only included for corn, soybean oil and sugar (where it is proxied by the oil price) since only these three commodities are used for biofuel production. log !" and log !$ are the natural logarithms of the GDP growth rates for advanced

(18)

17

normally distributed, some prices show some skewness to the right. This is reduced when the natural log is taken. This is in line with other researchers that use regression analysis to study commodity prices (e.g. Frankel and Rose, 2010; Baffes, 2007)

Method of the long run estimation

As a starting point, I will estimate the long run regression of the annual data by OLS. As table 3 shows, this is an approach commonly used in research of commodity prices when annual data are used. There is a number of researchers that investigates agricultural commodities behavior on financial markets with high frequency data. Using GARCH models they investigate return correlations and spot and future prices (e.g. Carpantier, 2010; Perales, 2010). Since this is not the subject of this study, table 3 only reports relevant papers.

Table 3. Methodologies used in commodity price research

Paper Method

Deaton and Laroque (1992) Generalized Method of Moments

Chambers (1996) Generalized Method of Moments

Frankel (2006) OLS

Baffes (2007) OLS

Frankel and Rose (2010) OLS (robust)

Saadi-Sedik and Cevik (2011) OLS + Generalized Method of Moments

(19)

18

Model of the short run estimation

To test if macroeconomic conditions or perhaps loose monetary policy can explain the price behavior in the short run, I specify the following model for the monthly data. Autocorrelation plots showed that there is significant autocorrelation in the monthly prices of the commodities of 6 to 9 months, depending on the commodity. I therefore specify an ARMAX model, an autoregressive moving average with exogenous inputs model (Box, Jenkins and Reinsel, 2008). This captures the dynamic aspect of the data, while it is still able to incorporate explanatory variables. 2 & 3 4 5 46 7₃ 4 589 4658 :7_'

Where is the price of commodity x at time t. 2 is the freight rate at time t and & is the interest rate at time t. ∑ 5₄₆ ₄7 specifies p autoregressive terms, where p is determined for each commodity by plotting the partial autocorrelation functions and testing which lags are significant. ∑589₄₆₅₈₄:7 specifies q moving average terms, where q is determined for each commodity by plotting the autocorrelation function and testing which lags are significant. This ARMA process is applied to all variables in first differences, an approach mentioned by Deaton and Laroque (1992). I introduce the explanatory variables to this process which makes it an ARMAX model. The betas are parameters to be estimated and ' is an error term. This regression on monthly observations can estimate the influence of

worldwide demand growth, as proxied by the dry cargo freight rates and liquidity conditions, as proxied by the prevailing interest rate whilst simultaneously acknowledging that trending, mean-reversion and unobserved shocks can be of influence.

Method of the short run estimation

(20)

19

5. Empirical results

Descriptive statistics

Before running the regressions, it is useful to have a look at the data. Figure 2 shows the price developments of maize, wheat and soybeans. The prices of the other commodities follow a similar pattern but are not shown for illustrative reasons. What immediately springs into attention are the large price rises of the early 1970s, which coincided with the 1973 oil crisis and a boom in mineral commodities. Factors mentioned to have been of influence then are also mentioned for the present boom, including low inventory levels, high-energy intensity of production and low interest rates. Since the mid 1970s, prices have declined during the 1980s and remained stable at low levels in the 1990s and early 2000s. In 2005 however, prices began to rise again, and at a very fast pace. The price rise looks less pronounced than in the 1970’s, but rising from an index of around 50 to well over 100 is a more than doubling of prices, even larger than in the early 1970s. What is noteworthy is that prices have a large upward volatility and little downward volatility (Cashin and McDermott, 2002).

Source: World Bank (2011) 0 50 100 150 200 250 1 9 6 0 1 9 6 2 1 9 6 4 1 9 6 6 1 9 6 8 1 9 7 0 1 9 7 2 1 9 7 4 1 9 7 6 1 9 7 8 1 9 8 0 1 9 8 2 1 9 8 4 1 9 8 6 1 9 8 8 1 9 9 0 1 9 9 2 1 9 9 4 1 9 9 6 1 9 9 8 2 0 0 0 2 0 0 2 2 0 0 4 2 0 0 6 2 0 0 8 2 0 1 0

Figure 2. Index of agricultural commodity prices 1960-2010

(21)

20

The price developments for the shorter time span, from 2000-2010, can be seen in figure 3. Sugar is left out of this figure, but was discussed before (see figure 1). Prices rise somewhat from 2000 until 2006. Then prices rise dramatically until 2008, when the financial crisis hit the markets. In 2009 prices start to rise again and have continued to rise until spring 2011, when they stabilized at high levels.

Source: World Bank (2011)

What is interesting is the strong correlation between the prices of the different commodities. Table 4 shows the pairwise correlations for the commodities for the period 2000-2010, based on monthly data. Logically, the highest correlations are between soybeans and its products soymeal and soybean oil. But the correlations for the other commodities are also very high, with the exception of sugar. This can be due to the fact that wheat, corn and soybeans are to a large extent substitutable as human or animal feed, and sugar is not. Also, as mentioned before, the sugar market has its very own dynamic with extensive government involvement.

0 50 100 150 200 250 300 350 2 0 0 0 M 0 1 2 0 0 0 M 0 5 2 0 0 0 M 0 9 2 0 0 1 M 0 1 2 0 0 1 M 0 5 2 0 0 1 M 0 9 2 0 0 2 M 0 1 2 0 0 2 M 0 5 2 0 0 2 M 0 9 2 0 0 3 M 0 1 2 0 0 3 M 0 5 2 0 0 3 M 0 9 2 0 0 4 M 0 1 2 0 0 4 M 0 5 2 0 0 4 M 0 9 2 0 0 5 M 0 1 2 0 0 5 M 0 5 2 0 0 5 M 0 9 2 0 0 6 M 0 1 2 0 0 6 M 0 5 2 0 0 6 M 0 9 2 0 0 7 M 0 1 2 0 0 7 M 0 5 2 0 0 7 M 0 9 2 0 0 8 M 0 1 2 0 0 8 M 0 5 2 0 0 8 M 0 9 2 0 0 9 M 0 1 2 0 0 9 M 0 5 2 0 0 9 M 0 9 2 0 1 0 M 0 1 2 0 1 0 M 0 5 2 0 1 0 M 0 9

Figure 3. Index of agricultural prices 2000-2010

(22)

21

Table 4. Correlations of prices in levels (2000-2010)

Corn Wheat Soybean Sugar Soymeal Soyoil Rice

Corn 1 - - - - Wheat 0.84 1 - - - - - Soybean 0.90 0.83 1 - - - - Sugar 0.44 0.30 0.45 1 - - - Soymeal 0.86 0.76 0.96 0.52 1 - - Soyoil 0.93 0.89 0.95 0.42 0.88 1 - Rice 0.86 0.74 0.84 0.47 0.82 0.88 1 Stationarity tests

An important assumption is stationarity of the time series. Commodity consumption and biofuel production are also likely to be non-stationary and show an upward trend. Therefore before running the regressions, the time series are subjected to an augmented Dicky-Fuller test to determine if they are stationary. The results of the stationarity tests can be found in appendix A. Most variables were nonstationary in levels. Therefore, I took the first difference of these variables and in all cases these proved to be stationary. Some variables were already stationary in levels, then they are included in levels in the long run regression. For the short run regression, it is also important that the series are stationary. Therefore augmented Dickey-Fuller tests were performed. These results can be found in appendix B. All series with monthly data were non-stationary at level, but stationary when the first difference was taken. Therefore in the ARMAX estimation, the first difference is used.

Multicollinearity

(23)

22

slight changes. Therefore, I conclude that multicollinearity is not a problem and the results with the inclusion of all explanatory variables will be presented below.

Table 5.a Long run regression results 1980-2010 Wheat price

Variable OLS GMM

Coefficiënt p-value Coefficiënt p-value Inventory level -0.421 0.033** -0.375 0.041** Total consumption -2.409 0.065* -2.676 0.289 GDP growth advanced 0.026 0.640 0.030 0.345 GDP growth emerging 0.053 0.491 0.064 0.368 Interest rates -0.079 0.396 -0.076 0.218 Constant -0.513 0.058* -0.0463 0.050** Adjusted R-squared 0.10 0.09 Corn price Variable OLS GMM

Coefficiënt p-value Coefficiënt p-value Inventory level -0.570 0.004*** -0.813 0.000*** Total consumption -2.896 0.031** -3.233 0.041** GDP growth advanced -0.072 0.232 -0.167 0.002*** GDP growth emerging 0.079 0.357 0.134 0.001*** Interest rates 0.011 0.912 0.108 0.149 Ethanol production -0.069 0.686 0.504 0.026** Constant 0.136 0.109 0.173 0.086* Adjusted R-squared 0.22 0.20 Soybean price Variable OLS GMM

Coefficiënt p-value Coefficiënt p-value

Inventory level 0.112 0.310 -0.08 0.556 Total consumption -1.163 0.153 -3.150 0.178 GDP growth advanced -0.019 0.677 0.011 0.741 GDP growth emerging 0.137 0.068* 0.085 0.399 Interest rates -0.048 0.567 -0.020 0.855 Constant 0.253 0.188 0.128 0.216 Adjusted R-squared 0.16 0.19 Soymeal price Variable OLS GMM

(24)

23

Table 5.b Long run regression results 1980-2010 Soyoil price

Variable OLS GMM

Coefficiënt p-value Coefficiënt p-value Inventory level -0.073 0.787 0.067 0.777 Total consumption -0.359 0.746 -1.199 0.166 GDP growth advanced -0.081 0.393 -0.079 0.126 GDP growth emerging 0.142 0.265 0.130 0.076* Interest rates 0.037 0.820 0.067 0.526 Biodiesel production -0.308 0.176 -0.325 0.021** Constant 0.133 0.223 0.155 0.016** Adjusted R-squared -0.02 0.16 Sugar price Variable OLS GMM

Inventory level 0.561 0.156 0.224 0.528 Total consumption 0.857 0.030** 0.809 0.006*** GDP growth advanced 0.064 0.513 0.102 0.114 GDP growth emerging -0.028 0.861 0.053 0.683 Interest rates 0.185 0.308 0.161 0.174 Oil price -0.203 0.441 -0.135 0.595 Constant -4.153 0.027** -3.973 0.004*** Adjusted R-squared 0.10 0.24 Rice price Variable OLS GMM

Coefficiënt p-value Coefficiënt p-value Inventory level 0.182 0.074* 0.108 0.224 Total consumption 0.498 0.271 0.158 0.388 GDP growth advanced 0.043 0.590 -0.064 0.538 GDP growth emerging -0.11 0.322 -0.08 0.888 Interest rates -0.058 0.660 0.122 0.144 Constant -3.210 0.055* -1.008 0.387 Adjusted R-squared 0.01 0.03

(25)

24

Results of the long run estimation

For all seven commodities, the results can be found in table 5a and table 5b. I will first discuss the differences between the OLS and GMM estimation, before focusing on the most important results of the GMM estimation. In most cases, the results of the GMM estimation are similar to the OLS estimation. The coefficient estimates do not differ much in most estimations and the signs are the same in all but three cases. An important difference between the OLS and GMM estimation is the effect of ethanol production from corn. In the OLS estimation this is insignificant and the coefficent has a negative sign, in the GMM estimation it is significant and the parameter has a positive sign. This last result support hypothesis 3, that more biofuel production from corn leads to higher prices. In the GMM estimation, consumption of corn for ethanol production shows a coëfficient of 0.504, indicating that a 1% increase in ethanol production from corn yields a 0,5% price rise. Given the fact that production of ethanol from corn has grown very fast since 2004, this can explain part of the price rise observed and is in line with the evidence from Perrin (2008). My estimate however is higher than found by Perrin and other researchers he reviews in his study. For corn, the GMM estimation also shows significant values for GDP growth in advanced and emerging countries, which are insignificant in the OLS estimation. Finally, the GMM estimation for soyoil shows three significant variables, whereas the OLS estimation shows no significant variables. The model as a whole for soyoil is also insignificant when the OLS estimation is used, therefore I focus on the GMM results.

(26)

25

support for hypothesis 3, that biofuel production and other consumption growth fueled by emerging economies contributed to price rises. The result for soyoil shows that emerging country growth has a significant and positive effect on soyoil prices, supporting hypothesis 3, but biodiesel production from soyoil has a significant negative effect, contradicting hypothesis 3.

The results for soybean and rice show no significant variables. For soybeans, this may be due to the fact that soybeans often are crushed before they are traded, by that process becoming soymeal and soyoil. Inventories and consumption of soybeans than do not say much, as one should also include inventories of soymeal and soyoil before something can be said about the soymarket as a whole. In a previous section, the peculiarities of the rice market were already considered. Only a very small volume of total rice production is traded internationally. The GMM regression shows no significant variables and the model has very little predictive power. Government policy exerts considerable influence over the international market as export restrictions are often in place to ensure food security (Piesse and Thirtle, 2009). The determinants of the rice market warrant further investigation tailored to its specific circumstances.

Finally, the results do not support hypothesis 4, that monetary policy working through interest rates influences commodity prices. In none of the results is the interest rate significant. This also means that low interest rates per se do not, via speculation fueled by liquidity chasing assets with higher returns, cause higher agricultural commodity prices. So when you are purely looking at the interest rate, this is not supportive for hypothesis 1, that speculation caused by easy liquidity conditions causes price rises in the long run. Perhaps it has an effect on the short run, to which we turn now.

Results of the short run estimation

The results of the estimation for the short run for all six commodities can be found in table 6. For all variables, the partial autocorrelation function (results not included) suggested including 2 lagged autoregressive terms. The number of moving average terms differed per commodity and was determined by the number of lags that were significant in the autocorrelation function of the prices. Depending on the commodity, this ranges from lags from 6 to 9 months. All variables are included at first differences in order to attain stationarity

of the time series.

(27)

26

the dry cargo freight rate, influence agricultural commodity prices. Given the peculiarities of the sugar and rice market, as outlined before, it is not surprising that no significant results are found for these prices. This result for corn, wheat and soy products is in line with the finding of Kilian (2009) for the oil price. Using industrial production indices instead of freight rates, Saadi-Sedik and Cevik (2011) also find that global demand conditions are significant in explaining oil and fine wine prices. These results therefore lend support to hypothesis 3; in the short run demand conditions have a significant effect on prices. Interestingly, the interest rate is insignificant for all commodities. This is supportive for neither hypotheses 1 nor hypothesis 4, which predict that monetary influences or speculation are of influence. This result is in line with the results from Frankel and Rose (2010) for agricultural and mineral commodities. It is not in line with Saadi-Sedik and Cevik (2011), who find that in the short run liquidity conditions are of influence in commodity prices. However, they use a measure of excess liquidity in their regression and not the interest rate. It is possible that the interest rate as a measure of liquidity conditions is a flawed indicator on speculative pressures or monetary conditions. More conclusive evidence of the role of speculation on the price of agricultural commodities has to await future research.

Table 6. Short run regression results 2000-2010

Variable Corn Wheat Soybeans Soymeal Soyoil Sugar Rice

Freight rate 0.19 (0.06)* 0.39 (0.03)** 0.84 (0.00)*** 0.72 (0.00)*** 2.16 (0.00)*** -0.02 (0.35) 0.22 (0.54) Interest rate 1.04 (0.79) -5.42 (0.52) 2.66 (0.72) -0.29 (0.98) 0.41 (0.97) 1.57 (0.15) -8.49 (0.29) No. AR terms 2 2 2 2 2 2 2 No. MA terms 8 7 8 9 8 8 6 P>Χ² 0.00 0.00 0.00 0.00 0.00 0.00 0.00

(28)

27

6. Conclusion

The recent developments in the agricultural commodity markets spurred a wave of new research into the determinants of commodity prices. Numerous explanations were investigated as drivers of commodity prices. Previous research found evidence for the role of speculators driving up prices and creating asset bubbes as well as for the role of biofuel production, which gave birth to the food versus fuel debate. The role of strong economic growth in emerging countries, which goes together with a rising demand for commodities was also found significant in previous research. Moreover, some argue that monetary policy is of great influence for commodity prices. Most of this research looks at one of these factors in isolation, I take all these explanations together in a framework for the long run. In the short run, I look at the role of the interest rate and global demand conditions.

(29)

28

production. This might also explain why the result for soyoil is negative. Biodiesel has known a slower start than ethanol in terms of usage and the data available now are perhaps of a too short horizon to pick up the effects.

The results from the short-run analysis also support the third hypothesis. Worldwide demand conditions, as proxied by the dry cargo freight rate, are in the short run significant. Financial observers are right to look at shipping indices such as the Baltic Dry index as an indicator for worldwide demand growth and increasing trade. For investors in agricultural commodities, it might be worthwhile to do the same.

Summing up, this thesis finds that in the long run fundamentals rule. Pressures on supply and demand are of influence, which in the end work through inventory levels. Buoyant economic growth in emerging economies adds to the demand side pressure for those commodities used as animal feed. In the short run, booms and slumps in world economic growth influence price behavior. In both the long and short run, interest rates do not seem to play a role when other factors are considered.

(30)

29

References

Baffes, J. 2007, Oil spills on other commodities, Resources Policy, vol. 32, no. 3, pp. 126-134.

Barsky, R.B. & Kilian, L. 2001, Do We Really Know That Oil Caused the Great Stagflation? A Monetary Alternative, NBER Macroeconomics Annual, vol.16, pp. 137-183

Box, G.E.P., Jenkins, G.M. & Reinsel, G.C. 2008, Time series analysis: forecasting and control, 4th edition, Wiley, Hoboken, New Jersey.

Cabellero, R., Farhi, E. & Gourinchas, P. 2008, Financial crash, commodity prices and global imbalances, NBER Working Paper Series, Working Paper 14521.

Cashin, P. & McDermott, C.J. 2002, The Long-Run Behavior of Commodity Prices: Small Trends and Big Variability, IMF Staff Papers, vol. 49, no. 2, pp. 175-199.

Carpantier, J.F. 2010, Commodities Inventory Effect, Center for Operations Research and Econometrics, CORE Discussion Paper 2010/40

Chambers, M.J. & Bailey, R.E. 1996, A Theory of Commodity Price Fluctuations, Journal of Political Economy, vol. 105, no.5, pp. 924-957

Chinn, M & Frankel, J. 2005, The Euro Area and World Interest Rates, Harvard University and University of Wisconsin (Madison), Unpublished manuscript.

Deaton, A. & Laroque, G. 1992, On the Behaviour of Commodity Prices, Review of Economic Studies, vol. 59, no. 198, pp. 1-23

Elobeid, A. & Beghin, J. 2006, Multilateral Trade and Agricultural Policy Reforms in Sugar Markets, Journal of Agricultural Economics, vol. 57, no. 1, pp. 23-48

European Commission. 2007, Communication from the Commission concerning proposals to amend Council Regulation (EC) No 318/2006 on the common market organization of the markets in the sugar sector and Council Regulation (EC) No 320/2006 establishing a temporary scheme for the restructuring of the sugar industry in the Community, Commission of the European Communities, Brussels.

FAO, 2011. FAO Statistics database, http://faostat.fao.org/site/567/default.aspx#ancor , accessed on 02/05/2011.

Federal Reserve, 2011. Federal Reserve Economic Data (FRED),

http://research.stlouisfed.org/fred2/categories/116, downloaded on 09/05/2011.

Frankel, J. 2006, The Effect of Monetary Policy on Real Commodity Prices, NBER Working Paper Series, Working Paper 12713.

(31)

30

Groen, J.J. & Pesenti, P.A. 2010, Commodity Prices, Commodity Currencies, And Global Economic Developments, NBER Working Paper Series, Working Paper 15743 Hertel, T.W. & Beckman, J. 2011, Commodity Price Volatility in the Biofuel Era: An

Examination of the Linkage Between Energy and Agricultural Markets, NBER Working Paper Series, Working Paper 16824.

Hotelling, H. 1931, The Economics of Exhaustible Resources, The Journal of Political Economy, vol. 39, no. 2, pp. 137-175.

Kaldor, N. 1939, Speculation and Economic Stability, Review of Economic Studies, vol. 7, pp. 1-27.

Kilian, L. 2009. Not All Oil Price Shocks Are Alike: Disentangling Demand and Supply Shocks in the Crude Oil Market, American Economic Review, vol.99, no.3, pp. 1053-1069

IMF. 2011a, World Economic Outlook database April 2011,

http://www.imf.org/external/pubs/ft/weo/2011/01/weodata/index.aspx, downloaded on 10/05/2011

IMF, 2011b. International Financial Statistics database, http://www.imfstatistics.org/imf/ , downloaded on 09/05/2011

OECD-FAO. 2007, Agricultural Outlook 2007-2016, OECD-FAO, Paris.

Perales, G.B. 2010, The Theory of Storage and Price Dynamics of Agricultural Commodity Futures: the Case of Corn and Wheat, Ensayos, vol. 29, no.1, pp. 1-12

Perrin, K. 2008, Ethanol and Food Prices - Preliminary Assessment, University of Nebraska - Lincoln, Faculty Publications: Agricultural Economics

Piesse J. & Thirtle, C. 2009, Three bubbles and a panic: An explanatory review of recent food commodity price events, Food Policy, vol. 34, pp. 119-129.

Pindyck, R.S. & Rotemberg, J.J. 1990, The Excess Co-Movement of Commodity Prices, Economic Journal, vol. 100, no. 403, pp. 1173-1189.

Rapsomanikis, G. & Hallam, D. 2006, Treshold cointegration in the sugar-ethanol-oil price system in Brazil: evidence from nonlinear vector error correction models, FAO, Commodity and Trade Policy Research Working Paper No. 22

Saadi-Sedik, T. & Cevik, S. 2011, A Barrel of Oil or a Bottle of Wine: How Do Global Growth Dynamics Affect Commodity Prices?, International Monetary Fund, IMF Working Papers: 11/1.

(32)

31

Samuelson, P.A. 1971, Stochastic Speculative Price, Proceedings of the National Academy of Sciences, vol. 68, no. 2, pp. 335-37

Sarianidis, N. 2010, The Impact of Stock, Energy and Foreign Exchange Markets on the Sugar Market, International Journal of Economic Sciences and Applied Research, vol. 3, no.1, pp. 109-117

Tang, K. & Xiong, W. 2010, Index investment and financialization of commodities, NBER Working Paper Series, Working Paper 16385.

The Economist. Feb 10th 2011, Wheat prices: grim reaping, Print edition, The Economist. The Economist. Apr 17th 2008, The new face of hunger: Global food shortages have taken

everyone by surprise. What is to be done? Print edition, The Economist.

Tokgoz, S. & Elobeid, A. 2006, An Analysis of the Link between Ethanol, Energy, and Crop Markets, Center for Agricultural and Rural Development Iowa State University,

Working Paper 06-WP 435

Timmer, C.P. 2009, Did Speculation Affect World Rice Prices?, FAO, ESA Working Paper No.09-07

UNICA (Brazilian Sugarcan Industry Association). 2011,

http://english.unica.com/br/dadosCotacao/estatistica/, accessed on 05/04/2011. US Department of Agriculture. 2011, Production, Supply and Distribution dataset,

http://www.fas.usda.gov/psdonline/, accessed on 04/04/2011.

US Department of Agriculture Economic Research Service. 2011, Feed Grains Database, http://www.ers.usda.gov/Data/FeedGrains/ , accessed on 04/04/2011.

US Department of Energy. 2010, Biomass Energy Data Book: Edition 3, http://cta.ornl.gov/bedb/index.html, downloaded on 07/04/2011.

US Department of Energy. 2011, http://www.afdc.energy.gov/afdc/fuels/index.html, accessed on 07/04/2011.

US Senate Permanent Subcommittee on Investigations. 2009, Excessive speculation in the wheat market, Committee on Homeland Security and Governmental Affairs.

Westhoff, P.C. 2010, The Economics of Food: how feeding and fueling the planet affects food prices, 1st edition, FT Press, New Jersey.

(33)

32

Appendix A. Stationarity tests for annual data in logarithms

Variable Measure Test statistic MacKinnon approx.

p-value

Result

Corn prices Price level -1.747 0.4068 I(1)

First difference -6.411 0.0000

Corn total consumption Level -0.872 0.7969 I(1)

Corn inventory Level -2.077 0.2540 I(1)

Ethanol consumption Level -1.742 0.4097 I(1)

Wheat prices Price level -2.051 0.2648 I(1)

Wheat total consumption Level -2.551 0.1036 I(1)

Wheat inventory Level -3.545 0.0069

Soybean prices Level -1.879 0.3419 I(1)

Soybean total consumption Level -2.148 0.2257 I(1)

Soybean inventory Level -3.720 0.0038

Soymeal price Level -2.240 0.1921 I(1)

Soymeal total consumption Level -2.040 0.2691 I(1)

Soymeal inventory Level -2.693 0.0752 I(1)

Soyoil price Level -2.670 0.0794 I(1)

Soyoil total consumption Level -1.216 0.6667 I(1)

Soyoil biodiesel Level 1.773 0.9983 I(1)

Soyoil inventory Level -2.242 0.1913 I(1)

Sugar price Level -2.838 0.0530 I(1)

Sugar total consumption Level -3.492 0.0082

Sugar inventory Level -2.586 0.0958 I(1)

Oil price (Brendt) Level -0.967 0.7651 I(1)

Rice price Level -2.385 0.1461 I(1)

Rice total consumption Level -2.944 0.0405

Rice inventory Level -3.029 0.0323

GDP Advanced Growth rate -3.603 0.0057

GDP Emerging Growth rate -2.849 0.0516 I(1)

Interest rate Level 0.312 0.9779 I(1)

(34)

33

Appendix B. Stationarity tests for monthly data

Variable Measure Test statistic MacKinnon approx. p-value

Result Corn prices Price level -0.496 0.8929 I(1)

Wheat prices Level -1.499 0.5341 I(1) First difference -9.204 0.0000

Soybean prices Level -0.786 0.8233 I(1) First difference -8.135 0.0000

Soymeal prices Level -1.039 0.7386 I(1) First difference -8.951 0.0000

Soyoil prices Price level -0.271 0.9296 I(1) First difference -7.234 0.0000

Sugar prices Level 0.174 0.9708 I(1) First difference -7.351 0.0000

Rice prices Level -1.633 0.4661 I(1) First difference -5.678 0.0000

Freight rate Level -1.951 0.3085 I(1) First difference -7.083 0.0000