University of Groningen
Faculty of Economics and Business
Master thesis
The role of speculations in the surge of the corn, wheat and
soybeans prices
Groningen 2011
Supervisors:
D.J. (Dirk) Bezemer
Prof. Dr. habil. Csaba Forgács Student name: Borbála Nóra Szököl (s2050447)
1
The role of speculations in the surge of the corn, wheat and
soybeans prices
Abstract
A rapid surge in the international prices of agricultural commodities was observed between 2006 and mid 2008. Because the rising food prices threaten the food security world attention has turned to the food crisis. The aim of this thesis is to analyse the relationship existing between the agricultural commodity prices and potential driving factors during the period of January 2002–December 2010 by using a time-series econometrics and data at monthly frequency. Special attention is placed on the financial activities in futures markets. Overall, my empirical analysis provides evidence that high oil prices and financial activities in futures markets can help explain the observed surge in agricultural commodity prices.
2
Table of content
1. Introduction ... 3
2. Literature Review ... 6
2.1. What might be the main factors of the surging food prices? ... 6
2.2 The role of financial activities ... 8
3. Methodology ... 14 3.1. Variables ... 14 3.1.1. Independent variables ... 14 3.2. The model ... 17 4. Results... 20 4.1. Summary Statistic ... 20 4.2. Wheat prices ... 24 4.3. Corn prices ... 27 4.4. Soybean prices ... 29 4.5. Average prices ... 31 4.6. Discussion ... 33 5. Conclusion ... 35 References... 37
Appendix A - Price volatility ... 42
Appendix B - Extra Tables ... 56
Appendix C – ADF test ... 59
3
1. Introduction
After the 1973/74 food crisis the food prices drastically fell in real terms until 2000. However, an increase in international food prices has been observed since 2002, especially since late 2006. In many cases the prices of agricultural commodities more than doubled within a few years (see Figure 1). Moreover, the food prices tend to be more volatile which can exacerbate the effects of high food prices.
Figure 1
Source: FAO International Commodity Price Database
4
Experts give several explanations for the observed run-up in agricultural commodities prices. However, one factor may not be the sole cause of the upward trend, furthermore some authors such as Trostle (2008) and Headey and Fan (2008) notice the “perfect storm” hypothesis, which means a conflagration and interaction among the main factors. While some explanations stress the role of the supply factors others enhance the demand-size changes. On the one hand the supply-side factors may be the high oil prices, low R&D investments in the agriculture, export restrictions and adverse weather (Abbott, et al., 2008; Mitchell, 2008, OECD, 2008a; Trostle, 2008; von Braun, 2007).
On the other hand the main demand driven factors may be rising demand for a variety of food products in developing countries (Mitchell, 2008; Trostle, 2008), increasing biofuels production (Mitchell, 2008; OECD, 2008a; Trostle, 2008; von Braun, 2007). Furthermore, some authors argue that the dollar devaluation and the easy monetary policy in the US were important causes of the increased food prices (Abbott, et al., 2008; Mitchell, 2008). Additionally, some experts emphasize that the increased future markets’ activities might have a significant role in the price surge in 2006–2008 (Abbott, et al., 2008; Irwin, 2008; Mitchell, 2008; Robles, et al., 2009). However, others find that there is a big hole in this “bubble” theory, which claims futures market activity had a severe impact on the agricultural prices and may have played a significant role in the price run-up (Irwin, et al., 2009; Headey and Fan, 2010; Irwin and Sanders, 2010).
The aim of my paper is to validate empirically the relative magnitude of the above mentioned potential factors in the recent surge in food prices. Furthermore, during my research I would like to put special emphasis on the role of the futures markets in the high food prices, especially in the high corn, wheat, soybeans prices and their average.
The objective of this paper is to answer the main research question:
5
Moreover, I also try to answer the following sub-questions during my research:
“Does the growing demand agricultural commodities from the developing countries affect positively the agricultural spot prices?”
“Do the oil prices push up the agricultural commodity prices through the production costs?”
“Does the increasing biofuels production contribute to the high agricultural spot prices?” “Does the US monetary policy influence significantly the agricultural spot prices?”
“May a supply shock generated by adverse weather and trade restriction be a cause of the high agricultural prices?”
6
2. Literature Review
Since 2007 researchers have given several explanations for the source of rise in food prices, however, they emphasize different factors which were the sources of the upward trend. However, most experts assert that there might have been more driving factors of the peak in agricultural prices supporting the “perfect storm” theory (Abbott, et al. 2008; Headey and Fan, 2010; Trostle, 2008). In this chapter I summarize these explanations with a special attention to the role of futures markets’ activities.
2.1. What might be the main factors of the surging food prices?
In this section I would like to show in short the relevant literature about what the driving factors of the food prices may be in our days. The literature implies that long/short term and supply/demand side factors can be distinguished. Table 1 summarizes the potential driven factors of the food prices.
Table 1
Supply Factors Demand Factors
Long-term factors -agricultural productivity,
-investment in agriculture
-increasing demand (by rising population and income growth) - biofuels production
Short term factors -high energy (oil) prices
-weather (bad/ good harvest) -export restrictions
- stockholding (short and medium term)
- fluctations of U.S. dollar exchange rate
-speculation, futures markets’ activities
The table is based on Wahl (2009) and Timmer (2009)
7
grains but also an increasing demand for animal feed due to the changing diet in these countries. Wahl (2009) also claims that the rapid economic growth of the emerging countries (e.g. China) generates increasing demands for cereal crops, particularly through the adaption of western consumption habits. This changing diet means that dairy products and meat consumption is growing in these countries. Thus, it generates a growing demand for animal feed grains. In contrast with this theory Headey and Fan (2010) find that the growing demand from the developing countries such as India and China is playing a more indirect role since there is no clear evidence that these countries are becoming more dependent on agricultural commodities exports. In addition, Baffes and Haniotis (2010) do not find any evidence that increased demand from developing countries affected the world food prices.
Furthermore, Trostle (2008) emphasizes the role of high oil prices and increasing biofuels production. Others also claim that the high energy prices can affect the food prices through the supply and the demand channels. High energy prices can cause a surge in fertilizer prices and transportation costs, therefore, increasing oil prices may boost the input costs of agricultural production (Baffes, 2007; Headey and Fan, 2010). Additionally, experts also emphasize that high energy prices can incite the biofuels production which is mostly made from agricultural commodities, therefore, it may affect the food prices (Headey and Fan, 2010; Schmidhuber, 2006). Nevertheless, biofuels production is mainly driven by government policies and strategies (OECD, 2008b).
Headey and Fan (2010) conclude that high energy related production costs might be regarded as a strong factor and biofuels production has an effect on at least corn and likely soybeans markets in U.S. and on wheat prices in the EU. Mitchel (2008) concludes that the most important factor was the biofuels production from oilseeds and grains in the food crisis. He also estimates that 70-75% of the run-up in food prices was due to the biofuels production and its consequences, such as low stocks, exports ban, speculations and land use shift. Moreover, according to his estimation the high energy -, fertilizer - and transportation prices and dollar’s weakness can explain the remaining 25-30% of the price increase. However, Baffes and Haniotis (2010) evolve that biofuels production played a role but less than initially researchers thought.
8
and Fan (2010) discuss that money flows into emerging markets stocks, other securities and commodities when the interest rate is low. Thus, monetary policy may inchoate that real commodity prices rise “more than other prices because other prices are sticky” (Headey and Fan, 2010, pp. 39). Frankel (2006, 2008) also argues that low real interest rate can induce a general increase in commodity prices and he also presents some econometric evidence that there is an inverse relationship between the interest rate and commodity prices in United States. He also emphasizes that data from before and after the run-up in food prices are in the line with this theory and the historical evidence (Frankel, 2008). Although Headey and Fan (2010) mention that the main inconsistency with this theory is that the level of agricultural stocks is low.
Trostle (2008) and Abbot, et al. (2008) plead that decreasing productivity growth, low rates of investment in agriculture and declining stocks can be regarded as principle causes of supply-demand imbalance and the run-up in food prices. The World Bank (2008) reports that the growth rate of yield of wheat, corn and rice has declined. However, Headey and Fan (2010) are sceptical whether productivity decline of agricultural production can cause an increasing pressure on international agricultural prices and declining stocks may be regarded as a principal source of the food crisis.
Dollive (2008), Headey (2010) and Mitra and Josling (2009) conclude that trade shocks such as export restrictions played an important role in the rise in food prices. Headey and Fan (2010) discuss that adverse weather and as a consequence poor harvest might have accounted for a large part of the peak of prices.
Additionally, Mitchell (2008) and Abbot, et al. (2008) also suggest that increased activity in the future markets may have an effect on agricultural prices and can raise the volatility of food prices in short run. Baffes and Haniotis (2010) suspect that index fund activity, which is one type of speculative activities in futures markets, had a key role in the surge of food prices during 2006 – 2008. However, Irwin (2008) is more skeptical about that other factors than economic fundamentals caused the rise in agricultural prices.
2.2 The role of financial activities
9
may have a severe role in the price surge. After that, I highlight the concerns about this “bubble” theory. Moreover, in Appendix B Table 2 summarizes the empirical findings of the experts.
Fabozzi and Modigliani (2003) allege that the major role of the futures markets is the price risk transferring from hedgers to speculators. In other words in the futures market “ risk is transferred from those willing to pay to avoid risk to those wanting to assume the risk as in the hope of gain” (Fabozzi and Modigliani, 2003, pp. 174). Robles, et al. (2009, pp. 2) define speculation as it „ is the assumption of the risk of loss in return for the uncertain possibility of a reward”. Furthermore, in the supply of storage model short-term spot prices’ behaviour for storable commodities is explained by the inter-related behaviour of hedgers and speculators when they judge their levels of stocks in relation to use (Timmer, 2009; Houthakker, 1987; Brennan, 1958). The model emphasizes that the formation of the price expectation has a key-role in this process (Timmer, 2009). Timmer (2009) also notes that the key-role of the outside speculators, who are not participating in the spot commodity markets, is more controversial in the formation of futures prices and price expectations. He asserts that corn, rice and wheat markets are related to financial markets in short run, thus, the surge in the corn and wheat prices was surely caused by financial speculators who flew to the commodities futures markets after the collapse of the real estate markets.
10
Index Funds’ activities in futures markets “was the principal channel through which monetary and financial activity have affected food prices over recent years” (Gilbert, 2010a, pp. 420), although he claims that main driver of index investment has been the rapid Chinese economic growth.
Robles, et al. (2009) analyze whether the speculative activity in the futures markets may be regarded as a principal source of the food price boom in 2007–2008. They use five indicators for the speculations: monthly volume of futures contracts, monthly open interest in future contracts, ratio of volume to open interest in future contracts, positions in futures contracts by non-commercial traders and Index traders’ net positions. With Granger causality test they test what extend these proxies of speculation may help forecast spot price changes. They find some evidences that the ratio of monthly volume to open interest in futures contracts, the positions in futures contracts by non-commercial traders and the Index traders’ net positions have an influence on forecasting price movements for agricultural commodities. Nevertheless, they notice that their results are far from conclusive, although they support the idea that speculations might be influential factor of food prices.
Andreosso-O’Callaghan and Zolin (2010) conducted a research about the linkages among grain prices and several variables such as export, income, population, the exchange rate of dollar and speculation. In their linear regression analysis great emphasis is placed on financial activities of the futures markets. Their dependent variable is the logarithmic of spot cereals price index and they conduct the regression over the period of mid 2001-2009. They find that there is a positive relationship between cereal prices and the long positions of traders. These results support the idea that speculation plays an important role in the surge in grains prices over the examined period. Furthermore, Andreosso-O’Callaghan and Zolin (2010) conclude that speculation is the most relevant explanatory variable in their regression analysis.
Cooke and Robles (2009) also do a time-series analysis about the recent price movements on the wheat, maize, rice and soybeans markets. The aim of their research is to validate the relative importance of the explanations that other –above mentioned- researchers give for the agricultural price increase. They also use more indexes to capture the speculations and they give the following interpretation for the relationship between speculation and prices:
11
market. However, when participation of non-commercial players and speculators increases significantly on both sides of the market, then we argue that speculation has a more active role in driving prices. As a group of speculators enters the market taking long positions betting on higher future spot prices, futures prices might be driven upward to the point of attracting more conservative (with respect to their expectations about future spot prices) non-commercial traders. As this happens in futures markets, signals of higher prices are transmitted to the spot market in such a way that initial expectations are confirmed and provide feedback on further expectations” (Cooke and Robles, 2009, pp. 15). They deduce
that speculation and/or financial activities in futures markets may help explain the recent movements of rice, wheat, maize and soybeans prices. However, they do not find significant evidence which supports other explanations of the run up in prices.
Moreover, Gilbert (2010b) analyses whether high futures prices over 2006-2008 caused by bubble behavior. He find no evidence for bubbles in the futures markets of corn and wheat, although he finds some evidence that there was a soybean bubble over the period of December 2007 and March 2008. Gilbert (2010b) concludes that index investment in commodity futures may have had a modest effect on the agricultural futures prices. He also assesses that index investors amplified the price movements driven by fundamental factors, albeit, agricultural futures prices were not driven by index based investment over the examined period.
12
Irwin and Sanders (2010, 2011) analyse the relationship between commodity index traders’ activities and grain futures prices with Granger causality test over the period of 2007–
2008 and 2004 –2005. They fail to find any casual link between futures prices and commodity index activities, furthermore, their long horizon regression does not support the index-induced price bubble theory.
Brunetti and Büyükşahin (2009) conclude that speculation in the futures market does not cause any price movement. Moreover, it lessens risk. They also claim that speculation activity has an essential role in futures market since hedgers can find counterparties to close their positions, thus, speculations allows futures markets to fulfil their institutional role.
My hypotheses based on the above mentioned literature are the following:
Hypothesis 1:
The observed high agricultural prices were mainly caused by the increased financial
activities in the futures market during 2006–2008. However, there were other factors which
attributed to the peak in food prices, thus, there was a “perfect storm”.
Hypothesis 2:
The adverse weather and the trade restrictions, which were introduced as a reaction to the
increasing food prices, also pushed up the agricultural commodities prices during 2006–
2008, thus, these supply shocks made the prices even higher.
Hypothesis 3:
The increased oil prices raised the input costs of agricultural production, therefore, the surge in oil prices pushes up the world food prices.
Hypothesis 4:
13
Hypothesis 5:
The increasing biofuels production has also generated an additional demand for agricultural commodities. This is another reason why agricultural commodity prices have been higher than historical standards since the beginning of 2000s.
My conceptual model is the following:
14
3. Methodology
In order to answer my research questions I conduct a time-series analysis about monthly world food prices, respectively corn, wheat, soybean and their average prices over the period of 2002 January–2010 December. Due to using monthly data I am able to capture the effects of financial markets activities, particularly the impacts of speculations. My main dependent variables are the individual food prices and in one regression the average of the monthly corn, wheat and soybean prices and the independent variables are capturing the factors that may influence the food prices according to my conceptual framework.
3.1. Variables
In this section I will describe the data which are collected and used in my time-series regression of agricultural commodity prices. In Appendix B Table 1 summarizes the data collected.
As mentioned above my dependent variables are the individual agricultural prices and the average of them. Therefore, I run my model four times and my dependent variables of my regressions are respectively, the monthly US yellow no. 2 corn at the Gulf of Mexico, US no. 1 yellow soybean at the Gulf of Mexico and US no. 2 soft red winter wheat at the Gulf of Mexico prices in US$/ton and the monthly average of them. I collected monthly, individual agricultural prices from the FAO International Commodity Prices Dataset.1
3.1.1. Independent variables
In my model six explanatory variables are applied, in this section the short description of them is given.
The demand and the demand_percapita explanatory variables are capturing the growing real world demand for -or aggregate expenditure on – agricultural commodities. Real world money supply is applied as a proxy for the growing real world demand for cereals. I use two indexes as a proxy based on Cooke and Robles (2009). I construct real world M2 indexes
1
15
using individual monthly real M2 series for the tenth biggest economies and the EMU (euro area) members2. Thus, my data contain Brazil, China and Russia which are the biggest emerging countries with rapid economic growth and fast growing money supply. Therefore, I believe that my index can also capture the growing demand for cereals generated by the changing diet in emerging countries.
Each M23 is divided by each country’s CPI to create a real M2 series and from them I construct individual real M2 index (January 2002 = 1) for each country. After that, each real M2 index is used to construct two weighted indexes. One of the weighted indexes is based on the size of the country’s GDP in U.S. dollars adjusted by purchasing power parity in 2002. In my regressions this index is the demand variable, and this is created in the same way as Cooke and Robles (2009). Below I present the formula:
The other weighted real world M2 index is based on the size of the country’s GDP per capita in U.S. dollars adjusted by purchasing power parity in 2002. I use the GDP per capita to get the proportions of the countries in the weighted index because I believe that the cereals consumptions is highly related to the people‘s financial situation. This index is the demand_percapita in my specifications.
The data for M2 and CPI are available at the IMF’s International Financial Statistics (IFS) online database and at the United Kingdom Central Bank online database.
2
The real worl M2 indexes contain : China, Brazil, Russia, United Kingdom, United States, Japan, Canada, Mexico, Republic of Korea, and the EMU members except Estonia. Estonia is excluded since it joined the Euro zone in 1 January 2011.
3
Generally, M2 is one measure of money supply which is composed of currency, overnight deposits, deposits
redeemable at notice of up to three months and deposits with an agreedmaturity of up to two years (OECD
16
The oilp variable measures the monthly crude oil prices. The monthly data is constructed from the weekly all countries spot oil prices FOB weighted by estimated export volume. The data are collected from the webpage of U.S. Energy Information Administration.
The biofuel variable captures the monthly biofuels – especially, ethanol and biodiesel- production in the US. I use only the US biofuels production data since the data of the Brazilian and European Union biofuels production are not available on monthly basis. The data are also collected for the biofuel variable from the webpage of U.S. Energy Information Administration.
My supply variable captures the supply shocks. On the base on Cooke and Robles (2009) my supply variable combines the top exporters’ monthly export quantity data for each agricultural commodity markets. I use export data since I think it can capture the effects of supply shocks, such as low yields, bad weather and the trade restrictions too. Moreover, other supply proxy such as yields is unavailable at monthly frequency. In case of wheat I collected data from USA, Canada, EU and Argentina, which accounted for 75% of the worldwide export in 2002. In the case of corn market USA, EU and Argentina made up nearly 80% of the export in 2002. Furthermore, in case of soybean market USA, Brazil and Argentina were accounted for 90% of the export in 2002. Since the monthly US export data is only available in $ value, I divided the monthly value export data by the monthly world price so as to get (approximately) the US monthly export data in quantity. In case of the average prices the average of the corn, soybean and wheat export data is applied as the supply variable. EU data is available at Eurostat Comext, the US data can be found at www.usatradeonline.gov, Brazil export data is available at AliceWeb and the Canadian data at CanSim. The Argentinean data were collected from the Argentinean Ministry of Agriculture.
My proxy for easy monetary policies (monpol variable) will be the monthly average dollar-euro exchange rate. I collect the data from IMF’s International Financial Statistics (IFS) online database.
17
traders4 as the commercial or non-commercial trader (CFTC, ca.2011). When the traders use futures contracts for hedging purposes their positions will be defined as commercial otherwise, the traders’ futures positions will be regarded as non-commercial (CFTC, ca.2011). Therefore commercial positions represent hedging activity and non-commercial positions are held mainly for speculative purposes, e.g.: searching of financial profits (Robles, et al. 2009, Cooke and Robles, 2009). Thus, the ratio of non-commercial positions to total positions for short and long positions can show the importance of speculation activity relative to hedging (Robles, et al., 2009; and Cooke and Robles, 2009.) In case of average prices I will use an average of the corn, wheat and soybeans ratios. In one specification I use the ratio of long positions and in the other case I use the ratio of the short position.
3.2. The model
As it has already been mentioned a time series analysis is conducted over the period of 2002 January – 2010 December. Using monthly data my sample is composed of 108 observations.
The log-log functional model has several advantages, e.g.: it can mitigate the problem of heteroskedasticity, the coefficients are easy to interpret (Hill, et al., 2008). Moreover taking natural logarithm can narrow the range of the variables which makes the estimates less sensitive to outliers (Wooldridge, 2009). Wooldridge (2009) mentions that natural logarithm of the variable is often taken when the variable is positive dollar amount or when the variable has the common feature of being large integer values. One limitation of the logarithm is that the variables have to be positive values. Since all my variable are positive value and some (agricultural commodities prices, oil prices) are in dollar amount and other (export data) has the feature of being large value the log-log form is chosen. Furthermore, both Cooke and Robles (2009) and Andreosso-O’Callaghan and Zolin (2010) apply log-log model when they are modelling the agricultural prices.
Since I work with monthly data seasonality might be a problem (Wooldridge, 2009). However, seasonal dummies5 are used in to control the seasonal effects in one specification
4
Reportable traders hold positions in options and futures at or above specific reporting levels determined by the Commodity Futures Trading Commission (CFTC, ca. 2011). The aggregate of all reportable traders’ position accounts for more than 70% of the total open interest in any given market (CFTC, ca.2011).
5
18
and monthly dummies6 are applied in other. However, Cooke and Robles (2009) and Andreosso-O’Callaghan and Zolin (2010) conclude that there is no evidence of seasonality.
Furthermore, it is important to know whether the time series data are stationary or nonstationary. Time series are regarded to stationary when it has a property of mean reversion. When the time-series are nonstationary there is a danger that the regressions are spurious which means that significant regression results can be obtained from unrelated data (Hill, et al., 2008). To detect whether my data are stationary or nonstationary after taking the natural logarithm of all my variables I conduct an augmented Dickey-Fuller test (ADF test). In the Appendix C Figure 1 depicts the time plots of my variables and their first differences in order to decide which type of ADF tests has to be applied. In Appendix C the Table 1 shows the results from the augmented Dickey-Fuller test. Regarding to the test results most of my data series are nonstationary. Therefore, I check whether my dependent variables and independent variables are cointegrated or not. Cointegration means that my dependent variable and my independent variables share the similar stochastic trends, therefore, the error term is stationary. Hence, if there is a cointegration there is no the fear of spurious regressions when my time-series variables are nonstationary. In order to detect whether there is a cointegration I will test the least squares residuals with augmented Dickey-Fuller test for stationary (Hill, et al., 2008). In Table 2 the results from the cointegration test are shown. Since I do not find cointegration between the variables I have to convert nonstationary series to stationary series. The variables do not show any signs of trend, therefore, I conclude that they are difference stationary and first differences are taken so that the nonstationary variables are converted to stationary (Hill, et al., 2008). Additionally, looking at the time plots of the first difference no sign of nonstationary can be found. Moreover, Cooke and Robles (2009) use the first differences to handle the nonstationary problem. However, using the first differences mean that I lose the first observation, therefore I have 107 observations only.
Moreover, because of using time-series analysis and highly frequently data it is highly likely to face with autocorrelation which means that the error terms are correlated. To detect whether I have autocorrelation the Lagrange multiplier (LM) test are conducted. The tests’ results can be found in Appendix D in Table 1. When I reject the null-hypothesis of no autocorrelation at the 5% significance level I conclude that I have autocorrelation. However,
6
19
when there is autocorrelation the standard errors of OLS are not correct, since one of the least squares assumptions is violated. Therefore, the hypothesis tests and confidence intervals that use these standard errors may not be correct. I use autoregressive distributed lag (ARDL) model to eliminate the autocorrelation problem. The ARDL model is a linear regression model in which the time-series dependent variable is expressed as a function of lags of independent variables and of dependent variables (Wooldridge, 2009; Stock and Watson, 2008). Cooke and Robles (2009) also use an ARDL model. Furthermore, in all specifications I test the first, second and third lags of the independent variable to detect whether there are slow, delayed effects of the independent variables. However, I do not find statistically significant evidence that there are delayed effects of the explanatory variables. Thus, when there is no autocorrelation OLS model without lag variables is applied.
In addition, with White test I detect whether there is heterokedasticity which means that the variances of the residuals are not the same, thus one of the least squares assumptions is violated. If I reject the null hypothesis of homoskedasticity robust standard errors will be applied to mitigate the problem. The test results are shown in the Table 1 in Appendix D. Regarding the test results only in case of soybeans I face with heteroskedastictiy in some specifications.
Last but not least, with the one of the automatic normality test of Stata7 (sktest) I detect whether the residuals are normally distributed. The sktest is really similar to the Jarque-Bera test. The results can be found in the Table 1 in Appendix D. I can conclude that the residuals are normally distributed in most cases, albeit in some specifications, particularly in cases of soybean I reject the null-hypotheses. In some specifications of soybean standard robust errors are applied, which is less sensitive to the normality assumptions, thus, it is less likely that the trust in the model will be decreased.
Below I present my econometric model:
7
20
4. Results
Firstly, I publish my summary statistic of my data. After that, I interpret the results from my models and specifications, namely the results from the regressions of wheat, corn, soybean and the average prices. Finally, I discuss my results of my different models.
4.1. Summary Statistic
Table 2 shows the data description of the whole sample. As mentioned above I apply the first differences of the natural logarithm of the variables. Therefore, I have 107 observations, and I do not have missing values. The growth rates of agricultural commodity prices, oil prices, export data and speculation indexes look more volatile. These facts are in the line with the observed surge in commodity prices, supply shocks and the increasing financial markets activities of commodity futures markets during the period of 2002–2008.
Moreover, I can conclude that I do not have outliers in my sample.
Table 2
Variable name Obs. Mean Standard
deviation
Variance Min. Max. Skewness Kurtosis
first difference of the natural logarithm of wheatprice
107 0.0085405 0.0800772 0.006412 -0.250 0.256 0.1962606 4.389803
first difference of the natural logarithm of cornprice
107 0.0092496 0.0657493 0.004323 -0.226 0.163 -0.6277891 4.828876
first difference of the natural logarithm of soyaprice
107 0.0101027 0.0812962 0.006609 -0.244 0.415 0.3123293 9.08442
first difference of the natural logarithm of average price
107 0.0094342 0.0624911 0.003905 -0.223 0.195 -0.3737083 4.637515
first difference of the natural logarithm of demand
107 0.006183 0.0063859 0.000040 -.0083 0.028 0.9418571 4.646239
first difference of the natural logarithm of demand_percapita
107 0.004359 0.0059382 0.000035 -0.012 0.026 0.8492874 4.833269
first difference of the natural logarithm of Oil price
107 0.0148312 0.0958079 0.009179 -0.373 0.209 -1.188785 5.360734
first difference of the natural logarithm of Biofuel 107 0.0176507 0.0647761 0.004195 -0.216 dlbiofuel 111100007777 ....0000111177776666555500007777 ....0000666644447777777766661111 ----....222211116666444488885555 ....1111999911118888111111111111 0.191 -0.6027355 4.846792
first difference of the natural logarithm of wheatsupply
107 0.0006907 0.1185757 0.014060 -0.392 0.332 -0.061173 3.552891
first difference of the natural logarithm of cornsupply
107 0.001202 0.139944 0.019584 -0.271 0.383 0.5827013 3.000439
21
natural logarithm of soyasupply
first difference of the natural logarithm of avergasupply
107 0.0011026 0.0946964 0.008967 -0.185 0.252 0.3876003 2.375163
first difference of the natural logarithm of monpol
107 0.0037676 0.0264651 0.000700 -0.076 0.067 -0.2035716 3.39282
first difference of the natural logarithm of wheat_ncom_long
107 -0.0024259 0.0913879 0.008351 -0.289 0.235 -0.0260118 3.959528
first difference of the natural logarithm of wheat_ncom_short
107 0.0019364 0.1528158 0.023352 -0.729 0.564 -0.8220906 8.651063
first difference of the natural logarithm of corn_ncom_long
107 0.0036497 0.0782301 0.006119 -0.225 0.211 0.1968491 3.784194
first difference of the natural logarithm of corn_ncom_short
107 -0.005772 0.1799584 0.032385 -0.617 0.608 -0.3714139 4.809266
first difference of the natural logarithm of soya_ncom_long
107 0.0050599 0.104757 0.010974 -0.418 0.401 0.2121483 7.257536
first difference of the natural logarithm of soya_ncom_short
107 -0.0039464 0.1497176 0.022415 -0.417 0.399 -0.1564465 3.421653
first difference of the natural logarithm of av_ncom_long
107 0.0020506 0.0679883 0.004622 -0.241 0.201 -0.0202428 4.002686
first difference of the natural logarithm of av_ncom_short
107 -0.0023359 0.1271422 0.016165 -0.499 0.326 -0.9314745 5.653614
Table 3 and Table 4 show the correlation matrix of my variables. Correlations above 0.3 between the variables are highlighted in the tables. The agricultural commodity prices and export volumes are correlated which is not surprising since the agricultural commodities are substitute and complementary goods.
. The ratios of the short and long positions by noncommercial traders are negatively correlated. This is in the line with the expectations. Moreover, the ratios of noncommercial positions to total positions are also correlated with each other. This fact indicates that the futures markets of the agricultural commodities are also related to each other. In addition, the prices and the speculation proxies are also correlated. This fact may indicate that there is a relationship between them.
Since my two real world money supply variables are slightly different they are highly correlated. However, above mentioned correlations do not cause a problem because I do not apply these variables in the same regressions.
22
24
4.2. Wheat prices
In Table 5 I show the results from my regression model when the dependent variable is the wheat prices. In case of wheat prices there is autocorrelation, thus, ARDL(1,1), ARDL (2,2) and ARDL(3,3) models are applied to mitigate this problem. In regression 1, 2 and 3 the real World M2 index based on country’s GDP on PPP (demand) is applied and in regression 4, 5 and 6 the real World M2 index based on country’s GDP on PPP per capita (demand_percapita) is used. In regression 1, and 4 the ratio of long positions held by non-commercial traders is used and in the other specification the ratio of short positions held by non-commercial traders is applied. Fortunately, I fail to reject the null hypothesis of homoskedasticity in all cases, and only in one specification (regression 4) I reject the null hypothesis of normality. Thus, I can conclude that it is less likely that the trust in the model will be lower due to the violation of normality assumptions.
Table 5
(1) (2) (3) (4) (5) (6)
VARIABLES dlwheatprice dlwheatprice dlwheatprice dlwheatprice dlwheatprice dlwheatprice
25 L2.dldemand_percapita -3.211 -0.887 1.226 (2.470) (2.119) (1.403) L2.dloilp 0.115 0.0914 0.168 0.143 (0.108) (0.111) (0.102) (0.0974) L2.dlbiofuel -0.106 -0.0787 -0.228 0.0287 (0.180) (0.218) (0.168) (0.133) L2.dlwheatsupply -0.0281 -0.0958 -0.0152 0.0395 (0.0855) (0.0913) (0.0797) (0.0705) L2.dlmonpol -0.274 -0.345 -0.403 -0.235 (0.347) (0.400) (0.336) (0.312) L2.dlwheat_ncom_long 0.194* 0.181 (0.104) (0.112) L2.dlwheat_ncom_short -0.0858 -0.0894 (0.0559) (0.0550) L3.dlwheatprice -0.247* (0.132) L3.dldemand_percapita 0.559 (2.906) L3.dloilp -0.0236 (0.119) L3.dlbiofuel -0.0803 (0.188) L3.dlwheatsupply -0.123 (0.0877) L3.dlmonpol -0.157 (0.392) L3.dlwheat_ncom_long 0.192* (0.112)
Monthly Dummies Yes Yes No Yes Yes No
Seasonal Dummies No No No NO No No
Constant -0.0176 -0.00178 0.0128 -0.0513 -0.00565 -0.00396
(0.0698) (0.0600) (0.0139) (0.0575) (0.0472) (0.0142)
Observations 105 106 106 104 105 105
R-squared 0.466 0.489 0.372 0.509 0.531 0.411
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
The seasonal dummy variables are insignificant in all specification. However, the monthly dummies show some explanatory power and with the F-test I fail to reject the null-hypothesis at the level of 10% that their joint effect is zero. Thus, in these cases I present the results with and without the dummies. Moreover, in some specifications (regression 1 and 4) I fail also to reject the null hypothesis at the level of 5 %, therefore, I only publish the results with the monthly dummies.
My regressions show low R-squared, which can be regarded as a measure of “goodness of fit”. However, it is not surprising as I am modelling monthly prices, and due to the nonsationary series the first differences of the variables are used. Cooke and Robles (2009) also have low “goodness of fit” when they analyse the agricultural commodity prices.
26
The coefficients of the growth rate of the real world money supply and their lag variables which are capturing the growing demand are insignificant in all of my regressions. This is partly not in line with Cooke and Robles (2009) findings because they find that their real world M2 index has an explanatory power in half of the specifications when they analyse wheat prices.
Additionally, the growth rate of wheat export is insignificant in all specifications. Albeit, its first lag has an explanatory power in some regressions since it is significant with a positive sign. It is hard to explain the positive sign. Following Cooke and Robles (2009) explanation the export data may capture the demand pressure since the export and import data are quite similar in international markets. Thus, I can conclude my findings do not support the theory that supply shocks such as trade restrictions and weather shocks can push up the agricultural commodity prices (Dollive, 2008; Headey, 2010).
Furthermore the growth rate of the dollar-euro exchange rate and their lags are also insignificant in all regressions. Cooke and Robles (2009) also find that the dollar-euro exchange rate does not have a significant effect on the wheat prices, however Andreosso-O’Callaghan and Zolin (2010) find that the dollar devaluation has a significant effect on the agricultural commodity prices. Moreover, my results are not consistent with the findings of Abbot, et al. (2008), Gilbert (1989) and Mitchel (2008) who claim that depreciation of dollar increases the dollar (agricultural) commodity prices.
My results show that the growth rate of oil prices has an explanatory power, since it is significant in most specifications at the level of 5%. Its sign is positive which is in line with the theory of Trostle (2008), Baffes (2007) and Headey and Fan (2010
) and
the findings of Cooke and Robles (2009). Although its lag variables are always insignificant, thus, my results do not support that the oil prices have a slow economic effect on the wheat price. In addition, the coefficients of growth rate of biofuels production are insignificant in all regressions. Hence, I can conclude that the higher oil prices affect mainly the production costs of wheat and the rising biofuel production does not influence the wheat prices significantly. I believe that my findings do not support the idea that the oil prices can also influence the grain prices via the demand side, namely through the biofuels production, because I do not find evidence that biofuels production has a significant influence.27
positively significant in some cases. Furthermore, in regression 2, 3, 5 and 6 the growth rate of the ratio of short positions is also significant with a negative sing at the level of 1% but its first lag has a positive effect on the growth rate of the wheat prices. However, their joint effect is still negative. My findings are in consistent with my previous expectations and they are consistent with the results of Cooke and Robles (2009) and Andreosso-O’Callaghan and Zolin (2010). Consequently, I deduce that financial activities in futures market by noncommercial traders have an explanatory power on the growth rate of wheat prices. However, the findings suggest that their role in the food price changes is more ambiguous.
4.3. Corn prices
Table 6 presents the results from the regressions of the growth rate of corn prices. In regression 1, 2 the real World M2 index based on country’s GDP on PPP (demand) is applied and in regression 3 and 4 the real World M2 index based on country’s GDP on PPP per capita (demand_percapita) is used. In regression 1 and 3 the ratio of non-commercial positions to total positions for long positions is used and in regression 2 and 4 the ratio of non-commercial positions to total positions for short positions is applied to proxy for the speculation variable.
Table 6
(1) (2) (3) (4)
VARIABLES dlcornprice dlcornprice dlcornprice dlcornprice
dldemand 0.620 -0.870 (0.929) (0.882) dldemand_percapita -0.0998 -0.903 (1.015) (0.950) dloilp 0.124* 0.116* 0.110* 0.114* (0.0642) (0.0601) (0.0655) (0.0611) dlbiofuel 0.00791 -0.0782 0.0203 -0.0780 (0.0999) (0.0936) (0.100) (0.0938) dlcornsupply -0.0902** -0.0676 -0.0927** -0.0725* (0.0450) (0.0424) (0.0456) (0.0428) dlmonpol 0.417* 0.325 0.435* 0.328 (0.223) (0.210) (0.223) (0.210) dlcorn_ncom_long 0.314*** 0.308*** (0.0729) (0.0728) dlcorn_ncom_short -0.182*** -0.178*** (0.0305) (0.0300) Monthly dummies No No No No Seasonal Dummies No No No No Constant 0.000827 0.0121 0.00505 0.0107 (0.00816) (0.00769) (0.00735) (0.00686) Observations 107 107 107 107 R-squared 0.278 0.368 0.275 0.368
28
In case of corn I do not find that there is an autocorrelation or delayed, economic effects of my explanatory variables, hence, OLS without lag variables is applied. Moreover, in all specifications I fail to reject the null hypotheses of homoskedasticity and normality.
Moreover, I do not find any evidence that my seasonal or monthly dummies can capture any effects, therefore, I include neither the seasonal dummy variables nor the monthly dummies in the regressions in the case of corn.
The regressions also show low “goodness of fit” which is not surprising due to above mentioned reasons. It is also in the line with Cooke and Robles (2009).
The growth rate of the real world money supply is again insignificant in all of the regressions regardless which indexes are applied as a proxy. It is in line with Cooke and Robles (2009) who also do not find any statistical evidence that their real M2 world index have any explanatory power in case of corn price. Moreover, the growth rate of biofuels production is also insignificant in all regressions. These results do not support the theory that biofuels production was the main cause of the price surge (Mitchell, 2008; Schmidhuber, 2006; Andreosso-O’Callaghan and Zolin, 2010). However, my result is partly in line with Cooke and Robles (2009). They claim that the growth rate of biofuel production has a slightly explanatory power in corn price changes8.
However, the growth rate of the oil prices has an explanatory power in all my regressions, it is always significant with a positive sing at the level of 10%. Thus, my findings are consistent with the theory that oil prices can positively affect the agricultural commodity prices (Trostle, 2008; Baffes, 2007; Headey and Fan, (2010). I think my results support that the high oil prices influence significantly the input costs of the corn prices, and the increased production costs raise the corn prices.
The growth rate of corn export is also significant with a negative sign in three of my specifications. As follows, I conclude that the growth rate of export negatively affects the growth rate of corn price. These findings are in line with the theory that trade restriction can raise the prices (Dollive, 2008; Headey, 2010; Mitra and Josling, 2009). Additionally, there is also some evidence that the changes of the exchange rate positively affect the growth rate of corn price. These results do support the theory that the devaluation of dollar has an explanatory power over the food prices (Gilbert, 1989; Mitchell 2008).
8
29
Looking at the speculation variables both of proxies enter as significant in all specifications at the level of 1% in case of corn. The growth rate of the ratio of short positions has always a negative influence on the growth rate of the corn price, whereas, growth rate of the ratio of long positions positively affects the growth rate of corn price. My findings are consistent with my former expectations, the theory and findings that futures market activities have an effect on agricultural spot prices (Robles, et al., 2009; Cooke and Robles, 2009; Andreosso-O’Callaghan and Zolin, 2010). However, I would like to emphasize that my results suggest that the activities of non-commercial traders can influence positively or negatively the spot corn prices.
4.4. Soybean prices
Focusing on soybean prices, the results from the regressions are published in Table 7. In case of soybean I fail to reject the null hypothesis of no autocorrelation with the LM test, furthermore I do not find that there are slow economic effects of the independent variables, therefore OLS without lag variables is applied. However, in some specifications I reject the null hypothesis of homoskedasticity with the White test. In these cases (regression 2 and 4) the robust standard errors are applied so as to mitigate this problem. Moreover, there is also a normality problem in all specifications, although robust standard errors are less sensitive to the violation of the normality assumptions.
In the regressions 1 and 2 the real world M2 index based on GDP on PPP (demand) is used and in regression 3 and 4 in the regressions the real world M2 index based on GDP on PPP per capita is applied (demand_percapita). In regression 1 and 3 the ratio of non-commercial positions to total positions for long positions is used and in regression 2 and 4 the ratio of non-commercial positions to total positions for short positions is applied to proxy for the speculation variable.
30
Table 7
(1) (2) (3) (4)
VARIABLES dlsoyaprice dlsoyaprice dlsoyaprice dlsoyaprice
dldemand 1.413 1.619 (1.241) (1.054) dldemand_percapita 2.227* 2.234* (1.308) (1.243) dloilp 0.169** 0.194** 0.188** 0.209** (0.0819) (0.0829) (0.0823) (0.0811) dlbiofuel -0.183 -0.255*** -0.181 -0.249*** (0.115) (0.0945) (0.113) (0.0905) dlsoyasupply 0.0243 0.00712 0.0258 0.00746 (0.0294) (0.0363) (0.0287) (0.0386) dlmonpol 0.492* 0.450 0.479* 0.443 (0.288) (0.274) (0.286) (0.279) dlsoya_ncom_long 0.243*** 0.242*** (0.0696) (0.0690) dlsoya_ncom_short -0.220*** -0.217*** (0.0659) (0.0648) Monthly dummies No No No No Seasonal Dummies No No No No Constant -0.00104 -0.000869 -0.00225 -0.000878 (0.0106) (0.0115) (0.00937) (0.00905) Observations 107 107 107 107 R-squared 0.220 0.276 0.232 0.285
Standard errors (regression 1 and 3) and robust standard errors (regression 2 and 4) in parentheses *** p<0.01, ** p<0.05, * p<0.1
I find no evidence that there is any effect of the monthly dummies or the seasonal dummies, so thus the regressions without the dummies are presented.
As in earlier cases the regressions of soybean show low “goodness of fit” which is not surprising due to above mentioned reasons. It is also consistent with Cooke and Robles (2009) results.
I find some evidence that the growth rate of real world money supply has an explanatory power over the growth rate of soybean price. The real world M2 index based on country’s GDP on PPP per capita is significant with a positive sign at the level of 10% in both specifications. These results party support the theory that demand pressure has pushed up the agricultural prices.
31
Cooke and Robles (2009) also find in some cases that the biofuels production have a negative effect on agricultural food prices. However, they cannot give a convincing interpretation for it. I would like to emphasize that my biofuels production data cover only the U.S. production which is mostly made from corn. However, soybean can be also a source of the biofuels production. Thus, I think my biofuels production variable might mainly pick up the growing demand for corn. This increasing demand for the corn might mean that the demand for soybean decreased in U.S. However, I have to highlight that there is no research or theory which can strongly support this explanation.
The growth rate of the soybean export data is insignificant in all the specifications. These findings are not in the line with the theory of trade restriction (Dollive, 2008; Headey, 2010; Mitra and Josling, 2009).
Moreover, there is some evidence that the growth rate of the exchange rate positively influences the growth rate of soybean price. This result does support the theory that the devaluation of dollar has a positive effect on agricultural prices (Gilbert, 1989; Mitchell, 2008). Furthermore, Cooke and Robles (2009) can conclude that half of the dollar depreciation is transmitted, therefore, it indicates that the “determination of soybean prices takes place in a different currency than the U.S. dollar” (Cooke and Robles, 2009, pp. 22). My finding is consistent with it and supports this explanation.
Analysing the speculation variables I conclude that both of them have an explanatory power on the growth rate of soybean prices. Both ratios are highly significant in all specifications and their signs are in line with the expectation. The growth rate of the ratio of short positions has always a negative influence on the growth rate of the corn price, whereas, growth rate of the ratio of long positions positively affects the growth rate of corn price. My findings are consistent with the theory and findings that futures market activities have an effect on agricultural spot prices (Robles, et al., 2009; Cooke and Robles, 2009; Andreosso-O’Callaghan and Zolin, 2010). However, my findings again induce that the effect of activities by non-commercial traders in futures market is less clear.
4.5. Average prices
32
index based on GDP on PPP per capita (demand_percapita) is used. In regression 1 and 4 the ratio of long positions held by non-commercial traders is used, whereas, in other specifications the ratio of short positions held by non-commercial is applied.
Fortunately, in case of average prices with the LM test I fail to reject the null hypothesis of no autocorrelation, and I do not find that the lag variable has a significant effect, thus, OLS without lag variables is applied. Moreover, I fail to reject the null hypothesis of homoskedasticity, thus I can conclude that I do not face heteroskedasticity problem. However, in one specification (regression 1) I reject the null hypothesis of normality, although I can conclude that it is less likely that the trust in my results will be lower.
Table 8
(1) (2) (3) (4) (5) (6)
VARIABLES dlavprice dlavprice dlavprice dlavprice dlavprice dlavprice
dldemand 1.244 0.244 0.259 (0.928) (0.879) (0.889) dldemand_percapita 1.337 0.645 0.819 (1.012) (0.946) (0.957) dloilp 0.149** 0.200*** 0.159*** 0.152** 0.207*** 0.170*** (0.0609) (0.0604) (0.0576) (0.0618) (0.0605) (0.0582) dlbiofuel -0.108 -0.132 -0.172* -0.102 -0.140 -0.182** (0.0948) (0.0897) (0.0899) (0.0937) (0.0889) (0.0887) dlavsupply -0.0282 -0.0279 -0.0269 -0.0264 -0.0189 -0.0149 (0.0648) (0.0663) (0.0614) (0.0654) (0.0667) (0.0617) dlmonpol 0.411* 0.200 0.253 0.413* 0.194 0.244 (0.212) (0.202) (0.204) (0.212) (0.202) (0.204) dlav_ncom_long 0.346*** 0.343*** (0.0792) (0.0791) dlav_ncom_short -0.241*** -0.238*** -0.241*** -0.237*** (0.0414) (0.0416) (0.0409) (0.0410) Monthly dummies No No No No No No
Seasonal dummies No Yes No No Yes No
Constant -0.000782 0.0163 0.00703 0.000918 0.0149 0.00510
(0.00783) (0.0122) (0.00748) (0.00704) (0.0116) (0.00666)
Observations 107 107 107 107 107 107
R-squared 0.271 0.389 0.346 0.271 0.391 0.350
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
In some specification I find that the seasonal dummies have an effect on the average prices at the level of 10%, thus, in these cases the results from the regressions with and without the dummies are presented in the tables.
33
The growth rate of the biofuels production is again significant with a negative sign in some specifications at the level of 10%. This finding does not support the theory that biofuels production has a positive effect on food prices and it is not consistent with the results of Andreosso-O’Callaghan and Zolin (2010) and Mitchel (2008). However, as already mentioned above it is difficult to explain why the biofuel production can have a negative effect on food prices. I believe that it might be caused by the fact that my biofuels production data only cover the U.S. production.
Furthermore, the growth rate of oil prices is strongly significant with a positive sign in all regression. Thus, my findings are consistent with the theory that oil prices positively affect the agricultural commodity prices (Trostle, 2008; Baffes, 2007; Headey and Fan, 2010). I believe that the high oil prices raise the agricultural prices mainly through the supply channel in which the increased oil prices make the transportation, fertilizer and other production costs higher. Moreover, my findings do again not support the idea that the increased oil prices also push up the food prices through the demand-side, namely via the biofuels production, since there is no evidence that the biofuels production has a positive pressure on the food prices,
The growth rate of the dollar-euro exchange rate is significant with a positive sign at the level of 10% in some specifications. Hence, I conclude that the changes of the exchange rate positively affect the growth rate of average grains price. These results do support the theory of devaluation of dollar (Gilbert, 1989; Mitchell, 2008).
Turn the attention to the speculation variables I evolve that both of them have a strong explanatory power on the growth rate of average prices. The growth rate of the ratio of short position negatively affects the growth rate of the average prices, whereas, growth rate of the ratio of long positions positively influences the growth rate of average prices. My findings are consistent with the theory and findings that futures market activities have an effect on agricultural spot prices (Robles, et al., 2009; Cooke and Robles 2009; Andreosso-O’Callaghan and Zolin, 2010). However, I can assert that the aggregate effect of activities by non-commercial traders in futures market is more ambiguous.
4.6. Discussion
34
rate of the export volume has some explanatory power on the cereals prices, particularly on the corn prices. This finding is in line with the theory that the export restriction and other supply shocks can push up the prices (Dollive, 2008; Headey, 2010; Mitra and Josling, 2009). However, in case of other agricultural commodities and their average price I do not find evidence that export volume has an explanatory power.
Thirdly, my results support the theory that the high oil prices raise the food prices. Moreover, my findings do not promote that the increasing biofuels production has a positive effect on food prices since when I find a significant effect that is negative. It is hard to give a convincing economic interpretation since there are only a few studies which examine the effects of biofuels production on food prices. Moreover, I do not find any explanation for this sign in the literature. My personal view is that my data cover only the U.S. biofuels production which is mostly made from corn, therefore, it might be an explanation for the negative sign. Thus, I can conclude that it is highly possible that the oil prices influence the agricultural commodity prices mainly via the production costs, particularly through the transportation and fertilizer costs. My results do not support the theory that the high oil prices can affect cereals prices through the demand side, namely the biofuels production because the biofuels production does not have a positive effect on the food prices in my models.
Fourthly, my results party support the theory that the dollar devaluation raises the food prices (Gilbert, 1989; Mitchell, 2008). In the international cereals market most of the transaction is in US dollar, therefore, the strength of the US dollar can influence the prices. When the dollar is weak, there is more incentive to import cereals which can raise the prices (Headey and Fan, 2010; Abbot, et al., 2008).
35
5. Conclusion
During 2006–2008 and in recent time a rapid run-up in agricultural commodities prices was observed. The surge in food prices turned public attention to food security questions because poor people spend a large part of their household budget on food, hence, they have few opportunities to adjust the rising food prices. In addition, high food prices may have bad macroeconomic effects. Because of above mentioned reasons it is crucial to understand what the driving factors of the agricultural commodity prices are. Understanding the factors may help to prevent future rapid run-up in food prices and maintain the food security in all over the world.
Since 2007 experts and researchers have given several explanations for the food prices surge. Some enhance the role of the supply driven factors such as oil prices, export restrictions, bad weather, and agricultural productivity. Others assert the demand driven factors such as growing demand, biofuels production, dollar devaluation or futures markets’ activities. Some experts claim also that there is no one factors behind the surge in agricultural prices, although there is conflagration and interaction of several factors.
The aim of my research is to give an empirical evidence for the magnitude of the potential driven factors of the food prices with a special attention to the futures market activities of the non-commercial traders. I conduct a time-series analysis for the corn, wheat and soybean prices and their average prices over the period of 2002–2010.
As mentioned above my research does not prove that the theory of the growing demand which is cause by changing diet in developing countries may be a driven factor. In addition, my analysis partly supports that export restrictions, other supply shocks and dollar devaluation might be the causes of surging prices. In case of biofuels production I find negative effects which are not in line with the theory and expectation. However, I do not affirm that above mentioned factors should be disregarded I can only claim that my results do not prove that they have significant effects on the behaviour of food prices.
36
conclude that the non-commercial traders’ activities have an influence over the spot prices, however, their overall effects are unclear and the mechanism through which activities of futures markets affect the spot prices is more ambiguous.
I would like to assert that policy makers should be more aware of the effects of the oil price and futures markets when they maintain and enhance the food security. Better international cooperation and more transparency may mitigate these effects on food prices.
I have to mention that due to monthly data my research has some limitations because some data are not available in monthly frequency, for example world population data, imports data of the main importers, biofuels production data about EU and Brazil. Thus, I apply real world M2 index based on Cooke and Robbles (2009) as a proxy for demand instead of imports data. Furthermore, only the U.S. biofuels production data is applied.
In addition, the ratio of the long and short positions held by non-commercial traders are applied to capture the speculation activities. Although some researchers also claim that the category of non-commercial traders is highly aggregated and more detailed information about the traders needs to be analysed. Furthermore, my research focuses only on the corn, wheat and soybean prices and markets, and do not examine other agricultural commodity markets such as rice markets.
Regarding my findings the biofuels production has a negative effect in some cases, which is not consistent with the theory that increased biofuels production pushes up the food prices. However, I do think that the effects of the biofuels production and strategies should also be analysed in the future since more and more governments introduce biofuels strategies to mitigate the climate change problem.
37
References
Andreosso-O’Callaghan, B. and Zolin, M.B., 2010. Long-term cereal price changes: how important is the speculative element. Working Papers Department of Economics of Ca
Foscari University of Venice.
Abbott, P.C., Hurt, C. and Tyner, W.E., 2008. What’s Driving Food Prices? Farm
Foundation Issue Report, July 2008.
Baffes, J., 2007. Oil spills over into other commodities. Policy Research Working Paper
of World Bank, 4333.
Baffes, J. and Haniotis, T., 2010. Placing the 2006/08 commodity price boom into perspective, Policy Research Working Paper of World Bank, 5371.
Brennan, M.J., 1958. The Supply of Storage. American Economic Review, 48(1) pp.50-72, [online] Available at:
<http://www.econ.kuleuven.be/public/NDBAE55/Brennan%201958.pdf.> [Accessed 27.04.2011].
Brunetti, C., Büyüksahin B., 2009. Is speculations destabilizing? CFTC Working papers, [online] Available at: <https://editorialexpress.com/cgi
bin/conference/download.cgi?db_name=CEF2010&paper_id=218> [Accessed 20.05.2011]. Cooke, B, and Robles, M., 2009. Recent Food Movements. A Time Series Analysis.
Discussion Paper of International Food Policy Research Institute (IFPR), 0942, [online]
Available at: <http://www.ifpri.org/publication/recent-food-prices-movements > [Accessed 07.03.2011].
CFTC, ca. 2011. Explanatory notes, [online] Available at:
< http://www.cftc.gov/MarketReports/CommitmentsofTraders/ExplanatoryNotes/index.htm> [Accessed 07.04.2011].
Dollive, K., 2008. The impact of export restraints on rising grain prices. Economics
Working Paper of U.S. International Trade Commission, 08-A.
Fabozzi, F.J. and Modigliani, F., 2003. Capital Markets: institutions and instruments, 3rd ed. Upper Sanddle River, NJ: Prentince Hall, Pearson Education International.
Frankel, J., 2006. The effect of monetary policy on real commodity prices. Working Paper
of U.S.A.: National Bureau of Economic Research, 12713.
Frankel J., 2008. Monetary policy and commodity prices. VOX, [online] 29 March 2008, Available at : <http://www.voxeu.org/index.php?q=node/1178> [Accessed18.03.2011].
