Electronic copy available at: http://ssrn.com/abstract=2534753

Institute of Economics of the Polish Academy of Science

(INE PAN)

INE PAN WORKING PAPER SERIES

Did the Fertilizer Cartel Cause

the Food Crisis?

Hinnerk Gnutzmann Piotr Śpiewanowski

Working Paper #14-03

December 2014

Electronic copy available at: http://ssrn.com/abstract=2534753

Institute of Economics of the Polish Academy of Science (INE PAN)

Ul. Nowy Świat 72

00-330 Warsaw, POLAND

Tel: +48 22 657 27 07

Email: inepan@inepan.waw.pl

INE PAN Working Papers are preliminary materials circulated to stimulate discussion and critical comment.

Results may be in a preliminary or advanced stage. Any reproduction, publication and reprint in the form of a

different publication, whether printed or produced electronically, in whole or in part, is permitted only with the

explicit written authorisation of the INE PAN or the authors. Information on all of the papers published in the

INE PAN Working Paper Series can be found on the INE PAN’s website,

http://www.inepan.waw.pl/en/publications/institute-s-publications

Did the Fertilizer Cartel Cause the Food Crisis?

Hinnerk Gnutzmanna, Piotr �piewanowskib,c

aDepartment of Economics, Universita Cattolica del Sacro Cuore, 20123 Milan, ItalybInstitute of Economics of the Polish Academy of Sciences, Warsaw, Poland

cEducational Research Institute, Warsaw, Poland

Abstract

Food commodity prices escalated during the 2007/2008 food crisis, and have scarcelyfallen since. We show that high fertilizer prices, driven by the formation of aninternational export cartel as well as high energy prices, explains the majority ofthe recent price spikes. In particular, we estimate the pure fertilizer cartel e�ectexplains more than 60% of crisis food price increases. While population growth,biofuels, high energy prices and �nancial speculation doubtlessly put stress on foodmarkets, our results help to understand the severity and sudden emergence of thecrisis and suggest avenues to prevent its repetition.

Keywords: food crisis, fertilizer, cartel, competition policyJEL: F10, L40, Q02

1. Introduction

During the 2007�2008 food crisis, commodity prices escalated leading to foodinsecurity and political instability in the developing world (Timmer, 2010). Simulta-neously an export cartel for fertilizer, an essential input to industrial crop production,established itself and fertilizer prices tripled (Jenny, 2012). This paper estimates thatthe cartel directly led to a 60% overcharge in the fertilizer market, which � throughcost passthrough to agriculture � translated into a 25% increase in food prices. Sincefood prices rose on average by 40% during the crisis, we can attribute nearly twothirds of the price increase to the formation of the fertilizer cartel. Moreover, whilethe cartel channel has apparently been overlooked in the literature so far, it canexplain not only the severity but also sudden emergence of the food crisis which

Email addresses: hinnerk.gnutzmann@unicatt.it (Hinnerk Gnutzmann),piotr.spiewanowski@gmail.com (Piotr �piewanowski)

1

other explanations � such as population growth or high energy prices � do not. Moregenerally, our results indicate an urgent need to better understand the role of exportcartels in the working of commodity markets and containment of potential futurecrises.

Commensurate to the importance of the food crisis, a large literature is developingto estimate its causes1. On the demand side, analysts emphasize population growthand rising per capita meat consumption (Nazlioglu and Soytas, 2012). Mitchell(2008) focuses on the role of biofuel subsidies in raising demand for crops, althoughsubsequent research is ambiguous at best (Serra and Zilberman, 2013). Adversesupply shocks due to high energy prices are a potential explanation (Harri et al.,2009), although challenged by Zhang et al. (2010). Bad weather in some regionscertainly did not help (Headey and Fan, 2010, chapter 2), although harvests werenot unusually poor during the crisis period. Current research and debate largelyfocuses on determining the relative importance of these causal factors to the foodcrisis2. Additionally, a growing body of research investigates the role of �nancialspeculation in the food crisis, e.g. Irwin and Sanders (2012); Fattouh et al. (2013);Sanders and Irwin (2010). Finally, trade shocks (Headey, 2011) may have been afactor.

Fertilizer has so far been at best at the sidelines of the discussion. While highfertilizer prices during the crisis are sometimes noted, they are often attributed tohigh energy prices (e.g. Headey and Fan (2010, p. 25)). Mitchell (2008) notes that�energy�intensive� components of total production costs account for 6.7%�13.4% inUSDA farm survey data; but we know from economic theory that prices dependon marginal cost in a competitive industry. According to USDA (2013), fertilizeraccounts for 37% of the marginal cost of wheat3. Dorward and Chirwa (2011) pointout that high fertilizer prices increased Malawi's fertilizer subsidy expenditures in2008/2009. Yet we are not aware of previous literature taking fertilizer seriously as adriver of the food crisis, or estimating a formal econometric model in this direction.

An appreciation of the role of fertilizer may be crucial to understanding the past

1Headey and Fan (2010) provide a very comprehensive survey2One view is succinctly expressed by then-President George W. Bush in a White House press

conference on 29.04.2008: �I thought it was 85 percent of the world's food prices are caused byweather, increased demand and energy prices ... 15 percent has been caused by ethanol�

3In the 2012 wheat farmer survey reported in USDA (2013), fertilizer cost per acre is indicatedas $46.08; total operating costs per acre stand at $126.72. To these marginal costs, one must addan overhead � largely the rental cost of capital and land, since US agriculture is not labor intensive� of $175.28 per acre to obtain average costs. These overheads can be interpreted as a largely �xedcost, which should not a�ect competitive market prices.

2

food crisis and mitigating future ones. Given the large share of agricultural marginalcost accounted for by fertilizer, strong price passthrough from fertilizer to food is pre-dicted by standard economic theory4; careful econometric analysis is thus needed.Second, fertilizer prices are at times determined by an international export cartel(Hoekman and Saggi, 2007) or the degree of market concentration (Hernandez andTorero (2013)) and not market forces. Understanding the extent of the cartel over-charge is crucial to estimating the potential e�ects international competition enforce-ment against such cartels could have on food markets� particularly on developingcountries, where households spend roughly half their income on food (Mitchell, 2008).

Contribution: We contribute to the literature in three ways. First, using establishedtime serious methodology (cf. Ba�es (2010); Nazlioglu and Soytas (2011); Zhanget al. (2010)), we establish the strong cost passthrough from fertilizer to food. Sec-ond, we estimate the cartel overcharge (Connor and Bolotova, 2006; Connor, 2010)in the fertilizer market during the crisis period. Finally, we conduct a simulationof food prices in the counter�factual case that no cartel would have been formedand show that the majority of the crisis food price increase can be explained by thefertilizer cartel alone.

This paper proceeds by introducing data and methods in section 2, discussing theeconomic motivation of the regression equation and our the estimation strategy. Wepresent results in section 3, and place them in the context of the literature. Section4 simulates the development of food prices had no cartel been in place. Robustnesschecks are collected in section 5. Finally, we conclude.

2. Data and Methods

Model Speci�cation: Our interest is in determining the price passthrough fromfertilizer and energy indices on food prices. Thus, based on annual time series data,we estimate reduced form regressions explaining food commodity prices as a functionof fertilizer and energy indicies, controlling for in�ation and allowing a linear timetrend. This yields the regression model:

log(Xt) = µ+ β1log(IFERTILIZERt) + β2log(IENERGYt) (1)

+ β3log(MUVt) + τt+ εt

4See section 2 for details

3

where t denotes a year between 1960 and 2013, IFTERTILIZERt is the index offertilizer prices and IENERGYt the index of energy prices; moreover, a price de�ator,denoted MUVt and linear time trend t are added as controls. The dependent variableis denoted Xt; in our analysis, we consider a general food price index IFOOD as wellas prices for the two main agricultural commodities, wheat and maize. To facilitatecomparison with the existing literature, we also estimate a reduced model wherethe restriction β1 = 0 is imposed. Finally, we estimate the reduced model with thefertilizer index as the dependent variable. Throughout, we use OLS as our estimationmethod and carefully check for nonstationarities using appropriate versions of theAugmented Dickey Fuller (ADF) test. We assume that unobserved variables (e.g.other cost�push factors such as wages) are uncorrelated with the error term; furtherproperties of the error time series will be subjected to statistical testing below.

This statistical model can be given a structural economic interpretation. Considera competitive industry, where each �rm produces according to a constant returns toscale Cobb�Douglas production function. Then, the unit marginal cost function hasthe functional form of (1), allowing for a time trend in total factor producticity andpossibly non�neutrality of money (unless β3 = −β1−β2). Coe�cients β1 and β2 canthus be interpreted as the factor weights in the production function; moreover, thepredicted cost share of each factor corresponds to the coe�cients βi.

Data Sources: Our commodity price series is based on the World Bank Global Eco-

nomic Monitor Commodities database (World Bank, 2014). We study the broad�based food price index (series IFOOD), which includes grains, cereals and oils amongother food items as our dependent variable of interest; furthermore, two key com-modities � maize and wheat � are studied separately. Our explanatory variables arethe IFERTILIZER index, which includes all widely used fertilizer types: potassium,nitrogen and phosphate rock, and energy prices (IENERGY index). Individual com-modity prices are given in nominal USD; indices are also of nominal prices, with2010 as the base year. The Manufactured Unit Value (MUV) index, also providedby the World Bank, is included in the regressions as a de�ator. Data were current asof December 2014. The dataset is closely related to Ba�es (2010), who studies the1960�2008 period; however, his analysis omits consideration of fertilizer prices as anexplanatory variable.

Descriptive Analysis: The starkness of the ongoing food crisis is illustrated in �gure1. Food prices are drawn in the top panel; in 2010, they were 40% higher than twoyears previously; moreover, after a brief recovery in 2009, prices at the end of thesample period exceed the peak of the crisis level. Simultaneously with the food crisis,

4

Figure 1: Trends in food and energy/fertilizer prices

energy prices also rose � yet more mildly, by merely 25% over the same period. Butby far the most drastic change in the crisis period took place in the fertilizer market:by 2010, fertilizer prices more than tripled compared to 2006 levels.

All time series are trending upwards; this includes those in the data set that arenot drawn in the �gures. By visual inspection, one may suspect nonstationarity inthe form of a unit root; this intuition is supported through autocorrelation functions.Below, we conduct an ADF test to test this view against the alternative hypothesisof a trend stationary process.

Fertilizer�Energy Link: Fertilizer and energy price indices are closely linked. Thisis apparent from visual inspection: from the mid�1970s to 2005 or so, energy andfertilizer indices move almost entirely in parallel. This is explained by the fact thatfertilizer production is highly energy�intensive, be it due to mining (potassium, phos-phate rock) or as a direct input (e.g., natural gas in the Haber�Bosch process). Yetthere is independent variation in fertilizer prices: in both 1973/1974 and 2007/2008,fertilizer prices are much higher than one would expect from energy prices. More-over, in both of these periods food prices were abnormally high � indeed, at crisislevels. It is this variation that underlies the more formal analysis below and allowsus to estimate the direct impact of fertilizer prices on food commodities.

5

3. Results

Unit Root Tests: We use the Augmented Dickey�Fuller (ADF) test testing the nullhypothesis of a unit root against the alternative of a stationary autoregressive process.The optimal lag order is determined through the Akaike Information Criterion (AIC);results are presented in table 1. It is evident that for the levels series � displayed inthe left panel � the ADF test largely fails to reject the unit root hypothesis. For thefertilizer and wheat series, there is weak evidence for stationarity at the 10% level;we consider this signi�cance level insu�cient to reject the null. Thus, all levels seriesare considered to have a unit root. In the right panel, the corresponding ADF testfor the �rst�di�erenced series is displayed. Here, results are rather straightforward:the null hypothesis is rejected at the 1% level for all series. Hence, the evidencesuggests that our data follows an I(1) process.

tau3 sig lags

IFERTILIZERS -3.29 * 1.00

IENERGY -2.79 5.00

IFOOD -2.16 3.00

WHEAT_US_HRW -3.3 * 1.00

MAIZE -1.75 2.00

tau3 sig lags

D.IFERTILIZERS -6.16 *** 1.00

D.IENERGY -4.77 *** 1.00

D.IFOOD -6.48 *** 1.00

D.WHEAT_US_HRW -5.97 *** 1.00

D.MAIZE -6.24 *** 1.00

Table 1: Unit Root Tests

The I(1) �nding has profound implications for our econometric analysis of model(1). On the one hand, standard errors and the R2 statistic are biased towards ex-cessive signi�cance (the �spurious regression� problem); hence signi�cance tests needto be interpreted with care. On the other hand, as we show below, the model infact provides a cointegrating regression; hence the error term εt will be stationaryand, importantly, OLS parameter estimates are super consistent, i.e. they convergeto the population at a faster rate than in an I(0) regression with the same sample size.

Estimation Results: Table 2 collects estimation results. Given the I(1) data, univer-sally high values of adjusted R2 (ranging from 0.80 to 0.94) are expected; encourag-ingly, we are able to reject statistical signi�cance of some parameters. In particular,there appears to be essentially no time trend in commodity prices. Moreover, thede�ator MUVt is mostly insigni�cant, reaching the 5% level only in few speci�ca-tions. In regressions (1-3), which include the fertilizer index, there is robust evidencein favor of cointegration. The Augmented Dickey Fuller test, computed on residualsof the estimated equation, is able to reject the null hypothesis of a unit root in all

6

Dependent Variable β1 β2 β3 100× τ Adj. R2 ADF

IFOOD 0·420∗∗∗(0·057) 0·047 (0·052) 0·256∗∗ (0·107) −0·376 (0·003) 0·939 −3·596∗∗WHEAT_US_HRW 0·474∗∗∗(0·061) 0·051 (0·055) 0·238∗∗ (0·115) −0·542∗ (0·003) 0·934 −5·371∗∗∗MAIZE 0·473∗∗∗(0·077) 0·062 (0·070) 0·082 (0·145) −0·317 (0·004) 0·882 −3·801∗∗IFOOD 0·308∗∗∗(0·054) 0·005 (0·146) 0·121 (0·004) 0·874 −2·83WHEAT_US_HRW 0·347∗∗∗(0·059) −0·045 (0·161) 0·020 (0·004) 0·856 −3·385∗MAIZE 0·358∗∗∗(0·067) −0·201 (0·181) 0·244 (0·005) 0·795 −2·949IFERTILIZERS 0·624∗∗∗(0·093) −0·597∗∗ (0·253) 1·186∗ (0·007) 0·849 −3·386∗

Table 2: Estimation Results

cases at the 5% level; in the case of wheat, the ADF very robustly rejects the nulleven at a one�percent signi�cance level.

There is mixed evidence regarding cointegration in the restricted model whichomits the fertilizer index. Two equations rejecting no cointegration at the 5% level,and two equations o�ering no evidence of cointegration. Notably, for food prices,there is no evidence of cointegration with energy. Relatedly, Zhang et al. (2010)also reject both cointegration and Granger causality from fuel to food in a di�erentdataset. Ba�es (2010) found cointegration for a similar dataset to ours, but with ashorter time period. In line with Harri et al. (2009), our results indicate lack of arobust cointegration relationship between energy and food across di�erent commodi-ties.

Fertilizer is a consistently statistically and economically signi�cant price driver infood markets. Estimates for β1 range from 0.42 to 0.47 depending on the commodityor index considered; this implies that a doubling in fertilizer prices would be predictedto lead to at least a 42% increase in food prices. Given the structural interpretation ofthe model, this would imply a 42% cost share of fertilizers. Using a producer survey,USDA (2013) estimates a fertilizer operating cost share of 37% for the US. Kenkel(2014), based on data from wheat production budgets of Oklahoma State Universityfarms, �nds that fertilizer costs consistently exceed 30 percent of total costs. In thecorn belt, larger cost shares (38 percent to 40 percent) are evident. Our estimatedcost shares � between 42 and 47 percent � are somewhat higher, but quantitatively inline. This is especially reassuring since the methodologies are clearly very di�erent.

Energy does not have a statistically signi�cant e�ect on food prices once fertilizeris included. This �nding is especially striking in light of the spurious regressionproblem: due to biased standard errors, it may be di�cult to reject truly insigni�cantparameters. Moreover, hovering around 5%, estimated e�ect sizes are economically

7

minor. Even in the USDA (2013) producer survey, combined costs for �fuel, lube andelectricity� account for merely 15% of operating costs. Note that since commodityprices are generally quoted �free on board�, the bulk of transportation costs does notenter directly in the data.

The energy�food price link reappears once fertilizer prices are excluded. Ourestimates suggest an energy index elasticity in the range 0.31�0.36 of food priceswith respect to the energy index; these �ndings are close to earlier estimates in theliterature5. Combined with the lack of cointegation between energy and food, thissuggests that earlier studies � which omitted fertilizer � may have identi�ed only theindirect impact of energy on food, channeled through higher fertilizer prices.

At an estimated 62% cost share, energy prices have a very strong cost passthroughto fertilizer prices. Due to the energy�intensive production of fertilizers � with energycosts accounting for up to 90% of nitrogen fertilizer cost (Headey and Fan, 2010, p.25) � the estimate is plausible.

4. Did the Fertilizer Cartel cause the Food Crisis?

During the �perfect storm� of 2007/2008, the food crisis coincided not only withan energy crisis but also the establishment of a fertilizer cartel. Given our estima-tion results above, we are interested in isolating the e�ect of the cartel formation onfertilizer prices, and hence, ultimately on food. To this end, we �rst seek to datethe recent cartel period in the fertilizer industry and then an augmented fertilizerpricing equation in the spirit of model (1). Based on predicted fertilizer prices in thecounter�factual case of no cartel formation, we then simulate what food prices hada cartel not been formed.

Dating the Cartel: The propensity of the fertilizer industry to establish cartels iswell known (al Rawashdeh and Maxwell, 2014). However, competitive periods � af-ter breakdown of a cartel � also occur. This makes it imperative to precisely datecartel periods for empirical analysis. Taylor and Moss (2013) conduct a detailedmicro�level study to distinguish di�erent phases of market competition in fertiliz-ers. They conclude, based on qualitative and quantitative evidence, that the cartelexercised in market power in the 2008�2012 period6. Building on their work, we

5Ba�es (2010), using a similar data set until 2008, estimates an elasticity of 0.27; due to persis-tently high energy and food prices, the correlation appears to be stronger in the extended sampleused in the present analysis. Relatedly, Gilbert (1989) estimates an elasticity of 0.25 using quarterlydata from 1965.Q1 to 1986.Q2

6Since the fertilizer export cartel is not constrained by usual anti�trust enforcement, the cartel is

8

introduce a new dummy variable CARTELt, which takes value 1 in those years andzero otherwise (i.e. 1960�2007 and 2013).

Estimating Cartel Overcharges: We seek to estimate a model that disentangles theimpact of energy prices and cartel formation on fertilizer prices. To this end, weaugment the version of model (1) already used to estimate potash prices with thecartel variable. Suppressing for brevity the intercept and usual controls into a vectorXt of controls and associated coe�cients γ, we obtain:

IFERTILIZERt = 0.51∗∗∗IENERGYt+ 0.63∗∗∗∗CARTELt+ γ′Xt + εt(0.08) (0.15) (Adj.R2 = 0.89)

(2)

The cartel impact on fertilizer prices is very strong. According to the model,cartel formation was associated with a 63% overcharge in the fertilizer market. Sta-tistical signi�cance at the 1% level, in spite of only few years of observations withthe cartel in place, is a re�ection of the magnitude of the economic e�ect. Suchlarge overcharges are not uncommon: in a meta�study of cartel overcharges, Connorand Bolotova (2006, table 3) report an average overcharge of 54.2% for internationalcartels based on a sample size of 365 cartel episodes. Allowing for the cartel reducesthe earlier estimate of energy passthrough on fertilizers; this is intuitive, since thecartel period coincided with high energy prices.

Impact on the Food Crisis: The fertilizer cartel can potentially explain nearly two�thirds of food price increases during the crisis time. According to our earlier es-timates, the 63% fertilizer cartel overcharge would be expected to lead to a 26%increase in food prices. Given the observed increase in the food price index of 40%,our model attributes 65% of the crisis price hike to the fertilizer cartel. Figure 2 showsthe counter�factual food and fertilizer prices without cartel formation side�by�sidewith actual values: it is apparent that, according to our estimates, food prices couldhave remained nearly stable over recent years in the absence of the cartel7.

relatively transparent. For example, according to the Potash Investing News, potassium producerssimultaneously announced in 2009 an intention to further reduce output (Toovey, 2009). Followinga rapid collapse in fertilizer prices, Uralkali Chief Executive O�cer Vladislav Baumgertner wasdetained in a former KGB prison in Belarus after talks on restoring the cartel were apparentlyunsuccessful (Fedorinova and Kudrytski, 2013)

7Interestingly, results also indicate the 1973/1974 food crisis would have been considerably milderin the simulation. This result can be attributed partly to weaker energy passthrough in the enhancedmodel; but it also suggests further research is needed to understand the structure of fertilizer marketduring the earlier crisis

9

Figure 2: Actual and counter�factual food and fertilizer prices without cartel formation

5. Robustness Checks

A possible threat to the estimation strategy outlined in the previous section isthe impact of the oil crises on the relation between energy and fertilizer prices andthe market turbulences in the aftermath. Therefore, to check the robustness of theresults from the previous section we introduce a dummy variable called fakeCartel

that similarly to the cartel dummy captures 5 years starting with the crisis year andalso takes value 1 for years 1974-1978. In addition dummy variables for the crisisyears 1974 and 2008 oil crises are introduced. The results are presented in table 3.

These regressions show that the cartel overcharge is consistently estimated ataround 60% in all speci�cations. Unsurprisingly addition of a 2008 crisis dummy de-creases the initial estimate. The placebo test with the fakeCartel dummy shows thatthe increase in potash prices is not caused by the oil crises. Although in speci�cation(4) the dummy is signi�cant at 10% level, as speci�cation (5) shows it is only due tothe fact it includes the 1974 crisis year.

6. Conclusion

Since the Green Revolution, the increased application of fertilizer has been instru-mental in raising crop yields worldwide. But, as this paper shows, this dependence

10

(1) (2) (3) (4) (5) (6)

cartel 0.63∗∗∗ 0.54∗∗∗ 0.56∗∗∗ 0.56∗∗∗

(0.15) (0.16) (0.14) (0.14)IENERGY 0.51∗∗∗ 0.50∗∗∗ 0.44∗∗∗ 0.57∗∗∗ 0.56∗∗∗ 0.42∗∗∗

(0.08) (0.08) (0.07) (0.10) (0.09) (0.08)MUV −0.16 −0.15 −0.05 −0.57∗∗ −0.51∗∗ −0.05

(0.24) (0.24) (0.21) (0.25) (0.23) (0.21)t −0.00 0.00 0.00 0.02∗∗ 0.02∗∗ 0.00

(0.01) (0.01) (0.01) (0.01) (0.01) (0.01)crisis08 0.43 0.45∗ 0.45∗

(0.28) (0.24) (0.24)crisis74 0.94∗∗∗ 0.81∗∗ 0.85∗∗∗

(0.22) (0.30) (0.24)fakeCartel 0.26∗ 0.11 0.12

(0.15) (0.15) (0.12)(Intercept) 2.60∗∗∗ 2.60∗∗∗ 2.30∗∗∗ 3.66∗∗∗ 3.50∗∗∗ 2.28∗∗∗

(0.68) (0.67) (0.58) (0.71) (0.67) (0.58)

R2 0.90 0.90 0.93 0.87 0.88 0.93Adj. R2 0.89 0.89 0.92 0.86 0.87 0.92Num. obs. 54 54 54 54 54 54∗∗∗p < 0.01, ∗∗p < 0.05, ∗p < 0.1

Table 3: Regression Results: Cartel

on fertilizer is not merely technological, but economic: high fertilizer prices directlytranslate into high food prices. This link is crucial to our understanding of the recentfood crisis, during which prices rose by 40% on average. We estimate this was largelycaused by the formation of an international export cartel for fertilizers. Accordingto our model, the cartel overcharge in the fertilizer market amounted to 63%. Thisdirectly led to a 26% increase in the food price index. In other words, the formationof the fertilizer cartel explains nearly two�thirds of the crisis price increase.

While many factors doubtlessly contributed to the food crisis, we believe therole of fertilizer should be taken seriously. First, while our results are stark, theyare in line with the related literature. We estimate fertilizer cost passthrough fromaggregate time series data running many years; but the share of fertilizer in marginalcost implied by our model is close to estimates obtained from production budgets and

11

farmer surveys. Our estimated cartel overcharge of 63% is typical for internationalexport cartels. Finally, we control for energy prices; according to our estimates, foodprices would certainly have been high in 2007/2008 � due to the simultaneous energycrisis � but they would not have reached crisis levels.

These results highlight the importance of addressing fertilizer cartel � the �OPECof world potash markets� (Scherer, 1996) � in combating food crises. In recent history,various export cartels have at di�erent times dominated the fertilizer market (cf.al Rawashdeh and Maxwell (2014); e.g. Newman (1948) discusses the role of theGerman potassium cartel in the Nazi economy). These cartels �ourished, with eitherexplicit or implicit state backing, because of the absence of e�ective internationalcompetition authorities and enforcement (Marquis, 2014)8. Sokol (2008) discussespossible institutional reforms to mitigate export cartels; Taylor and Moss (2013)make the case for global antitrust enforcement in the fertilizer industry.

Perhaps above all, this paper is a call for further research. Since the �fertilizerhypothesis� is new, it is in need of further corroboration for additional food commodi-ties and from other datasets. The recommended steps include data disaggregation.The analysis performed on the indices may miss some important subtleties of thefertilizer and energy markets. It would be very valuable for the agricultural policyto understand the co-movement of prices of nitrogen, potassium and phosphorus fer-tilizers. Furthermore, the recent dramatic shocks to the fertilizers markets causedby the cartel formation and the signi�cant decrease of natural gas prices may helpto identify the links between the di�erent fertilizer types. Furthermore, a focus onthe oil and natural gas prices rather than the analysis compound index would allowto separate the e�ect of cartel formation from the recent processes that change theshape of the relation between food, energy and fertilizer prices - the introduction ofthe fracking technology and the recent increase in the use of biofuels. Finally, wetreated fertilizer cartel formation as exogenous; but a better understanding of thereasons for success and breakdowns in fertilizer cartels over time is sorely needed.

7. References

al Rawashdeh, R. and Maxwell, P. (2014). Analysing the world potash industry.Resources Policy, 41:143�151.

Ba�es, J. (2010). More on the energy/nonenergy price link. Applied Economics

Letters, 17(16):1555�1558.

8See in particular the discussion on pp. 65.

12

Connor, J. M. (2010). Price �xing overcharges: Revised 2nd edition. Available at

SSRN 1610262.

Connor, J. M. and Bolotova, Y. (2006). Cartel overcharges: survey and meta-analysis. International Journal of Industrial Organization, 24(6):1109�1137.

Dorward, A. and Chirwa, E. (2011). The malawi agricultural input subsidy pro-gramme: 2005/06 to 2008/09. International Journal of Agricultural Sustainability,9(1):232�247.

Fattouh, B., Kilian, L., and Mahadeva, L. (2013). The role of speculation in oilmarkets: What have we learned so far? The Energy Journal, 34(3).

Fedorinova, Y. and Kudrytski, A. (2013). Uralkali ceo recently ar-rested in belarus over potash cartel bust-up moved from kgb prisonto house arrest. http://business.financialpost.com/2013/09/26/

uralkali-ceo-recently-arrested-in-belarus-over-potash-cartel-bust-up-moved-from-kgb-prison-to-house-arrest/.

Gilbert, C. L. (1989). The impact of exchange rates and developing country debt oncommodity prices. The Economic Journal, pages 773�784.

Harri, A., Nalley, L., Hudson, D., et al. (2009). The relationship between oil, ex-change rates, and commodity prices. Journal of Agricultural and Applied Eco-

nomics, 41(2):501�510.

Headey, D. (2011). Rethinking the global food crisis: The role of trade shocks. FoodPolicy, 36(2):136�146.

Headey, D. and Fan, S. (2010). Re�ections on the global food crisis: How did ithappen? how has it hurt? and how can we prevent the next one? TechnicalReport 165, International Food Policy Research Institute.

Hernandez, M. A. and Torero, M. (2013). Market concentration and pricing behaviorin the fertilizer industry: a global approach. Agricultural Economics, 44(6):723�734.

Hoekman, B. and Saggi, K. (2007). Tari� bindings and bilateral cooperation onexport cartels. Journal of Development Economics, 83(1):141�156.

Irwin, S. H. and Sanders, D. R. (2012). Testing the masters hypothesis in commodityfutures markets. Energy Economics, 34(1):256�269.

13

Jenny, F. (2012). Export cartels in primary products: The potash case in perspective.Trade, Competition, and the Pricing of Commodities, page 99.

Kenkel, P. (2014). Understanding fertilizer price risk. http://agecon.okstate.

edu/coops/files/agec-260.pdf.

Marquis, M. (2014). Idea merchants and paradigm peddlers in global antitrust. InBailleux, J. and Vauchez, A., editors, Exploring the Transnational Circulation of

Policy Paradigms: Law Firms, Legal Networks and the Production of Expertise in

the Field of Competition Policies. Robert Schuman Centre for Advanced Studies.

Mitchell, D. (2008). A note on rising food prices. World Bank Policy Research

Working Paper Series, Vol.

Nazlioglu, S. and Soytas, U. (2011). World oil prices and agricultural commodityprices: evidence from an emerging market. Energy Economics, 33(3):488�496.

Nazlioglu, S. and Soytas, U. (2012). Oil price, agricultural commodity prices, andthe dollar: A panel cointegration and causality analysis. Energy Economics,34(4):1098�1104.

Newman, P. C. (1948). Key german cartels under the nazi regime. The Quarterly

Journal of Economics, pages 576�595.

Sanders, D. R. and Irwin, S. H. (2010). A speculative bubble in commodity futuresprices? cross-sectional evidence. Agricultural Economics, 41(1):25�32.

Scherer, F. M. (1996). International trade and competition policy. Technical report,ZEW Discussion Papers.

Serra, T. and Zilberman, D. (2013). Biofuel-related price transmission literature: Areview. Energy Economics, 37:141�151.

Sokol, D. D. (2008). What do we really know about export cartels and what is theappropriate solution? Journal of Competition Law and Economics, 4(4):967�982.

Taylor, C. R. and Moss, D. L. (2013). The fertilizer oligopoly: The case for globalantitrust enforcement. Technical report, American Antitrust Institute.

Timmer, C. P. (2010). Re�ections on food crises past. Food policy, 35(1):1�11.

Toovey, L. M. (2009). Potash prices at record high. http://potashinvestingnews.com/354-potash-prices-at-record-high.html.

14

USDA (2013). Wheat production costs and returns per planted acre.http://www.ers.usda.gov/datafiles/Commodity_Costs_and_Returns/Data/

Current_Costs_and_Returns_All_commodities/cwhea.xls.

World Bank (2014). Global economic monitor: Commodities database.http://databank.worldbank.org/data/views/variableselection/

selectvariables.aspx?source=global-economic-monitor-(gem)

-commodities.

Zhang, Z., Lohr, L., Escalante, C., and Wetzstein, M. (2010). Food versus fuel: Whatdo prices tell us? Energy Policy, 38(1):445�451.

15