A forecasting model is not viable without taking into account the volatile nature of agricultural prices
The liberalization of international agricultural trade, which has been a subject of fierce negotiation within the WTO, is advocated by economic models such as that of the World Bank, which put forward their supposedly positive impact on developing countries.
As we have already shown, however, they are based on unrealistic hypotheses such as the perfect fit of supply with demand, which implies that the prices of agricultural raw materials do not fluctuate and that world hunger does not exist.
Such a viewpoint is dangerous when more than 800 million people are affected by famine and when there is speculation with regard to most agricultural products, which leads to the impoverishment of farmers all over the world.
This is why it is vital that the volatile nature of agricultural prices be integrated within economic forecasting models in order to gauge the potential benefits of liberalizing international agricultural trade. Major international models, however, having reached an impasse over this important agricultural specificity, can neither explain, nor simulate, nor forecast.
This is according to Jean-Marc Boussard, Françoise Gérard and Marie Gabrielle Piketty, in their most recent work entitled "Libéraliser l’agriculture mondiale? Théories, modèles et réalités" ("Liberalizing World Agriculture? Theories, models and realities") an extract of which is given below.
"…We may wonder whether the current trend to liberalize everything is not an effect rather of a passing trend than of the result of an actual deepening of knowledge about social arithmetic. In short, we may simply be talking about an effect brought about by the urge felt from time to time by political leaders before reality makes them see more sense. Such an interpretation is largely suggested by analysis of the facts with regard to two crucial aspects which set apart the differing theories in favour of agricultural liberalization: while some see it as possible to disregard price fluctuations when calculating the benefits of liberalization, others see in expanding the global market the specific possibility of stabilising prices.
It has been seen that, with regard to the thinking behind liberalization, price fluctuations are largely ignored, except in respect of the affirmation that liberalization will lower them. Those who are against liberalization, on the other hand, emphasise the fact that they are generated by the market itself and that they play a vital part in forming supply. Two points need to be verified in order to attempt to decide between these two analyses:
- Are these fluctuations important with regard to production-related decisions? The answer to this is without doubt "Yes", as we shall see;
- Are these fluctuations generated by random phenomena outside the system, or do they derive from market operations, in such a way as to make it impossible for a market solution to eliminate them? Economists are unable to agree on this.
Price fluctuation and production
The thinking of agricultural economists is that the role played by risk in decisions made by producers is strangely ambiguous. On the one hand, everybody maintains that it is important. On the other, it is usually very difficult to reveal this role and to measure its effect on the global economy.
The importance of decision-related risk derives initially from surveys conducted among producers: when asked, it is only rarely that they do not express uncertainty with regard to the future and with regard to reaching decisions in this context. Sociological surveys1 are, however, difficult. One can never be totally sure that responses are sincere, particularly within a field such as this one, which is very difficult to apprehend because of the concepts involved, containing, as they do, nothing by way of self-evidence2 . There are therefore few conclusions to be drawn from this type of source.
There are other reasons to believe in the importance attached to price risk, which derive from the habit models of farmers.
Almost fifty years ago, an American researcher named Rudolf J. Freund (1956) conducted studies into optimal rotation for farmers in South Carolina, taking the prices at the time into consideration. He used a revolutionary calculation method which made it possible to research a solution involving simultaneous equation systems and "linear programming", which would be destined for a bright future. The solution to the problem which was thereby submitted to computer was very different to the one that was being adopted by farmers: rather than corn and potatoes, which then made up the bulk of local production, his model recommended setting aside almost all of the region to "beef" and to "autumn cabbage".
Freund did not, however, believe that farmers chose a rotation method which was so different to the one which would have made it possible for them to maximise their profits simply by chance or foolishness. There must have been reasons for this. He considered the risks associated with cabbages and beef, which were, in fact, much greater than those associated with corn, as they were not given support by US policy at the time. He then used the specifications of Markowitz3 in his model, in order to take the effects of risk into account. The result was an optimal rotation which was very similar to actual rotations.
Freund’s experiment has since then been reproduced thousands of times throughout the world with the same success. It demonstrates two facts:
- Firstly, that it is impossible suitably to reproduce habits of farmers if one is unaware of the risks to which they are subject;
- Secondly, and most importantly, that the risk, which differs between cultures, shapes agricultural supply to at least as great an extent as the average price level.
The latter is extremely important. We know that the role of price in a market economy is to transmit information: from the producer to the consumer with regard to production difficulty and from the consumer to the producer with regard to what is desired by the public. But it is not, or is not only, the average price level which plays this role. Variability (or volatility as we now refer to it) is also very important. In addition to this, the average level and volatility do not necessarily push in the same direction.
Let us imagine that the price of wheat were to double suddenly, after a long period of stability. Farmers would be able to draw two opposed conclusions: they could either go by the average price level, which would lead them to increase production as a response to market indicators; or they could see it as a change without any particular long-term significance and they would only consider the fact that price volatility has suddenly increased, which would lead them to decrease production of a product deemed to be "dangerous". Of course, given that the first reaction is likely to "re-establish market equilibrium" (increased production will restore the price towards its long-term balance), this does not go for the second, which, on the contrary, will aggravate the shortage.
This example shows that the major disadvantage of risk is "blurring market signals". These can be mis-interpreted and lead to a worse situation than that which needed to be corrected. This makes up a very serious criticism of the market economy. However, if markets are functioning badly, this should show up in macroeconomic data. With the means of econometric investigation available today, it should be possible to reveal these phenomena by using statistical tests.
Strangely, however, it is not that simple. Most surveys conducted in this area only allow price volatility a small or negligible influence on the total volume of agricultural produce in a country or region. One of the reasons that explains this is that statistical tests probably involve hypotheses that are different from the one which could be expressed as "price volatility lowers (and stability increases) the supply quantity of any agricultural product". In fact the hypothesis which is usually tested in statistical surveys is along the lines of: "last year’s average prices" or "volatility measured last year" affects "production this year". Such a hypothesis, however, is in a way too precise to be verified. In particular, volatility is a long-term concept, with a slow rate of adaptation. In these conditions, it is not surprising that hypotheses involving a very rapid reaction by the decision-makers are unverified.
In order to take this phenomenon into account, it is necessary to conduct qualitatively-based tests, with prices being classified as "stable" or as "unstable". In this way, it would be possible to discern very significant differences among the different series categories (Gérard et Boussard, 1994).
Price stabilisation is therefore of fundamental importance, with particular regard to expanding production in developing countries. […].
Are price fluctuations endogenous?
We have established above the importance of the stakes involved: if price fluctuations are "exogenous", brought on for example by climate variations, then liberalization should have a beneficial and absorbent effect. If they are endogenous, generated by a cobweb system, then only well-conducted policies which isolate agriculture from the market, such as those put into force by Roosevelt in 1935, have any chance of lowering them.
Sadly, it is very difficult to decide one way or the other.
Many mathematicians have attacked the creation of statistical tests for "chaos4 " existence. As these series give every appearance of being "random", it is extremely tricky to make a distinction. While in general, attempts are made to reveal the "sensitivity to initial conditions", given that these initial conditions are themselves random, this random nature exists, as it were, in the statistical tests.
Despite this, and though they are not without flaws, tests have been finalised. Using them for a series of prices of agricultural raw materials usually results in mitigated conclusions. It cannot be discounted that these series are subject to chaotic dynamics, but it equally cannot be categorically stated that they are 5 .
Are price fluctuations climate-based?
Conversely, can it be asserted that agricultural price fluctuations derive only from random causes outside the system, in particular climate-based ones?
Here again, and very strangely, empirical surveys are very scarce. The difficulty lies not in revealing the relationship that exists in a given environment between climate and agricultural production. We know perfectly well how to do this, and the relationship is both unquestionable and important. Models for crop behaviour even make it possible to forecast quite precisely what the yield will be for a given cultivation method in a given climate environment. The real question is whether climate fluctuations exert an influence on areas which are large enough that the variations in production that they cause can disrupt the markets. This, however, is not at all evident.
The drought of 1976 provides a revealing anecdote. Everybody in France remembers the disaster which led to an exceptional tax known as "drought tax" to assist farmers affected by the cataclysm. At the time there was a French agriculture model in use by the INRA (France’s national institute for agronomic research) which included a "climate index" (Boussard 1975). It was therefore tempting to make use of it in order to evaluate the damage very precisely. However, applying the 1976 parameters to the model in question showed almost no decrease in production. The reason for this was that, while the meteorological indexes were in fact very bad in the west and centre-west of the country, they were excellent elsewhere, and the one made up for the other.
Agricultural production actually did decrease in 1976, but probably for other reasons: for two years, because of the increased price of oil-based products associated with the Six Day War, farmers had been reducing the amount of fertilizers they used, as statistics from the time show. Lowering the amount of fertilisers given to crops does not necessarily have immediate effects, since there are stores within the ground and plenty of "buffer phenomena". In the long term, however, it was inevitable that it would lead to a decrease in production, and this was what happened in 1976.
This story is instructive in more than one way. Firstly, it is comical that public opinion readily swallowed the drought story, while it probably would have balked at saddling the emirs with responsibility for another calamity since it had been they who had increased oil prices6 . The truth is that, in all instances of catastrophic decreases in agricultural production, an explanation involving natural catastrophe is always well-accepted. This does not mean that it is pertinent. Authors such as Amartya Sen7 have clearly demonstrated that, on the contrary, famines are usually caused by phenomena which are not "technical" in nature.
It also shows that, even in a relatively small country such as France, meteorological events have a tendency to compensate themselves. This means that benefits anticipated from liberalization in this regard are probably slender since, in reality, they have already been reaped by liberalization in most countries of medium importance.
Finally, it makes it possible to situate the difficulty of conducting many of these types of survey: the contingent and questionable nature of any average meteorological index for a region that is fairly enormous. In fact, a superficial look at one of the meteorological maps shown in all of the dailies gives the impression of immense phenomena, covering vast areas such as the North Atlantic, Siberia and the Sahel. This is not a false impression at a given moment; if "climate-based anomalies" covered such areas, they would quite clearly be the cause of serious market disruption.
For crops, however, what counts is not the rainfall or temperature at a particular moment8 . It is what happens over a longer or shorter period, according to the species and particularly to the stage of development, and which can be measured in days and sometimes weeks. Changing meteorological phenomena, on the other hand, are measured in hours. None is similar to the preceding one and none affects the same area.
In this way, in order to construct a meteorological index, differences from the average that have been calculated over various periods need to be evaluated. Ideally, light should be shed on the links between instantaneous fluctuations observed and those that have been observed over recent periods. The "correlation distance" should then be evaluated from this - the distance beyond which statistical links between anomalies become so weak that that they become negligible and at which point one can contemplate applying them to the Law of Large Numbers. Finally, an examination should be carried out to see whether the areas involved for a given hazard, in this context, are big enough to ensure that the corresponding production variations will have the ability to disrupt the markets … such a programme and such an accumulation of difficulties are enough to quench the enthusiasm of the young doctorand who is looking to confront the problem!
The almost total lack of reliable surveys relating to this matter is probably due to these difficulties – and to the fact that social structure taken as a whole "does not want to know". Some isolated surveys do exist. For example, Martineu and Tissot (1993), in a survey which, sadly, was never published, attempted the task and examined whether "drought" could actually have been at the root of the Sahel difficulties – which is an obligingly widespread idea, particularly among NGOs. The "correlation distance" to which we referred previously, would be in the order of 100 kilometres. This excludes the idea of drought from Dakar to Djibouti, but does not reject the possibility of local market disruption in this zone where regions are very much compartmentalised, and where transport is difficult. Such a result, however, if it were also valid in France, would explain the lack of impact of the 1976 drought of which we spoke earlier. The question may be tackled in a different way. A market could be selected for which the collection area covers a reduced section and research carried out into whether the climate in this area explains the fluctuations observed. This is what was done with great care by Richard Roll (1984) in a survey on orange juice in Florida. In fact, the citrus fruit area in this state is situated in a zone in which the correlation distance is exactly in the order of 100 kilometres. Besides this, orange juice must surely be one of the agricultural products that is subject to the least state intervention. It was therefore tempting to use it as model experimental material. The results are very interesting:
"The market price of concentrated orange juice is affected by the weather, in particular by lower temperatures. There is a mystery, however … despite the fact that the weather is the clearest determining factor as regards harvest volume, climate-related anomalies only explain a small fraction of final price volatility … There is a large area of uncertainty with regard to explaining price volatility".
If climate plays such a minor role in the case of a product for which one has clear expectations, we could assume that it plays almost no role at all in primary products covering the entire globe. However, we could also assume that mechanisms involving "endogenous fluctuations" which we referred to earlier are well-placed to play the role of "missing link" as indicated by Roll.
Whatever the case, it is clear that these price fluctuations cannot ever be ignored when evaluating the benefits of liberalization. […]. We now see that the risk, by reducing the space available for production possibilities, quite clearly plays the role of negative technical progress. If liberalization reduces this risk, then that is a good thing, […]. But if it turns out that it increases risks, then the question arises as to whether the disadvantages associated with the risk might not prevail over the benefits gained from exploiting comparative advantages.
It would be quite reasonable to wonder whether the controversy surrounding the liberalization of agricultural trade, which has been raging for the past few centuries, did not originate there. Liberalization was probably a tempting proposition during each era, since the usefulness of exploiting comparative advantages was apparent. Each time an attempt was made, however, there was a clash with an increase in price volatility, which wiped out the anticipated benefits, leading to a speedy retreat.
Is it not possible that the same thing will happen with the Doha negotiations9?
To respond to these concerns and to examine its validity, two things need to be done:
- An attempt must be made to calculate, in order of magnitude, the benefits anticipated from liberalization, while disregarding the objections which have recently been raised. Over the past few years, a number of international agencies have attempted to do so by using a version of an "economic model" that we shall refer to as "standard";
- An alternative to the standard model must be constructed, which will take into account all of the phenomena discussed above 10 .»
1Pierre Bourdieu, the famous sociologist who recently died, and who began his career in rural sociology, even referred to "combat sport". More seriously, the major problem here is how correctly to interpret the responses provided by those concerned. A good example of this is provided by Morlon (1987).
2 Is it not therefore obvious to ask a farmer "Do you have any idea about price variations of potatoes?", when he knows the exact price for the previous week and is fully aware that the price is speculative?
3 Markowitz studied the influence of taking risk into account in the decision-making process by introducing the notion of "optimal portfolio" as a "high potential" but risky mix of values, and of values that show small profitability but which are safe – with a measure dependent on "aversion to risk" by the holder.
4 In mathematics, the word "chaos" has a very precise meaning, referring to phenomena which are similar to those which have been described in developments on cobweb.
5 See Leuthold and Wei (1998), Hommes et al. (1998), Hölzer and Precht (1993), Burton (1993), Lücke (1992), Chavras and Holt (1993).
6 We could also certainly ask whether they were actually responsible for it, which would lead us in a very different direction!
7 See Drèze and Sen (1989) in particular
8 The only exception, from this point of view, is frost. It nevertheless affects different crop species in different ways.
9 The Doha negotiations concern the liberalization of international trade (agriculture and services in particular). Having been initiated in 2000, these gave rise to a declaration following the fourth ministerial conference organized by the World Trade Organization in November 2001 in Qatar. This conference placed emphasis on the importance of development objectives for southern countries.
10 This has been undertaken by WOAgri through the construction of the new agricultural regulations model, with the participation of Jean-Marc Boussard and Françoise Gérard (editor’s note).