Technical paper

DAIRY TRADE LIBERALISATION WITH AGLINK

(Aglink changes required for the international dairy trade liberalisation scenario)

January 2004

OECD

Directorate for Food, Agriculture and Fisheries, Committee for Agriculture

Paris

Introduction

The objective of this technical paper is to identify how to produce trade liberalisation scenarios with AGLINK rather than to arrive at definitive conclusions about the market impact of these hypothetical reforms. For this reason this document is much more a reference manual on how to produce these scenarios with AGLINK than a document explaining the economic and political implications of such reform. Since the results of the alternative scenarios are baseline dependent, it is important to describe rapidly the key elements of the baseline published by the OECD in Agricultural Outlook 2003-2008.

Baseline

The baseline from which these scenarios will be compared is the OECD 2003 Medium Term Outlook (MTO). The simulation period of the MTO is from 2002 to 2008 and the key elements as it regards the dairy sector are the following;

-  EU subsidized exports of cheese and other dairy products (including whole milk powder (WMP)) are always at the World Trade Organisation (WTO) limits while exports of butter and skim milk powder (SMP) are always much below.

-  Canadian export of dairy products is limited by the WTO value limit and the ruling against commercial export of milk (CEM).

-  The rapid growth in SMP stocks in the USA resulting mostly from unconstrained imports of certain dairy products stops at the beginning of the outlook and return to more normal level through increased subsidized exports.

-  Milk production in Oceania continues to grow especially in New Zealand.

-  World prices of dairy products over the outlook period are lower than the average of the last seven years especially for butter.

Introducing milk supply functions

For countries which operate a quota system, the Aglink model has not specified the underlying milk supply function and milk production has been fixed at the quota level. In order to assess the impacts of trade liberalisation in the dairy sector supply functions for quota countries needed to be built into the model. The results of such scenario will depend critically on parameter assumptions of those supply functions and in particular on the true marginal cost, the elasticity of the milk supply function and the speed of adjustment of supply since AGLINK is a dynamic model. Discovering these parameters in the context of regulated production systems is not an easy task.

The economics of production regulated markets in the dairy sector has been studied in some detail for almost two decades and was very well summarized in a document by Larivière and Meilke presented at the OECD meat and dairy meeting in the fall of 1998. The problem comes from the difficulty of empirically identifying the true supply function. “For empirical purposes, if the rental rate of quota is known, the supply function can be estimated directly using information on output, price of output and price of other inputs. However, in most cases quota is traded only as a capital asset. Thus the rental value of quota is not directly observable. Two approaches can be used to overcome this lack of information: 1) estimate the marginal cost directly from micro data; or 2) determine the rental price of quota prior to estimating supply response.”

The marginal cost approach has already been used in an OECD document “An analysis of impacts of the relaxation of dairy quotas in OECD Member countries” presented at the same meeting. The marginal cost information used was from Larivière and Meilke. According to their survey of micro data, the marginal cost of industrial milk production in Canada was 25 US$/hl in 1995 and 29 US$/hl in the case of milk in France and Germany in 1994. In national currency and in the unit of the Canadian and European milk price used in AGLINK these numbers become 34.9 C$/hl and 23.7 euro/100 kg respectively. These costs of production implied quota rents of 35, and 21 percent respectively for Canada and the EU.

These marginal costs are merely a point on the supply curve of the respective country/region at a given moment in time. To identify the supply function it is also necessary to know (rather to assume) the elasticity. According to Larivière, one seems to be reasonable long term elasticity for Canada and the EU. Because of the multi-commodity and dynamic aspect of AGLINK short term elasticity and cross-price elasticities in respect to feed and beef also had to be assumed. In order to be consistent with the previous study, the same elasticities will be used initially in this analysis. Because of the interrelationship with the beef market the distinction between dairy cow and milk yield had to be made. Finally we had to assume that technological improvement, which affect yield, will be the same under the supply function than in the context of milk production quotas.

On the basis of these assumptions the elasticities presented in Table 1 were used in the previous OECD analysis (see “An analysis of…” for detailed explanations).

Table 1: Elasticities

Canada / European Union
Short term
Milk cow/milk price / 0.1 / 0.06
Milk cow/feed price / -0.023 / -0.015
Milk cow/beef price / -0.015 / 0.06
Partial adjustment coefficient / 0.885 / 0.93
Yield/milk price / 0.1 / 0.12
Yield/feed price / -0.04 / -0.05
Production/milk price / 0.2 / 0.18
Production/feed price / -0.063 / -0.065
Long term
Production/milk price / 1 / 1
Production/feed price / -0.24 / -0.264

In the context of the analysis of dairy policy reform, the Secretariat decided to employ in addition to the Aglink model also its PEM model, in order to examine the impact of dairy policy reform on production, consumption, income, welfare, trade, and world prices. In PEM the quota rent is assumed to be 20 percent of the milk market price in the EU which is slightly smaller than in the first scenario. The long term elasticity of milk supply in PEM is however somewhat larger at 1.23. Assuming the same partial adjustment coefficient and the same milk yield own price elasticity as in the previous version (1998) of the analysis, the implied short term elasticity of milk cow in respect to milk price in PEM is 0.078 for the EU. For Canada, the quota rent in PEM (and by construction in Aglink) is assumed to correspond to 23% of the price and the long run milk supply elasticity is 0.81. For Canada, assuming the same partial adjustment coefficient and the same milk yield own price elasticity as in the previous version (1998) the implied short term elasticity of milk cow in respect to milk price is 0.082.

An interesting question is whether or not the quota rent has increased in recent years. A study produced by Institute National de Recherches Agronomiques (INRA) and the University Wageningen suggests that the weighted average quota rent in the EU was around 39 percent in 1998. This is higher than all the numbers used in the analyses presented in this report. In Canada the unit average value of quota has more than doubled since 1995. The discount rates (used to capture the risk of the asset) has probably not increased by the same percentage and as result the quota rent in recent years is probably higher than the numbers used in this analysis. However, in the context of dairy policy reform, quota rent should be evaluated with the long run cost curve not the short run as it was the case for the analysis done by INRA. In this context the farm family labour and the return to farm land should also be included. The estimation of the quota rent from the quota value is not easier. Researchers should consider five points. First, quota value is often strongly influenced by short term market conditions reflecting the need of expanding producers. For this type of analysis a segregated quota market for new entrants would be much more adequate. They should also consider the number of years of depreciation, the discount rate, the impact of the reform on cost of production and in particular on the price of land and finally, they should factor in the anticipation of producers about compensation. These considerations suggest lower quota rent than what was used in the 1998 AGLINK analysis especially for Canada.

Trade liberalisation and dairy policy reform

This section illustrates the necessary modifications to Aglink required for the trade liberalisation scenarios and the difficulties of producing such scenarios with AGLINK. The analysis will be done in two steps. Scenarios of a complete elimination of all market price support policies will be produced initially on a component by component basis. These scenarios will be merged one by one in the second step. Only market price support policies will be eliminated, direct payments will be kept to their baseline value if they are exogenous or will be calculated by the model if they are endogenous. More specifically and depending on the country, border protection, tariffs, production quota, support prices, fluid milk premium and consumer subsidy will be eliminated. Producing this scenario on a component by component basis in AGLINK is technically “safer” and the contribution of each country reform to the overall impact on world markets can be isolated.

European Union

The first modification made to the EU component was the introduction of the milk supply function using PEM quota rent and long term elasticity. The methodology used has already been explained in details in the second part of this report. By simply replacing the quotas with supply functions, we are implicitly assuming that the quotas were not integrated into the cost structure or that government will offer a full compensation to the producers. All the dairy products subsidized export equations were eliminated. All the market clearing price identities were replaced by price transmission equations. Since the world prices in AGLINK are based Northern Europe, adjustments to reflect transport cost were not necessary. An adjustment was however necessary in the price transmission equation of cheese to reflect the different type of cheese used to denominate the EU and the world prices. The cheese price in the EU is for an emmental while the world price indicator is for a cheddar. The difference in the price of these two cheeses has been consistently 30 percent (emmental in Germany versus cheddar in Belgium) lately in the EU. We have used this factor in the price transmission equation. The market clearing identities are now used to calculate trade; imports in the case of cheese, butter and SMP and exports in the case of WMP. The speciality cheeses exported without subsidy, which is an exogenous variable in the baseline, is extrapolated according to the growth in the imports of cheese from Canada, Japan and the USA in these scenarios. The consumer subsidy for milk and SMP are eliminated.

In the first simulation the model generated negative SMP production mostly because of a reduction in milk production and a strong increase in consumption of fresh dairy products (including yoghurt). The decline in the internal price of butter was much stronger than the reduction in the internal price of SMP. In spite of this strong market driven tilt, the percentage of fat in milk stayed the same than in the baseline simply because that variable is solely driven by a time trend in AGLINK. In the context of relatively stable butter/SMP price ratio as we normally have in baselines, this specification is probably sufficient. However, in the context of such a shock, a decline in the fat content of milk is expected. This variable was re-estimated with an additional explanatory variable, i.e. a two years moving average of the butter/SMP price ratio. With this revised equation the amount of solids-non-fat increased but not sufficiently to prevent negative SMP production in all the years of the baseline.

In the absence of other solution, a conditional statement was introduced to prevent a decline in SMP production below a minimum level (i.e. an allocation for the very specialized plants, further research is needed) and the difference between this level and the number calculated by the model was considered as solids-non-fat import (i.e. whey powder concentrate or milk protein concentrate). This last variable was introduced on a SMP equivalent basis in the SMP world market clearing price identity. The strong decline in the EU internal price of butter had a negative impact on the New Zealand price of butter because of the butter sales done under tariff quota allocation. The increase in world prices associated with much higher volume of exports in non-EU markets by New Zealand resulted however in an overall increase in the price of butter in that country.

Central and Eastern European countries

Three countries of central and eastern Europe are present in AGLINK; Hungary, Poland and Russia. According to AGLINK these three countries gives support to their milk producers mostly through the milk fat market. In the case of Poland and Hungary it is done through non-tariff barriers while in Russia, support is given through the butter and cheese tariff.

The Hungarian component is characterized by an internal market clearing price for milk above world level, no market for SMP and no supply function and price for cheese, butter and WMP. The market clearing price was replaced by a price transmission equation linking the Hungarian milk price to the world price. Since the proportion of industrial and fluid milk in Hungary will vary very much in this scenario, milk price in Hungary was specified as the weighted average of industrial and fluid milk price. Fifty percent of the USA fluid milk premium is assumed to prevail in Hungary and is used in this calculation. In the absence of supply function and product prices we have assumed that production of cheese, butter and WMP would be calculated as their baseline share of milk production minus fluid milk and other dairy product production generated in this scenario. Milk consumed on the farm is specified as a demand function in AGLINK. A similar specification than the industrial dairy products seemed more reasonable in the context of this scenario. The former identities which were used to calculate production of cheese, butter and WMP are now used to calculate imports in this scenario. By 2007 all dairy product exports are set to zero.