DEFRA FEED IMPORT PROJECT: SUMMARY OF MODELLING WORK (paper by Defra) This note aims provides a summary of the various modelling tasks feeding into the project, looking at the economic Impact of unapproved GMOs on EU feed imports and livestock production. There were two parallel workstreams: one undertaken in-house, using the OECD/FAO’s Aglink-Cosimo model (which will be referred to as Aglink), and a second using GTAP, a world trade model. This note will (1) summarise the method used to generate the results in Aglink, then (2) discuss the GTAP methodology. Section (3) will compare the results across models. A conclusion (4) will follow. (1) Aglink Method The aim of the workstream was to update the earlier European Commission report from 20071 which gave limited results and used a modified version of the 2006 Aglink model. The earlier report aimed to quantify the impact of main soya exporting countries growing GMO soybeans which are not approved by the EU. Background information can be found in that report. The key changes compared to the 2007 European Commission report (which will be referred to as EC paper) were: 1. The use of more recent data. Data was taken from Eurostat to calculate reductions in supply if soya became unavailable to the EU due to the production of unapproved GMOs. Other data, such as the capacity for substitution was not recalculated, but taken from the 2007 report. 2. The use of more up to date versions of the Aglink model. The 2008 version of Aglink was primarily used, with Defra’s Trade Improved 2007 model (TIM 2007) for further comparison. The Trade Improvements have not yet been included in the 2008 model. Three scenarios were shown in the EC paper, which were used for the Defra analysis. These all relate to a fall in the availability of soya over 2009 and 2010: 1. ‘Minimal Impact’: This scenario envisages the US growing unapproved GM soya, but not Brazil or Argentina. No EU imports would be available from the US. Changes to trade patterns were expected, but a reduction in supply to the EU was not considered: no further analysis was undertaken. 2. ‘Medium Impact’: The US and Argentina grow unapproved GM soya, removing the availability of EU purchases of soya from those countries. The reduction is partly compensated by increasing imports from Brazil. 3. ‘Worst Case Scenario’: The three major exporters of soya start growing unapproved GMO soya (US/Argentina/Brazil), therefore there are no 1 http://ec.europa.eu/agriculture/envir/gmo/economic_impactGMOs_en.pdf soya imports to the EU from those three countries, with no compensation from the rest of the world. It should be noted that the Aglink model is not primarily concerned with trade: flows into and out of each country are modelled, but not the origin/destination countries. As bilateral trade is not modelled, it is not possible to show the impact of a reduction in imports from a particular country to the EU – this can be shown, however, by exogenously reducing the level of imports to a specific amount. Some agricultural commodities are aggregated together for trade flows, of relevance here is that soybeans make up a part of the trade in oilseeds and trade in soybean meal makes up a part of the trade in oilseed meals. This reduces the effectiveness of the model where trade in other oilseeds and meals and/or trade from other countries would be expected to partially offset the decline in available soybeans and meal. To get around these limitations the EC analysis exogenously adjusted the reduction in imports of oil meals to account for increased production and trade in other oilseed meals. The model was also recalibrated to account for a substitution of animal feeds away from oil meal. In this analysis the recalibration values were the same as those used in the EC report. Some modifications to Aglink were required, further to the recalibration undertaken in the EC paper. Of particular note, the EU oilseed and oil meal prices were set equal to the market clearing EU price. (2) GTAP Method Unlike Aglink, which models net trade, GTAP explicitly models bilateral trade flows. This allows the modelling of a reduction in EU imports from individual countries. The version of GTAP used is known as Defra-TAP and includes a number of changes from the standard model to improve the usefulness in modelling the interrelationships between agricultural commodities. For this work Defra-TAP was updated to show the link between biofuels and agricultural markets and to include new member states in EU enlargement. Further to this, the baseline oil price assumption was updated. Similar scenarios were considered to those modelled in Aglink. For Scenario 1, imports to the EU of soybeans and cereals from the USA were reduced to zero. Scenario 2 reduced imports from the USA and Argentina to zero. Scenario 3 was, again, a ‘worst case’ scenario which modelled the impact of no soya/cereal imports from Brazil as well as Argentina and the USA. A key benefit to using GTAP here is that, with particular relevance for Scenario 2, feedback effects can be shown implicitly. Using Aglink required an assumption at the outset about how much soya Brazil would supply to the EU if supplies from the USA and Argentina were not available; the GTAP model is able to calculate the impact on Brazil-EU trade. Contrary to the Aglink model GTAP required input values for feed costs. The change in feed costs could not be adequately reflected in the model because trade in soybean meal is not modelled separately but is combined with many other commodities. The values shown for the high scenario in the EC paper were considered too great a shock for the model to converge satisfactorily, so an upper limit of a threefold increase in feed costs was used for Scenario 3. (3) Results Scenario 1 - Comparison with Baselines Defra-TAP showed increases in feed costs of less than 0.1% in the EU, leading to a similarly negligible impact on output. No modelling was undertaken in Aglink. Scenario 2 - Comparison with Baselines Aglink 2006 (from EC paper) Aglink 2007 TIM Aglink 2008 Defra TAP EU-27 EU-27 EU-27 UK Feed Cost Index 23% 11% 4% 7-23% Pig Meat Production -1% -1% 0% Poultry Meat Production -2% -7% -2% -3%* Pig Meat Imports 29% 51% 0% Poultry Meat Imports 7% 103% 29% -2%* Pig Meat Exports 0% -2% -1% Poultry Meat Exports -3% -4% -1% -5%* Pig Meat Prices n/k 3% 1% Poultry Meat Prices n/k 1% 0% 4%* Beef/Veal Production 0% 0% 0% 0%* Beef/Veal Imports 13% 4% 3% 0%* Beef/Veal Exports -41% 0% 0% -3%* Beef/Veal Prices n/k 2% 1% 1%* * Defra TAP values are for white meat and red meat. The medium impact scenario is projected to have fairly limited impacts. With only around 10% of the import supply unavailable, the substitution of other crops, as well as additional soy imports from Brazil, allow prices, production and consumption to remain at a similar level to the baseline scenario. Feed costs would rise slightly, leading to a slight fall in the production of white meat. There could be a small substitution of beef for white meat. Due to the small reduction in production costs in the rest of the world (from Aglink 2008, the feed cost index for the USA falls by 2-3%), and the slight decline in domestic supply, meat imports increase and/or exports fall slightly. Using Aglink 2008, following the reduction in the availability of soy imports, prices, consumption and production are projected to return to the baseline levels within one or two years. Scenario 3 – Comparison with Baselines Aglink 2006 (from EC paper) Aglink 2007 TIM Aglink 2008 Defra TAP EU-27 EU-27 EU-27 UK Feed Cost Index 2068% 1725% 1975% 300% Pig Meat Production -29% -26% -24% Poultry Meat Production -29% -68% -55% -10%* Pig Meat Imports 637% 8801% 0% Poultry Meat Imports 93% 1298% 1007% 16%* Pig Meat Exports -86% -23% -28% Poultry Meat Exports -100% -49% -42% -20%* Pig Meat Prices n/k 64% 175% Poultry Meat Prices n/k 19% 9% 20%* Beef/Veal Production -1% -1% 0% -2%* Beef/Veal Imports 397% 19% 84% -3%* Beef/Veal Exports -100% -100% -100% -38%* Beef/Veal Prices n/k 28% 22% 4%* * Defra TAP values are for white meat and red meat. The high impact scenario projects much more significant impacts than the medium impact scenario. As the availability of oil meal for animal feed falls by around three quarters the price of feed is likely to increase substantially – perhaps by an order of magnitude (although with such a large shock there is substantial uncertainty). At the production level our results are similar across different versions of Aglink: pork production falls by around a third and poultry by around half, with little change in beef production. The results from Defra- TAP are less significant, with only a ten percent decline in white meat production. For both poultry and beef there are a range of price increases between TIM 2007 and the 2008 Aglink model. For 2009 these range from around 10% to around 30%. It is interesting to see that the changes in pig and poultry meat prices differ. This is most likely due to differences in how trade is modelled. The 2007 TIM model shows the EU moving from a net exporter of pig meat to a net importer. The 2008 model (without the trade modelling improvements) does not show this switch to net-importer of pork leading to much higher price for pig meat. In Aglink 2008 feed costs outside of the EU fall. For example feed costs in the USA are 15% below the baseline values following the soybean trade reduction. Defra-TAP suggests significant decreases in meat trade balances. The Aglink models show similar impacts, although the extent and product mix varies according to each version. Aglink 2008 was also used to project the impact in the years following approval of the GMO: the prices of oilseeds and oil meals are projected to fall back close to the baseline. This is also true in meat products, although the impact may take a few years longer to feed through. (4) Conclusion Defra-TAP and Aglink are both sophisticated models, able to show us the impact of changing feed costs on the agricultural sector through to the meat sector. However, these models are designed to assess the impacts of marginal changes, and as such the magnitude of changes in scenario 3 do push the limits of such models. The modelling results are very useful in giving us a good idea of the direction and scale of any change – but beyond this extent percentage deviations from the baseline should not be considered as robust. The conclusion of the modelling work is that there would be negligible impact of US soya being unavailable to the EU (scenario 1). With Argentina included (scenario 2) there is a noticeable impact, with prices expected to rise somewhat and trade balances to worsen. The large drop in soya imports following the introduction of unapproved GM soya in Brazil, Argentina and the USA would have much more notable impacts. With large feed cost increases the EU livestock sector’s competitiveness would decline substantially and large drops in exports and increases in imports could be expected. If closer to the time a complete decline in the availability on imports of soya from the US, Brazil and Argentina does look likely, changes in trade flows might be expected. In this case, action from the EU, in an attempt to counteract the problems this scenario would cause, might be expected.