6    Implications for trade: Projections to 2005

6.1 Trade model and database

6.2 Macro-economic projections

6.3 Results

6.1 Trade model and database

Following the study of Rae and Hertel (2000) we incorporate the previous projections of productivity growth into a slightly modified version of the Global Trade Analysis Project (GTAP) applied general equilibrium model (Hertel 1997) to project national and regional production, consumption and trade flows between 1995 and 2005. This is a relatively standard, multiregion model built on a complete set of economic accounts and detailed inter-industry linkages for each of the economies represented. The GTAP production system distinguishes sectors by their intensities in five primary production factors: land (agricultural sectors only), natural resources (extractive sectors only), capital, and skilled and unskilled labour. In trade, products are differentiated by country of origin, allowing bilateral trade to be modelled, and bilateral international transport margins are incorporated and supplied by a global transport sector. The model is solved using GEMPACK (Harrison and Pearson 1996).

The 50 commodities in the version 4 GTAP database have been combined up to 14 commodity groups, of which 6 commodities (rice, wheat, other grains, oil crops, other crops and processed food) compete for use in the feedstuffs composite. (We modified the model to incorporate feedstuff substitution into the livestock production functions.) Livestock farming is represented by three aggregates: beef cattle (i.e. ruminant livestock), other livestock (i.e. non-ruminants) and raw milk production. These farming sectors provide inputs to the beef processing (ruminant meat), other meat (non-ruminant meat) and dairy products industries in each region. All remaining production sectors are aggregated into manufactures and services, or other natural resource based commodities. Regions are aggregated to match the regions reported in previous tables.

6.2 Macro-economic projections

The productivity catch-up, which we have projected here, is only part of the story of what will be happening in the world economy in the coming years. Other sectors will also be experiencing technological change. Income growth will tend to boost the demand for livestock products relative to grains, and in some regions there will be a strong shift away from food products altogether. On the supply side, the accumulation of skilled labour and capital in China can be expected to continue to promote the shift of activity away from agriculture, in favour of manufacturing and services.

As has become standard with the GTAP model, following the work of Gehlhar et al. (1994) projections are made through exogenous shocks to each region’s endowments of physical capital, skilled and unskilled labour, population, and technology.¹ Table 11 reports the shocks to population, endowments and productivity that we assume in this paper. Forecasts for population, investment (capital stock), and labour force are based on the latest forecasts from the World Bank as of spring, 1999. Projected changes in skilled labour are based on expected increases in the stock of tertiary educated labour and are taken from Ahuja and Filmer (1995) for developing countries while projections for the Organization for Economic Cooperation and Development (OECD) countries are based on World Bank (1997) report. The stock of farmland in each region is simply held constant.

 1. We also follow Gehlhar et al. (1994) suggestion that increasing the standard trade elasticities is appropriate in longer run simulations. For this eleven-year period, we double the standard GTAP values for the elasticities of substitution between imports and domestic goods and among imports from different sources.

Table 11. Annual growth rates of exogenous variables used in the projections and gross domestic production growth.

Forecasting productivity growth is notably difficult. Therefore, we adopt a rather simple approach which is transparent and which can be easily modified. First of all, based on the work of Bernard and Jones (1996), we observe that productivity growth tends to be more rapid in agriculture than in manufacturing, which in turn has a higher productivity growth rate than services. (They find virtually no evidence of productivity growth in mining where quality of reserves confounds the usually difficult measurement problems.) Based on their averages for the OECD as a whole (Bernard and Jones 1996, Table 1), we obtain the following multiples of the manufacturing productivity growth rate for the other sectors: (non-livestock) agriculture = 1.4 * manufactures, services = 0.5 * manufactures, and mining = 0 * manufactures. In this way, we are able to link productivity growth in each sector of the economy to a common metric—namely the rate of manufacture’s productivity growth.

We then divide economies into four groups according to their overall rate of productivity growth: low, medium, high and very high. The assumed annual growth rates productivity in manufacturing value-added for these groups are as follows: 0.25, 0.75, 1.25 and 1.75% per year. As can be seen from Table 11, the low growth group includes Japan, South-East Asia, and New Zealand. The medium group includes the US, sub-Saharan Africa and the rest of the world. Higher productivity growth rates are foreseen for Australia, the EU and South America. Finally, Korea and China’s productivity growth rates are expected to remain quite high—although somewhat lower than implied by the period prior to the Asian crisis. As a check on the plausibility of these assumptions, we compare our baseline cumulative gross domestic product (GDP) growth to that forecast by the World Bank (Table 11). Apart from China and Korea, all of these GDP projections are reasonably close. In order to hit the World Bank targets for these regions, we would have to raise the very high growth category still further. In light of the current macroeconomic uncertainty in that region, we opt for our more conservative projections.

Forecast distributions presented before are used to project livestock productivity in the different regions. Following Rae and Hertel (2000)² we apply these productivity shocks to both value-added and to the feed composite, to maintain a constant ratio of feed use per animal. Provided these shocks are positive, feed consumption per unit of output (the feed conversion ratio) will decrease. If this is the case, then the implications for feed demand, and hence for trade in grains and oilseeds as well as livestock products could be substantial. There is considerable evidence to support this assumption. A recent survey conducted by Wailes et al. (1998) gathered data on feed use across a range of enterprise and livestock types in seven provinces of China where the trend is towards development of specialised livestock production units and larger, more intensive management systems. They concluded that such structural changes would contribute to a declining demand for feed grains per kg of meat production. Another set of livestock and feeds projections for China are those of Simpson et al. (1994, Tables 7.6, 7.7 and 8.1), covering the period 1989–2000. Their projections imply little increase in feed inputs per animal so feed per unit output (the feed conversion ratio) shows negative growth, indicating increases in feed efficiency especially for poultry. This is consistent with the projections of Wang et al. (1998) who assume improvements in feed efficiency for all animal types and technologies. Finally, Tweeten (1998) reported projected annual USA growth rates in output per feed of 0.2% (beef and pigs), 0.6% (milk) and 2.0% (poultry). If USA is the source of much of the new livestock production technology that is transferred to China, then such improvements will eventually be felt in China.

 2. Sub-Saharan Africa was omitted and the historical trends are used.

6.3 Results

We focus here on the impact of alternative livestock productivity scenarios on the changes of regional trade balances. Table 12 reports the change in sectoral trade balances for each region in our global simulation of the period 1995–2005. For convenience, Table 13 compares the trade balance of livestock products in 1995 with the projected trade balance of 2005. Even though productivity growth in livestock products is very high for China, there is little change in its trade balance between 1995 and 2005. This is because China’s demand is also increasing sharply. All other Asian countries show negative impacts on the trade balance of livestock products. Among the developing regions, South America appears as a major exporter of beef and other meats with a five-fold increase in the trade for other meats and a two-and-a-half-fold increase in the trade for beef. On the other hand, sub-Saharan Africa shows deterioration in the trade balance for all livestock products. Developing regions all show negative trade balances in dairy products.

Table 12. Change in trade balance (US$ × 10⁶), 1995–2005.

Table 13. Trade balance in meat products (US$ × 106).

Table 14 compares trade balance of grains in 1995 and 2005. The most important result here is the projected increase in net grain imports to China. In general for the Asian countries we can see the trend toward increasing imports relative to exports in most of the agriculture-related sectors. This is particularly striking in the case of grains and other crops. It conforms to the findings of Delgado et al. (1999) who estimate that China will be a 46 million tonnes net importer of cereals by 2020.

Table 14. Trade balance for grains (US$ × 10⁶).

There are many uncertainties implicit in the productivity forecasts (Tables 6, 7 and 8) and in the macro-economic forecasts (Table 11). We now focus on the uncertainty associated with productivity growth in livestock production. This analysis revolves around the uncertainty associated with the change in sectoral trade balance. The average productivity shock, standard deviation, minimum and maximum shocks for non-ruminants and beef production are shown in Table 15. Mean and standard deviations are derived from the forecast distributions generated using the bootstrapping procedure. The maximum and minimum values are calculated as the mean ± 4.5 times the standard deviation and a triangular distribution is assumed for the shocks. We use the Gaussian Quadrature approach to Systematic Sensitivity Analysis (SSA) proposed by de Vuyst and Preckel (1997) and automated by Arndt (1996) and Arndt and Pearson (1998) to draw a weighted sample from this distribution and generate standard deviations for our simulation results. Using the standard deviation associated with the simulated change in trade balances we can obtain Chebychev’s 95% confidence intervals on the projected trade balance in 2005. These are reported in Tables 16, 17 and 18. The results for China suggest that it is not likely to be a net importer of livestock products in the year 2005. Results for other countries confirm that Asian countries will mostly be importers and the developed countries plus South America will be net exporters of livestock products.

Table 15. Mean, standard deviation, maximum and minimum values for the productivity shocks as derived from the bootstrapped productivity forecasts.

Table 16. Chebychev’s 95% confidence interval for the trade balance of Asian countries.

Table 17. Chebychev’s 95% confidence interval for the trade balance of developed countries.

Table 18. Chebychev’s 95% confidence interval for the trade balance of other countries.