Lecture 27.
Measuring trade liberalisation effects

(download powerpoint slides here)

What we want to learn about this topic

This lecture summary should be read in conjunction with the later lecture summary which examines specifically the impact of agricultural protectionism and liberalisation on developing countries.

Short introduction to the issues

Wide range of projected welfare changes from agricultural liberalization

Examples are given in the Powerpoint slides. Such differences in results are disconcerting to policymakers. Why such differences? How much weight should be put on the results of one individual model? Are some models superior to others? What are main characteristics of models which mean some might be taken more seriously than others?

Different types of model are used

Effects of simultaneous liberalisation are different from the sum of the effects of liberalising each commodity separately because of the important interdependencies which exist between agricultural commodities. These interdependencies occur with respect to both the supply and demand sides, and can be of both a competitive and complementary kind.

For example, a relative reduction in the price of an individual grain crop will cause farmers to substitute other crops consistent with rotation and other technical requirements. However, if the prices of all crops are reduced simultaneously, substitution effects will be negligible and the main effect on supply will come through changes in farmers' use of variable inputs and subsequent changes in yields.

Another important example is the consequence of feed/livestock interdependence. If both feed and livestock prices are simultaneously reduced, the impact of the livestock price reduction on livestock supply will be much reduced compared to the situation where feed prices are assumed to remain unaltered. Similarly, the impact of grain price reductions on the export surplus of grains will be indeterminate when livestock prices and thus livestock demand for feed can vary.
Low computing power meant that in the past most models were partial equilibrium. Computable general equilibrium (CGE) models are now much more practicable but when should one be preferred? PE models often allow much more detailed implementation of policies, CGE models often much cruder. CGE models may be difficult to interpret. But if the change is likely to have economy-wide effects, then CGE models are to be preferred.

Example: suppose EU liberalises its ag policy. Ag exports fall and EU runs a larger balance of payments deficit. Value of euro falls. This boosts growth in the nonfarm sector which, if unemployment exists, may produce much bigger welfare gains than those observed in the agricultural markets alone (i.e. the classic welfare triangle effects). Yet such effects would be ignored in partial equilibrium models.
Partial equilibrium Modellers often confine their model coverage to the main temperate zone commodities, thus mexcluding tropical, Mediterranean and processed products. If these products are important for some countries, then the effects of agricultural support or trade liberalisation may be over- or under-estimated. EU studies which look at the effect of CAP reform on individual Member States often suffer from this problem. By definition, general equilibrium models do not suffer from this problem as all economic sectors are included.

Econometric models are usually applied to single commodity or multi commodity partial equilibrium studies. Responses to price changes are based on ex post observations of what happened in the past. Most of the recent global analyses are based on computable general equilibrium models. These models use databases which capture all flows and interdependencies in the global economy, and assume that neoclassical optimising behaviour drives responses to price changes. The magnitude of these responses is estimated from econometric studies of past behaviour summarised in the size of the elasticitiy values used as parameters by the model (e.g., demand elasticities, substitution elasticities between inputs in production, substitution between different sources of supply such as domestic production vs. imports).

Most CGE models use the GTAP database, a global database maintained by a consortium of government agencies and international research institutions. There is also a GTAP model which has been widely applied to analyse the implications of further agricultural trade liberalisation in the Doha Round. CGE models are calibrated models, that is, they are deliberately made consistent with a single year's data which is assumed to represent a long-run equilibrium. It is more difficult to validate the results of a CGE model because unlike econometric models we have no information about the distribution of errors.

Many models are comparative static. They are built on data for an historical base period, and the policy experiment is to ask what would have happened in that period if some one or more particular policies had been different (e.g. what would have happened if the rate of tariff protection had been 0 per cent rather than 50 per cent?). The usefulness of these comparative static experiments is that the results, when expressed in percentage terms, can give us an idea of what the impact of a similar policy if it were introduced today or in the future might be. Dynamic models attempt to project a baseline, i.e. an assumption about the state of the world under particular assumptions into the future, and then ask how that baseline would be different if a particular policy change were introduced at a given time. Dynamic models have the advantage that they can trace the path of change over time, which may involve overshooting and then convergence to a new equilibrium. They have the drawback of the added complexity needed to be able to model the way the world changes from one time period to the next. The FAPRI-Ireland model is an example of a dynamic (partial equilibrium) model which is used for the simulation of policy changes affecting Irish agriculture. See their analysis of the impact of decoupling on Irish agriculture for an example of price overshooting in the beef market.

Model and parameter assumptions may be different

The key parameters in partial equilibrium studies include the base period used which can influence the initial levels of protection assumed, the size of the behavioural responses by producers and consumers (as measured by price elasticities) as well as price transmission elasticities. Nearly all global models assume perfectly competitive markets, although imperfectly competitive behaviour may be more realistic and might yield quite different results.

It is the reduction in protection as a result of policy liberalisation which drives the model results, and the initial level of protection is clearly important.

Given the importance of non-tariff barriers to trade, most studies use tariff equivalent measures derived by taking the difference between domestic and world price levels such as the PSE or the Nominal Assistance Coefficient . Even for the same time period, there is a surprising diversity in the values used by different researchers, which reflect in part the complications in making such comparisons.

Because world prices fluctuate much more than domestic prices, protection levels vary considerably from year to year, so the base period chosen for the study will also influence the results.
The magnitude of the world price effects of any given liberalisation policy will depend on the responses of producers, consumers and other governments. The behaviour of producers and consumers is summarised in the supply and demand elasticities used by the researcher. Because accurate information is lacking, a common approach is to parameterise them and to test the sensitivity of the results to these changes.
Estimation of producer supply response is further complicated in countries where price support is accompanied by supply controls. In principle, it is possible that the removal of price support and supply controls at the same time could lead to an increase rather than the expected decrease in supply. If the modeller has not tried to incorporate this additional effect into his or her model, these results will differ from someone who has.
The response of other governments to the initial increase in world prices is given by the price transmission elasticities assumed. A price transmission elasticity of one implies that other governments permit changes in world prices to be reflected in domestic prices on a one-for one basis or, in other words, it assumes that whatever policy stance they originally adopted (as measured by the level of protection in place) remains unchanged. The other extreme is a price transmission elasticity of zero, implying that other governments hold their domestic prices unchanged even when world prices change. The higher the weighted price transmission elasticity of those governments not taking part in the initial liberalisation, the lower the world price effect and the higher the trade effect of the policy change. Price transmission elasticities can either be assumed or estimated from time series data.
If the modeller attempts a future-oriented projection, this may include attempting to project not only future prices and quantities but also the future stance of government policy in response to the implied changes in budget costs or farm incomes. Where the level of protection in a model is not assumed exogenously (for example, by assuming that the same rate of protection will apply in the future as in the base year) but is itself a function of model variables, this is referred to as endogenising policy. But it is clear that if an attempt is first made to project the likely level of protection in the future, and then the model is solved to reflect the removal of this likely level, then the results will differ from a standard comparative static experiment in which the level of protection is assumed to be equal to the base year period.

Scenario assumptions may be different

Even where modellers make use of the same model and the same parameter and modelling assumptions, the results may differ because the scenarios being modelled (i.e. the questions being asked) may differ. Even where the question being asked seems the same, the way the scenario is implemented may differ. Thus it is very important to interpret the results of a study in terms of the scenario being modelled. For example:

Another important issue in all simulations is the counterfactual, i.e. what assumption is the modeller making about the reference scenario against which the policy scenario is being compared? Specifying the counterfactual is a very important part of policy modelling. For example, suppose you were asked to estimate the impact on Irish agriculture if EU agricultural policy eliminated market price support under the Common Agricultural Policy. Your answer would be very different depending on whether you assumed that the CAP status quo could continue or whether you believed that, in the absence of CAP market price support, some alternative support mechanism for farmers would be put in place. In comparative static models, the counterfactual is implicity the policy structure which was in place in the base period of the model. In dynamic models, the counterfactual is more explicit and is built into the baseline against which the outcome of the policy simulation is compared.

Reading suggestions

Empirical studies

Martin, W. and Anderson, K., 2006. The Doha Agenda negotiations on agriculture: what could they deliver?, American Journal of Agricultural Economics, 88, 5 (2006), 1211-1218 (access through Trinity Library only)

Taylor, L. 2006, Modelling the Impact of Trade Liberalisation: a Critique of General Equilibrium Models, Oxfam.

Surveys and commentaries

Bouet, A., 2006, How much will agricultural trade liberalization help the poor? Comparing global trade models, IFPRI Research Brief No. 5, Washington, D.C., International Food Policy Research Institute.

Elliott, Kimberley Ann, 2006. Can Doha Still Deliver on the Development Agenda? Policy Briefs in International Economics No. PBO6-5, Washington, D.C., Institute for International Economics.

Ackerman, F., 2005. The Shrinking Gains from Trade: A critical assessment of Doha Round projections, Global Development and Environment Institute, Working Paper 05-01, Tufts University.

The Economist, 'Weighed in the balance: global trade', 8 December 2005.
(short survey of how the estimated gains from liberalisation have shrunk over time)

FAO 2005. Trade policy simulation models: Estimating global impacts of agricultural trade policy reform in the Doha Round, FAO Trade Policy Technical Note # 13, FAO, Rome. [download]
(20-page briefing which explores why model results differ and explains the key assumptions made by modellers in generating simulation results).

Supplementary reading

Anderson , K. and Martin W., eds., 2006. Agricultural Trade Reform and the Doha Development Agenda, World Bank and Palgrave Macmillan
(see especially Anderson, K., Martin, W. and van der Mensbrugge, D., Market and welfare implications of Doha Reform Scenarios, Chapter 12 and Hertel, T. and Keeney, R., 2006. What is at Stake: The relative importance of import barrie rs, export subsidies and domestic support, Chapter 2)
(Note these readings cover similar ground to the AJAE article above at greater length. To download the tables and figures, you will need to click here and look for Chapters 2 and 12)

Congressional Budget Office, 2005, The Effects of Liberalizing World Agricultural Trade: A Survey, Congress of the United States.
(short, 33-page paper which discusses in order the total cost of policies that distort agricultural trade, considerations with partial liberalisation, and distributional issues).

Web resources

The Global Subsidies Initiative has a very useful page which brings together a series of links on why model results have changed.