Avishay Braverman and Jeffrey S. Hammer
In agriculture as in other sectors, the policy content of the Bank’s work has increased substantially. In today’s more difficult world economic environment, sustained development depends crucially on a supportive policy framework, and the design of all types of lending operations raises issues of country strategy and policy.
On many occasions agricultural policy changes are perforce introduced without a clear idea of their consequences. Though this is sometimes a question of timing, where reforms appear so urgent that governments consider they have little choice but to act and risk the consequences, another important reason is a lack of convenient, quick-to-use, analytical methods for identifying and evaluating these consequences in advance.
Traditional techniques such as the calculation of effective protection coefficients cannot answer questions about the consequences of proposed agricultural policies. Analysis based on producer and consumer surpluses in individual markets often misses crucial features of economies where substitution between markets and complementarity between policy tools are important. Highly sophisticated modeling exercises, for their part, are capable of addressing these issues but are often too complex and time-consuming, and require such a level of technical training for those designing and using them that they are unsuitable for most operational work.
To help make explicit the effects of agricultural price, tax, and subsidy policies, the Bank has developed an approach which uses simple simulation models tailored to fit a country’s circumstances and policy interests. These “multi-market” models have been used so far in twelve countries. In most of these applications, the models have been developed jointly by staff of the Bank and the borrower government, for use in connection with Bank lending and subsequently by the borrower country on its own.
The models are flexible enough to be applied in different and changing economic environments and institutional structures. They show the effects of various policy instruments on production, consumption, foreign exchange earnings, government def-icits, and real incomes. They are called “multi-market” models because they take account of interactions among markets by allowing for the substitution effects of changes in relative prices, both among competing crops on the production side and among final goods on the consumption side. The models indicate the costs and benefits of proposed policy changes for different groups in the economy. In cases where poverty alleviation and food security are of major concern, they can be particularly useful in assessing the policies’ impact on the real incomes of urban and rural people and of rich and poor within a sector, and, based on the income effects, the nutritional intake of these groups.
Perhaps most important, the models are easy to explain and use and can thus serve as a basis for structuring a richer dialogue on agricultural policy. They give results in a form that can easily be understood and acted upon by nonspecialists. This article describes the method and some of its applications.
The multi-market method
Main elements. The method is the same in all the applications. A simple simulation model is built of the agricultural sector or country economy in question, concentrating on the specific policy instruments of interest (see box). The method has as its centerpiece easy-to-use software for use on personal computers. This provides a common ground for discussions among Bank staff and between the Bank and the economists and policymakers of a country initiating a reform program.
In constructing the model the analyst identifies the markets that are most affected by the policy under review. The determinants of supply in that market (usually prices of inputs, outputs, and competing crops) and demand (usually prices of consumer goods and income) are specified. Substitution possibilities in supply and demand are incorporated.
A critical aspect is the specification of the way markets operate. In fixed-price markets, imports, exports, or inventories adjust, while flexible-price markets are cleared by price movements. There are many possible variants of these basic types, however. For example, some goods may be rationed by government at prices too low to clear the market. If a free market exists in the same good, as was found in the application of this method in Indonesia, the rationing program may simply be a transfer of resources to the recipients. If there is no alternative source of the good, as is the case in some socialist economies, changes in the ration price or quantity may directly affect the demand for other goods. Other applications may have to simulate the actions of a marketing board not under the direct control of policymakers.
A model used in any given country is the simplest one capable of addressing the questions at issue. This simplicity is its strength. Usually the approach will not require a complete general equilibrium representation of the economy, though it is possible to “close” the model to make it so if the key questions being asked are essentially economy-wide.
Bank staff working with the method have always stressed the importance of taking into account the detailed institutional structure of each country in which it is used. Economists advising on policy options need to be able to deal with the messy reality of policy making where abstract theoretical rules and prescriptions rarely apply. They also need to know the limitations imposed by the basic structure of the economy and the limitations imposed on the use of different policy instruments by political or administrative factors. For example, even if income or land taxes are known to cause fewer distortions than indirect taxes, a government seeking to raise revenue may find such instruments politically or logistically impossible; similarly, even if high subsidies on bread are an important cause of a high public deficit, the government may find it too politically risky to remove them. The politically difficult policy options need to be analyzed along with the alternatives, if only to make explicit the consequences of following one policy course over another.
The models allow policymakers to evaluate the trade-offs between different policy options. These arise because only a limited number of policy instruments are available to further a wide range of policy objectives. Thus one objective may have to be sacrificed to further another. For example, export taxes are frequently an important source of revenue for domestic expenditure; they can also be used to influence income distribution. But if the taxes are a disincentive to domestic production, they have costs as well as benefits; they will depress output, and hence incomes, in the affected sectors and will reduce foreign exchange earnings. The explicit evaluation of trade-offs is very important, particularly as many countries are facing unsustainable fiscal deficits and external debt burdens.
The results of the models are expressed in terms of changes in the policy variables of interest, given some appropriate values for parameters such as elasticities of supply and demand, cross-elasticities and base levels of supplies, demands, factor inputs, and prices. Experiments are undertaken with different policy options and with different assumptions concerning the parameters.
Applications in Africa
In Sierra Leone and much of French West Africa, export crops such as cocoa, cotton, or oil seeds are taxed, as are imports of a food crop, usually rice. The imported food crop is also produced locally, alongside traditional food crops such as coarse grains or tubers that do not generally enter international trade. Several governments faced with the need to collect more revenue or generate more foreign exchange earnings (net of import costs), have considered increasing the tariff rate on imported food. At first glance, this could both yield extra revenue and reduce the foreign exchange spent on imports. (For members of the franc zone, where devaluation of the currency is not an option, increases in the tariff on imported food are consistent with the common policy recommendation to increase the prices of traded goods, by raising import duties and export subsidies. Where higher export subsidies are ruled out for budgetary reasons, increases in the import tariff rate are often pursued alone.)
The multimarket method has been applied in West African countries to examine the consequences of proposed tariff increases on imported foods, taking into account the possibilities for substitution: by producers, between rice and cash crops and between nontraded foods and cash crops; and by consumers, between demand for rice and for nontraded foods. In the models, the chain of logic is as follows: the increase in domestic rice prices caused by the higher import tariff would induce farmers to shift production from export crops and nontraded foods into rice, while consumer demand would shift from rice toward nontraded staple foods. These two effects would tend to raise the prices of the nontraded foods. However, the price rise for nontraded foods would lessen producers’ shift out of these foods.
The model results showed that in response to an increase in the import tariff, the net increase in domestic rice production would be smaller than that implied by the price elasticity of rice supply when taken alone (as would probably be done in a single-market analysis). The results also showed that the increase in the prices of nontraded goods would exacerbate producers’ shift out of export crops. The net effects, then, of an increase in the rice import tariff would be a decrease in the supply of the export crop; an increase in the prices of nontraded foods, and only an attenuated supply response in rice.
Since the main purpose of the tariff increase would have been to raise government revenue and save foreign exchange, the analysis took account of the secondary effects of reduced production of the export crops being taxed. These indirect effects were quite large relative to the expected direct effects. Some of the exports taxed were tree crops, the supply of which takes several years to adjust fully to price changes. Results from the models showed that even if the anticipated gains from a higher rice import tariff were forthcoming in the short run, they might be eroded in the longer run, as tree crop producers adjusted their plantings, with resulting reductions in tree crop exports and the associated tax revenues.
Data in Africa are often very limited and for the cases just described, the relevant elasticities were not known with certainty. However, sensitivity analysis with the models showed that, within the likely range of elasticities, losses in exports and export tax revenue may offset the direct effects of the tariff increase on import and tariff revenues.
Republic of Korea, Senegal, and Sierra Leone) did not stress the “user-friendly” aspect of the current procedure. These applications were more in the spirit of elaborate modeling techniques and relied on specific algebraic functional forms to characterize supply and demand systems. The implementation required fairly complicated calibration techniques which were difficult to explain and to transfer to other users. The new version does not require calibration, nor does it require the use of specific functional forms. The method proceeds by arranging information in a set of market-clearing equations which is totally differentiated so that changes in the outcomes of interest can be solved in terms of changes in the available policy options. Since the resulting models are linear, they can easily be solved on a personal computer. Initially the analysis is restricted to small changes, but simplified methods of analyzing larger changes have been devised as well.
Special characteristics. For any economic policy, at least in the short run, there are groups who lose and groups who gain. The multi-market approach identifies the groups who are likely to lose and gain from a policy change, and by how much. This is vital information both for the choice of policies and for the design of measures, or packages, to make policy changes acceptable to the public (it can help, for example, with the design of measures to compensate the losers).
Four characteristics of the approach are noteworthy. First, the models aim to assess trade-offs between alternative policies. They facilitate analysis of the consequences of policies, but they do not impose a particular recommendation or a particular calculation of social welfare. Many practical economists and public officials are suspicious of recommendations from social welfare maximization exercises and prefer to discuss the scenarios emerging from alternative price reforms. Besides changes in people’s private incomes, governments have other objectives in pursuing price reforms. For example, government deficit reduction, particularly in the presence of high inflation, is often a high policy priority. To attribute a shadow price or ex-ante “welfare weight” to reduction in the government deficit is difficult. Changes in government revenue, foreign exchange, and other objectives which indirectly affect people’s welfare should be looked at specifically.
Second, like most simple models, the multi-market method is not a pure forecasting tool. Rather it traces out the effects of different policies given the schematic representation of how markets work.
Third, these are not monetary models of the economy. The policies that are appropriately analyzed are those which change rela-tive prices between goods. Though there is a good deal of interaction between the agricultural sector and the macroeconomy, macroeconomic policies are not fully accounted for in this framework.
Fourth, the method is a deterministic one. Currently it is being extended to incorporate uncertainty issues and further macroeconomic aspects.
Questions sometimes arise concerning the advisability of using the multi-market method when data are poor. The argument may take one of two forms: (1) that data are poor in many developing countries, making any quantitative analysis of questionable value; and (2) since parameter values can be chosen at will, the scope for abuse is large.
As to the first point, the main contribution of the method is to improve thinking about economies, rather than to obtain precise figures. The basic structure of the model should reflect those aspects of reality the analyst feels are important. Whether data are good or not, these aspects should be brought together in a consistent manner to show the logical consequences of the assumptions being held. Especially when data are lacking, the interactive computer program makes sensitivity analyses easy to perform. Sensitivity analysis allows one to jump gaps in one’s knowledge by using informed judgments about elasticities and the like. It also reveals how important these gaps are for policy conclusions, and thus helps to guide research priorities and show where caution should be exercised in giving policy advice.
There is no reply to the second argument.
The only check on abuse of the models is that all assumptions must be explicit. The model structure is transparent in each case and parameter values are clearly visible, rather than being hidden in a “black box” or in the implicit assumptions of smaller models. But the possibility of abuse makes it advisable to transfer a model and the capacity to create alternatives to more than one user group within a country, and perhaps to some whose interests conflict. This way there is less risk of giving a monopoly of claims to information to any one group.
The method has been applied in Argentina, Brazil, Cyprus, Hungary, Indonesia, Republic of Korea, Senegal, and Sierra Leone, and work on Cameroon, Cote D’lvoire, Mexico, Morocco, Tunisia, and Turkey is under way. The approach has been used to examine taxes, prices set by government, subsidies, devaluation, tariffs, and quotas; it has also been used to analyze the effects of nonpolicy variables such as falling world prices, technological change, or changes in transport costs. Many of the applications have addressed the consequences of liberalizing markets and attempted to quantify the costs and benefits that would result—for example for tax revenues, subsidy costs, foreign exchange earnings, and levels of production and consumption.
In each of the applications, the use of the multi-market method helped to clarify the trade-offs inherent in policy decisions. One of the earliest applications was in 1981 in the Republic of Korea. The government had been pursuing the goal of self sufficiency in rice for some time, maintaining high producer prices and subsidizing fertilizer inputs. Consumer prices of rice were subsidized, though even with the subsidy they were higher than international prices. For nutritional reasons there was a similar subsidy program for barley. The multi-market method was used to help quantify the costs of maintaining self-sufficiency in terms of the drain on the government budget and the costs of living of consumers, differentiating among urban and rural consumer groups at different income levels. The standard account of the problem in terms of rural versus urban interests turned out to be too simplistic. The model was important in capturing the indirect effects of (1) additional rice production on the cost of the fertilizer subsidy; (2) rice price increases for producers on the cost of free-market goods for consumers; and (3) changes in rice subsidies on the cost of the barley subsidies.
Indonesia faced a similar set of issues: a goal of self-sufficiency in rice supported by high prices and fertilizer subsidies. Again the main trade-off was between self-sufficiency and the incomes of larger farmers, on one hand, and government deficits, urban incomes, and incomes of smaller farmers, on the other. The software for the program is being installed in various government and research offices in Indonesia; training has begun to allow researchers and policy analysts there to adapt the model as market conditions change.
In both Argentina and Hungary, where the method was used to analyze the effects of differential taxes on competing export crops, the essential trade-off was between foreign exchange earnings and government revenues. In both countries the Bank is supporting training in this method so that it can be used to evaluate policy changes in future. In Brazil, the method was used to evaluate the benefits and costs associated with reducing the consumer subsidy on bread and the producer subsidy on wheat. The results showed how the impact of the policies would differ by region and by income group, the effects on the prices and production of other foods, and the secondary effects on government tax revenues.
Improving policy dialogue
The approach was conceived to promote a better understanding of issues and choices, a better dialogue between the Bank and governments, and better decisions on the policies to be implemented. The most successful applications have been those in which staff of the Bank and the government concerned worked jointly from the start. In these instances, preliminary interviews with a variety of government officials were conducted to identify the key policy options and to determine the structure of relevant markets. Asking people to state their opinions and assumptions about how markets work and about potential substitution possibilities encourages thinking in system-wide terms and helps to develop intuitions that go beyond the actual numerical results. Clarification of the model structure, either in seminars in which representatives of different governments offices are present, or by simply testing one office’s opinions out on another, encourages constructive debates over goals and how to meet them.
Has the multi-market method actually changed governments’ policy decisions? That is difficult to answer. While policy is ultimately a political matter that cannot be determined outside political processes, its formulation can be influenced for the better by more systematic thinking about inevitable trade-offs. The purpose of the multi-market method is not solely to generate numbers. It is to help governments focus policy discussions on the short- to medium-run consequences of the various policy reform options. In this it fills a badly needed gap in operational work, supplementing the quick single-market approach, while complementing the elaborate models which require more in-depth research.
A further benefit of the approach is the development of institutional capacity for policy evaluation within the borrowing coun-try. The Bank frequently recommends that developing countries improve their capacity for policy analysis, and the multi-market method has proved a good means of encouraging this.