Prioritizing agricultural investments across commodities for income growth and poverty reduction: Methods and applications
Some agricultural investments are commodity-specific, meaning that they increase the productivity of production, processing, or marketing of a single agricultural commodity or a set of closely-related commodities. Examples include investment in cassava breeding, expanding cotton ginning capacity, irrigation for rice production, expansion of cold storage capacity for horticultural exports, or road investment to a region whose main product is maize. Traditional cost-benefit analysis estimates the effect of investments on net income assuming that the investment is not large enough to influence market prices. However, a different approach is needed when the investment affects market prices and/or there is an interest in other outcomes such as poverty reduction.
This report describes an approach to estimating the impact of commodity-specific agricultural investments on income, poverty, and other measures of welfare. This approach can be extended to identify the optimal allocation of an investment budget across commodities subject to a given objective function. For example, it could be used to allocate agricultural research funds across commodities to maximize income, poverty reduction, or a weighted average of the two.
The method can be divided into four steps.
- We start with an estimate or an assumption about the impact of the investment on productivity, where productivity could refer to production (an increase in yield) or marketing (a reduction in the marketing margin).
- Second, we use a model to estimate the effect of the productivity increase on prices. This could be a partial equilibrium model, for which commodity price changes can be simulated with information about supply and demand elasticities. Or it could be a general equilibrium model, in which all sectors of the economy and factor markets are included and income is endogenous.
- Third, the results of the simulation are used to estimate the change in income for each household in a nationally-representative household survey. In the case of the partial equilibrium model, the change in income is estimated using the changes in yield and commodity prices, along with elasticities and household-level information on the importance of the commodities in consumption and production. In the case of the general equilibrium model, household incomes are also influenced by factor prices (such as wages) and household-level factor endowments.
- And fourth, the household-level results are aggregated to the national level, which gives both aggregate changes in income and changes in poverty indicators. The latter would typically be the incidence of headcount poverty, but could also include other indicators such as the poverty gap and the depth of poverty.
In order to find an optimal allocation of funds across commodities, the first step is replaced by identifying relationship between the investment in each commodity and the increase in productivity. In addition, we repeat steps 1-4 in a search for the allocation that maximizes the objective function. For a small number of commodities (less than 8), this could be a simple grid search, but for a larger number a more efficient search procedure, such as Newton-Raphson, is needed.
The partial equilibrium approach is demonstrated on investments in agricultural research in the Philippines. We use data on five important crops: rice, maize, cassava, coffee, and vegetables and assume a quadratic relationship between funding and yield increases, reflecting diminishing marginal returns. The distributional effects are estimated using the 2012 Family Income and Expenditure Survey. To find iv the optimal allocation, a two-stage grid search is used. In the first stage, the allocations are in increments of 10 percentage points, and in the second, the allocations are in 1 percentage point increments in the vicinity of the best results from the first stage. We find a strong correlation (R2=0.99) between income growth and poverty reduction, suggesting little trade-off between the two objectives. In addition, the best allocations are those in which a large share of the budget (83-90%) is allocated to rice, with the remaining being allocated to vegetables and sometimes coffee and maize. If not for the diminishing returns to investment, all funds would be allocated to rice research. Third, the optimal allocations raise income by 0.8 percent (US$ 950 million/year) and reduce poverty by 0.9 percentage points (lifting 800 thousand people from poverty). Finally, the optimal allocation generates US$ 79 million/year more than using the rule of thumb that one should allocate research funds in proportion to the value of production.
The general equilibrium approach is first introduced and then applied to a problem of research resource allocation in Rwanda. Applied general equilibrium models look much more different from partial equilibrium models than they are in reality and these differences are explained. The approach is applied using the well-known MIRAGRODEP model linked to a survey of almost 15,000 households with data on their income sources and expenditure patterns. A nonlinear programming approach is used to solve for the allocations of research resources that optimize objective functions that include combinations of average income and income inequality. With a zero weight on income inequality, the model solution implies allocating almost all resources to the commodity with the largest initial output share. Once greater weight is applied to income inequality, the recommended research resource allocations shift to a more diversified set of staple food commodities. The impact of increased agricultural productivity on poverty in this small poor country is very large, with enough resources for a 50 percent increase in productivity resulting in a halving of poverty if allocated effectively for that goal. Increasing the resources available to promote productivity growth allowed a further dramatic decline in poverty to around 16 percent.
Photo credit:Gunnar Salvarsson