Modelling regional market integration : application to policy simulation in Eastern and Southern African maize markets

Abstract

The need for a vibrant and sustainable agricultural sector in Africa was recognised by the African Union in its Maputo declaration on agriculture and food security in 2003. Member states committed to allocating at least 10% of national budgetary expenditure towards implementation of the Comprehensive African Agricultural Development Program (CAADP). Despite these efforts, Africa remains the largest recipient of food aid in the world and over the coming decade, population growth in Sub-Saharan Africa is projected to exceed any other region globally. Consequently, the need for efficient policies that promote growth in the agricultural sector has been reaffirmed in the recent Malabo declaration presented by the African Union in 2014. Pledging to end hunger in Africa by 2025, it outlines ambitious targets such as doubling productivity and tripling intra-regional trade in agricultural products. Maize represents the principal staple in Eastern and Southern Africa (ESA) and consequently it has been prioritised in much of the historic agricultural policy initiatives. Despite international pressure to liberalise markets, the need to stabilise prices at tolerable levels has been offered as justification for continued intervention in the sector. Contrary to these objectives however, observed volatility over the past decade has often been higher in markets where governments intervene most actively and with few exceptions, maize prices in the region remain high in the global context. As such, literature evaluating policies in the region has questioned the efficiency of historic interventions in achieving stated objectives. Most of the policy analysis in the region to date has been retrospective in nature and focused in specific countries where policies have been employed. As the region moves toward implementation of the ambitious targets outlined in the Malabo declaration, this study presents a partial equilibrium simulation model as a tool for forward looking, region-wide analysis of policy options prior to implementation. After evaluating price transmission between different markets in the region, it raised concern regarding the mismatch between the structure of maize markets in ESA and the traditional structure of partial equilibrium models. Underpinned by the law of one price, such models are typically non-spatial, relying on pooled net trade with a single representative world price transmitted to domestic markets through price transmission elasticities. This implies that trade elasticities are infinitely large, while a number of factors such as the time required to exploit arbitrage conditions, policy implementation, infrastructural restraints and imperfect information point to the need for finite elasticities. Maize markets in the region remain isolated from the global market and, with the exception of yellow maize in South Africa, the bulk of trade occurs within the region. This results in complex interactions between multiple regional markets, but limited interaction with the world reference price. Prices in any one country are influenced not only by domestic supply and demand dynamics, but also by availability of tradable product (mainly non-GM white maize) in a number of potential trading partners. Hence any model utilised for forward looking policy analysis should incorporate this combination of factors. The model outlined in this study specified a system of behavioural trade flow equations based on spatial arbitrage conditions and includes threshold variables that render trade-flow more elastic when breached. Hence it accounts for non-linearity and multiple regimes identified in price transmission analysis, which have largely remained absent from simulation models with the ability to project trade flow into the future under alternative assumption. The model applied in this study was shown to provide a more accurate representation of prices in ESA through a number of validation tests. Firstly, a range of statistical measures related to goodness of fit suggested that it improved the accuracy of simulating historic prices from 2013 – 2016 relative to a traditional price linkage approach. It was also shown to simulate a plausible outlook for maize prices in ESA over a ten-year horizon and provided responses to simple fluctuations in world prices and domestic supply that are more in line with literature and prior expectation. Furthermore, application of historic volatility in domestic yield levels and world prices resulted in an improved replication of past price volatility in the trade-linkage model relative to traditional price linkage approaches. Application of the modelling framework to the simulation of two different policy related future scenarios provided a final validation of its usefulness to answer relevant questions. In a situation where domestic supply is reduced by climatic variation, imposition of export controls in Zambia were shown to have the desired effect of reducing domestic prices for consumers, but the loss to producers outweighed the gain to consumers resulting in a net loss to society. Conversely, accelerated productivity gains in Tanzania were shown to provide a net benefit by reducing the price of maize for consumers. While the negative impact of lower prices on producers was noted, it was partially offset by higher output volumes and outweighed by consumer gains. The study’s contribution is therefore twofold: Firstly, it provided empirical evidence of the benefits attained from prioritising long-term productivity gains over short term reactions to inherently volatile prices. Secondly, it validated a tool for future policy analysis that can be applied in support of strategic decision making. The model structure essentially allows pricing to switch not only between import parity, export parity and autarkic regimes, but also between different markets as trade fluctuates. As such, it resembles actual market conditions more closely and contributes to narrowing the gap between retrospective econometric analysis of price transmission and the simplified structure of simulation models often used for ex ante analysis.