An Agent-Based Model of the Global Salmon Market
MetadataShow full item record
While the global production of Atlantic salmon has grown substantially in the last few decades, the wholesale spot market is still organised around over-the-counter transactions negotiated between sellers and buyers. This mode of organisation may lead to a suboptimal price discovery process and high price volatility, which exposes market participants to risk. The volatility is imperfectly understood, making price forecasting difficult. For example, practitioners report that volatility often increases near the end of the week, a phenomenon that does not appear to be described in existing literature. We present an agent-based model of the market in which the distinctive features of the salmon market, notably its bilateral organisation and the constraints imposed on suppliers by salmon biology, are represented explicitly. The model replicates the real market by simulating a succession of weeks of agent interactions, where each week is structured as a set of iterative negotiations between sellers and buyers. Agents position themselves strategically so as to maximise profits and minimise risk. The supply and demand sides of the market are estimated based on a combination of historical data and the previous events of the simulation. The price of salmon arises as an emergent property from the interactions of agents in the market. This price forecast is shown to match historical fact in some aspects, and will at times outperform the futures market in terms of accurately forecasting the price. The model also accurately predicts the market's reaction to a selection of exogenous shocks.