Keywords
Agent-Based Model; Food Security; Calibration
Start Date
5-7-2022 1:40 PM
End Date
5-7-2022 2:00 PM
Abstract
Global food and nutrition security has long been an issue of international concern, with ‘zero hunger’ prominent as one of the United Nations’ (UN) Sustainable Development Goals and the importance of safeguarding food security acknowledged in the Paris Agreement. In previously published work (Ge et al. 2021), the calibration and validation of an agent-based model of global food and nutrition trade was demonstrated. The model focused on the role of relational factors in shaping countries’ trade behaviour, with each country represented as a discrete agent. It was designed to allow a wide variety of scenarios to be simulated, including the effects of changes to production and breakdowns of trading relationships. UN Food and Agriculture Organization (FAO) Food Balance Sheet (FBS) data was used to calibrate parameters pertaining to the weighting given by prospective trading partners to Gross Domestic Product, geographic distance, historical trade volumes, and emergent (i.e., modelled) trade relationships. The relative values of these weights determine how countries prioritize each other for buying and selling food. The model was calibrated and validated using a pareto analysis with two criteria: matching the trade volumes of the 91 FBS food goods and matching the trading partners. On the pareto front, the model performance on one of these dimensions cannot be improved without lowering the performance on the other dimension. The parameter combinations that form this front are therefore considered nondominant solutions. The historical trade volumes and geographic distance weights were acting to some extent as proxies for plurinational trade agreements such as the European Union or East African Community, but in a revised version of the model, we have added an additional weight that allows common membership of such agreements to be prioritized by countries explicitly. The data used for this task was drawn from the World Trade Organisation database of Regional Trade Agreements. We are interested in whether, and if so by how much, this additional detail improves the quantitative calibration performance of the model.
Recalibrating an agent-based model of global food and nutrition trade
Global food and nutrition security has long been an issue of international concern, with ‘zero hunger’ prominent as one of the United Nations’ (UN) Sustainable Development Goals and the importance of safeguarding food security acknowledged in the Paris Agreement. In previously published work (Ge et al. 2021), the calibration and validation of an agent-based model of global food and nutrition trade was demonstrated. The model focused on the role of relational factors in shaping countries’ trade behaviour, with each country represented as a discrete agent. It was designed to allow a wide variety of scenarios to be simulated, including the effects of changes to production and breakdowns of trading relationships. UN Food and Agriculture Organization (FAO) Food Balance Sheet (FBS) data was used to calibrate parameters pertaining to the weighting given by prospective trading partners to Gross Domestic Product, geographic distance, historical trade volumes, and emergent (i.e., modelled) trade relationships. The relative values of these weights determine how countries prioritize each other for buying and selling food. The model was calibrated and validated using a pareto analysis with two criteria: matching the trade volumes of the 91 FBS food goods and matching the trading partners. On the pareto front, the model performance on one of these dimensions cannot be improved without lowering the performance on the other dimension. The parameter combinations that form this front are therefore considered nondominant solutions. The historical trade volumes and geographic distance weights were acting to some extent as proxies for plurinational trade agreements such as the European Union or East African Community, but in a revised version of the model, we have added an additional weight that allows common membership of such agreements to be prioritized by countries explicitly. The data used for this task was drawn from the World Trade Organisation database of Regional Trade Agreements. We are interested in whether, and if so by how much, this additional detail improves the quantitative calibration performance of the model.
Stream and Session
false