A General Conceptual Approach to Modeling the Livestock Sector: Application to the Japanese Swine-Pork Sector Model with Analysis of the "No Gate Price Policy Scenario"

Abstract

This study develops a generic conceptual approach to modeling the livestock sector that provides better coverage of variables to ensure the consistency in the underlying biologics of the model and also provides consistent rules of specification. This approach is applied to modeling the swine-pork sector of Japan and then used to analyze the impact of removing Japan's gate price policy and variable levy for pork imports. The new approach departs significantly from existing models. For consistency in the biologics of the model both live animals and meat components of the sector are fully covered. The structure clearly differentiates stock and flow variables and investment and production decisions. A standard rule of specification is established that only flow variables that correspond to the actual decisions faced by producers and consumers are specified with a behavioral equation, while stock variables are derived from changes in the relevant flow variables using an accounting identity. The flow variables are expressed in rates rather than levels and specified with logistic functions to automatically impose biological-technological constraints. Swine slaughter number and weight are disaggregated into sow and barrow-gilt. Market-clearing price is endogenously solved. The estimated 12-equation Japanese swine-pork model has good fit as shown by high R2, no serial correlation with Durbin-Watson (DW) statistics approaching 2, significant coefficients, correct signs, and reasonable magnitudes. All validation statistics from a within-sample simulation of the model suggest that the model with enriched structural specification is also able to capture both the mean and variability of all endogenous variables. Except for swine death (2.3 percent) and pork stock (-1.9 percent) which are subject to unexpected technical and policy shocks, all other endogenous variables had mean percentage errors (MPE) of less than one percent. Prediction errors are mostly from random sources. Except for pork stock (5.3 percent), the share of bias for all other endogenous variables is less than 3 percent. Finally, the model was also used to analyze the No Gate Price Policy Scenario. The model generated expected changes both in terms of direction and magnitude that are within reported range from earlier studies. Specifically, imports increased significantly but were mostly absorbed into pork stocks to cushion its impact on domestic prices.