A New General Conceptual Approach to Modeling the Livestock Sector: An Application to the Japanese Swine-Pork Sector Jacinto F. Fabiosa and Xiaoting Qi Center for Agricultural and Rural Development

Many livestock sector models have limited coverage of relevant variables, and are somewhat ad hoc in their choice of what should be specified as behavioral equations. This study develops a generic conceptual approach to modeling the livestock sector that provides consistent rules of specification and better coverage of variables. This approach is then applied to the swine-pork sector of Japan. The new approach departs significantly from existing models. The structure clearly differentiates stock and flow variables; only flow variables have behavioral specifications and stock variables are accounting identities; flow variables are expressed in rates rather than levels; logistic functions are used in most flow variables to automatically impose biological-technological limits; and swine slaughter number and weight are disaggregated into sow and other swine (i.e., gilt-barrow). The estimated Japanese swine-pork model has good fit, no serial correlation, significant coefficients, correct signs, and is also able to capture both the mean and variability of all endogenous variables. A NEW GENERAL CONCEPTUAL APPROACH TO MODELING THE LIVESTOCK SECTOR: AN APPLICATION TO THE JAPANESE SWINE-PORK SECTOR This study develops a general conceptual approach to modeling the livestock sector. The specific features of this approach are then applied to the Japanese swine-pork sector. Many existing livestock sector models are limited in their variable coverage and ad hoc in the choice of what should be specified as behavioral equations and what should not be specified. For example, the Organization for Economic Cooperation and Development’s (OECD) model of the Japanese swine-pork sector is a three-equation model that includes behavioral specification of a pork import demand, pork domestic demand, and pork production. There is no coverage of any live animal variables. Most of the U.S. Department of Agriculture’s (USDA) country models are similarly specified. The Food and Agricultural Policy Research Institute’s (FAPRI) Japanese model, on the other hand, includes both meat and live animal inventory. It has six behavioral equations: sow inventory, swine inventory, swine slaughter, pork production, domestic demand, and import demand. We develop a new general conceptual approach in modeling the livestock sector to capture more amply the behavioral and biological processes unique in this sector. There are new contributions that depart from existing models in a significant way. First, stock and flow variables are clearly differentiated. This differentiation allows for proper accounting of live animals to ensure consistency in the underlying biologics of the model. It also allows the model to easily accommodate adjustments in inventory and production that may result from exogenous shocks such as the swine foot and mouth disease (FMD) in Taiwan and classical swine fever (CSF) in the European Union (EU). Second, only flow variables that correspond to the actual decisions facing producers and consumers are specified with behavioral equations. These variables may include number of animals A New General Approach to Modeling the Livestock Sector / 3 slaughtered, number of animals added to the breeding herd, slaughter weight, import, export, consumption, pork stock demand, pig crop, and mortality. On the other hand, stock variables such as sow inventory and other swine inventory are not directly specified as behavioral equations. Instead, their levels over time are simply derived, given their beginning levels, from the changes in the flow variables that impact them. Third, swine slaughter is disaggregated into sow and other swine slaughter to allow differential impacts of price changes in both the direction and magnitude of these variables. An econometric decomposition method is applied to derive a slaughter weight equation that can estimate a separate average slaughter weight for sows and other swine slaughtered. Pork production is a product of number slaughtered and slaughter weight for sows and other swine. A fourth new contribution is that the flow variables are modeled as rates rather than as levels. For example, the rate of sow slaughter, which is defined as the ratio of the number of sows slaughtered and the beginning sow inventory, is used rather than the number slaughtered. Based on historical data, the rates appear to have more stable behavior and are usually comparable over time and across countries. Fifth, biologicaltechnological restrictions are imposed using a logistic function on some of the rates. Model The specification of the Japanese swine-pork sector includes eight flow variables: domestic demand, pig crop, swine death, sow slaughter, other swine slaughter (giltbarrow), slaughter weight, sow addition, and pork import. The two stock variables are sow inventory and other swine inventory. There is also an equilibrium condition. A combination of linear, double log, and logistic functional forms is used. A linear function is specified for domestic demand and slaughter weight equation. A double log function is used for swine death, pork import, and pork stock. A logistic function is specified for pig crop, sow slaughter rate, other swine slaughter rate, and sow addition rate to