Predicting Young Bovine Slaughter Numbers Using Statistical Modelling

Nikolina Rizanovska, Aleš Stele, Andreja Smukavec
Statistika, 106(2): 214–227
https://doi.org/10.54694/stat.2025.37

Abstract
The Statistical Office of the Republic of Slovenia (SURS) developed a predictive model to estimate the intended slaughter or breeding of young bovine animals using administrative data from the Central Register of Bovine Animals (CRB). A binomial regression model with a logit link was employed to forecast slaughter rates, replacing the traditional, resource-intensive survey-based approach. Internal bootstrap validation and external calibration confirmed the model’s robustness, ensuring that predictions align with real-world occurrences and are suitable for future forecasting. The model demonstrated a significant improvement in predictive accuracy, with a difference of around 2% between the model's estimates and the survey results, equating to approximately 3 000 animals per year. The model is now closely aligned with observed values, demonstrating that administrative data can effectively replace costly telephone surveys. This shift promises both cost savings and enhanced accuracy in official agricultural statistics, with potential for broader application in other agricultural sectors or regions.

Keywords
Central Register of Bovine Animals, logistic regression, prediction, calibration, validation