Industry Class Stability Index (ICSI): A novel diagnostic metric to enhance the transferability of freight demand models

Abstract

Reliable freight demand models are essential for planning and policy, yet their development often depends on costly establishment-level surveys that are difficult to replicate across regions. This study introduces the Industry Class Stability Index (ICSI), a novel diagnostic metric for assessing the homogeneity and transferability of data-driven industry classifications. A four-stage framework is proposed: (i) deriving industry classes using latent class analysis on shipper characteristics (fleet ownership, commodity value density, and business age cohort); (ii) evaluating similarity with ICSI; (iii) developing industry class-specific freight production models using robust regression; and (iv) assessing spatial transferability through mean absolute errors ratio (R). The framework is demonstrated with establishment-based freight survey data from North and Central Kerala, India. Results show that derived industry classes exhibit high similarity (ICSI >0.97) and that freight production models achieve near-perfect transferability (R ≈ 1). Medium-duty fleet ownership emerges as the most influential driver of industry class stability. From a managerial perspective, the framework enables cost-efficient surveys and model development, reducing redundancy in data collection while ensuring robust forecasting. In the longer term, it supports scalable and sustainable freight planning, policy continuity across regions, and greater efficiency in logistics systems.