Overcompensating expatriate employees has obvious pecuniary costs, but undercompensating risks potentially more costly attrition. We propose an approach that incorporates employee and assignment characteristics to determine the compensation strategy that best balances employers’ cost and employees’ net income. Using data from 6
Global IQ, a full-suite HR data company, we train surrogate models to predict the salary and net-income under each of the five main expatriate compensation strategies: tree-based models, boosting models, and a stacked ensemble model perform best. The optimization engine uses a weighted sum with additional components to incorporate exchange rate fluctuations as the objective function and finds that salary comparison with allowances is the optimal strategy for a plurality of strategies in our data.
Watch the team present this project at 01:45:04 in the session recording here.
Keywords: expatriates, expatriate compensation, human resources, compensation strategy, machine learning, optimization
Faculty Advisor
Gizem Agar, PhD is an expert in data analytics, machine learning, and transformation with a passion for mentoring. With 15+ years of interdisciplinary academic and industry experience, her latest work is in manufacturing, pricing, logistics, and supply chain. She has received the CEO Award and Outstanding Achievement in Analytics awards for her contributions to Caterpillar.
Dr. Agar teaches Principles of Data Mining, Python for Analytics, Supply Chain Optimization, Capstone courses and advises students. She holds a PhD and MSc in Industrial Engineering from University of Oklahoma, and BSc in IE and a BSc in CE from Cankaya University, Turkiye. She was a visiting scholar at the Kuhne Logistics University (Hamburg, Germany) and at the Technical University of Vienna (Vienna, Austria).