Drug Demand Elasticity Estimation Based on Instrumental Variable Regression: A Case Study of Hospital X Transaction Data

Abstract

Introduction: Price determination is a key strategic decision affecting revenue and consumer behavior in the pharmaceutical sector. Demand elasticity is used to measure price sensitivity, but its estimation may be biased due to price endogeneity, requiring a more robust econometric approach. Research Methods: This study uses transaction data from Hospital X, including drug codes, prices, quantities, and HNA. Data are cleaned by removing missing values and trimming outliers at the 1^st–99^th percentile. Instrumental variables are constructed from HNA using lag, moving average, and differencing, and selected dynamically per product based on the highest first-stage F-statistic. Elasticity is estimated using a log-log model with Two-Stage Least Squares (2SLS) or Ordinary Least Squares (OLS), guided by the Durbin-Wu-Hausman test. Results: The results show that after preprocessing and structural selection criteria, 13 drug products were analyzed using elasticity estimation. The estimated price elasticity values ranged from -0.2249 to -4.3687, indicating different levels of consumer sensitivity toward price changes across products. Based on the comparison between estimated elasticity and break-even elasticity, four inelastic products were recommended for price increases, while nine elastic products were not recommended due to higher demand sensitivity. Conclusion: The study concludes that dynamic IV selection combined with structural diagnostics effectively handles price endogeneity, providing robust elasticity estimates to support targeted pharmaceutical pricing decisions.