Taming the Complexity of Distributed Lag Models: A Practical Approach to Multicollinearity, Outliers, and Auto-Correlation in Finance

This research investigates the application of robust estimators within the finite distributed lag model (DLM), a critical framework in finance research capturing temporal dependencies between lagged explanatory variables and a response variable. Traditional Ordinary Least Squares (OLS) estimation faces challenges when dealing with high lag counts, multicollinearity, and outliers, potentially compromising parameter estimates and model reliability. Employing real-world data from the RBI, spanning the years 2022–2023 encompassing budgetary and economic variables of Indian states and Union Territories, the study demonstrates that the MMS estimator emerges as the most efficient estimator, showcasing enhanced robustness against outliers and multicollinearity. Additionally, the study reveals positive autocorrelation in residuals, underscoring the importance of robust methods in addressing such issues in financial modeling. This research contributes valuable insights to financial analysts and offers a more accurate understanding of dynamic relationships in financial systems.