RIDGE REGRESSION AS A POLICY FORECASTING TOOL IN LOW-DATA ENVIRONMENTS: EVIDENCE FROM BANGLADESH

Abstract

Forecasting economic development outcomes in low-data environments is a significant challenge in many developing countries. This study presents ridge regression as a solid alternative to ordinary least squares (OLS) for policy forecasting in situations with small sample sizes and multicollinearity. Using Bangladesh as a case study, we apply ridge regression to evaluate how population growth, savings rates, and inflation affect GDP per capita growth from 1990 to 2025. The model shows a moderate fit (R² = 0.686) and delivers stable, interpretable coefficients despite multicollinearity. Policy simulations, such as increasing the savings rate or reducing population growth, indicate considerable potential to improve economic outcomes. These findings suggest that ridge regression can act as both a diagnostic tool and a practical method for forecasting and assessing policy interventions in data-limited contexts.