Unveiling the dynamics of unemployment in Bangladesh through non-linear modeling based on economic perspective

Unemployment is a significant challenge globally, and it threatens economic growth and stability, particularly in developing countries such as Bangladesh. Both economic and non-economic factors influence a country’s unemployment rate. Identifying and addressing the factors that contribute to the unemployment rate is crucial to reduce it. Therefore, this study aims to identify an effective model for depicting the unemployment scenario by evaluating its predictive performance and identifying its contributing factors. Yearly data from 1991 to 2023 were extracted from the World Bank and the Bangladesh Bureau of Statistics’ publicly accessible databases to model unemployment trends. The study employed Nadaraya-Watson kernel regression with Gaussian, Truncated Gaussian, and Epanechnikov kernels, along with the Generalized Additive Model (GAM) and Multiple Linear Regression (MLR). The results revealed a nonlinear pattern in unemployment rate data. Among the models, the Nadaraya-Watson Gaussian kernel demonstrated superior predictive accuracy based on most of the performance measures, achieving an R-square value of 0.9913, the lowest AIC (−102.1859) and BIC (−84.5971), RMSE of 0.0067, MAE of 0.0035, and MAPE of 0.6442. The study concludes that the Nadaraya-Watson Gaussian kernel is a highly effective model for modeling unemployment trends in Bangladesh for the data. The study also identifies significant potential factors associated with unemployment, such as GDP growth, literacy rate, and FDI. The Nadaraya-Watson Gaussian kernel model is the most robust and reliable approach for modeling unemployment dynamics in Bangladesh.