Unemployment is a critical factor in the global economy, influenced by both economic and non-economic variables. The complexity of trends and fluctuations makes forecasting unemployment rates challenging, hindering policymakers in implementing effective measures to mitigate economic impact. Traditional forecasting methods often struggle with capturing non-linear dependencies and sudden shifts in labor markets. This work explores the use of echo state networks for unemployment forecasting based on publicly available historical economic data. A modified optimizer is proposed to address the challenging task of hyperparameter selection in echo state networks, ensuring favorable performance and improved generalization. Evaluations on real-world data demonstrate promising results, with best generated model achieving a low mean squared error of 0.006788, highlighting the potential of reservoir computing in enhancing predictive accuracy and supporting data-driven economic decision-making.