A Review of Nowcasting, Forecasting and AI in Economic Indicator Prediction

Abstract Accurate and timely assessments of economic indicators are essential to provide an early analysis of the economic situations. Therefore, forecasting and nowcasting applications with the Artificial Intelligence (AI) and Machine Learning (ML) technologies, which are offering enhanced accuracy and timeliness in predictions, have been widely used. The emerging chatbot technologies provide new insights in searching for data and generating information in these applications. This study conducts a comprehensive literature review to explore the use of AI and chatbot technologies in forecasting and nowcasting applications for economic and financial indicators. The “nowcasting”, “forecasting”, “artificial intelligence”, “chatbot”, “chat”, and “bot” are searched on the Web of Science (WoS) database. The 111 studies on the topic are reviewed according to dependent variable, AI model, and main findings. The results indicated that the AI models increased the accuracy of the forecasting and nowcasting applications compared to the traditional time series models. However, the data availability remains a significant concern. The use of chatbots in predictions limited with the conceptual studies. Therefore, the forecasting performances of ChatGPT and Google Gemini in EU Gross Domestic Product (GDP- current prices) and unemployment rate were evaluated. They showed good predicting performances according to mean absolute percentage error (MAPE). The promising prediction accuracies provided by AI models can contribute to the decision making process of the policy makers, consumers, and investors. These findings can also be valuable guidance for the technical applications of AI and chatbot in forecasting and nowcasting.