The Prediction of GDP by Aggregate Accounting Information. A Neural Network Model

Abstract

Currently, accounting information plays a key role in the economic world and bridges the gap between macroeconomics and microeconomics. The existing literature corroborates that aggregate-level accounting earnings data encapsulates information regarding GDP growth. Nonetheless, accounting data derived from GDP components have not been given due consideration. Consequently, this research proposes an aggregate-level model grounded in the four components of the GDP income-based method, intending to assess the predictive power of aggregate-level accounting data concerning GDP. Furthermore, this paper scrutinizes the forecasting performance of a neural network model built upon the conventional linear regression framework. Finally, a comparison of the outcomes derived from both models is conducted. The findings reveal that both the present value model and the value-added model corroborate the notion that accounting information derived from the four components of the income-based GDP accounting framework encapsulates data pertinent to future GDP. Furthermore, the model demonstrates heightened sensitivity towards the performance of the subsequent second quarter's GDP. Among the variables, Depreciation and Income exhibit the most substantial impact, while Salaries exhibit the least impact. Concurrently, this research noted an improvement in the fitting performance of the neural network model as compared to that of the traditional linear model.