How does the economic structure break change the forecast effect of money and credit on output? Evidence based on machine learning algorithms

This study examines quarterly data from China spanning 1993 to 2016, employing machine learning algorithms to investigate the predictive ability of money and credit on output amidst significant structural changes, including the 2007 Global Financial Crisis and shifts in financial structure. The findings highlight that credit is a more effective predictor of output than money, capable of forecasting both short and long-term output trends. However, the predictive strength of both money and credit has diminished post-2007, with the advent of financial development further diminishing their forecasting effectiveness. The analysis demonstrates that machine learning offers more nuanced, long-term predictive insights compared to traditional Vector Autoregression (VAR) methods. For developing economies like China, the results indicate a significant reliance on the bank credit channel for monetary policy transmission. The study emphasizes the importance of market-oriented reforms to minimize financial market arbitrage and advocates for a refined classification of non-monetary assets into bank loans and other bonds to enhance the accuracy of general equilibrium analyses.