Trend-cycle decomposition is inherently sensitive to model specification. This paper utilizes the Bayesian model comparison framework of Grant and Chan (2017) to evaluate a suite of unobserved components (UC) models, identifying the optimal specification for Chinese aggregate output. We formally test for output hysteresis, the timing and magnitude of structural breaks, and the interdependence of trend and cyclical innovations. Our results provide robust evidence of output hysteresis in China, with the optimal specification incorporating both structural breakpoints and significant trend-cycle correlation. We identify a pivotal regime shift in 2013, characterized by a structural deceleration in the annualized trend growth rate from 9.8% to 5.76%. Furthermore, variance decomposition reveals that transitory cyclical shocks dominate output dynamics, outweighing the contribution of stochastic trend variations. These findings suggest that China’s transition to a New Normal growth path is coupled with a high degree of path-dependency, where cyclical volatility exerts persistent effects on long-run productive capacity.
