VECM estimation, structural break tests, and rolling regressions on Chinese household savings data
This chapter evaluates whether the dominant models of Chinese household savings — life-cycle (Modigliani & Cao, 2004), permanent income (Friedman, 1957), precautionary motives (Choi et al., 2014), and competitive saving via sex ratios (Wei & Zhang, 2011) — produce stable parameter estimates when extended to the most recent available data (2023–2024).
Structural break tests — Chow tests at candidate break dates (1998: housing reform; 2008: financial crisis; 2015: slowdown) and Bai-Perron tests for endogenous, multiple break detection.
Recursive tests — CUSUM and CUSUM-of-squares statistics to detect progressive coefficient instability over time (Page, 1954).
Rolling regressions — Estimation over 10–15 year moving windows to visualise the trajectory of key coefficients and their standard errors.
VECM framework — Vector Error Correction Models for non-stationary series, with Johansen cointegration tests (Engle & Granger, 1987; Johansen, 1991).
Configure and run stability tests on macroeconomic savings models. Currently inactive — awaiting data.
A synthetic table comparing stability results across models — indicating which coefficients are stable, which exhibit structural breaks, and the magnitude of parameter drift. This diagnostic will formally justify the search for an omitted cultural variable developed in Chapters 2 and 3.