A Two-Stage Framework for Detecting Conditional Lead-Lag Relationships: Evidence from the Onshore-Offshore Renminbi Dual-Listing

The Renminbi is unique among major currencies: it trades simultaneously in two segregated markets with distinct institutional rules.. This study maps the lead-lag structure of the USD/CNH–USD/CNY pair at the H1 timeframe using a two-stage framework: exhaustive combination testing and six-layer statistical validation. We test 1,280 configurations of tolerance window (±1 to ±5 bars), trading session (All, Tokyo, London, New York), and regime filter for Pair 2 (80 combinations of return, volatility, and stochastic state). The lead-lag relationship is session-dependent: it is strongest during Asia-Early (84.6% LEAD) and Asia-Mid (81.0% LEAD), and weakens during Asia-Late (37.5% LEAD). It is also regime-conditioned: the stochastic filter OVERBOUGHT_OVERSOLD appears in 6 of the top-7 configurations, suggesting that the signal is strongest when the onshore market is in an extreme state. Volume analysis reveals a substantial asymmetry—the median CNH/CNY volume ratio is 252.3×—yet neither volume ratio nor liquidity asymmetry significantly predicts offset magnitude (all p > 0.05 after FDR correction). Liquidity asymmetry is a descriptive characteristic of the dual-listing, not a predictive one for lead-lag timing. This study makes four contributions. Methodologically, it introduces a two-stage framework tailored to a single pre-determined pair, eliminating the need for screening and directing full analytical effort at parameter optimization and statistical validation. Empirically, it identifies one validated configuration that survives all six validation layers, providing a specific, actionable signal for practitioners. Negatively, it documents the rarity of validated structure (0.08% conversion rate) and the non-significance of volume and liquidity as timing predictors—both informative results. Practically, it offers a reproducible framework that can be applied to other dual-listed or cross-market pairs.