Reference

Answers to the questions that come up when you actually use Chinese equity market data. Each one spans several datasets, which is why none of them fits on a product page. Every figure comes from our own onboarding notes on the data we sell.

• Why a join to China A-share company data silently drops rows

  Entity resolution in Chinese company data is partial, and the resolved share differs by dataset and by relationship type. An inner join discards exactly the rows that were not resolved, and those rows are not a random sample.

  11 datasets

• Which China A-share datasets are point-in-time, and which only look like it

  A dataset is point-in-time only if it tells you when a figure became public, not just which period it covers. Several widely used China datasets carry the period and not the filing date, which reads as point-in-time and is not.

  10 datasets

• Missing, zero, or blank: three different states in China filing data

  A real zero, a blank cell the filer left empty, and a line with no row at all are three distinct states in filing data. Collapsing them to one fills a panel with zeros that were never reported.

  4 datasets

• Code and unit traps in China A-share data

  Sentiment where 2 means negative, ownership ratios in two different units inside one dataset, and standardized line items that hide the as-reported label. Each one returns plausible numbers when read wrongly.

  4 datasets

• Disclosed versus inferred: two kinds of China supply-chain and ownership data

  Disclosed data is what filers published about themselves and arrives already consolidated. Inferred data is constructed and is built to be aggregated. The two obey opposite rules, and treating one like the other double-counts or under-counts.

  7 datasets

• One company, many securities: why A-share company counts come out too high

  Company facts attach to each listed security, so a company with two listings appears twice. Grouping on the ticker instead of the company key inflates every company-level aggregate.

  6 datasets

• When to roll up subsidiaries in China company data, and when it double-counts

  Whether to roll up depends on whether the dataset is already consolidated. Some ChinaScope tables are pre-aggregated up the tree, some are deliberately not, and one column inside SAM is a subtree total that looks additive and is not.

  7 datasets

• Primary key versus grain: why de-duplicating China filing data deletes real rows

  The column that uniquely identifies a row and the columns that say what the row is about are not the same thing. De-duplicating on the second silently discards real disclosures.

  8 datasets

• Which industry tree should you use for China equities?

  ChinaScope Reference carries five independent industry schemes in one table. They are separate trees, not levels of one hierarchy, and mixing them in a single grouping is meaningless.

  5 datasets

• Daily delivery does not mean daily data

  Delivery cadence and data cadence are separate facts. Several China datasets are delivered daily and change semi-annually, and one table inside Affiliate runs a full period behind its siblings.

  2 datasets