Equity Transfer Disputes in Chinese Case Law: The Outcome Depends on Who Sues, and Two Courts Wrote Almost Half of One Category

Equity transfer disputes are the largest company-law cause label in our index: 148,500 records, more than three times the next one. They are also a poor fit for a single outcome number. A seller suing for an unpaid price and a buyer suing to undo the deal are both “equity transfer disputes”, and they do not end the same way. We measured the category against our own corpus today. In 42,745 first-instance cases decided on the merits, the court granted at least part of the claim in 80.8 per cent. For sellers claiming the price the figure was 86.0 per cent. For plaintiffs trying to unwind a transfer it was 67.2 per cent.

Two further findings matter to anyone building a model or a dashboard on this category. First, 3,219 first-instance cases in two basic-level courts concern the same asset, shares held through China Shanshui Investment. They make up 45.1 per cent of all merits judgments in the category where the plaintiff asks the court to confirm or register a transfer. Second, the category is not only about judgments. It contains 30,756 enforcement case numbers, and 41.6 per cent of the enforcement files we could classify end with a ruling that closes the current round of enforcement. Under the rules, courts issue that ruling when no enforceable property has been found.

This is the fourth piece in our company-law series, after capital acceleration under three names, shareholder inspection suits and judicial dissolution. The dissolution piece showed that the choice of denominator changes the outcome rate. This piece adds two more things that change it: what the plaintiff is asking for, and how concentrated the filings are.

Finding the cases

We took the union of two routes. The first is the cause-of-action label 股权转让纠纷 (equity transfer dispute), on 148,500 records. The second is the phrase 股权转让 in the title of records that carry no cause value at all, which added 105,089. That gives 253,589 records. We normalised case numbers (full-width and half-width brackets, whitespace) and kept one record per case number and document type, which removed 37,759 duplicate records (14.9 per cent) and left 215,830 documents.

By document type, 104,523 are rulings, 79,565 are judgments, 9,929 are mediation statements, 2,824 are notices, 420 are decisions and 18,569 have a title that names no type. We split case numbers by the procedure code inside the case number itself:

A further 3,581 documents carry no case number. The label column is the first warning. For first-instance cases the label reaches 82.3 per cent of cases filed up to 2016, 31.5 per cent of those filed in 2017 to 2019, 49.0 per cent for 2020 to 2022 and 100 per cent for 2023 to 2025. As in our earlier pieces, whole ingest batches were loaded without a cause value. A cause filter on its own would return little more than half of this category, and for the most litigated years less than a third.

What happened in 118,625 first-instance cases

We parsed the operative part of every first-instance document, the passage after 判决如下 or 裁定如下, and gave each case number one final disposition. A decision on the claim outranks a mediation statement, which outranks a withdrawal, which outranks a ruling that struck out or refused the suit. Judgments whose operative part mentions a counterclaim were set aside, because they do not say cleanly whether the plaintiff’s claim succeeded.

Of the 42,745 cases decided on the merits without a counterclaim, the plaintiff obtained at least part of the claim in 80.8 per cent and all of it in 45.1 per cent. Measured against all 90,801 cases with a disposition, at least partial success falls to 38.0 per cent. That gap comes from withdrawals, which ended more than a third of the cases. As in the dissolution piece, the number you quote depends on which question you ask.

The 6,000-character cap was heavier here than in dissolution suits. When we ran this analysis on 17 September 2026, 30,462 of the 253,589 records had a body of exactly 6,000 characters. Only 72 records were longer, so this was a storage cap, not the real length of the documents. The cap was backfilled later in September 2026; re-measured on 24 September 2026 it affected 0.003 per cent of the corpus, and the counts in this section describe the corpus as it stood on 17 September 2026. 8,095 first-instance case numbers with a judgment lost their operative part this way. If they were all dismissals, the merits figure would be 67.9 per cent; if all grants, 83.8 per cent.

The claim decides the rate

For every case decided by judgment, we read the plaintiff’s claims from the passage that follows 诉讼请求 (claims) near the start of the judgment and sorted them into five groups. The groups are: a price claim (payment of the transfer price, without asking to undo the deal); an unwinding claim (rescission, a declaration of invalidity or revocation, or return of money or shares, without a price claim); a confirm or register claim (declare the agreement valid, confirm who holds the shares, or order the registration change); mixed (a price claim together with an unwinding claim); and other.

The pattern makes legal sense. A seller who has transferred shares and not been paid usually holds a signed agreement and a registration record, and the defence is often simply that the buyer has no money. A plaintiff who wants to undo a completed transfer has to prove misrepresentation, a fundamental breach or a ground of invalidity, and courts are cautious about reversing changes to a company’s shareholding. The gap between the two, nearly 19 percentage points, is larger than most of the year-to-year movements that analytics products like to chart.

Here is how the classes look in the documents. In (2021)浙0727民初1758号 the Pan’an County People’s Court ordered the buyer to pay a transfer price of 100,000 yuan with interest. In (2021)皖0522民初2354号 the Hanshan County People’s Court ordered payment of 355,000 yuan to three sellers. In (2020)川1525民初456号 the Gao County People’s Court dismissed a claim that a transfer agreement was invalid.

We sampled each claim group at random to check the rules. Of 20 unwinding claims, 20 asked for rescission, invalidity, revocation or return. Of 20 price claims, 17 clearly asked for the price. In two the claim passage was not where the rule looked, and one was a claim to buy back shares. The confirm-or-register group is the weakest: only 11 of 20 were clean. The others were loan-style claims, claim passages the rule did not locate, or empty text. Treat its rates as indicative. Among judgments classed as granted in full, 15 of 15 sampled ordered relief for the plaintiff, and 15 of 15 sampled dismissals dismissed the claim.

Two courts, one asset, 3,219 cases

Counting filings by party exposed something no rate table shows. The Tianqiao District People’s Court in Jinan has 1,682 first-instance equity transfer cases docketed in 2018 that belong to one set; in almost all of them the plaintiff is Jinan Industrial Development Investment Group. The Ruzhou People’s Court in Henan has 1,537 docketed in 2018 and 2019, and 1,509 of their titles name the same individual plaintiff. Almost every judgment in both sets concerns agreements to transfer interests in shares held through China Shanshui Investment Company Limited, the holding structure of a cement group. Each suit asks the court to confirm an agreement and the plaintiff’s entitlement to the shares.

The outcomes are uniform: in Tianqiao 1,434 cases granted in full, 124 withdrawn and no merits dismissal, as in (2018)鲁0105民初5154号. In Ruzhou 590 were granted in full and 478 in part, as in (2018)豫0482民初11900号. The same court in Jinan then docketed 695 third-party revocation suits in 2019 and 2020 whose titles name that individual. We did not read the merits of these suits, and we draw no conclusion about the dispute itself. The data point is simpler. Two courts account for 2,502 of the 5,542 confirm-or-register judgments in the category (45.1 per cent), and removing the clustered cases moves that group’s rate from 89.2 to 80.4 per cent.

Two smaller clusters point the other way. The Mianzhu People’s Court in Sichuan has 455 cases docketed in 2017 against the same four companies of the Sichuan Hongda group. 402 were struck out, as in (2017)川0683民初1004号, and 53 were withdrawn. The Jinniu District People’s Court in Chengdu has 297 price claims against one company, all dismissed in 2014, as in (2014)金牛民初字第3144号. Across all four clusters, 2,802 merits judgments are 6.6 per cent of the category. They make up 17.5 per cent of merits judgments in cases filed in 2017 to 2019. A court-level or year-level chart built on this category without a cluster check is partly a chart of four dockets.

By filing year

With the clusters removed and cases bucketed by the year in the case number, the merits rate barely moves. What changes is how cases end:

The last row should not be read as behaviour. Cases filed recently are incomplete in the same two ways we described for dissolution suits. A withdrawal ends a case quickly and a contested judgment takes longer, so recent cohorts are missing the cases still being fought. Publication of judgments also fell sharply from 2023. The near-disappearance of mediation statements is most likely a publication effect: mediation statements are not routinely published, and the earlier share may come from particular courts or batches. We have not separated these effects.

On appeal, and after judgment

We found 35,704 second-instance case numbers. For 10,589 (29.7 per cent) the text stopped at the 6,000-character cap before the operative part, because appellate judgments recite the first-instance decision at length. Of the 23,215 with a readable outcome, the appeal was dismissed in 12,736 (54.9 per cent), withdrawn in 4,944 (21.3 per cent), led to reversal or variation in 3,017 (13.0 per cent) and to remand in 2,518 (10.8 per cent). We did not determine which party appealed.

The enforcement files deserve their own attention. Of 24,141 implementation files (codes 执 and 执恢), 10,037 (41.6 per cent) contain a ruling closing the current round of enforcement, as in (2020)苏0115执1085号 from the Jiangning District People’s Court and (2020)豫0106执恢201号, a resumed enforcement file that was closed again. 5,017 (20.8 per cent) contain a ruling ending enforcement outright, often after the parties settled. 3,962 show only freezing, seizure or auction measures, 2,499 (10.4 per cent) record that the obligation was performed, and 2,626 are other or empty.

Those shares are shares of published enforcement documents. Rulings closing a round of enforcement are published as a matter of course, while many completed files end with a closure notice that is less often published. The true completion rate is therefore unknown and probably higher than 10.4 per cent. What the data does show is that a model trained only on first-instance judgments sees the 86 per cent of price claims that succeed. It does not see how often that judgment was later collected.

How to reproduce this against any corpus

1. Take the union of the cause label and the dispute name in titles, and de-duplicate on normalised case number plus document type.
2. Split by the procedure code in the case number and report enforcement files separately. They are part of the category, not noise.
3. Classify the plaintiff’s claim before computing any outcome rate. At minimum, separate price claims from unwinding claims.
4. Count filings per court and per repeat party, and publish rates with and without clusters above a few hundred cases.
5. Count records whose body ends at a round number of characters, and report bounds for them.

What a buyer should ask

If a vendor quotes an outcome rate for Chinese equity transfer disputes, ask what the plaintiff was claiming. Ask whether a court with 1,600 near-identical judgments is weighted the same as everywhere else. Ask whether enforcement files are in the dataset and what the product does with them. And ask how many records end at a fixed length. A single “win rate” for this category mixes sellers, buyers and four large dockets. It describes none of them well.

What we are not claiming

We did not measure any other provider’s data. These are rates among published documents in one corpus, not among all suits filed in China, and nothing here predicts a particular case. “Granted in part” covers anything from nearly full relief to a small fraction of the claim; we did not measure amounts. Claim groups come from rules with the precision reported above, and the confirm-or-register group is only indicative. We did not examine the merits of the clustered suits, identify which party appealed, or link enforcement files to the judgments they enforce. The title route adds records that mention equity transfers but carry no label, so a small number may belong to other disputes. We have not corrected any defect described here: the capped texts, empty records and duplicates are still in the index.