Disclosure and Scale: HK Connected Transactions Under Chapter 14A

A 74 Percent Capped Cohort

In the cohort of 1,403 Hong Kong-listed issuers^[1] for which the framework can compute the ratio of connected transactions to revenue, 73.8 percent — 1,036 firms — sit at the upper measurement bin. The ratio measures the aggregate scale of related-party transactions over a fiscal year against the issuer's reported revenue. A firm at the upper bin carries a related-party transaction volume that is substantially equivalent to, or in excess of, the revenue base of the listed vehicle.

The Hong Kong Stock Exchange's Listing Rules Chapter 14A defines a connected-transaction regime that surfaces these flows through one of the most detailed disclosure architectures among Asian listing venues^[2]. Connected persons are identified under Rule 14A.07. Percentage ratios under Rule 14.07 calibrate the disclosure obligation to transaction size. De minimis exemptions apply below low thresholds; written agreement and annual disclosure obligations engage between low and middle thresholds; full announcement plus shareholder approval apply above the higher thresholds, with Majority-of-Minority approval required for connected transactions exceeding the principal materiality bands. Continuing connected transactions under Rule 14A.31 are subject to annual caps, typically renewed on a three-year horizon, with INED confirmation and auditor reporting on cap compliance.

The architecture surfaces the transactions. It does not, by itself, constrain their absolute scale. The 74 percent capped distribution observed across the measurable universe is the framework's reading of where that distinction settles in practice. This Note examines what the distribution shows about the relationship between Chapter 14A disclosure architecture and the substantive Conflict-of-Interest Risk axis identified as the universe-level signature in Note 1^[3].

---

What Chapter 14A Compels

Chapter 14A operates as a graduated disclosure-and-approval regime calibrated to transaction size relative to the listed issuer. The percentage-ratio framework — assets, profits, revenue, consideration, and equity capital ratios — produces a single measure of materiality against which each connected transaction is classified. Below a de minimis floor, transactions are exempt. Above ascending percentage-ratio thresholds, escalating disclosure and approval obligations engage: written agreement and annual report disclosure for smaller transactions, full announcement and shareholder circular for larger ones, and Majority-of-Minority shareholder approval at the materiality bands above the principal disclosure threshold.

For continuing connected transactions — recurring related-party transaction streams that span multiple fiscal periods — Rule 14A.31 introduces an annual cap mechanism. The issuer must establish an annual cap on the aggregate value of the recurring transactions, obtain shareholder approval if the cap exceeds the disclosure threshold, and renew the cap on a typical three-year cycle. Auditor and INED confirmation are required annually that the actual transaction volume has not exceeded the cap.

The architecture has two functional purposes. First, disclosure: every connected transaction above the de minimis floor enters the public record with terms, parties, and INED rationale on continued independence. Second, approval: above the principal materiality thresholds, minority shareholders have a binding vote, with controlling shareholders required to abstain. The combination is structurally robust by Asian-market comparison; few listing venues impose comparable Majority-of-Minority approval architecture on related-party transactions.

What the architecture does not do is constrain the absolute scale of the transactions. A controlling shareholder whose connected transactions clear the disclosure and approval gates may carry connected-transaction volumes that are substantial relative to the listed vehicle's underlying business, provided the disclosure is complete and the approval is obtained. The 74 percent capped distribution is the empirical signature of this distinction.

---

The Universe-Wide Capped Distribution

Of the 2,768 issuers in the universe, the framework's connected-transaction scale indicator is measurable for 1,403 firms (50.7 percent). The remaining 1,365 firms either lack disclosed connected-transaction data in a parser-readable format or carry zero or unmeasurable revenue figures that make the ratio undefined. The findings below are bounded to the 1,403 measurable cohort.

RPT-to-revenue ratio	n	% of measured
Below 5%	220	15.7%
5% – 10%	42	3.0%
10% – 20%	55	3.9%
20% – 50%	50	3.6%
50% – 100% (upper measurement bin)	1,036	73.8%
Total measured	1,403	100.0%

The distribution is concentrated at the endpoints. The upper bin captures 73.8 percent of measured firms; the below-5-percent bin captures 15.7 percent. The middle bands (5 percent through 50 percent) account for 10.5 percent collectively. Universe summary statistics confirm the shape: mean ratio 0.750, median 1.000 (at the upper bin), IQR 0.380 to 1.000.

The 74 percent at the upper bin is a universe-baseline pattern, not a Poison Apple cohort exception. Note 3 examined the Poison Apple cohort's R-02 distribution in detail^[4] — 90 percent of PA firms with measurable ratio sit at the upper bin. The universe baseline is already 74 percent. The PA cohort is more extreme on this dimension than the universe; the universe is not far behind.

The structural implication is direct. Chapter 14A disclosure is producing transaction-level disclosure across the issuer base, but the aggregate scale of related-party transactions disclosed under that architecture runs up against the framework's measurement ceiling for three-quarters of the firms whose ratio can be computed. The disclosure is achieving its surfacing function. The substantive transaction streams that the disclosure surfaces are large.

---

R-axis Sub-component Asymmetry

The R-axis universe mean of 56.6 (Note 2) does not reflect uniform weakness across the axis's underlying components. The framework's R-axis aggregates several sub-components measuring different aspects of conflict-of-interest exposure. Their universe-level distributions show where the load-bearing weakness sits and where the axis is structurally clean.

R-axis sub-component (functional)	Universe coverage	% at max	% at zero
Connected transaction scale	1,678	23.2%	64.6%
Related-party loans / guarantees	2,252	59.8%	31.9%
Auditor change direction	2,443	0.0%	0.1%
Director interlock	2,443	1.4%	1.2%
Controlling-shareholder + business change	2,443	99.2%	0.8%
Whistleblower policy disclosure	2,443	71.8%	28.2%
Cash-vs-interest yield consistency	1,006	53.0%	29.7%

Two patterns dominate the universe-level signal.

The load-bearing weak sub-component is connected transaction scale. Sixty-four point six percent of measurable firms score at zero; only 23.2 percent reach the maximum. The bimodal distribution — most firms either at very low connected-transaction exposure or near the measurement ceiling — produces the universe-level R-axis depression. This is the indicator pathway through which the 74 percent capped distribution from the prior section enters the composite score.

The load-bearing strong sub-component is related-party loans and guarantees. Fifty-nine point eight percent of measured firms reach the maximum, indicating the absence of measurable related-party loan exposure. The measurement reflects parser-extracted disclosure of related-party loan amounts, not an audit of off-balance-sheet exposures; a measured-zero score reflects no disclosed related-party loan amount, which the framework reads as cleaner than partial or non-zero disclosure.

The remaining sub-components contribute less universe-level variance. The director interlock measure clusters at the cohort midpoint (98.6 percent of firms score in the middle band). The controlling-shareholder-plus-business-change measure — which captures reverse takeover patterns — registers at the maximum for 99.2 percent of firms, reflecting the relative rarity of concurrent controlling-shareholder change and business pivot at the universe scale. The auditor change direction measure clusters at a stable midpoint, with the cohort's variability low.

The R-axis weakness story, in this reading, is fundamentally a connected-transaction story. The 56.6 universe mean reflects the bimodal distribution of the connected-transaction sub-component, mediated by stronger but less variable contributions from related-party loans and the structurally rare reverse-takeover pattern.

---

Origin and Archetype Variation

The within-origin breakdown of the connected-transaction scale distribution shows a uniform pattern across the three dominant origins.

Origin	Measured n	Mean ratio	Capped (% at upper bin)	Mean CCT count
HKL Local	1,214	0.752	70.7%	2.96
PCH P-chip	153	0.756	69.3%	3.07
HSH H-share	18	0.744	66.7%	6.11
RCH Red Chip	17	0.630	47.1%	6.71
FOR Foreign	1	0.031	0.0%	6.00

HKL, PCH, and HSH converge on capped shares between 67 and 71 percent. The cross-origin range on capped share is 4.0 percentage points; the cross-origin range on mean ratio is 0.012. Connected-transaction scale ceiling pressure is a uniform pattern across the three principal Hong Kong, mainland-private, and mainland-state-listed cohorts.

RCH presents a different profile. Capped share 47.1 percent (the lowest among cohorts with non-trivial sample); mean ratio 0.630 (also the lowest). The cohort's connected-transaction scale is meaningfully above zero but distributes across mid-range bins rather than concentrating at the ceiling. The compensating signal is the connected-transaction count: RCH issuers carry an average of 6.71 enumerated continuing connected transactions per firm — more than double the HKL/PCH baseline of approximately three. RCH issuers, on a measurable cohort of 17 firms (out of the 20 in the universe RCH cohort), exhibit a many-small-transactions pattern rather than a few-large-transactions pattern.

HSH issuers, on a similarly small base of 18 measurable firms, share RCH's elevated CCT count (mean 6.11) but follow HKL/PCH on the capped share (66.7 percent). The H-share cohort carries both more enumerated CCT items and similar RPT-to-revenue ratios. The mainland-incorporated, state-supervised issuer regimes produce a higher count of distinct connected-transaction items than the HK-incorporated or mainland-private regimes do, even where the aggregate revenue ratio is comparable.

The archetype-level breakdown shows a sharper structural pattern.

Archetype	Measured n	Mean ratio	Capped %
Celestial	67	0.034	1.5%
Hidden Gem	66	0.471	45.5%
Chameleon	662	0.728	67.7%
Poison Apple	519	0.905	85.4%
Kill Switch	89	0.756	69.7%

The Celestial cohort exhibits an extraordinary cleanness on connected-transaction scale: mean ratio 0.034, only 1.5 percent capped. Celestial issuers carry minimal related-party transaction volume relative to revenue. The Hidden Gem cohort sits at an intermediate position — mean ratio 0.471, 45.5 percent capped — substantially below the universe baseline but above Celestial. The Chameleon cohort tracks the universe baseline at 67.7 percent capped. The Poison Apple cohort confirms Note 3's reading at 85.4 percent capped (the previously cited figure of 90 percent reflected the indicator's tighter cap-only criterion). The Kill Switch cohort sits at 69.7 percent capped, statistically indistinguishable from Chameleon.

The Kill Switch finding is informative. KS firms — the override classification for issuers triggering audit-opinion failure, prolonged suspension, or related-party loan exposure thresholds — are not particularly distinct on connected-transaction scale. The KS classification does not run through the R-02 indicator. It runs through the substantive triggers that define the Kill Switch override taxonomy, which sit elsewhere in the framework's architecture.

---

The R-weak Mechanical Link

Note 4 examined the relationship between the framework's [B-weak] Chameleon sub-tag and the INED tenure indicator^[5]. The [B-weak] Chameleon cohort carried a mean long-tenured INED rate of 39.1 percent, against the [R-weak] cohort's 23.9 percent — a 15.2-point gap that the framework's three-axis architecture identifies as the structural expression of B-axis weakness. This Note's connected-transaction data reproduces the pattern in mirror.

The [R-weak] Chameleon cohort numbers 697 firms in the universe by classifier output (Note 1); the 480-firm count in the table below is the subset for which the connected-transaction scale indicator is measurable to ratio resolution. Findings are bounded to that measurable cohort.

Chameleon sub-tag	n measured (R-02)	Mean ratio	Capped %
[R-weak]	480	0.837	78.3%
[B-weak]	166	0.408	36.1%
[T-weak]	15	0.811	80.0% (small-n)
[balanced]	1	0.127	0.0% (n=1)

The [R-weak] Chameleon cohort sits at 78.3 percent capped — 4.5 points above the universe baseline and 42 points above the [B-weak] cohort. The mean ratio 0.837 places R-weak firms substantively closer to the upper bin than the universe mean of 0.750. The [B-weak] cohort, by contrast, sits at 36.1 percent capped with mean ratio 0.408 — substantially below the universe baseline on connected-transaction scale.

The pattern is structurally mirror-symmetric. Note 4 showed [B-weak] firms carrying long-tenured INED retention as the substantive expression of B-axis weakness. This Note shows [R-weak] firms carrying connected-transaction scale ceiling-pressure as the substantive expression of R-axis weakness. The sub-tag taxonomy identified in Note 1 is not a free-floating classification label. It is a mechanical signal-locator: each sub-tag points to the specific substantive governance dimension where the framework reads the cohort's primary stress.

[R-weak] Chameleons carry their structural weakness through Chapter 14A connected-transaction exposure. The 78.3 percent capped share is the cohort's mechanical signature on the R-axis. [B-weak] Chameleons carry their weakness through INED tenure retention. The two sub-tags identify two distinct substantive pathways into the same archetype's middle composite band, and the framework's three-axis architecture surfaces the distinction.

---

HKL versus RCH Poison Apple: Two Sub-Mechanisms

The Poison Apple cohort's R-axis weakness — examined at the universe level in Note 3 — does not run through a single mechanism across origins. The within-origin breakdown of connected-transaction scale and CCT count, restricted to the Poison Apple subset, separates two distinct patterns.

PA × Origin	Measured n	Mean ratio	Capped %	Mean CCT count
HKL PA	420	0.924	87.6%	3.8
PCH PA	77	0.874	81.8%	3.6
HSH PA	14	0.728	64.3%	3.7
RCH PA	8	0.520	37.5%	8.5

HKL Poison Apple — the largest subset at 420 measurable firms — carries the strongest scale-capped pattern at 87.6 percent capped, with the lowest mean CCT count among the four cohorts at 3.8 enumerated transactions per firm. The mechanical reading is a few-large-transactions pattern. Each firm in this cohort carries a small number of related-party transactions, but the aggregate volume of those transactions is substantial relative to revenue. The pattern is consistent with HK family-conglomerate intra-group transactions in which a single parent or affiliated entity dominates the listed vehicle's connected-transaction exposure.

RCH Poison Apple — on a small subset of eight measurable firms within the 20-firm RCH origin — carries the opposite profile. Capped share 37.5 percent, mean ratio 0.520, mean CCT count 8.5. Many enumerated connected transactions per firm, none individually dominating the revenue ratio. The pattern is consistent with a state-shareholder cohort where the listed Red Chip vehicle conducts numerous discrete CCT items across the parent group's affiliated SOEs and joint ventures, with each item moderate relative to the listed vehicle's revenue but the cohort's CCT enumeration density running materially above the universe baseline.

The PCH and HSH Poison Apple subsets sit between these two patterns. PCH PA carries the HKL-adjacent few-large pattern (mean CCT count 3.6, capped share 81.8 percent). HSH PA, on a small base of fourteen measurable firms, carries an intermediate profile: capped share 64.3 percent (well below HKL PA) with CCT count similar to HKL PA. The H-share Poison Apple subset — which Note 3 identified as carrying the panel's lowest within-cohort B-axis mean (41.1) — does not, on this evidence, also carry the universe's most extreme R-axis pattern. Its R-axis weakness is the cohort's secondary signal; the primary distortion is on the board axis.

The framework's archetype label combines both HKL PA and RCH PA into a single Poison Apple classification. The connected-transaction architecture below the label separates them. The HKL pattern — few-large CCT items in family-conglomerate intra-group structures — and the RCH pattern — many-small CCT items in state-affiliated networks — are mechanically distinct but produce the same archetype assignment, the same R-axis weakness signature, and the same eventual position in the framework's composite distribution. The distinction matters for any reform leverage analysis: the regulatory mechanism that constrains few-large transactions is different from the mechanism that constrains many-small ones, and Chapter 14A's percentage-ratio architecture engages each pattern asymmetrically.

The RCH cohort caveat applies. The eight-firm measurable base limits inferential weight; the directional pattern is consistent with the broader RCH cohort's elevated CCT count finding (mean 6.71 across all measured RCH firms), but specific magnitudes on the eight-firm subset are subject to small-sample variability.

---

Year-on-Year Variation and What Comes Next

The findings in this Note are observed in the framework's current measurement window. The framework's Phase 11 timeseries panel allows a direct read on how stable the R-axis signal is across consecutive fiscal years^[6]. The panel's 582-firm cohort produces a Pearson correlation of 0.655 between FY-prior and FY-current R-axis scores, against a composite-score correlation of 0.923. The R-axis is meaningfully more variable year-on-year than the composite or the other axes that the composite aggregates.

Two factors contribute. First, the connected-transaction scale indicator's measurement coverage has improved across recent parser releases — the v1.5 parser raised the indicator's universe coverage from 16 percent to 60.8 percent in a single release, with the measurable-with-ratio subset reaching 50.7 percent. Coverage gains shift firms from un-measured to measured, producing year-on-year R-axis movements that reflect parser reach rather than underlying disclosure change. Second, connected-transaction scale itself moves with each fiscal cycle: a single large transaction in one year can shift a firm's RPT-to-revenue ratio sharply, particularly for issuers near the boundary between mid-bin and upper-bin classification.

The implication for this Note's findings is bounded. The universe-wide patterns — 74 percent at the upper bin, R-axis sub-component asymmetry, [R-weak] mechanical link, HKL versus RCH Poison Apple sub-mechanism — are robust as cross-sectional readings of the framework's current cohort. They are not asserted as stable firm-level classifications across multiple fiscal cycles; the Phase 11 correlation indicates that meaningful firm-level R-axis movement is part of the data's structure. Subsequent panel refreshes will reveal which firms move and which do not.

This Note completes the framework's direct examination of the substantive Balance-of-Power and Conflict-of-Interest axes. Note 1 identified the universe-level R-weak Chameleon signature; Notes 2 and 3 mapped its origin-level and archetype-level decompositions; Note 4 examined the B-axis through the INED tenure cap; this Note has examined the R-axis through Chapter 14A connected-transaction architecture. The form-versus-substance dynamic that Note 4 traced for the board axis reproduces here for the conflict-of-interest axis. Chapter 14A surfaces connected transactions through one of the most detailed disclosure architectures in Asian listing rules. It does not constrain their absolute scale. The 74 percent capped distribution is the framework's reading of the gap.

Note 6 closes the series by establishing a baseline for the climate disclosure mandate that took effect within the panel's coverage period — the most recent regulatory architecture in the universe Note 1 introduced, and the architecture whose form-versus-substance dynamics will become observable as subsequent panel refreshes accumulate. The disclosure side of the Hong Kong governance architecture has been thoroughly developed across decades of HKEX listing-rule iteration. Its substantive reach, on the evidence the framework reads, has not kept pace.

---

The Apex G-Score framework currently covers 2,768 Hong Kong listed companies under v2.0 calibration. The connected-transaction scale indicator is parsed from issuer continuing-connected-transaction disclosures and related-party transaction notes under HKEX Listing Rules Chapter 14A; universe coverage with computable RPT-to-revenue ratio is approximately 50.7 percent (n=1,403). Granular threshold-trigger classification (de minimis, disclosure, Major transaction bands under Rule 14.07) is not currently in production; the analysis above relies on aggregate scale and CCT enumeration count as proxies. The Phase 11 timeseries panel covers 582 firms; R-axis findings are reported as cross-sectional readings of the FY2024-Dec / FY2025-Dec straddle cohort with year-on-year variability noted.