// analysis
Cum‑cum: measuring reform after the refunds
Cum-cum reform brings upfront withholding and a future EU fast track. Net receipts, treaty changes and refund timing shape any credible assessment.
The cum-cum investigation, part 6
For some overseas investors, France’s new regime first appears as a shortfall in a dividend payment. Since 1 January 2026, certain distributions that may qualify for a treaty exemption have been subject to withholding tax upfront. The recipient or paying institution can support a refund claim by establishing that all treaty conditions are met. Guidance published on 16 March explains the supporting evidence and the route through France’s tax directorate for non-residents. S01 S11
This reversal in the sequence gives the fight against cum-cum a tangible form. The state holds the money while it examines the evidence. That creates an opportunity to check a claim before releasing funds, while leaving a legitimate investor temporarily out of pocket. Evaluating the policy therefore requires two outcomes to be assessed together: preventing unwarranted tax benefits and returning tax that was never ultimately due. The initial cash receipt is only the first entry in a longer story.
At the end of this six-part investigation, the difficult question is whether the new rules actually reduce tax losses. French documents establish a specific change in the payment and refund process. The EU’s FASTER framework for faster and safer relief of excess withholding tax is due to apply from 2030. The public evidence assembled here does not isolate the recent reforms’ effect on fraud. It does, however, show what a credible assessment would need to measure. S01 S04
France moves the check ahead of the refund
The French provision that began applying in January 2026 addresses certain external cum-cum arrangements. An overseas intermediary may qualify for a treaty exemption unavailable to another holder. Earlier installments examined how a temporary transfer of shares, or associated contracts, can preserve the original investor’s economic exposure. The Conseil d’État’s January 2025 advisory opinion helps distinguish internal and external arrangements. The refund conditions discussed here are those in the enacted statute and the published guidance. S10 S02 S01
The refund procedure has a defined scope. The guidance covers treaties providing an exemption without a minimum shareholding requirement, while preserving exclusions for exemptions tied to the recipient’s status or level of ownership. At publication, it identified nine countries: Saudi Arabia, Bahrain, Egypt, the United Arab Emirates, Finland, Kuwait, Lebanon, Oman and Qatar. This is a description of treaty provisions, not a ranking of investors or countries by suspected misconduct. S01 (paragraph 10)
Treaty changes also alter this perimeter. The new France–Finland convention entered into force on 28 August 2026 and affects French withholding from 2027. For ordinary dividends, it replaces the general exemption with source tax generally capped at 15%, while retaining a zero rate subject to ownership and holding-period conditions. Future comparisons will need to account for that legal change alongside enforcement. S13
For a covered payment, the paying institution deducts withholding tax. Exemption is granted later, after examination. The claimant can also provide evidence equivalent to the documents described in the guidance, provided it demonstrates compliance with the treaty conditions. That flexibility matters. A procedure intended to establish the facts must be able to recognise a legitimate arrangement even when its documentation does not fit a standard presentation. S01 (paragraphs 20–50)
A deliberately fictional example makes the accounting clear. Take a gross dividend of 100 and an illustrative withholding of 25. The investor initially receives 75. If full exemption is established, the other 25 is returned. The state’s initial receipt exceeds the tax ultimately retained by 25. The example’s rate is an assumption, not a statement about any investor’s actual tax liability.
Method and scope
Withholding of 25 may be fully refunded on a dividend of 100. Fictional scenario, currency-neutral amounts and an assumed rate. Widths are proportional. A refund requires entitlement to full exemption to be established. Mechanism: S01, paragraphs 1 and 20–50.
Every document has to fit the same transaction
The French application brings together several kinds of evidence. Residence forms establish the treaty connection. Dividend and withholding records trace the payment through the full intermediary chain. A declaration addresses whether the recipient is legally or contractually required to dispose of the shares and return the associated income. A corporate recipient must also provide information about its business, resources, governance and reasons for acquiring the securities. The administration may ask for lending, repo or derivative contracts. S01 (paragraphs 40–50)
These documents answer different questions. Residence establishes a jurisdictional connection. Contracts and surrounding facts help establish whether the recipient can enjoy the income. Payment records help verify that the tax claimed back was actually borne. The useful test is whether all these pieces refer to the same shares, distribution, amounts and route to the beneficiary. Reconciliation can reveal an intermediary’s role that a single certificate would leave unexplained.
The guidance also addresses incomplete information. If the paying institution does not know the beneficial owner (the person entitled to use and enjoy the income under the applicable treaty rules), it must, on request, supply the information it holds about the intermediary or final investor, residence and investment account. An investigation can therefore begin with missing links. A useful performance measure would be the proportion of such chains eventually reconstructed, and the time required to do so. This is a proposed measure of investigative capacity, not a presumption about the correct tax outcome. S01 (paragraph 20)
Europe builds a conditional fast track
FASTER is designed to make legitimate relief easier while making payment chains easier to examine. The directive was adopted on 10 December 2024 and published in the Official Journal on 10 January 2025. Member states must transpose it by 31 December 2028, with application from 1 January 2030, subject to particular provisions for certain countries. Its legal adoption and its operational application belong to different stages of the timetable. S04 (Articles 2 and 25) S05
The framework includes certified financial intermediaries, standardised reporting and a digital tax residence certificate, or eTRC. The certificate is generally to be issued within fourteen calendar days. Where further checks are needed, the authority must explain the additional time required. The certificate establishes residence. Eligibility for relief, including beneficial ownership where the applicable rules require it, still has to be examined separately. S04 (Articles 4 and 12)
Reporting is intended to connect the investor, intermediaries, distribution and withholding. Annex II includes securities and event identifiers, dates and amounts. A country can require intermediaries to report directly or arrange sequential transmission along the payment chain to a designated reporting party. Certification does not approve every transaction: the intermediary remains responsible for the checks required of it. S04 (Articles 10–12 and Annex II)
Method and scope
Residence, contractual and payment evidence address complementary questions. Simplified reconciliation schematic under FASTER, not an actual transaction. Reporting may be direct or sequential. S04, Articles 4, 10–12 and Annex II.
The fast track has filters. Member states may exclude requests, wholly or partly, in specified circumstances, including where a share was acquired within five days before the ex-dividend date or a related financial arrangement has not been settled, expired or otherwise terminated before the ex-dividend date. These are discretionary procedural exclusions under the directive, not automatic findings of fraud. Ordinary refund channels retain a role for claims that cannot use accelerated procedures. S04 (Articles 11, 14 and 17)
The Commission also proposed targeted changes on 24 June 2026. Article 6 addresses FASTER access for certain payments exempt under the Parent-Subsidiary or Interest and Royalties Directives, and its interaction with national relief systems. The European Parliament’s file, checked on 28 October, still lists a proposal awaiting a committee decision. The deadlines examined here belong to the adopted directive. S14
National differences remain important. Countries with a comprehensive relief-at-source system may, under Article 2’s conditions concerning market capitalisation and other requirements, choose not to apply Chapter III. Extending the framework to bond interest is optional. Any European comparison must identify the countries, payments and procedures actually covered, rather than assume that every cross-border investment is treated identically. S04 (Article 2)
The quick-refund timetable
The documentation contains a useful warning against relying on summaries. In the section describing the original proposal, the Commission’s overview page still refers to a certificate within one working day and a refund within fifty days of payment. The final directive sets different terms. For an investor funding a temporary tax deduction, the difference between proposal and adopted law has a direct cash-flow consequence. S06 S04 (Articles 4 and 14)
Under Article 14’s quick-refund procedure, the necessary information must be supplied within the second month following the month of payment. The sixty-calendar-day processing period starts after that application period ends, rather than when the dividend is paid. Missing information or a risk-based tax examination may lead to rejection under the fast-track procedure. The directive also addresses late-payment interest and access to the ordinary refund system. S04 (Articles 14–17)
Consider a hypothetical dividend paid on 15 March 2030. The application period ends on 31 May. Adding sixty calendar days produces 30 July, which is 137 days after the original payment. This is an arithmetic illustration of the processing deadline for a complete, eligible request, excluding audit complications and particular national rules for computing time limits. An authority may act earlier. The example is neither an expected average waiting time nor a prediction of the date an investor’s account will be credited.
Method and scope
From 15 March to 31 May 2030: 77 days. Another 60 calendar days reaches 30 July, or 137 in total. Segment lengths are proportional to elapsed time. Article 14 processing-deadline illustration for an eligible, complete request, excluding audit and particular national time-computation rules. No average wait or actual credit date is predicted. The separate 14-day eTRC deadline allows additional verification. S04, Articles 4 and 14.
The amount of time cash remains tied up should form part of the evaluation. Even a fully valid refund can be costly if it arrives late. An average calculated only from completed cases would overlook requests still pending. A better account would show the median, the time within which 90% of closed cases were resolved, and the age of outstanding requests. These are suggested evaluation measures, not reported results.
More receipts can conceal a weaker yield
On 15 October 2026, the French government told the Senate that earlier cum-cum adjustments under the abuse-of-law procedure had resulted in €4.5 billion of additional tax assessments being notified. The figure describes tax assessed; final outcomes and collection are separate matters. The reply does not provide a series of net receipts attributable to the new provisions. Using it as a causal estimate of the recent reform would combine the wrong period, the wrong accounting category and the wrong mechanism. S07
The same problem affects a seemingly simpler measure: withholding collected. Larger dividend distributions can generate more revenue without any improvement in enforcement. Collecting tax first and returning it later can temporarily inflate cash receipts. A change in the investor population can also alter the average tax rate lawfully due. These explanations have to be separated before a revenue increase can be attributed to reduced fraud.
Take two fictional dividend cohorts, measured after their refunds. In the first, companies distribute €1,000 million and the state retains €120 million. In the second, distributions rise to €1,500 million and retained tax to €150 million. Revenue increases by 25%. Yet retained tax as a share of dividends falls from 12% to 10%. Growth in the underlying payments masks a two-percentage-point reduction in the observed yield.
The bridge is straightforward. At the initial 12% yield, the additional €500 million of dividends would generate €60 million of receipts. A two-point yield reduction applied to the final €1,500 million base subtracts €30 million. Together they produce the observed €30 million increase. Treaty entitlements, investor composition or other changes could explain the lower yield. This calculation assigns no amount to fraud.
Method and scope
Two fictional cohorts, after refunds, in € million. Observed yield is N/D, net receipts divided by gross dividends. The chosen bridge applies the initial yield to the base change, then the yield change to the final base. Effects reconcile exactly; this is an arithmetic decomposition, not causal attribution or a fraud estimate. l0g calculation.
l0g / FICTIONAL SCENARIOS
Separate the dividend base from the collection yield
Two fictional periods show how larger tax receipts can coexist with a lower collection yield.
Change the assumptions
Initial collection yield
Final collection yield
Change in receipts
Change in yield
Dividend-base effect
Collection-yield effect
Sum of both effects
Receipt growth
Calculation and interpretation
Collection yield = net receipts / dividends. Base effect = (final dividends − initial dividends) × initial yield. Yield effect = final dividends × (final yield − initial yield). Their sum equals the change in receipts. These are accounting scenarios, not measurements of fraud or compliance.
Results have two decimal places. Any presentation rounding remainder is assigned to the yield effect so that the displayed effects add up exactly to the change in receipts.
Applying this approach to real data would require refunds to be linked to the distributions that generated them. Recoveries relating to earlier years would have to be kept separate, and disputed amounts followed through to resolution. Here, a cohort means payments from one period whose subsequent refunds are tracked together. Comparing cohorts after the same amount of elapsed time would reduce the bias that otherwise favours the newest one, where refunds are still to come.
The receipts-to-dividends ratio would remain an observed yield, rather than a compliance rate. Its denominator includes income subject to different lawful treatments. A stronger assessment would reconstruct the mix of treaties, investor statuses and transaction types, then compare similar groups before and after the change. Other reforms or changes in trading activity could still complicate attribution. A credible method would make those rival explanations visible.
Supervision has to follow the contracts
The policy problem extends beyond the tax authority’s cash account. In 2020, the European Banking Authority (EBA) called for stronger governance, risk supervision and cooperation between prudential, anti-money-laundering and tax authorities. Its action plan also envisaged follow-up. S08
In February 2025, the EBA published a review carried out in 2024 across authorities in six countries. It found that most supervisors reviewed largely or fully applied the assessed benchmarks, while identifying improvements to make. The assessment concerns the integration of tax integrity into prudential and anti-money-laundering supervision. Its scope extends beyond cum-cum arrangements; the EBA explicitly leaves the overall effectiveness of national systems for identifying and investigating tax crimes outside the assessment. S12 (pages 6–10 and 61–63)
Useful evidence of progress would include the decisions taken after anomalies were detected: cases reopened, contracts reconciled, inconsistencies resolved, relief denied and then upheld or overturned. A count of forms submitted measures administrative output. Their ability to uncover an obligation to pass income on, or an economic exposure retained elsewhere, is closer to a measure of usefulness.
French guidance published in July 2025 covers linear instruments, often called delta one, and combinations producing sensitivity identical or very close to that of the underlying securities. It also provides safeguards for transactions whose main purpose and effect are other than avoiding withholding. Examination therefore has to consider the contracts together and allow for legitimate economic uses. S03 (questions 3, 5 and 6)
Monitoring share loans alone could overlook a change in the form of a transaction. Treating every hedge as evidence of misconduct would generate false positives. Movement into other instruments is a hypothesis to investigate with transaction data and contracts. This dossier establishes no post-reform migration of that kind.
ESMA had already highlighted the limitations of market volumes viewed in isolation. Its 2020 report found varied patterns around dividend dates and called for deeper evaluation of legislative effects. Those historical observations, largely covering years before 2020, provide a methodological warning rather than evidence of market behaviour in 2026. S09 (paragraphs 167, 173–174)
Deterrence and the quality of enforcement
The strongest argument for the new rules is that they may deter arrangements before execution. If that happens, fewer enforcement cases could coexist with better revenue protection. Conversely, more tax adjustments could reflect improved detection of an older problem. The number of penalties alone leaves both explanations open.
A credible assessment would bring together tax ultimately retained, the characteristics of the transactions and the treatment of eligible claimants. A high proportion of refusals later overturned, prolonged delays for legitimate refunds, or unusable information despite formal reporting would indicate weaknesses. More complete payment chains, accurate refunds and a documented reduction in unwarranted tax benefits would support a more favourable judgment. The criteria should be capable of producing a conclusion that conflicts with the policy’s stated intentions.
FASTER requires rules on intermediary liability for relevant failures and on penalties, as well as a European evaluation. By the end of 2032, the Commission must review reporting mechanisms and certain national exemptions. A broader review of the directive’s operation is due by the end of 2034. The annual statistics provided to the Commission are subject to confidentiality provisions. Their collection does not itself promise a public performance dashboard. S04 (Articles 18–19 and 22)
Method and scope
France’s reform and FASTER have different timetables. Future points are legal deadlines, not assumed outcomes. Wide version: calendar-year axis, with end-2034 at the right edge. Mobile version: chronological milestones with non-proportional spacing. National exceptions and particular procedures are explained in the text. S02–S04.
Aggregated publication could nevertheless inform the public without identifying taxpayers. Sufficiently broad cohorts could show processing times, amounts claimed and refunded, and how decisions change on review. The individual records used for enforcement require restricted access and justified retention periods. The directive expressly addresses supporting documentation and personal-data protection. Improving tax traceability also creates a responsibility to protect the resulting information chain. S04 (Article 10(7) and Article 23)
As of 28 October 2026, the conclusions can be stated precisely. France has changed collection and refund arrangements within a defined scope. The EU has adopted an architecture whose main application date is still ahead. The material reviewed does not separately measure how much fraud these reforms have prevented. The decisive step will be to follow each generation of dividend payments through to its final tax outcome, allowing for the rights actually applicable and the cost of administering them. That is the level at which the promises can be tested.
Documentary investigation with a research cut-off of 28 October 2026. No new interviews. All numerical scenarios are illustrative; individual tax records and proprietary market data are outside the evidence obtained. The source notes below set out the scope of the legal texts, institutional positions and measurement limits.
Earlier instalments
- The dividend mechanism.
- Warnings before the scandal.
- Estimated tax losses and recoveries.
- Accountability in court.
- The tax-guidance battle.
Sources
The notices identify adopted rules, a pending proposal and supervisory evaluations with their respective scopes.
- BOI-INT-DG-20-20-20-30 : modalités pratiques des restitutions conventionnelles
Passages and scope
Paragraphs 1, 10, 20, 30, 40 and 50; the two cumulative treaty-scope conditions and evidence requirements.
Payments from 1 January 2026; nine treaties within the stated scope; initial withholding, evidence and refunds. Dated administrative guidance, not observed refund delays or an evaluation of enforcement outcomes. The nine-country list belongs to this version; the Finnish treaty changes from 2027 are addressed in S13.
- Loi n° 2025-127 du 14 février 2025, article 96
Passages and scope
Article 96, I-B item 7 (new article 119 bis A-II) and II (application from 1 January 2026); safeguards under I-B.
Amendments to Articles 119 bis and 119 bis A; deferred application of the external-scheme provision from 1 January 2026. Original statute read with the dated guidance S01 and S03. The final external-limb refund condition is compliance with treaty requirements; the initial draft described in S10 is distinct.
- BOI-RES-RPPM-000203, version du 24 juillet 2025
Passages and scope
Questions 3, 5, 6, 7, 8, 9 and 10; linear instruments or combinations, safeguards and reservation of abuse-of-law procedures.
Linear instruments and combinations; safeguards; scope and commencement dates. The scope includes combinations. Being outside this specific provision does not determine treatment under every other legal basis; question 10 reserves the abuse-of-law procedure.
- Directive (UE) 2025/50 du 10 décembre 2024, FASTER
Passages and scope
Articles 2, 4, 10–14, 17–19, 22, 23 and 25; Annex II. Adopted act of 10 December 2024, published 10 January 2025.
Final legal text: a 14-day certificate deadline subject to checks; reporting and refund timing; exceptions; application in 2030; evaluations in 2032 and 2034. Adopted law with future operational application. Deadlines are conditional; the 137-day example is an arithmetic scenario, not an observed delay. The separate 2026 proposal is identified in S14.
- Taxation: Council adopts new rules for withholding tax procedures (FASTER)
Passages and scope
Adoption, relief at source, quick refund and transposition timetable.
Objectives, the two relief procedures and the transposition timetable. Adoption announcement; S04 controls the exact legal conditions and deadlines.
- FASTER Directive : page de présentation
Passages and scope
The section describing the original proposal, distinguished from the adopted act and its next steps.
The section on the proposal retains the one-working-day and 50-day figures; the final act S04 sets different timing. The overview still describes one-working-day and 50-day figures under the proposal heading. Those figures are not the deadlines in the adopted act.
- Réponse à la question écrite n° 07856 : contrôle des arbitrages de dividendes
Passages and scope
Government answer published 15 October 2026, distinguished from the question filed on 26 February.
€4.5 billion in additional tax assessments notified; no series here attributing net receipts to the reform. Government statement about historical notified tax assessments. No matching total cash collection or causal estimate of the recent reforms is supplied.
- EBA publishes its inquiry into dividend arbitrage trading schemes (Cum-Ex/Cum-Cum), and announces a 10-point action plan
Passages and scope
12 May 2020 release; supervisory findings and ten-point action plan
Original EBA source replacing the MNB relay. Governance, risk, cooperation with tax authorities and announced follow-up. Historical announcement, supplemented by the 2025 review S12; no measured causal effect of the French 2026 reform.
- Final report on Cum/Ex, Cum/Cum and withholding tax reclaim schemes
Passages and scope
Paragraphs 167, 173–174, printed pages 41 and 43.
Market volumes alone cannot support a definitive conclusion; legal changes require deeper evaluation. Historical study, chiefly covering 2014–2019. These observations are not measurements of market conduct in 2026.
- Avis relatif au renforcement de la retenue à la source sur les arbitrages de dividendes
Passages and scope
Introduction and paragraphs 3–6 and 11–15; advisory analysis of proposed legislation.
Distinguishes internal and external schemes and the economic structure targeted by the proposal. Consultative advice on a draft, not a judicial finding or an evaluation of the enacted reform. The final law and guidance determine the applied refund criteria.
- Actualité BOFiP du 16 mars 2026 : publication urgente
Passages and scope
Notice published on 16 March 2026 and the linked guidance for payments from 1 January 2026.
Confirms the March publication of practical guidance for payments covered since January. Publication is dated 16 March 2026 despite the 2025 identifier. Same institution as S01; it independently establishes no policy result.
- Peer review: Tax integrity and dividend arbitrage schemes, EBA/REP/2025/05
Passages and scope
Pages 6–7 (summary and limits), 8–10 (scope and methodology), 61–63 (conclusions and follow-up measures)
Review carried out in 2024 across authorities in six countries: Denmark, Germany, France, Ireland, the Netherlands and Spain. Most reviewed supervisors largely or fully applied the assessed benchmarks; improvements were prescribed. Institutional assessment of prudential and AML supervision, covering tax integrity beyond cum-cum arrangements. It explicitly excludes the overall effectiveness of national systems for identifying and investigating tax crimes. No measured causal effect of the 2026 reforms or FASTER 2030.
- Decree no. 2026-758 publishing the France–Finland tax convention signed on 4 April 2023
Passages and scope
Annexed convention: article 10(1)–(2), (4)–(6); article 29(1)–(3); decree footnote (1), entry into force on 28 August 2026.
The convention entered into force on 28 August 2026. Article 29 makes its French withholding provisions applicable to taxable amounts after that calendar year, hence from 2027. Ordinary dividends may generally bear source tax capped at 15%; the zero rate requires a resident beneficial-owner company to hold directly at least 5% of the payer’s capital for a 365-day period including payment. Permanent establishments and specified real-estate investment vehicles have separate rules. The nine-treaty list in the guidance dated 16 March 2026 should therefore not be projected unchanged into later years. Review limited to relevant dividend, commencement and transition provisions, rather than every pension or income rule. The official Légifrance indexed text supplied those provisions and the commencement footnote; later direct requests returned 403. Cross-checked against Finlex 33–34/2026 and the Finnish Tax Administration’s notice. No enforcement result or tax yield is inferred from the treaty change.
- COM(2026) 560 final, proposition du 24 juin 2026 ; procédure 2026/0163(CNS)
Passages and scope
Article 6, proposed amendments to Article 11(2)(d) and (7) of Directive 2025/50; OEIL procedure file
Proposal affecting, in particular, full-withholding-exemption filters and national systems. The OEIL file consulted on 28 October 2026 reports “Awaiting committee decision”. Legislative proposal and procedural status, not an adopted change to the law. Article 6 does not change the timing under Articles 4 and 14.
This analysis is not investment advice.
// cite this analysis
l0g, “Cum‑cum: measuring reform after the refunds”, l0g.fr, published October 28, 2026, updated October 28, 2026, https://l0g.fr/en/analysis/cumcum-06-measuring-reform-results/
$ cd ../analysis