Brazil's betting ban is the wrong move: tax and regulate the market

On 25 September 2026 Brazil issued Medida Provisória (MP) 1.394, which prohibits fixed-odds betting lotteries across the national territory. In practice that means sports betting and online casino-style games, and I use "the ban" for it. Banning is the wrong move. The government has a serious case that the regime Brazil built in 2018 and 2023 didn't protect people well enough, and that harm is real. It is also what regulation is for. A moderately taxed, tightly regulated market can target it, and the examples I lean on are alcohol and cannabis in the United States. The way the ban was done makes it worse: it took effect on publication, switched sites off after ten days and ended licences after thirty, before the joint committee that must examine it had even been installed. Translations from Portuguese are unofficial.

What the Brazilian betting ban does

When the ban applies: the timeline

The text of the MP appeared in an extra edition of the Official Gazette on 25 September, according to the Senate's record, and article 30 brings it into force on publication. Article 1 prohibits, throughout the national territory, the "exploração, oferta, intermediação e publicidade" (operation, offer, intermediation and advertising) of fixed-odds betting lotteries, on real sports events and on virtual events of online games, including when they are offered from abroad to people in Brazil. Other lottery modalities are excluded.

From publication, no new money may enter bettors' accounts. Article 7 gave operators ten days, and the Ministry of Finance's timeline puts the end of that period at 23:59 on 5 October, with sites and apps unavailable from 6 October. Refunds follow in the days after, and on 25 October, thirty days after publication, existing authorisations end under article 4. On 6 October the Ministry of Justice reported that of the 188 sites that had operated with authorisation until the day before, only one was still running.

Associations have challenged the ban in the Supreme Court.

What happens to bettors' balances and refunds

Article 7, paragraph 1, shuts the door on deposits: "fica vedado o ingresso de novos recursos nas contas transacionais" (the entry of new funds into transactional accounts is forbidden), except proceeds of assets in which bettors' funds were invested. Bets still unresolved at the deadline are void and the stake is refunded in full. Prizes on bets already resolved are honoured.

Refunds run through the institutions that hold bettors' accounts. Within two days after unavailability, operators must secure liquidity and send each bettor's details to those institutions, on pain of a fine of R$200,000 a day (article 8). The institutions refund within seven days to an account held by the bettor, identified by CPF. Money that cannot be refunded goes to a specific account at Caixa Econômica Federal (article 9 and paragraph 3). Agência Brasil, in a report carried by Seu Dinheiro on 12 October, put the balances already returned at more than R$1 billion, with R$159 million still to be refunded. Eight companies hadn't sent the banks the data needed, and the AGU said it would go to court to force them to provide it.

Wind-down duties of licensed operators

Ending the licences doesn't wipe out what was owed while they ran. The sole paragraph of article 29 says that the provisions of Lei 13.756 and Lei 14.790 "permanecerão aplicáveis aos fatos ocorridos durante a vigência das autorizações" (remain applicable to facts that occurred while the authorisations were in force). The Finance Ministry's note of 5 October says the companies "continuarão responsáveis pelas obrigações constituídas durante o período em que estavam em funcionamento" (will remain responsible for the obligations incurred while they were operating), with a deadline of 15 December 2026.

State and Federal District licences

Article 2, paragraph 1, ends state and Federal District concessions, permissions and authorisations on the same thirty-day deadline ("serão extintas no prazo previsto no art. 4º"). Article 4 gives the operator no right to a refund of the licence fee ("contraprestação de outorga") and no compensation from the State.

Advertising and sponsorship

Articles 14 to 23 ban communication, advertising, marketing and sponsorship. Article 16 requires advertising material and sponsorship signs to be removed within ten days of publication. It follows Supreme Court injunctions of 2024 on advertising to children, which the Attorney-General's office cites in defence of the MP.

How enforcement works

The MP uses several tools at once. It bans payment processing for betting, except what is needed to wind down and refund bettors (article 14). It imposes duties on digital platforms and app stores (articles 19 to 22) and provides for blocking of sites (article 23). The Ministry of Justice reported on 9 October that 20,318 sites and digital channels had been referred for blocking, 16,627 of them betting sites; how many are actually down is another matter.

Blocking has a poor record. An audit by the Federal Audit Court (TCU) counted 21,842 domain blocks processed by Anatel between October 2024 and September 2025 and 483 bank-account closures, and still found about 40 percent of the market with unauthorised operators. It wrote that "o bloqueio massivo de domínios mostrou-se ineficaz, pois novos portais ilegais surgem" (mass blocking of domains proved ineffective, because new illegal portals appear).

Article 50, paragraph 2, of the Law of Penal Contraventions provides a fine for a person found taking part in a game of chance as a bettor, "ainda que pela internet" (even over the internet). The offences in the companion bill described by the Health Ministry, by contrast, are all on the supply side.

The provision on enforcement that troubles me most is article 28. It lets funds blocked from unauthorised operators be forfeited to the Union through an administrative process, with full defence guaranteed, but it "não dependerá de processo judicial prévio" (will not depend on a prior court process). Decreto 13.033 of 2026 had required a judicial decision. Judicial review remains available afterwards, which is why it isn't my strongest objection.

The government's case

The harm it cites

The explanatory memorandum says existing instruments "não têm sido capazes de assegurar proteção adequada" (have not been able to ensure adequate protection) of health, consumers, children and adolescents, and the financial stability of families. It says the net transfer from households to betting platforms in 2025 may have been tens of billions of reais, links gambling disorder to anxiety, depression, indebtedness and suicide risk, and concludes that better regulation alone is not enough. It never names its studies.

The fuller argument is in the Attorney-General's (AGU) filing in ADI 7721, signed on 5 October. It cites an estimate by Comsefaz, the committee of state finance secretaries, that betting drew about R$62.5 billion net out of family budgets in 2025. It cites a register showing 1,294,328 people self-excluded from licensed platforms, 36.41 percent of whom gave loss of control as the reason. It reports a Datafolha poll in which 78 percent of 2,506 respondents favoured banning online betting and casinos.

The filing is candid about its limits. It concedes that the health-service data do not isolate fixed-odds betting, that a R$30.6 billion estimate of social costs "não corresponde a despesas efetivamente realizadas pelo SUS" (does not correspond to expenses actually incurred by the public health system), and that household deterioration has several causes, so betting should not be given "causalidade exclusiva" (exclusive causality). The size of the effect is still open.

The tax the state gives up

The government is paying to act. The AGU filing projects lost tax revenue of R$1.541 billion in 2026, R$5.150 billion in 2027 and R$5.330 billion in 2028, about R$12 billion over the three years, plus about R$268.8 million of lost supervision fees. It adds that the Treasury's estimates leave out possible compensation claims, because those depend on a legal obligation being recognised. A ban that costs that much is hard to call a cynical grab.

Was the ban's timetable too fast?

A provisional measure is itself a constitutional route to legislation. Article 62 lets the President adopt one "em caso de relevância e urgência" (in a case of relevance and urgency), "com força de lei" (with force of law), and the Union has competence over "sistemas de consórcios e sorteios" (article 22, XX). I accept the AGU's argument that the power to create a lottery modality includes the power to end it, and that anyone who took a licence under a regime written by the State couldn't expect it to stay frozen for ever. But changing the rules for the future is different from extinguishing authorisations whose fees have been paid.

The fee was capped by the old statute at R$30 million per authorisation. The Health Ministry's news item of 25 September reports 85 licences granted, each at R$30 million, R$2.55 billion in all, and says extinction carries no right to compensation or to a refund of the fee. Whether operators get that money back is a matter for a court, and the AGU argues they shouldn't. My complaint is the ten-day shutdown: a decision with large costs, taken before the committee could shape it.

The Senate's record shows the joint committee designated on 30 September and 99 amendments presented, but the measure "aguardando instalação da comissão" (awaiting committee installation) since 21 October. So as of 25 October neither the committee nor Congress had spoken, although the shutdown date had passed and the licences were about to end. Congress still has time. The initial 60-day period runs to 23 November, under article 62 it is extended once by a further 60 days if the vote is unfinished, and it is suspended during recess.

Nor is the damage easy to undo. If Congress rejects the MP or lets it lapse, the earlier law regains effect: the Supreme Court held in a 2019 ruling (ADIs 5.709, 5.716, 5.717 and 5.727) that a provisional measure does not repeal an earlier law but suspends its effects. But paragraph 11 of article 62 leaves relationships formed while the MP was in force governed by it unless Congress issues a legislative decree within 60 days. And the practical damage is harder to repair than the legal one: operators wind down, jobs and contracts tied to them end, customer accounts close, and some of that would stay done even if the law later reverted.

The strongest case for the government's route

The government's best argument is urgency. A provisional measure is a lawful instrument, the memorandum sets out its reasons, and the AGU cites Supreme Court case law that review of relevance and urgency is exceptional, limited to cases where the requirements are evidently absent. Congressional delay alone doesn't show urgency is lacking, and an earlier, narrower bill, PL 2.258/2026, was still awaiting the Speaker's referral.

Its second argument concerns the product. The memorandum argues that the harm is tied to permanent availability, instant betting, immediate payment, targeted advertising and personalisation, which expose people to "estímulos contínuos" (continuous stimuli), and that incremental rules can't fully remove a structural exposure. The AGU adds that the earlier injunctions on advertising to children show incremental regulation had been tried.

On migration, the AGU accepts that moving to unlicensed operators is a risk but says it doesn't follow that all demand would move. The MP aims at the financial institutions, payment arrangements and digital providers that can cut off irregular offers, and the AGU cites police investigations linking unlicensed platforms to money laundering. Add the revenue the government is giving up and the 78 percent poll, and the case is a serious one.

The urgency, product and revenue points show the government means it, and I take the harm seriously. They don't show that the legal market had to close. Every problem the memorandum names is something regulation can target: advertising limits, deposit and loss limits and session alerts for instant play, a central self-exclusion register and identity checks for vulnerable people, and payment controls for the money. Brazil's licensed regime already required several of these, and where they fell short the fix is to tighten them. The points also leave unexplained why the transition couldn't have been longer, or compensation legislated at the outset, since the same MP could have carried either. Anseja's action (ADI 8024) asks in the alternative for a transition of at least 180 days, and the AGU says ANJL and IBJR asked for at least six months. A ten-day shutdown and a thirty-day licence end were too short for a decision this large.

Will bettors go back to unlicensed sites?

Many of those who bet through licensed operators before 6 October will, I think, go straight back to the offshore sites they used before the regulation.

What the early numbers show

The early official numbers measure something else. The national public security secretary, Chico Lucas, said on 29 September that Pix transactions had fallen 10 percent since the MP was published, "de acordo com números do Banco Central" (according to Central Bank figures). On 2 October the Ministry of Justice reported Finance Minister Dario Durigan saying that daily flows to betting houses, around R$600 million a day before the MP, had ceased. Flows, compliance and referrals all measure the supply side. No one has yet published where former bettors went.

What the record shows about the unlicensed market

An unlicensed market survived the regulated one. Before 2023 the regulation the 2018 law required was overdue, its deadline having expired in December 2022, according to the Executive's own memorandum for the earlier MP 1.182. During the regime, the Central Bank estimated in a study of September 2024 that 56 companies it identified took in R$20.8 billion in August 2024 and that about 24 million individuals made at least one Pix transfer to the betting companies it analysed.

After that came the TCU audit, with its 40 percent. That figure comes from an industry-commissioned study by LCA Consultoria for the responsible-gaming institute IBJR, "Fora do Radar" of June 2025, which put the illegal share of the total market at 41 to 51 percent. It sets R$26 to 40 billion a year of unlicensed operator revenue against about R$38 billion for the regulated segment. The TCU also reports irregular operators taking payments in untraceable cryptocurrency and by international transfers to accounts in tax havens.

People will always gamble: the jogo do bicho and Argentina

People will always gamble. Brazil's own example is the jogo do bicho. It has been a contravention since the Law of Penal Contraventions of 1941, in force from 1 January 1942, and a decree-law of 1944 (Decreto-Lei 6.259, article 58) sets heavier penalties for the banker than for the buyer. A sociology article republished in 2012 by the Federal University of Espírito Santo's journal Simbiótica calls it a "fato social brasileiro" (a Brazilian social fact), and in April 2024 the Senate's news service still described it as "uma prática que hoje constitui contravenção" (a practice that today is a contravention). A game that has been illegal for more than eighty years is still a live practice.

Argentina tells the same story in a country that does regulate. Gonzalo Atanasof, head of the Buenos Aires provincial lottery institute, told a legislative committee on 19 October that, according to estimates mentioned at the meeting, 80 percent of gambling would be illegal and that illegal gambling in Argentina moves US$3.6 billion. In 2024 Daniel Di Lena, head of the Santa Fe lottery, said that by the estimates of the casino and bingo association together with ALEA, legal play in Argentina is "entre un 18 y un 20 por ciento" of what is played. None of these estimates states its method, but they all point the same way.

What a Curaçao, Anjouan or Costa Rica licence gives a player

Sites that stay reachable from Brazil are licensed, if at all, elsewhere, so it's worth setting out what three regulators' own texts promise on paper.

Curaçao

The Curaçao Gaming Authority (CGA) says the new National Ordinance on Games of Chance came into effect on 24 December 2024. Under its licence conditions, the licence goes to a company established under Curaçao law with its registered office there, player funds must sit in a segregated account, and the licensee must have enough funds at all times to pay prizes. The CGA's responsible-gaming policy of 21 February 2025 makes age verification, cooling-off periods, self-exclusion and deposit limits mandatory.

Set against Brazil, the gaps show quickly. Self-exclusion is with a specific operator, with no national register. Deposit limits are set by the player, a request to raise one waits seven days, and loss and time limits "may be considered" in line with the operator's target market. Brazil required a local company, funds in an institution authorised by the Central Bank and a centralised self-exclusion. On complaints, a player complains to the operator and can then go to an independent dispute-resolution provider at the operator's cost, but the CGA's policy states that it "will not resolve or make decisions on any player complaints regarding gambling-related transactions". The licence conditions also require licensees to ensure that the laws of the jurisdictions where they offer games are strictly adhered to, which, read literally, means a licensee offering games in Brazil after 25 September would be offering them where the law now prohibits it.

Anjouan

Anjouan Gaming calls itself the designated regulatory authority for internet gaming in the Autonomous Island of Anjouan, in the Union of the Comoros. It says it has issued more than 1,650 licences, with nearly 1,500 active. Its published standards list anti-money-laundering controls, age verification, self-exclusion, deposit and loss limits, and require operators to "segregate player funds from operational funds", but describe no compensation scheme if an operator fails. The Authority's policy says it "does not handle, mediate, or adjudicate player complaints". A player must first complain to the operator, which has 30 calendar days to respond.

Gambling.com, summarising a Guardian report of 19 October, says a spokesperson for the Comoros central bank called the regime "illegal activity by people abusively using the name of the Comoros Islands to give themselves apparent legitimacy", and that the Comoros government asked the UK for help. The Authority has issued that many licences, its first annual regulatory report is by its own account still in preparation, and the Comoros side is reported to dispute its mandate. That doesn't sound like close supervision.

Costa Rica

Costa Rica is thinner still on paper. Its Law 9050 of 2012 is a casino and electronic-betting tax law. Article 4 taxes "las empresas dedicadas a la recepción y el procesamiento de datos que generan apuestas electrónicas" (companies that receive and process data that generate electronic bets) by their number of employees, and article 6 leaves the permits to municipalities. The law sets up no gaming licence with player-protection supervision such as age checks, complaint handling or segregated funds. Bill 25.600, presented on 21 May 2026 and in committee according to the Legislative Assembly record as mirrored by Delfino.cr, would make the Junta de Protección Social the "autoridad reguladora, supervisora y fiscalizadora única" of the sector. Until something like it passes, operators in Costa Rica get very little supervision too.

What bettors and the state lose

Bettors who move to such sites give up the protections of the regime Brazil has just ended. Under Lei 14.790 and the Finance Ministry's Portaria 1.231 of 2024, licensed operators had to check identity using "a tecnologia de identificação e reconhecimento facial" (facial identification and recognition technology), and the law barred under-18s and people diagnosed with ludopatia, or gambling disorder, from betting. Self-exclusion was centralised: one request barred registration with every authorised operator. Operators had to let bettors set limits and offer session alerts, pauses and self-exclusion, and to require loss and time limits at registration.

The money was ring-fenced as well. Bettors' funds held at institutions authorised by the Central Bank "constituem patrimônio separado" (form a separate estate) that does not answer for the operator's debts. Deposits and withdrawals went only by electronic transfer between the bettor's registered bank account and the operator's, and bettors had the rights of the Consumer Defence Code and free customer service in Portuguese. Authorised operators used only the "bet.br" domain, a mark that told a bettor which sites were legal and now tells them nothing, because every authorisation has ended.

The state loses tax, supervision and visibility. Operators had to report well-founded suspicion of money laundering to the financial-intelligence unit, Coaf. The TCU says the persistence of unauthorised operators despite blocking creates an "ambiente de impunidade percebida" (environment of perceived impunity) that encourages their use for money laundering.

Who has lost their job since the ban?

The ten-day shutdown has a payroll attached to it, and the first counts are in. The national association of lotteries and games, ANJL (Associação Nacional de Jogos e Loterias), gave Poder360 a partial survey that the outlet reported on 7 October. ANJL says more than 3,500 workers had been dismissed since the MP, which it puts at 23 percent of the sector's workforce, and that another 4,167 were on collective leave (férias coletivas). The association describes its figures as partial and self-reported by companies.

A second count came from Amig, the Associação de Mulheres da Indústria do Gaming. According to Jornal de Brasília, in a report republished by Yogonet on 7 October, Amig's survey found at least 36 licensed operators had started layoff rounds by 5 October. H2Bet dismissed 170 of its 270 employees, according to its operations director, Diego Breda de Carvalho.

The operator KTO is the case with the least clear numbers. NEXT.io reported on 12 October that LinkedIn posts by former staff suggested more than 200 people were let go on 8 October, and said KTO had not confirmed the closure. On 16 October Intergame carried the word of CEO and co-founder Andreas Bardun, who said he had cut a "significant number" of jobs and called it an incredibly difficult decision. He added that "reports of KTO's death are greatly exaggerated".

These jobs are a cost of the ban itself, and the ten-day timetable made the loss sharper. The joint committee hadn't been installed when the first layoff rounds began, so it couldn't weigh what the ban would cost in jobs. A regulated market keeps those payrolls, and the tax and contributions that go with them.

Company and club accounts add a second set of figures. Flutter's 8-K filing of 28 September says "Flutter has ceased operating sports betting and iGaming in Brazil". The company expects a reduction of about US$70 million in 2026 revenue if it can't operate for the rest of the year. Football clubs are exposed too. Flamengo's president, Luiz Eduardo Baptista, estimated immediate revenue losses next year of R$400 million to R$430 million, InfoMoney and Estadão Conteúdo reported on 25 September. Poder360 reported on 18 October that 7 of the 14 Série A clubs that had betting sponsors were left without a master shirt sponsor. Flamengo wore the slogan "Proibir não é proteger" (banning is not protecting).

The Swedish evidence

Sweden is the comparison I find most useful, because it shows what a licensed market can and can't hold. The Swedish Gambling Authority's report of 20 March 2025 gives 86 percent as its 2023 estimate of channelisation for the competitive part of the market, lower for commercial online gaming. It says the risk of not being paid winnings is the most important reason players avoid unlicensed sites, according to a survey of November 2024, but that payments to and from unlicensed companies "är och kommer att vara relativt enkla" (are and will remain relatively easy).

The self-exclusion findings are sharper. In a survey for a 2023 government assignment on the national self-exclusion register, Spelpaus, 41 percent of respondents said they had gambled for money during their exclusion, so most didn't. The authority concluded that "ungefär 37 procent av alla respondenter med största sannolikhet har spelat hos olicensierade spelbolag" (about 37 percent of all respondents most probably played with unlicensed companies). These are people who asked to be stopped, a different group from the public. Most did stop, and a large minority found offshore play easy to reach.

The Netherlands adds a licensed-market data point. A letter of 3 July 2025 to the Dutch House of Representatives summarises the regulator's second measurement of the rules in force since 1 October 2024. The share of players netting more than €700 in deposits a month (€300 for those under 24) fell from 9.4 percent to 2.2 percent, while channelisation by players was 93 percent, down only slightly. The regulator's monitoring report of 16 April 2026 puts channelisation measured in money at 53 percent and suspects, without having measured it, that players at illegal operators lose much more, because they are protected less or not at all.

Will the ban do more harm than good?

The answer turns on how many people stop and how much harm that avoids, how many never start, how heavily those who stay bet, and who moves, how vulnerable they are and what protection they find where they go. None of the comparison evidence above concerns a ban: these are licensed regimes that were tightened or measured, so they say something about leakage when a licensed alternative exists and nothing about what happens when it is removed.

The case for the ban rests on exposure. The memorandum argues that the harm comes from a product that is always available, instant and heavily advertised, and if a meaningful share of people simply stop, and the advertising and payment routes that recruited them are cut, that is a gain the licensed regime didn't deliver. The case against rests on what movers lose and what the state can no longer see.

I'd frame the test as a balance: the harm avoided by those who stop, bet less or never start, set against the harm added by those who move, weighted by how vulnerable they are and how much they lose, and judged against a world in which the licensed regime had continued with tighter rules where it fell short. The data that would settle it, such as Pix flows to betting-type recipients, blocking records and surveys of former bettors, don't exist yet. I expect more harm than good. Unlicensed operators held about 40 percent of the market before the ban, they offer weaker protections than the regime Brazil just ended, and the state loses its tax and its view of the market.

What precedent the ban sets

At home, Brazil licensed a market in December 2023 and ended it by provisional measure less than three years later, with article 4 giving no right to a refund of the licence fee. The signal to anyone who pays the State for the right to operate is that the fee buys the right only while the policy lasts. The government's defence, that authorisations were discretionary acts open to review, is a coherent legal position, though hardly the message a licensing regime usually sends. The precedent isn't set yet. Congress can convert, amend or reject the MP, and the Supreme Court has the transition and compensation questions before it.

Abroad, Brazil is not the first. On 1 October 2025 the Council of Ministers of Timor-Leste approved a resolution that, in the government's own account, "determines the cancellation of licenses granted for the operation of online gaming and betting" and prohibits new licences. Trade press reports that Ethiopia revoked all licensed sports betting operators' licences from 15 December 2025, citing financial irregularities and national security risks. The stated grounds differ from Brazil's public-health case.

Two partial comparators come from a different setting. The Philippine regulator PAGCOR states that, following Executive Order No. 74, "ALL Internet Gaming Licensees and Authorized Providers were cancelled effective December 15, 2024". The UNODC's 2024 regional report says that in 2019 the Cambodian government announced it would issue no more online gambling licences, and that online gambling persisted in the region, with tighter controls "merely pushing operators underground". Both were offshore sectors in an organised-crime context, unlike Brazil's domestic market.

What alcohol and cannabis in the US teach about bans

US Prohibition is the usual comparison, and the record is more mixed than the slogan. The 18th Amendment, ratified on 16 January 1919, prohibited the manufacture, sale or transportation of intoxicating liquors for "beverage purposes" one year later, and Congress enacted the National Prohibition Act, called the Volstead Act, on 28 October 1919. The 21st Amendment repealed the 18th on 5 December 1933 after ratification by state conventions.

On consumption, Jeffrey Miron and Jeffrey Zwiebel estimated it from mortality, mental-health and crime statistics. They found it fell sharply at the start of Prohibition to about 30 percent of its pre-Prohibition level, rose over the next several years to about 60 to 70 percent, and returned to about its earlier level during the decade after repeal. The historian Jack Blocker gives a lower path after repeal, about 1.2 US gallons per capita against 2.6 at the pre-Prohibition peak. Blocker writes that "Prohibition did work in lowering per capita consumption" and calls it partly successful as a public health innovation.

The health gain was real. Angela Dills and Miron found that constitutional prohibition reduced cirrhosis deaths by about 10 to 20 percent, with state prohibitions having a minimal effect. The cost was a criminal supply. The National Archives' Prologue magazine says that after the first few years of Prohibition "illegal manufacturing, importation, and sale of alcohol became big business, run by organized crime". Blocker adds that "Although organized crime flourished under its sway, Prohibition was not responsible for its appearance", pointing to its persistence after repeal. Economists disagree on homicide. Miron argued that prohibitions create black markets in which participants use violence to settle commercial disputes, and found tougher enforcement associated with higher homicide rates. Brendan Livingston, using city data for 1911 to 1929, found that state alcohol prohibition reduced homicides immediately after enforcement began but had no measurable effect after three years.

Poisoned liquor added to the toll. A 2010 review in the Canadian Medical Association Journal of Deborah Blum's The Poisoner's Handbook says the Christmas celebrations of 1926 left 23 people dead and dozens blinded in New York City alone. The Tax Foundation argues that the income tax let the government ban alcohol without reducing tax revenue, and that when income-tax revenue later fell, people asked why the alcohol tax was being given up, which fed the case for repeal.

Repeal didn't bring a free-for-all. Prologue says there was no return to national prohibition, that the federal government again collected alcohol taxes, and that "a patchwork of local laws regulated aspects of alcohol policy such as Sunday sales, drinking age". Taxed, regulated sale is the model I want for betting. A ban can cut use, and Prohibition did. The country weighed that gain against a criminal market, poisonings and lost revenue, and chose regulation as the better trade.

The comparison has one limit that matters, and it strengthens the case against a ban. Alcohol needed physical production and distribution, including illegal importation across borders, whereas online betting can be supplied remotely. Article 1 of the MP bans fixed-odds betting lotteries "em meio físico ou virtual" (in physical or virtual form), including when carried out by an agent based abroad who offers them to a person in Brazil. That reach is what makes enforcement hard. The TCU found mass domain blocking ineffective, and the MP leans on payment and platform duties of a kind that didn't exist in 1920.

What happened where US states legalised cannabis

Cannabis is the recent test, and it points the same way. Colorado began taxed adult-use sales in 2014. Its Department of Revenue reports $236,449,405 in marijuana tax and fee revenue for 2025 and over $3.1 billion since sales began. The state's Division of Criminal Justice found in a 2021 report, as KKTV summarised it, that total marijuana arrests fell 68 percent between 2012 and 2019 and that youth use did not change significantly.

The usual fear is that legal sale pulls teenagers in. A study in JAMA Network Open by D. Mark Anderson and colleagues used national high-school survey data from 1993 to 2019, covering about 1.6 million students. It found no association between a state adopting recreational legalisation and current marijuana use, and a lower rate two or more years later. The authors concluded there was little evidence that legalisation encourages youth use.

The same record shows why the rate matters. California's Department of Cannabis Control wrote in March 2025 that "taxes, fees and other compliance costs can make it expensive to produce licensed cannabis". It put the licensed market's share of consumption at about 40 percent, up from 25 percent in 2019, and listed "decreasing costs for licensed businesses" among the ways to shrink the illicit market. The state then cut its retail excise tax from 19 percent to 15 percent on 1 October 2025, according to its Legislative Analyst's Office.

By a medium tax I mean one high enough to fund the state and the safeguards, low enough to keep buyers in the licensed market. Brazil's licensed regime sat in that range. Lei 14.790 sent 12 percent of the proceeds left after the statutory deductions to public destinations. Complementary Law 224 of 26 December 2025 cuts the operator's share from 88 percent to 87 percent in 2026 and 86 percent in 2027, which Focus Gaming News reports as a rise in the tax on gross gaming revenue from 12 percent to 13, 14 and then 15 percent in 2028. Even at that rate, the industry estimate the TCU cited put about 40 percent of the market with unauthorised operators, which argues for enforcing against them and for holding the rate where it is.

Predictions and verdict

Central Bank Pix data, blocking records and surveys of former bettors will show whether the migration I expect is happening, and a low rate among former licensed bettors would count against me. A large overall fall in play can coexist with many people moving, so a fall alone wouldn't settle it. I also expect the dispute to narrow from whether Brazil may ban fixed-odds betting to who bears the cost of a ban already in effect. As of 25 October the court's decisions page for ADI 7721 showed procedural orders, such as the 2 October order to notify the AGU, and no ruling deciding or suspending the ban, and the pages for ADIs 8024 and 8027 were empty.

Congress should decide expressly whether to convert, amend or reject the measure, and shouldn't treat 23 November as the only date. I would reject it and spend the time tightening the licensed regime on advertising, deposits and payments. If Congress converts the ban, it should write the transition and the treatment of licence fees into the statute instead of leaving both to litigation. The government should publish the studies behind its "tens of billions" figure, and the Central Bank data behind the 10 percent fall in Pix transactions. The Supreme Court should decide the transition and compensation questions quickly, whatever it does with the prohibition itself.

The harm the government cites is real and deserves an answer. The best one is what the United States settled on for alcohol after 1933 and several states chose for cannabis: legal sale, a moderate tax and rules that bite. Brazil had built most of that. It closed the market instead, with a ten-day shutdown, a thirty-day licence end and no refund of the licence fee, before the joint committee had been installed. I wouldn't call that unlawful. I'd call it the wrong policy, delivered too fast, with consequences that could be hard to undo even if Congress later rejects the measure.

Tax it, regulate it, and keep it where the state can see it.