The Unfinished Transition未竟的转型
The armed forces did not lose in 1985. They negotiated. The 1988 Constitution was written with the military’s pen, not against it. This assessment maps the institutional architecture, economic privileges, and structural immunity that the Brazilian military extracted as the price of democratization — and has collected ever since.军队在1985年并没有失败。他们是谈判退出的。1988年宪法是用军方的笔写成的，而非针对军方。本评估梳理了巴西军方作为民主化代价所攽取的制度架构、经济特权和结构性豁免——并且此后一直在兑现。
The military is the only institution in Brazil that has never been made to account for the dictatorship.军队是巴西唯一从未因独裁统治被追责的机构。
The military traded a return to barracks for permanent structural immunity. The amnesty law, Article 142, the military justice system, the pension architecture — these are not accidents of history. They are the institutional output of a negotiation in which the armed forces exchanged political power for institutional preservation.
Democracy arrived. The transition never did. Not because Brazilians forgot, but because the transition was structured to make accountability impossible. Every civilian government since Sarney has calculated that the cost of renegotiating the deal exceeds the cost of paying it. That calculation has never been wrong. It has also never been tested at full force.
Five structural claims五项结构性论断
| Instrument independence工具独立性 | The military chose when to leave power — and chose terms. No external force removed them. |
| Institutional preservation制度保全 | The 1988 Constitution preserved every military privilege — by inaction and deliberate ambiguity. |
| Economic position经济地位 | Military compensation and pensions exceed civilian equivalents by structural design, not market accident. R$50B+/year. |
| Accountability gap问责缺口 | Zero officers convicted for crimes committed 1964–1985 under any civilian government. Not one. |
| Democratic stress test民主压力测试 | 2018–2022 revealed conditional tolerance of democracy, not commitment to it. 6,157 military in government. January 8. |
Key numbers关键数字
The 1964–1985 regime didn’t just rule Brazil. It redesigned the institutional architecture — and the redesign survived the regime.1964–1985年的政权不仅仅统治了巴西。它重新设计了制度架构——而这一设计在政权结束后存活了下来。
The institutional transmission制度传递
Military Regime军事政权
ARENA party apparatus. SNI intelligence network. Officer corps wage premium. DOI-CODI + military justice over civilians. Defense industry (EMBRAER, ENGESA, Avibras).
Transition Deal转型交易
Lei da Anistia: no accountability. Military participates in constitution drafting. Services preserved (hospitals, schools, clubs). Controlled handover to civilian Sarney.
Constitutional Architecture宪法架构
Art. 142 “guarantor” clause. STM military justice autonomy. Pension system constitutionalized. Conscription infrastructure maintained. ARENA → PDS → Centrão continuity.
Ongoing Extraction持续攽取
Pension system: R$50B+/year. 6,000+ military positions under Bolsonaro. Prime urban real estate. Defense procurement without external audit. Art. 142 weaponized in 2018–2022.
The 1988 Constitution is celebrated as Brazil’s “Citizen Constitution” — and it genuinely expanded social rights. But its military provisions were not written against the dictatorship. They were written by its institutional heirs. Article 142, the STM, the pension architecture, the absence of transitional justice mechanisms — these survived the Constituent Assembly not because they were overlooked but because the military had the institutional leverage to ensure they survived. The design is the point.
The most concrete form of military privilege: a parallel welfare state funded by the public treasury, structurally immune from reform.军事特权最具体的形式：一个由公共财政资助、在结构上免受改革的平行福利国家。
The parallel welfare state平行福利国家
| Component组成部分 | Military军方 | Civilian equivalent文职同等 | Gap差距 |
|---|---|---|---|
| Base salary (General)基本工资（将军） | R$28,000–33,000/mo | Top civil servant: ~R$18,000 | 1.7× |
| Pension replacement养老金替代率 | Near 100% of final salary (integral) | INSS: ~70% max, capped at R$7,786 | Full vs. capped |
| Healthcare医疗保健 | Hospital das Forças Armadas (HFA) + FuSEx system | SUS (universal, underfunded) | Parallel system |
| Education教育 | 11 Colégios Militares (13,000 students), tuition-free | Public schools (variable quality) | Elite access |
| Housing住房 | Subsidized military housing, PNR system | Minha Casa Minha Vida (means-tested) | Guaranteed vs. lottery |
| Retirement age退休年龄 | 28–30 years of service (age ~48–50) | INSS: age 65 (men) | 15+ years earlier |
The 2019 pension reform excluded the military by design. Bolsonaro’s EC 103/2019 reformed INSS and civil servant pensions. The military was given a separate, gentler reform with a 5-year delayed implementation and terms negotiated directly between the Defense Ministry and the military command — not by Congress. The fiscal argument for inclusion was overwhelming: ~380,000 military pensioners cost R$50B+/year (vs. 35 million INSS pensioners costing ~R$800B). The per-capita cost is 7× higher. The exclusion was political, not actuarial.
The most dangerous clause in the Brazilian Constitution is also the most ambiguous. That ambiguity is the point.巴西宪法中最危险的条款也是最模糊的。这种模糊性正是关键。
STM — Military JusticeSTM——军事司法
A court system for the military, by the military由军方创建、为军方服务的法院体系
The Superior Tribunal Militar (STM) is Brazil’s military supreme court. 15 justices: 10 military officers, 5 civilians. Has jurisdiction over military crimes — but “military crime” is defined broadly enough to cover some acts against civilians. Post-1988 reform limited but did not eliminate civilian jurisdiction. The military polices itself.
GLO — Guarantee of Law & OrderGLO——法律与秩序的保障
The constitutional power that became a political instrument从宪法授权变为政治工具
Article 142 allows deployment of armed forces for “guarantee of law and order” (GLO). Under Bolsonaro: 130+ GLO operations vs. ~30 under Lula/Dilma combined. Rio de Janeiro GLO (2018): military occupied favelas with constitutional authorization. Each GLO normalizes military presence in civilian governance.
Poder Moderador调节权（Poder Moderador）
The most dangerous reading: a fourth branch of government最危险的解读：第四权力分支
The Imperial Constitution of 1824 established the Poder Moderador — the emperor as arbiter between branches. The Bolsonarista reading of Art. 142 resurrects this: the military as the “moderating power” that can intervene when civilian institutions “fail.” This reading has no legal basis after 1889. It was publicly endorsed by Bolsonaro and never formally repudiated by military command.
The ambiguity is the weapon. Article 142’s value to the military is not in any specific reading but in the fact that it admits multiple readings. As long as the clause remains ambiguous, the military retains interpretive option value — the ability to escalate from “instrument of the president” to “autonomous guarantor” when politically convenient. Clarification would eliminate this leverage. No government has attempted it.
The military is not just an institution. It is an economic actor with defense industries, real estate, procurement contracts, and a financial infrastructure.军方不仅仅是一个机构。它是一个拥有国防工业、房地产、采购合同和金融基础设施的经济行为体。
SIVAM — Amazon SurveillanceSIVAM——亚马逊监控
R$1.4 billion contract awarded to Raytheon (1994). Brazilian Air Force overrode civilian procurement rules. Operation Anaconda (2003) implicated military officials and a senator in kickbacks. The program continued. The procurement model continued. No structural reform followed.
PROSUB — Nuclear Submarine ProgramPROSUB——核潜艇计划
R$15.9 billion. DCN/Naval Group (France). No external audit permitted for classified components. The nuclear enrichment component involves the Navy’s parallel nuclear program — one of the few non-NPT-compliant military programs in a democracy. Strategic rationale is defensible. Oversight architecture is not.
Every Latin American country that transitioned from military rule made a deal. Brazil’s deal was the most generous to the military — and the least renegotiated.每个从军事统治转型的拉美国家都达成了一项交易。巴西的交易对军方最为慰慨——也最少被重新谈判。
Argentina Model: Prosecuted阿根廷模式：起诉
Juicio a las Juntas (1985): Videla and Galtieri convicted. ESMA trials continuing into the 2020s. Military institutional power dramatically reduced. The armed forces lost their political autonomy, budget control, and constitutional self-authorization.
Cost: 1987–1990 carapintada military rebellions. Menem pardons (1989–90). The cost was real but temporary.
Result: A military that accepts civilian authority because it lost. Argentine armed forces have no constitutional “guarantor” clause and no autonomous justice system.
Brazil Model: Amnestied巴西模式：特赦
Lei da Anistia (1979): Written by the military government. Mutual amnesty. No trials. No lustration. Upheld by STF in 2010 (ADPF 153). Inter-American Court of Human Rights ruled it violated the American Convention (2010). Brazil ignored the ruling.
Cost: Zero short-term political disruption. The transition was “smooth.”
Result: A military that tolerates civilian authority but has never accepted its legitimacy. Every privilege preserved. Conditional tolerance tested and found conditional in 2018–2022.
Six military figures who shaped — or tested — the institutional architecture.六位塑造——或检验——制度架构的军事人物。
Architect of abertura
Managed decline while locking in the deal
Most consequential post-1985 general
Moved spectrum rightward
Chief of Staff, Defense Minister
Alleged coup architect
Former UN peacekeeping commander
Institutional sabotage of GSI civilian role
Active-duty general
Zero accountability for civilian catastrophe
Senator 2023–
Present to stabilize, absent to disown
2019–2022 was the first full-spectrum test of the military’s institutional claims since re-democratization. The results are not reassuring.2019–2022年是再民主化以来对军方制度性诉求的首次全面检验。结果并不令人安心。
6,157 Military in Government6,157名军人进入政府
More military in civilian government positions than at any time since the dictatorship. Not a coincidence of qualified personnel. A deliberate colonization: Health Ministry, education secretariats, infrastructure agencies, Petrobras board, INEP, IBGE, Funai. The baseline under Lula/Dilma was ~1,000–1,300.
The COVID Military StateCOVID军事化国家
Pazuello’s Health Ministry run as a field operation. The Manaus oxygen crisis (January 2021): 69 people died of preventable asphyxiation while generals managed procurement logistics. Pazuello’s post-crisis trajectory: returned to active duty. No court-martial. No accountability. The system protected its own.
January 8 and the Question of Orders1月8日与命令之问
Bolsonaro supporters stormed the Planalto, Congress, and STF. Military forces in Brasília stood down. Who gave — or withheld — the order to protect the buildings? PF investigation implicated Braga Netto and others. Military command position: “we followed protocol.” Which protocol? Designed by whom?
The military’s selective silence on January 8 is not neutral institutional behavior. A military genuinely committed to constitutional order would have acted to protect it. The silence is itself a constitutional position — one that says: we tolerate civilian government when it serves our interests, and we withhold protection when it does not. The fact that civilian police (PF) — not the military — ultimately investigated and prosecuted the coup attempt tells you which institution defended democracy and which one hedged.
The SNI was abolished. Its institutional DNA was not. The intelligence apparatus is the vector through which authoritarian institutional memory travels.SNI被废除了。它的制度DNA没有。情报机器是威权制度记忆传播的载体。
Serviço Nacional de Informações. Created by the coup government. Answers directly to the president. Unlimited domestic surveillance authority.
Divisional intelligence expanded to every ministry. DOI-CODI torture centers. CIE (Army), CIM (Navy), CISA (Air Force). Intelligence penetrates all state institutions.
SNI survives the transition intact. Personnel retained. Files not transferred to civilian control. Institutional knowledge preserved within the military apparatus.
Formally dissolved. But: files not destroyed. Personnel not dispersed. Institutional memory transmitted to successor agencies and military intelligence branches.
Agência Brasileira de Inteligência. Civilian in name. Inherits SNI personnel, methods, and institutional culture. Military intelligence branches (CIE, CIM) continue independently.
ABIN under military-linked leadership. Used for political surveillance (alleged). “Abin paralela” scandal: parallel intelligence operation outside legal framework. PF investigation ongoing.
The SNI was abolished in 1990. Its files were not destroyed, not transferred to a civilian archive, and not made publicly accessible. The Lei de Acesso à Informação (2011) established a framework for declassification — but pre-1985 security documents face systematic obstruction. The Comissão Nacional da Verdade (2012–2014) documented 434 killings and disappearances but had no prosecutorial power. It recommended revoking the amnesty law. The recommendation was ignored.
The intelligence community is the institutional vector through which authoritarian institutional memory travels. Personnel rotate between ABIN and military intelligence branches. The “Abin paralela” scandal (2023–24) revealed that under Bolsonaro, ABIN was used for political surveillance operations outside any legal framework — a direct echo of SNI-era methods, executed by an agency that inherited SNI-era institutional culture.
The military’s constitutional position is effectively unreformable — not because of formal supermajority requirements, but because of the political cost architecture.军方的宪法地位实际上不可改革——不是因为正式的绝对多数要求，而是因为政治成本架构。
What needs reform需要改革的内容
1. Article 142 clarification: who interprets “guarantee of constitutional order”?
2. STM civilian jurisdiction: remove military justice over civilian cases
3. Pension parity: military on equal terms with civil servants
4. Transitional justice: revoke amnesty law, open files
5. Real estate audit: public valuation and divestiture of non-operational properties
Why reform cannot pass为何改革无法通过
Every government needs military passive consent. The Centrão’s ARENA lineage makes it structurally opposed. The cost is incurred immediately (military institutional resistance); the benefit is structural and diffuse. The military’s democratic threat capacity (Art. 142, GLO) is the enforcement mechanism that makes renegotiation costly. The cage enforces itself.
What has been tried已尝试的改革
Art. 142 reform: Never attempted.
STM jurisdiction: Partial STF rulings, reversed.
Pension reform: 2019 — military explicitly excluded.
Truth commission: 2012–14. Documented crimes. Zero accountability.
Real estate: Never attempted.
Score: 0 for 5.
The political trap loop政治陷阱循环
Reform requires改革需要
Congressional supermajority (3/5 for PEC) or STF jurisprudential shift. Both require sustained political will.
Centrão blocks中间派阻挡
ARENA institutional lineage. Centrão parties have no incentive to antagonize military. Military retains reserve bargaining power.
President trades总统交易
Every president needs Centrão for governing majority. Reform capacity is traded for coalition stability. Military reform is never the priority.
Preserved ↺保留 ↺
Military institutional position preserved by default. The loop repeats every administration. The cage enforces itself.
Assessment评估
| Dimension维度 | Assessment评估 | Rating评级 |
|---|---|---|
| Constitutional immunity宪法豁免 | Art. 142 ambiguity actively maintained. No STF clarification achieved. No PEC attempted. The interpretive option value is preserved. | LOCKED锁定 |
| Fiscal privilege财政特权 | Pension system explicitly excluded from 2019 reform. Per-capita cost 7× INSS. R$50B+/year growing. No equalization pathway. | ENTRENCHED根深蒂固 |
| Accountability问责 | Zero convictions for 1964–1985 crimes under any civilian government. CNV documented without consequence. Amnesty law upheld. | ZERO |
| Democratic fidelity民主忠诚 | 2019–2022: 6,157 military in government. Selective silence on Jan 8. Alleged coup planning (Braga Netto indicted). Conditional tolerance revealed. | CONDITIONAL |
| Reform pathway改革路径 | All five major reform vectors blocked or reversed. Zero major reforms achieved in 38 years. Political cost architecture prevents renegotiation. | CLOSED关闭 |
What the military’s institutional position costs Brazil军方的制度地位给巴西带来的代价
Fiscal: R$50B+ annual pension burden. 7× per-capita cost vs. INSS. Growing, not shrinking.
Accountability: 61 years since the coup. Zero convictions. The precedent: state violence has no consequences.
Democratic: Constitutional ambiguity that enables democratic stress. Art. 142 weaponization proven possible (2018–2022).
Institutional: Every government spends political capital placating rather than reforming military prerogatives. Reform capacity diverted to maintenance of the deal.
What sustains it是什么维持着它
The transition deal logic: The 1979–1988 bargain was rational for both sides. The military got preservation; civilians got stability. The calculation held.
Centrão continuity: ARENA’s institutional descendants control the coalition mathematics. They have no incentive to renegotiate.
Cost-of-renegotiation: Every president since Sarney has calculated that the cost of confrontation exceeds the cost of payment. That calculation has been correct every time. It may not always be.
Self-enforcement: The military’s democratic threat capacity (Art. 142, GLO, institutional leverage) is the enforcement mechanism that makes renegotiation costly. The cage enforces itself.
Project Brazil has now mapped five structural constraints on Brazilian democracy. None elected. All constructed.
The BCB constrains economic democracy — financial market discipline embedded in monetary architecture. The media oligopoly constrains informational democracy — seven families from the military era control the signal. The evangelical machine constrains cultural democracy — a vertically integrated religion-party pipeline that did not exist fifty years ago. The Centrão constrains legislative democracy — a behavioral coalition whose institutional DNA is ARENA, preserved by the same electoral design the military left behind.
And the military constrains sovereign democracy — the capacity of civilian government to hold a monopoly on violence, restructure power, and govern without negotiating the terms of that governance with the institution that last governed by force.
The other four cages are self-reinforcing but not inherently threatening to democratic existence. The military cage is different: it is the only one where the enforcer of the deal is also the institution being protected by it. The BCB cannot send tanks. The media cannot suspend Congress. Evangelical caucus members cannot arrest the STF. The military can do all three — and the fact that it has not done so since 1985 is evidence of the deal’s success, not its absence.
The transition deal worked. Brazil got democracy. The military kept its institutions. The renegotiation has never happened. The question is not whether it will. The question is what happens when it does, and who pays.
Sources & Data Caveats来源与数据说明
1Primary legal: Constituição Federal 1988, Art. 142, 50, 143; Lei da Anistia 6.683/1979; Lei 8.457/1992 (STM statute); Lei Complementar 97/1999 (GLO framework).
2Truth commission: Comissão Nacional da Verdade, Relatório Final Vol. I–III (2014). Documented 434 killings/disappearances, ~20,000 tortured.
3Pension data: TCU audit reports on military pension system (2018–2022). INSS ceiling: R$7,786 (2024). Military figures from STN and Defense Ministry transparency portal.
4Military in government: 6,157 figure (2021 peak) from Folha de S. Paulo investigation using ministerial transparency portal data.
5January 8: Publicly available federal indictments. PF investigation findings. Braga Netto arrest (December 2024). Full investigation record partially sealed.
6Analytical framework: Wendy Hunter, Eroding Military Influence in Brazil (1997); Thomas Skidmore, The Politics of Military Rule in Brazil (1988); Jorge Zaverucha, “Military Prerogatives and Political Transitions in Brazil” (2010); Alfred Stepan, Rethinking Military Politics (1988); Perry Anderson, “Lula’s Brazil” (LRB, 2011).
7IACHR: Inter-American Court of Human Rights, Gomes Lund et al. v. Brazil (2010). Ruled Lei da Anistia violates American Convention. Brazil has not complied.
8Caveats: Military pension figures vary by source and year. Real estate valuations are not publicly disclosed. Nuclear submarine program costs include classified components. ABIN personnel continuity estimates (40–60%) from investigative journalism, not official data.