The Removal That Cost Nobody Anything

Having found her guilty of a crime grave enough to strip her of the office, the Senate declined to bar her from holding office again. That motion got 42 votes, short of the 54 required. The same chamber, the same afternoon, the same evidence: guilty enough to remove, not guilty enough to disqualify. Constitutional lawyers called it fatiamento — slicing. It is as close to a signed confession as an institution gets. The senators knew exactly what they were doing, and they knew it was not a criminal finding, and they arranged the paperwork so that everyone could keep their options open.

Ten years on, the ledger is available. It is worth reading, because the sequence it records is not a Brazilian peculiarity. It is what happens when a political system reaches for an instrument that does not fit the job.

The Charge Was an Accounting Maneuver

Rousseff was removed for pedaladas fiscais — fiscal pedaling, the practice of delaying transfers to state banks so the banks covered social program outlays temporarily and the deficit looked smaller than it was. It is a real budgetary offense. It is also something Brazilian governments, federal and state, had done routinely for years, and something no one had previously suggested was a removable crime.

She was not accused of taking money. Ten years later, she still has not been.

Meanwhile, the people conducting the removal were carrying substantially heavier files. Eduardo Cunha, the Speaker of the Chamber of Deputies who accepted the impeachment petition and drove the process, was expelled from the Chamber twelve days after the Senate vote and jailed the following month. He drew a sentence of more than fifteen years for corruption, money laundering, and moving funds abroad. The man who opened the proceedings was in custody before the year was out.

Michel Temer, Rousseff's vice president, took the office. In May 2017 recordings surfaced in which Temer appeared to endorse hush-money payments to Cunha. The Chamber of Deputies voted twice that year to block charges against him. He was arrested in March 2019, after leaving office, and released within days.

That asymmetry is not hypocrisy, or not only hypocrisy. It is a structural signature, and it points at the actual mechanism.

A No-Confidence Vote Wearing a Criminal Costume

Brazil is a presidential system. There is no vote of no confidence. A coalition that has stopped supporting the president cannot simply say so and trigger an election; it has to wait out the term, or find a crime.

By 2016 the coalition had stopped supporting her. The economy had contracted for two consecutive years, the PMDB had left the government, and the votes existed to end the presidency. What did not exist was a mechanism to end a presidency for the reason it was actually being ended. So the reason was translated into the available vocabulary — a budget technique, retroactively promoted to a high crime — and processed through a body that then immediately signaled, in its second vote, that it did not believe its own finding.

Rousseff called it a coup in institutional clothing. That framing has always been more useful as accusation than as analysis, because the process really was legal: the forms were followed, the votes were real, the Supreme Court supervised. The interesting fact is not that the law was broken. It is that the law worked exactly as written and produced something nobody could describe honestly afterward.

Latin America has been running this experiment at scale. Carlos Andrés Pérez in Venezuela, 1993. Abdalá Bucaram in Ecuador in 1997, removed by Congress on grounds of "mental incapacity," a diagnosis rendered by legislators. Alberto Fujimori resigned from Japan by fax in 2000; the Peruvian Congress refused the resignation and declared him morally unfit instead, preferring the verdict to the vacancy. Fernando Lugo in Paraguay in 2012, tried and removed in under two days.

Peru has industrialized it. Article 113 of the Peruvian constitution permits removal for "permanent moral incapacity," a phrase the drafters left undefined. It has since become a functioning no-confidence vote with no election attached: Kuczynski, Vizcarra, Merino, Sagasti, Castillo, Boluarte — six presidents between 2016 and 2023, one of whom lasted five days. The clause was a gap in the drafting. Congress found the gap and moved in.

Across the region the twentieth century's method for removing an inconvenient president was the military. Since roughly 1990 the method has been the legislature, at close to the same frequency. The uniform came off. The function did not change.

What the Substitution Costs

The swap is not an improvement, and this is where the "at least it was constitutional" defense stops reaching.

A no-confidence vote in a parliamentary system says: we have lost confidence, go to the country. It is reversible. The removed party can win the next election and nothing about the removal marks them as criminal. Impeachment says something categorically different: you committed an offense. It converts a political dispute into a moral finding, and moral findings do not reverse. They accumulate.

So each use raises the stakes of losing. If defeat means not just leaving office but being adjudicated a criminal, then the losing side has no incentive to accept the result and every incentive to acquire the instrument itself. The tool is symmetrical. Whoever holds the legislature holds it.

Brazil ran that sequence in public. Sergio Moro, the judge presiding over Lava Jato, convicted Lula in 2017 and put him in prison in 2018, removing the polling leader from the presidential race. Jair Bolsonaro won that race. Moro joined his cabinet as justice minister. In 2019 The Intercept began publishing intercepted messages showing Moro coordinating with the prosecutors he was supposed to be judging. In 2021 the Supreme Federal Court annulled Lula's convictions and ruled Moro biased. Lula was elected in 2022. On January 8, 2023, a crowd stormed the Congress, the presidential palace, and the Supreme Court. Bolsonaro was declared ineligible until 2030 by the electoral court that June, and in 2025 he was convicted of plotting a coup and sentenced to more than twenty-seven years.

Seven years from a constitutional removal to an attempted unconstitutional one. I want to be careful about what that sequence proves, because the temptation is to read it as one causal chain and it isn't one. Lava Jato was running before the impeachment and would have reached Lula regardless. Bolsonaro's rise had drivers — the homicide numbers, the evangelical vote, a collapse in party identification — that had nothing to do with Rousseff.

What the impeachment supplied was narrower, and sufficient. It demonstrated that the presidency could be ended by whoever held the legislature, on a charge assembled after the decision to end it. From 2016 forward, every serious actor in Brazilian politics had that fact in the model. It does not explain January 8. It explains why the people who planned January 8 did not believe they were doing anything categorically new.

South Korea has been running a version of the same tape. Park Geun-hye impeached in December 2016, upheld by the Constitutional Court in March 2017, convicted, imprisoned, pardoned. Yoon Suk-yeol declared martial law in December 2024, was impeached within eleven days, removed by the court in April 2025, and tried for insurrection. Yoon's decree had its own local causes — a hostile legislature, investigations closing in — and I am not claiming Park's removal produced it. I am claiming that a president who had watched a predecessor travel from the Blue House to a cell had an unusually vivid picture of what losing looks like, and reached past the constitution rather than test it.

When a Removal Actually Holds

The pattern has an exception, and the exception is Brazilian, which makes it useful.

Fernando Collor de Mello was impeached in 1992 for a corruption scheme run through his campaign treasurer. He resigned during the Senate trial; the Senate proceeded anyway and barred him from office for eight years. And then Brazil was fine. Itamar Franco finished the term, the Real Plan stabilized a currency that had been in freefall, Cardoso won in 1994, and the country entered two decades of orderly alternation. Collor himself came back as a senator in 2006. Nobody built a movement around his innocence. Nobody stormed anything.

What separated 1992 from 2016 was not legality. Both were legal.

The obvious answer is substance — Collor took money through a named bagman, Rousseff moved transfer dates — and that answer is correct. It is also not available to me as a finding, because it is what this piece already argued, and using the thinness of the charge to explain the aftermath of the charge is a circle.

So here is what actually distinguishes them, and it is not a standard of legitimacy. It is a diagnostic. Watch what the removal cost the people who conducted it.

Collor was removed by his own side, at their own expense, with hundreds of thousands of students in the street wearing painted faces, and the coalition that removed him inherited nothing it wanted. Rousseff was removed by a coalition that inherited the presidency directly, led by a Speaker who was himself weeks from prison, and installed the vice president who had crossed the aisle to make it happen.

Cost and substance travel together, and not by coincidence. When the offense is real, the removal does not need to pay for itself — the evidence is the motive. When the offense is manufactured, something else commissioned it, and that something is usually visible in who ends up holding the office. Substance is the cause; cost is what you can see from outside while the substance is still being disputed. Which, in every case that matters, it is.

It is a diagnostic and emphatically not a test, and Nixon is the case that proves the difference. Removing him would have cost Republicans everything and handed Democrats the presidency outright; on a cost standard the proceeding was maximally suspect. It was also entirely warranted, because the tapes existed. Any rule that scores Watergate as illegitimate is not a rule about legitimacy at all. It is a rule about what to look at first when the evidence is contested — and nothing more than that.

Gerald Ford's pardon in 1974 is a different act from a removal, and I won't pretend the symmetry is exact. But it is the clearest instance on the record of the second vote's logic executed on time and on purpose: a successor declining to convert a political victory into a criminal finding, at a price he quite plausibly paid at the polls two years later. The American system absorbed the shock for roughly half a century.

The Brazilian Senate half-understood this on August 31, 2016. That is what the second vote was. Forty-two senators reached for a version of the Ford move — remove her, don't criminalize her, leave the door open — and it came too late and too small to matter, because the criminalization was already running two floors down in Curitiba, aimed at somebody else.

The Gap Is in the Drafting

A diagnostic is for observers. It tells you what you are looking at. It does not tell anyone what to do, and the standard it seems to imply — that removers ought to be willing to pay a price — asks politicians to behave better inside a structure that rewards them for not bothering. That is a wish, not a repair.

The repair is upstream, and Peru is the demonstration. Article 113 permits removal for "permanent moral incapacity" and never defines the phrase. That is not a moral failure of the Peruvian Congress. It is an unlocked door in a building with no other exits, and the Congress found it and moved in, six times in seven years, which is what any rational occupant would do.

Presidential systems have no lawful way to end a government that has lost its coalition. Parliamentary systems have one, use it constantly, and it carries no criminal charge — Italy has replaced governments dozens of times since the war and not once did the replacement require finding a crime. The presidential answer is to wait out the term or find one, and when waiting becomes intolerable, one gets found. Every case in this ledger is a coalition breakdown processed through a criminal instrument because that was the only instrument in the building.

Better instruments exist and are not exotic. Recall referenda, which hand the question back to the voters who granted the office. Snap elections triggered by legislative supermajority. Constructive no-confidence — the German version, which requires the removers to name the successor in the same motion, making the transfer of power the explicit subject of the vote rather than a side effect of it. Censure that removes without convicting. They are absent from most presidential constitutions because the drafters treated the fixed term as the feature, and never asked what a legislature would do when the term became unbearable and the only door was marked crime.

Ten years later, the thing the impeachment was supposed to resolve is unresolved and the country is further from resolving it. Rousseff was never charged with theft. Cunha went to prison. Temer was arrested. Moro was ruled biased. Lula was jailed, released, and re-elected. Bolsonaro was convicted of trying to seize the state.

It settled nothing. It was never built to settle anything — it was built for a different job entirely, and got handed this one because nothing else in the constitution could do it. It ended a presidency on the first ballot. Then the Senate voted again to make sure nobody mistook it for justice, which was the most honest thing it did all day.