Portugal’s Prime Minister António Costa resigned on Tuesday, ending nearly eight years at the head of government after prosecutors investigating alleged corruption surrounding lithium mining, hydrogen production and major investment projects searched government ministries and detained several people connected to his administration. Costa has denied wrongdoing, and at this stage it would be wrong to write as though criminal guilt has been established. His resignation therefore raises a more interesting institutional question than whether one politician is corrupt: what standard of responsibility should apply to political leaders when the evidence is serious enough to undermine confidence but not yet sufficient to establish criminal guilt?
Modern political debate often collapses criminal responsibility and political responsibility into one standard. A politician under pressure says they have not been convicted of anything and therefore have no reason to resign. Legally, that principle is essential. The presumption of innocence means an allegation, investigation or arrest cannot be treated as proof. Governments capable of removing opponents merely by opening investigations would turn law enforcement into an extraordinarily powerful political weapon. Yet political office has never depended solely on whether the office-holder can be imprisoned. A prime minister possesses responsibilities extending beyond the minimum requirements of criminal law.
This distinction is easiest to see outside corruption. A transport minister does not need personally to cause a railway accident before accepting political responsibility for systemic failures inside the ministry, as Greece demonstrated after the Tempi disaster. A defence minister does not need personally to commit an operational mistake before parliament can decide that leadership has failed. Political responsibility concerns fitness to continue exercising public authority, while criminal responsibility concerns whether specific conduct satisfies the elements of an offence under law. One can exist without the other.
Costa appears to have recognised that distinction explicitly. In announcing his resignation, he argued that the dignity of the prime minister’s office was incompatible with suspicion concerning his integrity, even while insisting he had committed no crime. That does not prove the investigation’s allegations are correct. Resignation should not be interpreted as confession. It means he concluded that governing effectively required a level of political confidence that the investigation had made difficult to maintain.
There is something institutionally healthy in that separation. If resignation automatically implied guilt, politicians would have strong incentives to remain in office until every legal process was exhausted, because leaving would amount politically to admitting the charge. Investigations can take years. Governments would then remain paralysed around leaders whose authority had become inseparable from their legal defence.
A system capable of saying “step aside while the institutions investigate, without assuming you are guilty” creates more room between impunity and premature condemnation.
That middle ground is difficult to maintain because political opponents naturally treat investigations as evidence. Opposition parties gain from portraying allegations in the strongest possible terms, while supporters gain from dismissing them as politically motivated. Courts operate much more slowly than electoral politics, so public opinion often reaches a verdict before investigators do.
This creates a mismatch between legal time and political time. Prosecutors can spend months examining documents and testimony. A prime minister needs to govern tomorrow morning. Parliament needs to pass budgets. Ministers need to make decisions. Investors and civil servants need to know whether the government will survive.
Political systems therefore sometimes need temporary decisions before courts can provide final answers.
Resignation is one such mechanism.
The alternative is not necessarily better. Imagine a rule stating that every politician must resign immediately whenever prosecutors open an investigation. Prosecutors would then indirectly acquire the power to remove elected governments without proving anything. Even independent prosecutors can make mistakes, and politically motivated legal institutions could weaponise investigations deliberately.
The correct threshold therefore cannot simply be “investigation equals resignation”.
Context matters.
How serious are the allegations?
How close are they to the political leader?
Does remaining in office create conflicts of interest?
Can the government function?
Is there evidence that institutions investigating the case remain independent?
Does stepping aside preserve the investigation’s credibility?
There is no mathematical formula capable of resolving those questions.
Political accountability is partly judgement.
Portugal’s response also demonstrates why institutional succession matters more than individual political survival. Costa’s resignation does not create a constitutional vacuum. President Marcelo Rebelo de Sousa consulted political parties and the Council of State before announcing that parliament would be dissolved and an early election held on 10 March 2024. Government continues in a caretaker capacity while constitutional procedures determine what follows.
This is democratic resilience in a relatively mundane form.
The prime minister leaves.
The state does not.
That distinction sounds obvious until we compare systems where political leaders become so central that their removal threatens the entire political order. Strong institutions make individual leaders replaceable precisely because rules exist for what happens when they disappear.
A democracy should therefore be judged partly by how boring leadership crises can remain.
Portugal is experiencing a serious political scandal. Yet nobody needs tanks outside parliament, emergency constitutional improvisation or mass mobilisation to determine who governs next. Investigators investigate, the prime minister resigns, the president exercises constitutional authority and voters will eventually decide the new parliamentary balance.
Institutional predictability lowers the cost of political failure.
This is the opposite of the dynamic visible in Pakistan, where losing political power can mean imprisonment, exile or long-term exclusion and where formal civilian politics operates alongside powerful unelected institutions. When leadership transitions become existential, politicians gain stronger incentives to manipulate institutions to survive.
Portugal’s system gives Costa another option: leave office while defending himself legally.
That makes resignation less personally catastrophic.
The same principle explains why democratic alternation matters. Politicians behave differently when opposition remains a legitimate future rather than political death. A leader who expects to retain rights, reputation and potential future influence after leaving government has fewer reasons to cling to executive power at any cost.
Making political defeat survivable therefore strengthens accountability.
Corruption itself creates a different institutional problem because modern states distribute enormous economic resources. Governments award infrastructure contracts, mining concessions, energy licences and regulatory approvals. Those decisions create concentrated financial benefits for private actors, giving businesses strong incentives to influence officials.
The Portuguese investigation reportedly concerns precisely these kinds of relationships around lithium projects, hydrogen production and a large data-centre development. Portugal possesses significant lithium deposits and hopes to become an important participant in Europe’s battery and electric-vehicle supply chains. The economic stakes are therefore substantial.
This is where the green transition intersects unexpectedly with governance.
Europe needs enormous investment in minerals, electricity networks, batteries, renewable energy and industrial infrastructure to decarbonise. Governments will increasingly determine which projects receive permits, subsidies and public support. That creates new opportunities for productive investment.
It also creates new opportunities for rent-seeking.
Whenever government decisions can make a project worth hundreds of millions of euros, firms have powerful incentives to influence those decisions. Some influence is legitimate: companies provide information, advocate policies and negotiate investment terms. The problem begins when access, political relationships or payments distort decisions away from public interest.
Industrial policy therefore requires unusually strong administrative institutions.
This connects directly to the India-US technology argument from June. Governments around the world are becoming more interventionist because they want resilient supply chains, domestic semiconductor capacity, clean energy and strategic industries. That can solve market failures and reduce geopolitical vulnerability.
It also increases the amount of economic value allocated through political discretion.
The more discretion officials possess, the more valuable access to officials becomes.
This is why state capacity and state integrity have to grow together.
A government capable of distributing billions in industrial subsidies but incapable of monitoring conflicts of interest can create politically connected capitalism rather than strategic development.
The lesson is not that governments should stop industrial policy.
It is that active government requires stronger institutions, not weaker ones.
Portugal’s lithium resources demonstrate the trade-off particularly clearly. Europe wants to reduce dependence on imported critical minerals, especially where processing and supply chains are concentrated in China. Developing domestic lithium production could therefore improve strategic resilience and support European battery manufacturing.
Mining also creates environmental costs and local opposition. Governments have to decide whether the national economic and strategic benefits justify those costs, how communities are compensated and which firms receive rights to develop deposits.
Those are legitimate political decisions.
They become much harder to trust when citizens suspect private relationships influence the process.
Corruption therefore imposes costs far beyond whatever money changes hands. It damages the informational credibility of government decisions. A project may be economically excellent and still face public opposition because citizens no longer believe the government selected it honestly.
Trust becomes a productive asset again.
When trust is high, governments can implement difficult infrastructure projects more easily because citizens begin with some presumption that procedures are legitimate.
When trust is low, every project becomes suspicious.
Transaction costs rise.
Litigation increases.
Protests expand.
Investment slows.
Corruption can therefore make the state less capable even when the direct financial loss is relatively small.
This is why independent prosecutors and courts possess economic value as well as constitutional value. Their purpose is not simply punishing wrongdoing after it occurs. The credible possibility of investigation changes behaviour before wrongdoing occurs.
Deterrence works through expectations.
An official considering favouring a politically connected firm behaves differently if they believe investigators can examine the decision independently.
A business considering bribery behaves differently if enforcement is credible.
Rule of law therefore lowers the expected return to corruption.
But prosecutorial independence creates another tension because prosecutors themselves possess enormous power. An investigation can destroy reputations and governments before charges are proven. Independent law enforcement therefore also needs procedural safeguards, evidentiary discipline and accountability.
Independence cannot mean absence of constraint.
This returns to the principle that strong institutions require power divided among institutions capable of checking one another. Prosecutors need enough autonomy to investigate government. Courts need enough independence to review prosecutors. Politicians need enough authority to govern without controlling either.
No institution receives complete trust.
The architecture assumes every institution can make mistakes.
That is institutional redundancy applied to constitutional government.
Costa’s resignation can therefore be interpreted as allowing those institutions more space to operate. A sitting prime minister inevitably possesses enormous influence over the executive branch. Even without interfering directly in an investigation, his continued presence could create perceptions of pressure or conflict.
Stepping aside reduces part of that problem.
Again, this does not establish guilt.
It protects the distinction between the office and the individual occupying it.
That distinction is particularly important for leaders who have governed for long periods. Costa became prime minister in 2015 and has dominated Portuguese politics for nearly eight years. Long-serving leaders naturally accumulate personal relationships across the state and governing party.
Democratic systems need mechanisms ensuring those relationships do not make the leader indispensable.
Leadership longevity becomes dangerous when institutions begin depending on the assumption that one person will remain.
Portugal’s transition will test whether Costa’s Socialist Party can function without him and whether political competition can reorganise normally around a new election.
That is what institutionalised parties are supposed to do.
The scandal also demonstrates why democratic accountability should not depend entirely on elections. Costa’s Socialist Party won an absolute parliamentary majority in 2022, giving the government a strong electoral mandate. If elections were the only legitimate constraint, the government could argue that voters authorised it to govern until the next scheduled election.
But electoral mandates do not suspend law between elections.
Courts, prosecutors, auditors and media provide continuous accountability precisely because voters cannot monitor government directly every day.
Democracy therefore combines vertical accountability, citizens removing governments through elections, with horizontal accountability, institutions constraining one another between elections.
The two mechanisms solve different problems.
Elections determine broad political authority.
Legal institutions investigate whether that authority was exercised within law.
Neither can substitute fully for the other.
This is why a politician saying “the voters elected me” is not a defence against legal scrutiny.
It is equally why prosecutors should not determine political programmes simply because they possess investigative authority.
Democratic government requires both institutions remaining inside their respective domains.
Portugal’s coming election will provide the final layer. Citizens will eventually judge not Costa’s criminal guilt, which belongs to courts, but the political meaning of the scandal and the broader performance of the Socialist government.
Some voters may punish the party.
Others may conclude Costa acted responsibly by resigning.
Still others may prioritise economic or social policy over the investigation.
The ballot aggregates those political judgements without pretending to resolve the legal one.
That separation is healthy.
Courts should not decide whether Portuguese voters prefer socialism, liberalism or conservatism.
Voters should not decide whether evidence satisfies a criminal statute.
Different institutions answer different questions.
The danger begins when political systems collapse those distinctions.
Governments interfere with prosecutors.
Prosecutors become political actors.
Parties declare convictions illegitimate automatically.
Opponents treat allegations as verdicts.
Eventually nobody trusts any institution outside their political camp.
Portugal has an opportunity to demonstrate another model.
A serious investigation can proceed.
A prime minister can resign.
His presumption of innocence can remain intact.
The government can continue temporarily.
An election can determine what happens politically.
None of those outcomes requires deciding today whether António Costa committed a crime.
That may be the most important distinction from this week.
Political responsibility does not always mean admitting guilt.
Sometimes it means recognising that public office belongs to the institution before it belongs to the person holding it.
A prime minister can therefore leave office while continuing to say, legitimately, that the courts must determine the facts.
The state survives both possibilities.
If Costa is eventually cleared, his resignation need not become evidence that leaving was a mistake.
If wrongdoing is eventually established, the investigation will have proceeded without the accused remaining at the centre of executive power.
Either way, Portugal’s institutions have an opportunity to demonstrate something increasingly rare in polarised politics:
A politician can be investigated without being presumed guilty.
A government can fall without the political system falling with it.
And accountability can exist before punishment without becoming punishment itself.
