Donald Trump’s transition back to the presidency is moving rapidly, and his first appointments already suggest that the second Trump administration may be structurally different from the first. Trump has selected figures including Marco Rubio for secretary of state, Pete Hegseth for defence secretary, Tulsi Gabbard for director of national intelligence and Robert F. Kennedy Jr. for health secretary, while announcing Matt Gaetz as his intended attorney general before the Senate confirmation process begins. Some selections bring conventional government experience; others are explicitly disruptive choices. The common political theme is loyalty to Trump’s broader project and scepticism towards parts of the federal bureaucracy he believes obstructed his first presidency. This makes the personnel question unusually important. Trump’s first administration tested how institutions constrain an unconventional president. His second may test what happens when a president enters office already knowing which constraints he wants to weaken.
The first Trump administration was often more institutionally conventional than Trump’s rhetoric suggested because personnel created internal constraints. Cabinet secretaries, military officials, lawyers and civil servants sometimes delayed, redirected or rejected presidential demands they believed unlawful, strategically dangerous or administratively impossible. Trump eventually came to view several of these officials not as legitimate advisers but as members of a hostile establishment.
That experience creates institutional learning.
Normally we discuss institutional learning positively: governments learn from mistakes and become more competent.
But every political actor learns.
A president frustrated by constraints can learn how to circumvent them.
An authoritarian government can learn which forms of repression produce the least international reaction.
A bureaucracy can learn how to resist political interference.
Institutions therefore participate in repeated games where each round changes the strategy of the next.
Trump enters 2025 with much more information than he possessed in 2017.
He knows how appointments work.
He knows where executive authority encounters statutory limits.
He knows which agencies possess independent organisational cultures.
He knows how courts can delay policy.
He knows that personnel willing to tell him no can become decisive.
That makes personnel selection a constitutional variable, not simply an administrative one.
Presidents are supposed to appoint people broadly aligned with their programme. Elections would mean little if permanent bureaucracies could simply ignore changes in government. A Trump administration therefore has a legitimate mandate to appoint conservative officials, reverse Biden policies and redirect executive agencies within the law.
The difficult boundary lies between policy loyalty and personal loyalty.
A secretary of state should implement the president’s foreign policy.
An attorney general should not treat the president’s political opponents as criminal targets merely because the president dislikes them.
A defence secretary should execute lawful military policy.
Military institutions should not become instruments of partisan political retaliation.
The difference is whether loyalty attaches to the elected administration’s lawful programme or to the individual leader independently of law.
Liberal democracies depend heavily on this distinction.
The bureaucracy is not supposed to govern against elected presidents.
It is also not supposed to become their personal apparatus.
The American system therefore deliberately creates tension between democratic responsiveness and professional independence.
Too much bureaucratic autonomy produces an unaccountable administrative state.
Too little produces politicised institutions whose behaviour changes according to the president’s personal interests.
The Justice Department sits directly on this boundary. Trump has repeatedly argued that prosecutions against him demonstrate that federal law enforcement was weaponised politically. His supporters therefore see restructuring as restoring neutrality.
Critics fear the opposite: that Trump may respond by using law enforcement against opponents.
Both narratives invoke the same principle.
Justice should not depend on political identity.
The test therefore cannot be whether prosecutions involve politicians. Politicians are not legally immune. The test is whether investigations begin from evidence of crimes and proceed through ordinary professional standards rather than beginning from the political desire to punish someone and searching afterwards for a legal mechanism.
This is why the attorney-general nomination matters disproportionately.
The Justice Department possesses coercive powers citizens cannot simply opt out of.
Tax policy can cost money.
Criminal prosecution can remove liberty.
Institutions possessing stronger coercive authority require correspondingly stronger procedural constraints.
The same principle applies to intelligence agencies. Tulsi Gabbard’s nomination as director of national intelligence is controversial partly because critics question some of her past positions on Russia, Syria and American intelligence institutions. Supporters argue that precisely this scepticism makes her useful in reforming agencies they believe became politically insulated.
Again, the relevant issue is not whether intelligence officials should be sceptical.
They should.
Intelligence agencies make enormous mistakes.
The Iraq War demonstrated the cost of flawed intelligence and political interpretation.
The institutional requirement is epistemic independence: analysts need to be able to tell political leaders information those leaders do not want to hear.
If intelligence becomes organised around confirming presidential preferences, decision quality deteriorates even when the president’s political objectives are legitimate.
This is the dictator’s information problem appearing inside democracy.
Leaders naturally prefer advisers who agree with them.
But a room containing only loyal agreement becomes informationally poor.
Good leadership therefore needs structured dissent.
This is one reason institutional friction can be productive.
A military commander asking whether a plan is feasible is not necessarily undermining civilian authority.
A lawyer saying an executive order exceeds statutory authority is not necessarily part of a “deep state”.
A scientist questioning a health policy is not necessarily resisting democracy.
Expert institutions add value partly by providing information political leaders would not generate themselves.
But expertise can become its own source of arrogance.
Officials can mistake technical knowledge for democratic authority and treat elected governments as temporary interruptions to bureaucratic preference.
The correct model therefore is neither bureaucratic supremacy nor presidential supremacy.
It is bounded hierarchy.
The president chooses objectives within constitutional authority.
Professional institutions provide evidence about consequences and implement lawful decisions.
Courts resolve disputes about legal boundaries.
Congress controls statutes and funding.
No actor receives complete control.
Trump’s second administration may challenge this model because his political movement increasingly views institutional resistance itself as evidence of illegitimate bureaucracy. If every constraint is interpreted as sabotage, the distinction between obstruction and constitutional limitation disappears.
That is dangerous.
But Trump’s critics can make the opposite error by treating every attempt to control executive agencies as authoritarian.
The president is constitutionally the head of the executive branch.
Elections are supposed to alter policy.
A bureaucracy completely insulated from presidential direction would itself create a democratic problem.
The real question is therefore which forms of independence are necessary for rule-bound government.
Central-bank independence is relatively strong because monetary policy needs insulation from election cycles.
Prosecutorial independence is partly conventional because presidents appoint attorneys general but are expected not to direct individual criminal cases for personal reasons.
Military independence is operational but not political: generals provide expertise while elected civilians determine strategy.
Civil-service protections prevent mass patronage while still requiring implementation of elected policy.
Different institutions require different balances.
A second Trump administration will test many of them simultaneously.
The Senate could become particularly important because presidential appointments require confirmation for many senior offices. Republicans will control the chamber, but senators still possess individual constitutional authority to reject nominees.
This makes confirmation hearings more than partisan theatre.
They are one of the mechanisms through which the American system tests whether party loyalty completely replaces institutional judgement.
A senator can support Trump’s policy agenda while deciding that one nominee is unsuitable.
Whether enough Republicans are willing to make that distinction will provide information about how independent Congress remains from presidential party leadership.
This is the broader institutional consequence of political polarisation.
When party identity becomes strong enough, constitutional branches stop behaving entirely like separate branches and begin behaving like teams.
A Republican Congress becomes more reluctant to constrain a Republican president.
A Democratic Congress behaves similarly towards a Democratic president.
Political scientists describe this as separation of parties rather than separation of powers.
The Constitution assumes institutional ambition will counter institutional ambition.
Modern parties can weaken that mechanism because legislators care about their party’s president succeeding.
This makes courts and federalism increasingly important as alternative constraints.
States can challenge executive policies.
Judges can review administrative action.
Independent media can expose misconduct.
Civil society can litigate.
The American system contains redundancy precisely because no one safeguard should need to work perfectly.
Yet redundancy works only if actors continue recognising the legitimacy of institutions when they lose before them.
If every adverse court ruling is described as political corruption, judicial authority erodes.
If every bureaucratic objection becomes sabotage, professional administration erodes.
If every congressional investigation becomes partisan warfare, legislative oversight erodes.
Institutional trust can therefore collapse through mutual delegitimation even without any one actor abolishing the Constitution.
This may be the largest risk of Trump’s return.
Not necessarily one dramatic authoritarian seizure.
A gradual reclassification of every independent institution as either loyal or hostile.
Once institutions are evaluated primarily according to whether they help the governing coalition, neutrality becomes
