The countries that chose autocracy democratically

SA men lined the walls of the Kroll Opera House. The Reichstag building had burned on February 27. Communist deputies were absent, having been arrested. Centre Party delegates had decided, after considerable internal deliberation, to vote yes. When the count came: 444 in favor, 94 against — all noes from the Social Democrats, the only party that held the line. The Enabling Act — formally, the Law to Remedy the Distress of the People and the Reich — passed with the required two-thirds supermajority and transferred legislative authority to Hitler’s cabinet for four years.

Nothing in the vote was constitutionally irregular. The Weimar constitution’s two-thirds threshold was met. Every procedural requirement was observed. The Act was, in the specific sense that procedural correctness is what confers democratic legitimacy, the most formally legitimate legislation the Reichstag had ever passed — and the last significant legislation a functioning Reichstag would ever pass.

Hold both of those facts at once. Not as moral judgment about fascism, which is a different essay, but as a procedural puzzle: a vote, conducted according to the rules, produced by parliamentary arithmetic the rules permitted, terminated the parliamentary system that conducted it. The mechanism consumed its own enabling conditions without technically violating them.

Was this the peculiarity of a singular historical moment? Or was it the most compressed, accelerated instance of a pattern that has since been slowed down, distributed across years, and regularized into something that barely looks like the same thing at all?

That question is this article’s subject. The answer is not reassuring.

The wrong mental model

When people imagine democratic collapse, they tend to picture a specific kind of event: tanks in the streets, a general on television, the declaration of martial law. The moment is recognizable. It’s dramatic. There’s a before and an after, and the line between them is unambiguous.

This describes the twentieth century’s dominant mode of autocratization tolerably well. During the Cold War, coups d’état accounted for nearly three out of every four democratic breakdowns worldwide, as Steven Levitsky and Daniel Ziblatt document in How Democracies Die (Crown, 2018). The coup-as-rupture was the normal mechanism, its diagnostic simple: no military intervention, no end of democracy.

The problem is that this metric stopped tracking the thing it was designed to measure. Since roughly 1990, the dominant mode has shifted. The V-Dem Institute’s 2024 Democracy Report — the most comprehensive global dataset on democratic health, produced annually by the University of Gothenburg — found 42 countries in active processes of autocratization, with executive overreach the dominant mechanism. Not military intervention. Elected leaders systematically dismantling the institutions designed to constrain them, using the powers of office to do so.

Elections, in this version of democratic collapse, keep happening. That’s partly the point.

In Hungary, Turkey, and Venezuela, elections continued throughout the periods of deepest institutional erosion. Voter turnout remained substantial. International observers noted irregularities. But irregularities are not coups, and anyone applying the coup model would conclude there was no moment of democratic failure — only the ongoing suspicion that something was happening that didn’t fit the category.

Something was happening, and the coup model has no category for it. The coup model looks for rupture. What these cases share is the meticulous avoidance of it. No decisive moment. No clear before-and-after. Each individual step defensible, the aggregate undeniable only when already irreversible.

The model’s cognitive cost is that it trains observers to ask the wrong question. “Are elections still happening?” is almost always yes. The question that matters — what elections mean when the institutions surrounding them have been systematically reshaped — is the one the coup model is not built to ask.

Competitive authoritarianism

Political scientists Steven Levitsky and Lucan Way introduced the concept of competitive authoritarianism in a 2002 paper in the Journal of Democracy to describe regimes that hold elections while systematically tilting the competition. The concept is related to, but distinct from, democratic backsliding: competitive authoritarian regimes often begin from a point of already-tilted competition — the field was never level — while the cases examined in this article involve leaders who dismantled genuinely democratic institutions they themselves inherited. The distinction matters analytically. The endpoint — elections that register popular preference through a system the government has structured to its own advantage — is similar in both cases.

The sequence

The playbook isn’t secret. It doesn’t need to be. Each step is defensible on its own terms. The problem is visible only in aggregate, and by the time the aggregate is visible, several subsequent steps have already been taken.

Step one: win with sufficient margin. Specifically, win enough to govern without coalition partners who would constrain subsequent steps, and ideally win enough to amend the constitution unilaterally. In Hungary’s April 2010 parliamentary election, Fidesz won 52.7 percent of the popular vote and received 263 of 386 parliamentary seats — a two-thirds supermajority sufficient to amend the Fundamental Law without any other party’s cooperation. The seat share substantially overstated the vote share, a distortion built into the existing electoral rules that Fidesz would subsequently amplify.

Step two: capture the judiciary. Courts are the institution most capable of halting subsequent steps; they go first. In Hungary, the 2011 Fundamental Law lowered the mandatory retirement age for judges from 70 to 62, forcing 274 judges into immediate early retirement — including 4 of 5 appeals court presidents and 20 of 80 Supreme Court judges, per legal scholar Gábor Halmai’s documented analysis. In Venezuela, Chávez’s government expanded the Supreme Court from 20 to 32 members in 2004 and filled the 12 new seats immediately with loyalists. Human Rights Watch documented the result as systematic deference: in every politically significant case from 2004 onward, the packed court declined to rule against the executive, while occasionally issuing minor pro-rights decisions in less visible matters — enough to maintain the appearance of independence, not enough to constitute one.

Step three: capture the media. Regulatory capture comes first, because it’s more procedurally defensible than confiscation. Hungary’s first major legislative act after 2010 restructured the media regulatory framework, placing government-aligned figures in charge of licensing and fines. State advertising — which became the principal revenue source for regional newspapers — was weaponized as an economic lever on editorial independence. Allied private buyers were facilitated. By 2018, 476 media outlets had been consolidated into the Central European Press and Media Foundation (KESMA), run by government-aligned figures, assembled through private commercial transactions that were individually unremarkable; across all government-aligned outlets, approximately 78 percent of available media market resources were in friendly hands. In Turkey, Zaman — at the time the country’s largest-circulation daily newspaper — was seized by an Istanbul court in March 2016 and placed under government administration.

Step four: use the protected legislative majority to change the electoral rules and make the next election harder to lose. Hungary’s 2011 Fundamental Law reduced parliamentary seats from 386 to 199 and introduced a “winner compensation” mechanism, recycling excess votes for winning candidates in single-member constituencies as national list credits — compounding the advantage of the largest party. The result in 2014: Fidesz won 44.87 percent of the national party-list vote and received 133 of 199 parliamentary seats — 66.8 percent of seats from under half the votes.

Step five: win the next election larger. On the tilted field, with a captured judiciary and reshaped media environment, the next election produces a bigger majority. That majority retroactively legitimizes steps one through four. Each subsequent mandate is presented as evidence that the population endorses what happened. In the technical sense that counts votes rather than examining the conditions under which they were cast, this is accurate. In Hungary in 2014, people voted for Fidesz.

The question the mechanism forecloses is not whether people voted. It’s what their vote meant, given what the information environment and the electoral architecture looked like when they made it.

The Hungarian winner compensation system

Hungary's mixed electoral system combines single-member constituency seats, won by plurality, with a national party list. Under the 2011 electoral reform, excess votes for winning candidates in single-member constituencies — votes beyond what was needed to secure a win — are recycled as fractional credits on the national party list. This mechanism is called winner compensation, and it amplifies the electoral advantage of the party that wins the most constituency seats. In 2014, Fidesz won the large majority of constituency seats, and the compensation votes compounded that advantage at the list level, producing the 44.87-to-66.8 percent disparity between vote share and seat share. The technical literature on electoral system design identifies this as a significant democratic anomaly. The relevant point here is that it was achieved through a constitutional process, voted through parliament, and is therefore defensible as procedurally legitimate in each formal respect.

The legality trap

Courts noticed. Suits were filed. The mechanism continued.

The reason is structural, not conspiratorial. Levitsky and Ziblatt draw a distinction in How Democracies Die between the written rules of democratic systems and the unwritten norms that sustain them: what they call mutual toleration — accepting the legitimacy of political opponents as valid participants in democratic life — and institutional forbearance, the restraint political actors exercise in not doing everything the law technically permits. Constitutions do not enforce themselves. They depend on political actors choosing not to use the full legal power they hold. When one actor stops observing forbearance while everyone else continues, the written rules cannot correct this — they were designed on the assumption of universal restraint.

Beyond the norms failure is a timing problem. Legal challenges to any individual step arrive after that step has already produced its institutional effect. Hungary’s Constitutional Court ruled the judicial retirement age change unconstitutional in 2012 — after the 274 judges had already been replaced. The ruling was formally respected: the dismissed judges were nominally offered reinstatement. Fewer than one in five accepted — their positions had already been filled. The institutional composition had changed. The legal ruling arrived one step behind the institutional fact it was adjudicating.

This is not bad luck. It’s the mechanism’s design.

The sequence runs faster than the adjudication timeline. By the time a court rules on step two, step three is complete. By the time step three is challenged, step four has passed. The mechanism doesn’t circumvent the legal system — it uses the legal system’s procedural speed against itself.

And the deeper point goes beyond what Levitsky and Ziblatt frame as constitutional hardball. Democratic institutions derive their authority from procedural legitimacy — from being the body that observes correct procedure and reaches decisions through the right channels. Any institution that is authoritative because it observes procedure can be captured by someone willing to use procedure strategically rather than being constrained by it. The attack surface is not a weakness in the design. It is the design. Procedural legitimacy — democracy’s operating principle — is precisely what the mechanism weaponizes.

Opposition parties in Hungary, Turkey, and Venezuela made correct constitutional arguments against each step. Those arguments were simultaneously true and insufficient, because by the time they were adjudicated, the institution that would have enforced them was no longer the same institution. Saying “this is unconstitutional” while the Constitutional Court is being stacked is accurate. It does not stop the stacking.

The population problem

Legal opposition failed. The harder question is why political opposition failed too.

The comfortable answer is deception: people didn’t understand what was happening, and by the time they did, it was too late. This is not false. It is incomplete. A more honest account looks at the structural conditions that made collective opposition difficult at every stage. The mechanism was engineered specifically to produce this difficulty.

Three factors, interlocking.

By the time the aggregate picture is undeniable, the media environment has already been reshaped to make it hard to circulate. A Hungarian voter relying on domestic television or regional newspapers in 2018 was operating in an information environment where approximately 78 percent of media market resources were government-aligned. The information capture was step three, executed while material conditions were still good enough that institutional concerns felt remote. The sequencing is not incidental. The degraded information environment has to precede the moment when the information would matter most.

And material conditions were, initially, good. Hungary under Orbán recorded consistent GDP growth through much of the 2010s. Venezuela under Chávez saw documented reduction in household poverty rates from 1999 into the mid-2000s, as rising oil revenues funded expanding social programs — a trend that would reverse sharply, but not until the institutional restructuring was already well advanced. The population’s trade-off under these conditions was between concrete and immediate gains — social spending, employment, visible economic growth — and abstract, deferred, institutionally complex goods: independent courts, editorial freedom, the procedural integrity of future elections. That’s not a trade-off that reveals civic failure. It’s a trade-off with a clear structural logic, made more legible still when the information environment has already been shifted to frame institutional concerns as opposition propaganda.

The third factor is ideological, and the Argentine political scientist Guillermo O’Donnell identified it in 1994 as delegative democracy: a governing logic in which the election winner is understood to have received a mandate to govern without institutional constraint — “constrained only by the hard facts of existing power relations and by a constitutionally limited term of office,” in his own formulation. A population that has internalized the delegative premise has already conceded the autocrat’s core argument before he acts on it. Electoral victory is authorization to use available power. The institutions that would constrain that use are positioned, in this frame, not as safeguards but as obstruction.

The mechanism works not because specific populations were uniquely credulous, but because it was engineered to make collective opposition structurally difficult before the moment when opposition would be most needed.

The limits of delegative democracy as a concept

O'Donnell developed his concept of delegative democracy in the specific context of Latin American post-authoritarian transitions of the late 1980s and early 1990s — societies with structurally weak horizontal accountability institutions inherited from preceding autocracies. Whether the concept applies cleanly to Hungary or Turkey — societies with stronger institutional legacies — is genuinely contested in the political science literature. Some scholars argue the framework requires the prior institutional weakness to be applicable; others find the delegative premise operative wherever leaders cultivate a mandate-as-mandate-for-power framing. The concept is used here for its analytical core — the ideological premise that electoral victory is authorization to govern without constraint — not as a claim that Hungary or Turkey matches O'Donnell's original Latin American case set.

The pattern, not the exception

Hungary completed the sequence most precisely. From 2010 to 2014, Orbán’s government worked almost entirely within the constitutional system it inherited, using its rules against itself. The judiciary was challenged; the Constitutional Court partially reversed the retirement age change in formal terms; the institutional composition had already changed. The media consolidation proceeded through commercial transactions. The electoral reform was debated in parliament and voted through by the majority parliament then contained. At no point was any individual step beyond legal contest. In aggregate, the steps were self-protecting.

Turkey’s version was slower, partly because Erdoğan’s AKP was working against genuinely independent institutional constraints — the military and the Kemalist judiciary — that required careful sequencing over approximately fifteen years. The 2010 referendum on constitutional amendments opened the door to court packing. The process compressed dramatically after the failed July 15, 2016 coup attempt, which gave the government emergency decree authority: more than 4,000 judges and prosecutors were dismissed within months. The 2017 presidential referendum then eliminated the parliamentary system itself, replacing it with a strong executive presidency. Turkey’s method was more exposed than Hungary’s, especially post-2016 — the emergency powers made the compression visible — but the constitutional mechanisms granting those powers remained formally intact.

Venezuela shows what the mechanism looks like without its most important feature. Chávez ran the same five-step sequence with markedly less procedural discipline. The 1999 constitutional assembly bypassed the National Congress rather than working through it; the resulting constitutional changes were revolutionary rather than reformatory, explicitly breaking the “technically legal at each step” constraint that makes the mechanism so difficult to oppose in its more refined form. This departure carries analytic weight: Venezuela without procedural discipline produced faster and more complete authoritarianism, recognized as such by international observers far more quickly than Hungary or Turkey. The contrast reveals what the procedural discipline in those other cases is actually doing. It’s not cosmetic cover. It buys deniability — the ability to operate in a legally contested environment, to maintain the democratic façade that makes each individual step so hard to oppose, to generate the legitimate-looking mandate that step five requires. Venezuela stripped of that discipline shows the mechanism without its most insidious feature, which is why it looks so different and functions through the same logic.

The Weimar case is the extreme. The sequence that took Orbán four years compressed into a single legislative session, because prior institutional collapse — economic catastrophe, street violence, normalized political terror, the Centre Party’s calculation that Hitler could be managed from within a coalition — had already done what ordinarily requires years of step-by-step capture. The March 23 vote was formally correct. The conditions under which it occurred made a free vote structurally impossible. This is different from saying it was coerced, but it is incompatible with describing it as an unencumbered democratic choice.

Levitsky and Ziblatt’s comparative framework accounts for exactly this: the same sequence, variation in pace and starting conditions rather than in logic. A government beginning with weaker institutions — Venezuela — moves faster and less carefully. A government beginning with stronger ones — Hungary’s post-communist constitutional structure, Turkey’s Kemalist establishment — must be more deliberate, and more deliberate means more durable. The sequence adapts to starting conditions. The logic does not change.

The Acerbo Law, 1923

The earliest clear modern European instance of step four — changing electoral rules to entrench the leading party — is Mussolini's 1923 electoral reform, named after the deputy who proposed it. The Acerbo Law guaranteed the party receiving the largest share of the vote (provided it exceeded 25 percent) automatic receipt of two-thirds of parliamentary seats. At the 1924 general election, the Fascist-led listone coalition — assembled under conditions of significant intimidation and political violence — received 65.7 percent of the vote. The law was repealed in 1928, by which point the parliamentary system it had been designed to appear to preserve was no longer operational. The Italian case is the mechanism stripped to its logical core: win election, change electoral rules, win next election decisively, then dispense with the machinery of plausible deniability altogether. The progression from step four to electoral theater took five years.

What remains

The empirical record on reversibility is thin.

Fully completed democratic capture — judicial, media, electoral, and informational environments substantially reshaped — has rarely been reversed through purely electoral means. The mechanism is specifically designed to prevent this: once the field is sufficiently tilted, an opposition victory is less likely, and when it occurs, less decisive, because the institutions the new government would need to implement its program remain staffed with the previous government’s loyalists.

Poland’s 2023 election is the best current test case, and what it shows is instructive in ways that don’t resolve the question cleanly. Donald Tusk’s coalition won decisively, ending eight years of Law and Justice rule during which Poland’s courts, prosecution services, and public media were systematically captured through means that repeatedly violated the constitution. The new government pledged institutional repair. What it found was that repair through legal means runs into the same timing asymmetry that made the damage possible in the first place: some reversals require the kind of decisive executive action that resembles the illiberal playbook itself; others require constitutional majorities the government doesn’t hold. Twenty-nine months in, reform is partial, contested, and constrained. A change of government is not a restoration of democratic institutions. The distance between those two outcomes is not small.

But the mechanism contains a fragility its designers didn’t build in. The ongoing requirement to win elections — even on a tilted field — keeps the government accountable to popular performance in ways closed autocracies aren’t. Performing electoral legitimacy requires, at the margin, actually delivering something material to the population whose votes are being sought. When delivery collapses — when oil revenues dry up, when GDP growth stalls, when social spending runs out — the mechanism encounters pressure it was not designed to handle. Venezuela after Chávez: complete institutional capture, every procedural framework in place, and the government still losing control because the collapse of oil revenues destroyed the material foundation that popular acquiescence had depended on.

O’Donnell’s passive audience can stop cheering.

Whether that’s sufficient to reverse the institutional damage — or only sufficient to replace the personnel within a permanently degraded institutional structure — is the question the empirical record has not yet answered cleanly. Poland suggests it’s the harder of those two problems. The mechanism’s real durability is not in the specific laws it passes or the specific courts it reshapes. It’s in the gap between winning an election and restoring what the previous government spent years methodically dismantling.

The Kroll Opera House vote didn’t begin the sequence. It completed it. The prior institutional degradation, the normalization of constitutional hardball, the parties’ calculations about available alternatives — all of that preceded March 23. What happened that afternoon was the endpoint of a process already largely done, not the first step of one about to begin. Which is why it’s the wrong place to look for the warning.

Locate the equivalent endpoint in Hungary, and it doesn’t look like one. It looks like a media regulatory restructuring. Or a constitutional amendment debated in parliament. Or a judicial retirement provision that the Constitutional Court will later review. In Turkey, the endpoint looks like a referendum, or an emergency decree, each with its own legal instrument, each individually defensible in its own formal terms. The mechanism requires only legibility in retrospect to function. Sufficient procedural cover — slow enough, distributed across enough years, individually defensible enough at each step — means the aggregate becomes undeniable only at the point where the institutions that would have addressed it have already been reshaped.

The V-Dem Institute counted 42 countries in active autocratization in 2023. Not coups, not revolutions. Elected governments running the same sequence at different speeds, with local adaptations.

What follows from knowing the sequence in advance is not this article’s claim to settle. But that something follows — that a recognizable mechanism, running in real time, creates a pressure that recognition alone cannot discharge — is what the argument has been building toward since the first sentence. That pressure doesn’t resolve here. The mechanism was designed to ensure it wouldn’t.

Zastrzeżenie dotyczące Gen AI

Niektóre treści tej strony zostały wygenerowane i/lub edytowane przy pomocy generatywnej sztucznej inteligencji.

Media

Beyazıt Meydanı önünde üniversite öğrencilerinin başlattığı prostestolar. – Wikipedia

Kluczowe źródła i odniesienia

Steven Levitsky and Daniel Ziblatt, How Democracies Die, Crown, 2018

V-Dem Institute, Democracy Report 2024: Democracy Winning and Losing at the Ballot, University of Gothenburg, 2024. Available at: https://www.v-dem.net/documents/43/v-dem_dr2024_lowres.pdf

Gábor Halmai, “The Early Retirement Age of the Hungarian Judges,” SSRN Working Paper, 2017. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2985219

Human Rights Watch, “Venezuela: Chávez Allies Pack Supreme Court,” December 13, 2004. Available at: https://www.hrw.org/news/2004/12/13/venezuela-chavez-allies-pack-supreme-court

Human Rights Watch, “Rigging the Rule of Law: Judicial Independence Under Siege in Venezuela,” June 16, 2004. Available at: https://www.hrw.org/report/2004/06/16/rigging-rule-law/judicial-independence-under-siege-venezuela

Human Rights Watch, “A Decade Under Chávez: Political Intolerance and Lost Opportunities for Advancing Human Rights in Venezuela,” September 18, 2008. Available at: https://www.hrw.org/report/2008/09/18/decade-under-chavez/political-intolerance-and-lost-opportunities-advancing-human

Guillermo O’Donnell, “Delegative Democracy,” Journal of Democracy, Vol. 5, No. 1, January 1994, pp. 55–69. Available at: https://www.journalofdemocracy.org/articles/delegative-democracy/

Steven Levitsky and Lucan A. Way, “Elections Without Democracy: The Rise of Competitive Authoritarianism,” Journal of Democracy, Vol. 13, No. 2, April 2002, pp. 51–65

Ágnes Urbán / Mérték Media Monitor, “Fidesz-friendly media dominate everywhere,” May 2, 2019. Available at: https://mertek.eu/en/2019/05/02/fidesz-friendly-media-dominate-everywhere/

Human Rights Watch, “Turkey: Judges, Prosecutors Unfairly Jailed,” August 5, 2016. Available at: https://www.hrw.org/news/2016/08/05/turkey-judges-prosecutors-unfairly-jailed

Stockholm Center for Freedom, “Only 95 of 4,156 Judges and Prosecutors Fired After Coup in Turkey Reinstated: Report.” Available at: https://stockholmcf.org/only-95-of-4156-judges-and-prosecutors-fired-after-coup-in-turkey-reinstated-report/

International IDEA, “Is Poland’s Democratic Backsliding Over? History Shows It Takes More Than an Election,” December 13, 2023. Available at: https://www.idea.int/blog/polands-democratic-backsliding-over-history-shows-it-takes-more-election