How Democracies Erode, and How They Hold

Follows: How Political Change Actually Happens

In this deep dive
1. From tanks to paperwork: how breakdowns changed
2. How democracy gets measured
3. Can we trust the measurements?
4. The guardrails: norms that never appear in a constitution
5. Weimar Germany, 1920–1933: survived a putsch, then dismantled legally
6. Chile, 1970–1973: when both sides stop trusting the rules
7. Venezuela, 1989–2009: an elected path
8. How democracies hold

A note on approach: this piece describes what comparative-politics research has found, using measurable indicators and historical cases at least twenty years old. It names no current parties and judges no living leaders. Where researchers disagree, including about whether a global decline is happening at all, both sides are given. The measures described here are the judgments of the organisations that make them, and Part 3 explains why those judgments are themselves debated.

On 23 March 1933, Germany's parliament voted 444 to 94 to give the government the power to make laws without it. The vote met the constitution's requirement of a two-thirds majority. The law ran for four years and was renewed three times. It stayed the legal basis for German lawmaking until the Nazi regime collapsed in May 1945.

The vote was not free. Every Communist deputy had already been barred from the chamber, and the government had suspended civil liberties four weeks earlier. But the form mattered: Germany's democracy was ended by a vote, not stormed by soldiers. The pattern it set, using a democracy's own procedures to dismantle it, has become the most common way democracies fail.

How Political Change Actually Happens described how movements and institutions change societies, usually slowly and through many channels. This piece looks at the same machinery running in reverse: how democracies lose their democratic qualities, how researchers measure that, what the historical cases show, and what the evidence says about the democracies that survive.

Part 1 — From tanks to paperwork: how breakdowns changed

For most of the twentieth century, a democracy usually died in a recognisable way: soldiers took the radio station, the president fled or was killed, parliament was dissolved. You knew the date.

In 2016 the political scientist Nancy Bermeo published an article in the Journal of Democracy that reorganised how researchers think about this. "On Democratic Backsliding" defined backsliding as the state-led weakening or elimination of the institutions that sustain a democracy, and argued that its forms had changed dramatically since the end of the Cold War.

The key word is legitimated. Bermeo argued that today's forms of backsliding are especially hard to resist because they are carried out through the very institutions democracy promoters care about: elections, courts, parliaments and referendums. Each step can be defended as legal, popular or both.

Bermeo's conclusion was more hopeful than the headlines it inspired. She argued that backsliding today reflects democracy's spread rather than its retreat (would-be autocrats now need democratic cover), that the gradual forms are easier to reverse, and that the regimes they produce tend to be shorter-lived and less repressive than old-style dictatorships.

A coup happens on a date. Executive aggrandizement happens across a hundred dates, each of which can be defended as normal politics.

Researchers at the Varieties of Democracy (V-Dem) project in Sweden reached a similar conclusion from a different direction. In a 2019 article, Anna Lührmann and Staffan Lindberg mapped every "episode of autocratization" since 1900 and announced "a third wave of autocratization," mainly affecting democracies through "gradual setbacks under a legal façade." They also cautioned that "panic is not warranted": at that point, they wrote, the declines were relatively mild and the share of democracies was still close to its all-time high.

What does executive aggrandizement look like in practice? V-Dem's 2026 report, looking at the 44 countries it classified as autocratizing in 2025, found that the most common tactic was media censorship, used in 32 of them (73 percent), followed by repression of civil society organisations, in 30 (68 percent). Many of the other steps researchers document are aimed at the institutions that act as referees: courts, election commissions, prosecutors, public broadcasters, auditors. None of these steps needs to be illegal. A court can be enlarged by statute, a broadcaster's board replaced by a lawful vote, a prosecutor's office restructured by decree. Each step on its own looks like the ordinary business of a government with a mandate. What matters is the cumulative direction: whether, at the end, the people in power can still be checked, criticised and voted out.

That's why researchers now talk about episodes rather than events. An "episode of autocratization," in V-Dem's usage, is a sustained and substantial decline in democratic qualities over several years, identified after the fact from the data, not a single date.

This creates a measurement problem. If democracy now erodes through many small, legal steps, then there's no single event to count. Someone has to judge, year by year, whether elections were fair, courts independent and journalists free. That is what the large democracy indices try to do.

Part 2 — How democracy gets measured

Two projects dominate the field. They use very different methods, and both are cited in news reports almost every week.

Freedom House, a US-based organisation founded in 1941, has published Freedom in the World since its research programme began in 1972 (the report first appeared in book form in 1978). Its current method, as described on its methodology page, works like this:

• Every country and territory is scored on 25 indicators: 10 for political rights (elections, pluralism, how government functions) and 15 for civil liberties (expression, association, rule of law, personal rights). The 2026 edition covers 195 countries and 13 territories.
• Each indicator scores 0 to 4, giving a maximum of 40 for political rights and 60 for civil liberties, 100 in total. An extra "discretionary" question can subtract up to 4 points where a government deliberately changes a region's ethnic composition.
• Analysts draft the scores, which are then reviewed at meetings with outside advisers and staff; Freedom House says the end product represents their consensus. The 2026 edition involved 134 analysts and 28 advisers.
• Countries are labelled Free, Partly Free or Not Free. Freedom House stresses that it scores what happens in practice, not what the laws say.

V-Dem, based at the University of Gothenburg in Sweden, takes a different approach:

• It measures more than 600 indicators for every country, going back to 1789.
• Most indicators are judgments by more than 4,000 country experts, usually several per country per question. V-Dem reports that 61 percent of its coders were born in or live in the main country they code.
• A statistical measurement model combines the experts' answers into a single scale and reports each score with a range of uncertainty rather than a single point.
• It splits democracy into five principles: electoral, liberal, participatory, deliberative and egalitarian. Its flagship Liberal Democracy Index (LDI) combines the electoral and liberal principles: free and fair elections plus limits on the power of those elected.

What they currently say

Both projects report a long global decline, with different framings. As of early 2026:

• Freedom House (Freedom in the World 2026): global freedom "declined for the 20th consecutive year in 2025," with 54 countries registering declines and 35 improvements.
• V-Dem (Democracy Report 2026, covering 2025): the level of democracy enjoyed by the average person in the world is "back to 1978 levels"; it counts 92 autocracies and 87 democracies, with 44 countries autocratizing and 18 democratizing.

A worked example: Hungary through the indicators

Hungary is the case researchers most often use to illustrate gradual erosion inside the European Union. Here is what the indicators themselves say, without commentary:

V-Dem's report attributes the change to the gradual removal of checks on the executive, interference with the quality of elections and restrictions on press freedom. Freedom House's 2019 report gave its reason for the change in status as "sustained attacks on the country's democratic institutions." These are the judgments of the measuring organisations, and they are exactly the kind of expert judgment whose reliability is debated in the next part.

Part 3 — Can we trust the measurements?

The indices are indispensable, and they are also contested. A careful reader should know the main criticisms, because some of them come from the people who build the indices.

Expert judgment carries bias. Long before the current debate, the sociologists Kenneth Bollen and Pamela Paxton (2000) tested the democracy ratings of the 1970s and 1980s and found "unambiguous evidence of judge-specific measurement errors," errors related to traits of the countries being judged. Gerardo Munck and Jay Verkuilen (2002) published an influential critique of how the major indices define and combine their components. The problem isn't that the raters are dishonest; it's that "how free is this country's press?" has no thermometer.

Who does the rating matters. Freedom House is based in the United States, and researchers have asked whether that shapes its scores. The political scientist Nils Steiner tested whether US allies are rated as more democratic by Freedom House (published 2016). Sarah Bush (2017) studied how the ratings gained authority, using Freedom House's internal archive, interviews and a statistical analysis of bias, and argued that their influence rests partly on "ideological affinity" between the raters and the people who use the ratings.

The objective-indicator challenge. The sharpest recent challenge came from Andrew Little and Anne Meng, published in PS: Political Science & Politics in 2024. They looked at "objective" indicators, measurable facts such as how often incumbents lose elections, and found "little evidence of global democratic decline during the past decade." Their explanation for the gap: either the experts' standards have shifted over time, or leaders are using subtler methods that only expert judgment can detect. They concluded that they could not rule out the second explanation, but that the claim of major global backsliding "is not justified by existing evidence."

The reply. In the same journal issue, ten V-Dem researchers led by Carl Henrik Knutsen replied, engaging both with the broader debate about measuring democracy and with Little and Meng's specific data and conclusions, and defending the use of expert judgment. The underlying point in the defenders' favour is real: much of what gradual erosion involves, such as pressure on judges or self-censorship by journalists, leaves few traces in election results. This is a genuine, unresolved scholarly debate.

The historical-perspective challenge. The political scientist Daniel Treisman (2023), using several measures, argued that "the global proportion of democracies remains close to an all-time high," that the current rate of decline "is not historically unusual," and that it's largely explained by the lower incomes and newer institutions of countries that democratised in the 1980s and 1990s. "The fear of a global slide into autocracy," he concluded, "appears premature."

Every democracy index is a set of informed human judgments. That makes it useful, not useless. It also means its numbers deserve the same scrutiny as any other expert opinion.

The index-makers themselves acknowledge much of this. V-Dem publishes every score with an uncertainty range rather than a single point, and a country's range in a given year can be wide. Freedom House stresses that its scores reflect what happens in practice rather than what the law says, which is exactly what makes them both valuable and dependent on judgment. Neither project claims to be a thermometer.

What should a careful reader take from this? Three things. First, the indices broadly agree on which countries changed most and in which direction; the debate is mostly over the size and universality of the global trend. Second, sudden, dramatic changes (coups, cancelled elections) are easy to measure, and gradual ones are hard, which is exactly why the gradual path is attractive to anyone who wants to avoid being measured. Third, it's worth checking several indices before trusting a strong claim from any one of them.

Part 4 — The guardrails: norms that never appear in a constitution

If erosion happens legally, then laws alone can't prevent it. That's the central argument of How Democracies Die (2018) by the Harvard political scientists Steven Levitsky and Daniel Ziblatt, the most widely read book on the subject.

Their claim is that working democracies depend on two unwritten norms, which they call democracy's guardrails:

1. Mutual toleration: treating your rivals as legitimate competitors who have an equal right to exist, compete and govern if they win, not as enemies or traitors.
2. Institutional forbearance: not using your legal powers to their full extent when doing so would violate the spirit of the system. Their definition is "avoiding actions that, while respecting the letter of the law, obviously violate its spirit."

Forbearance is the subtler idea, and the more important one. Almost every constitution gives officials powers that would wreck the system if used aggressively: the power to pardon, to delay appointments, to pack courts by adding seats, to call emergencies, to block every bill. Democracies function because those powers are mostly held back. When one side stops holding back, the other usually retaliates, and the cycle can escalate until the rules themselves become weapons.

Some generic examples of the difference between a legal power and a forbearing use of it:

• A legislature may have the power to change the size of the highest court. Forbearance means not doing it to create a friendly majority.
• An executive may have the power to declare an emergency. Forbearance means reserving it for real emergencies, and ending it when they end.
• A majority may have the power to change electoral rules. Forbearance means not redrawing them shortly before an election to keep itself in office.
• Officials may have the power to investigate anyone. Forbearance means not aiming that power at political opponents and critical journalists.

In each case the rule-breaker can say, truthfully, that nothing illegal happened. That is the point: the guardrails are needed precisely where the law runs out.

Levitsky and Ziblatt also emphasise gatekeepers: the people and institutions, traditionally party leaders, who keep extremists from reaching power in the first place. And they offer four warning signs of authoritarian behaviour in a politician:

The strongest case for the guardrails framework: it explains why democracies with very different constitutions survive or fail. Written rules are always incomplete; what keeps them working is the shared willingness to lose by them.

The main criticisms, raised by reviewers and other scholars: norms are hard to measure, so the framework can be applied after the fact to explain almost anything. "Forbearance" can also protect unjust arrangements: some norms of restraint historically rested on excluding whole groups from politics, and breaking them was part of democratisation. And the framework can be used by every side to accuse every other side, since nearly everyone believes their opponents broke the norms first.

Are voters a check?

A natural hope is that voters will punish politicians who break democratic rules. The evidence is sobering. In a 2020 study in the American Political Science Review, Matthew Graham and Milan Svolik ran a large candidate-choice experiment in the United States and found that "only a small fraction" of voters prioritised democratic principles when those principles conflicted with partisanship or policy, and that the fraction shrank as polarisation rose. The study is of one country, but its logic travels: when voters see the other side as an existential threat, they forgive their own side a great deal.

Part 5 — Weimar Germany, 1920–1933: survived a putsch, then dismantled legally

Weimar is the case everyone cites, and it's worth knowing that it is really two stories.

The first story is resilience. In March 1920, a group of right-wing officers and politicians tried to seize power in Berlin, an attempt known as the Kapp-Lüttwitz Putsch. The response was a general strike beginning on 15 March. According to the German Historical Museum, around 12 million workers stopped work, and the civil servants' association joined the unions in backing the republic. Without water, gas, electricity or the cooperation of the state's own employees, the putschists could not govern. They gave up on 17 March. A democracy barely two years old had defeated a direct attack.

The second story is erosion. By the early 1930s, in the depths of the Depression, Germany's parliament was deadlocked, and governments increasingly ruled by presidential emergency decree under Article 48 of the constitution. That article let the president take "measures necessary" to restore public order and suspend civil liberties. President Paul von Hindenburg invoked it sixty times in 1932 alone, according to the United States Holocaust Memorial Museum.

A sharper test came on 20 July 1932. By emergency decree, Hindenburg removed the government of Prussia, Germany's largest state and a stronghold of the democratic parties, and installed the chancellor, Franz von Papen, as Reich commissioner in its place. The stated reason was street violence. This time there was no general strike. The German Historical Museum's account explains why: with mass unemployment, the Social Democrats and the unions judged a strike pointless, and with the army backing the move, resistance by the Prussian police looked hopeless. They put their hopes in the national election eleven days later instead. The Prussian government sued, and in October 1932 the state court in Leipzig largely upheld the commissioners' appointment. Control over Prussia, and especially its police, would greatly ease the creation of the dictatorship a few months later.

Meanwhile, support for the Nazi party rose fast:

Two things in that table are easy to miss. The Nazis never won a majority in a free election. And when Hitler was appointed chancellor, on 30 January 1933, their vote had just fallen. He came to office not by sweeping an election but through a deal. The conservative politician Franz von Papen and others close to Hindenburg persuaded the reluctant president to appoint him, believing, as the Holocaust Memorial Museum puts it, that they could control him and limit his power.

What followed took less than two months:

1. 27 February 1933: the Reichstag building burns.
2. 28 February: the Reichstag Fire Decree, issued under Article 48, suspends freedom of expression and of the press, the rights of assembly and association, and the privacy of post and telephone, and allows searches without warrants.
3. 8 March: the 81 Communist deputies' mandates are annulled; many are held in "protective detention."
4. 23 March: the Enabling Act passes 444 to 94. The Nazis, the conservative nationalists, the Catholic Centre Party, the Bavarian People's Party and the German State Party vote yes. Only the 94 Social Democrats present vote no.

In 1920, a putsch failed because millions refused to cooperate. In 1933, the decisive steps were votes and decrees, each with a legal form.

What the Weimar case suggests to researchers is less "democracies are fragile" than something more specific: emergency powers built for crises can become the road out of democracy; established elites who think they can use an extremist movement can badly misjudge it; and a society's capacity to resist a frontal assault is no guarantee it will recognise a legal one.

Part 6 — Chile, 1970–1973: when both sides stop trusting the rules

Chile had one of Latin America's longest records of constitutional government. Its breakdown is studied because it shows something different from Weimar: a crisis in which both government and opposition came to doubt the other's commitment to the rules.

The election. In September 1970 the socialist Salvador Allende, leading the Popular Unity coalition, won a three-way presidential race with a plurality of about 36.6 percent, narrowly ahead of the conservative Jorge Alessandri. Because no one had a majority, Congress had to choose. Allende was confirmed on 24 October by 153 votes to 35, after agreeing to a "Statute of Constitutional Guarantees" in which he committed to respect the constitution.

Foreign interference. The United States government had worked to stop him taking office. A CIA report released in 2000 under a law requiring disclosure acknowledged that the agency paid $35,000 to a group of coup plotters after they murdered the army's commander-in-chief, General René Schneider, in October 1970, a general who had opposed a coup. The same body of declassified records documents US covert intervention in Chile running from 1962 to 1975. Historians still debate how decisive outside pressure was compared with Chile's internal crisis; nearly all agree it was not the whole story.

Polarisation. Allende's government pursued rapid nationalisation and land reform, and the economy deteriorated badly, with soaring inflation. Supporters and critics alike describe the period as intensely polarised: on the radical edges of both camps, some wanted to move beyond legal limits, whether towards faster revolutionary change or towards military intervention, and each side increasingly described the other as an existential threat.

The resolution. On 23 August 1973, the Chamber of Deputies passed a resolution, by 81 votes to 47, accusing the government of repeatedly violating the constitution and the law. It had no direct legal force, and the opposition lacked the two-thirds majority needed in the Senate to remove the president constitutionally. But it was widely read as giving the armed forces political cover.

The coup. On 11 September 1973, the military, led by General Augusto Pinochet, overthrew the government. Allende died in the presidential palace that day; his death is generally accepted as suicide. The military ruled until 1990. Chile's official truth commission of 1991 (the Rettig commission) documented more than 2,000 people killed or disappeared, and a later commission (Valech, 2004) recognised tens of thousands of victims of political imprisonment, most of whom reported torture. The CIA's 2000 report also acknowledged that the agency "actively supported the military Junta after the overthrow of Allende."

In Weimar, one side used the rules to destroy them. In Chile, each side came to believe the other already had, and that belief helped make it true.

How people interpret 1973 still divides along political lines, in Chile and elsewhere. One interpretation emphasises a constitutional government overthrown by its military, with foreign support, followed by long and brutal repression. Another emphasises a government that, its critics argued, was itself straining or breaking constitutional limits amid economic collapse. Comparative-politics researchers have mostly treated it as a case of mutual loss of trust in democratic rules, the opposite of Levitsky and Ziblatt's mutual toleration. Those interpretations aren't mutually exclusive, and none of them excuses what the dictatorship did afterwards.

For comparative researchers, the Chilean case is a warning about a specific dynamic rather than about any one side. When each camp becomes convinced that the other will not accept defeat, each starts to treat the rules as optional for itself, as a kind of self-defence. Graham and Svolik's experimental finding from Part 4, that partisans forgive their own side's democratic violations when polarisation is high, describes the same logic at the level of voters. Chile's institutions were old and, by regional standards, strong. They still could not hold once enough people on both sides stopped believing they would be honoured.

The story also has an end that matters for Part 8: the military regime lost a plebiscite in 1988 and handed power to an elected civilian government in 1990. Democracies can come back.

Part 7 — Venezuela, 1989–2009: an elected path

Venezuela is the case most often used to illustrate executive aggrandizement in Latin America. It's also a case where the government's supporters and its critics tell sharply different stories, and both stories contain real facts.

The background. After the fall of a military dictatorship, Venezuela's main parties signed the Punto Fijo Pact on 31 October 1958, agreeing to accept election results and share power. For decades two parties dominated politics, funded by oil wealth. By the 1980s, falling oil revenues had badly eroded public trust in both main parties. In 1989, sudden price rises for fuel and transport set off riots in Caracas, the Caracazo, put down violently.

The rise. On 4 February 1992, a young army officer, Hugo Chávez, led a failed coup attempt against the elected government. He was jailed, and later turned to electoral politics. In December 1998 he won the presidency with about 56 percent of the vote, campaigning against the old two-party system.

The new constitution. In April 1999, voters approved electing a constituent assembly, though turnout was under 40 percent. Chávez's supporters won about 95 percent of its seats with a much smaller share of the vote (sources put it between about half and two-thirds, depending on how the multi-vote ballots are counted), reflecting the electoral rules used. A new constitution was approved by referendum in December 1999. Human Rights Watch, a critic of the government, nonetheless noted that this constitution significantly expanded human-rights guarantees. The assembly also declared a judicial emergency and removed hundreds of judges.

The confrontation. In April 2002, a coup briefly removed Chávez. The interim leader dissolved democratic institutions, and the coup collapsed within about 40 hours. A two-month oil strike followed in late 2002 and early 2003, after which close to half the state oil company's workforce was fired. In August 2004, the opposition forced a recall referendum; Chávez won it with about 59 percent, and observers from the Carter Center and the Organization of American States found no evidence of fraud.

The institutional changes. This is where executive aggrandizement shows up most clearly in the record, as documented by Human Rights Watch's 2008 report A Decade Under Chávez:

The supporters' case, as they would make it: these governments won election after election, enfranchised millions of poor Venezuelans whom the old system had ignored, funded large social programmes from oil revenue, accepted the defeat of 2007, and faced an opposition that had backed a coup and an economy-crippling strike. World Bank figures show the national poverty rate falling from 54 percent in 2003 to about 27 percent in 2007 during the oil boom, and inequality falling too, though poverty was rising again by the mid-2010s.

The critics' case: each step was legal or popular, but together they removed the independent courts, broadcasters and term limits that let voters change their minds safely. Winning elections, on this view, is necessary for democracy but not sufficient.

Both cases describe the same events. That's what makes executive aggrandizement hard to call in real time: the warning signs are visible mainly in retrospect, and the people living through it can reasonably disagree about what they're seeing.

Part 8 — How democracies hold

After four parts on failure, the counter-evidence matters, and there's a good deal of it.

Most democracies survive most of the time. A 2021 study by V-Dem researchers Vanessa Boese, Amanda Edgell, Sebastian Hellmeier, Seraphine Maerz and Staffan Lindberg ("How democracies prevail") treated resilience as a two-stage process: avoiding decline at all, and avoiding breakdown once decline has started. They found that democracies have had "a high level of resilience" to the start of autocratization since 1900. The bad news is in the second stage. Resilience "has become substantially weaker" since the Cold War: they counted 59 episodes of substantial decline since 1993, leading to the breakdown of 36 democratic regimes, and found that once autocratization begins, "only one in five democracies manage to avert breakdown."

Courts matter most. In the same study, the factor most associated with resilience was strong judicial constraints on the executive, together with a history of strong democratic institutions. Economic development helped democracies avoid starting to decline but not stop a decline already under way. That's consistent with a classic finding: Adam Przeworski and Fernando Limongi (1997), studying 135 countries from 1950 to 1990, concluded that wealth doesn't cause countries to become democracies but does make democracies more stable once they exist: "democracies survive in countries that are modern."

Erosion can be reversed. V-Dem now tracks U-turns, cases where autocratization is stopped and reversed within about five years. Citing a 2025 study by Marina Nord and colleagues, its 2026 report says roughly 70 percent of all autocratization episodes in the current "third wave" have ended in U-turns. Of the ten U-turns under way in 2025, five avoided a breakdown of democracy altogether, and three quickly restored democracy after one. Bermeo's 2016 argument that the gradual forms of backsliding are easier to reverse than coups is, so far, broadly supported by these cases.

Erosion is common. Collapse is not inevitable. The strongest predictor of which way a democracy goes is whether its courts and other independent institutions can still say no.

What tends to stop erosion? Across these studies and the historical cases, a few factors recur:

1. Independent courts that rule against the government and are obeyed. The single strongest correlate of resilience in the V-Dem research.
2. Elections that can still be lost. V-Dem's summary of the U-turn research finds that elections were often the pivotal moment, that the first electoral cycle was often decisive, and that a unified opposition, an active civil society and independent media mattered. Acting early is part of the pattern.
3. Mass, broad and nonviolent public pressure. Weimar in 1920 is the classic case. How Political Change Actually Happens described the evidence that broad, nonviolent movements succeed more often than violent ones, and the same logic seems to apply to defending democracy as to winning reforms.
4. Refusals from inside. Civil servants, judges, officers and members of the governing camp who decline to go along. In 1920 Germany, the civil servants' association sided with the strikers against the putsch.
5. Restraint by the opposition. Levitsky and Ziblatt argue that answering norm-breaking with norm-breaking tends to accelerate the cycle. This is the hardest lesson to apply, because it asks people who feel their democracy is under attack to keep to rules their opponents may be ignoring.

What remains open

• How big is the global decline? The debate between expert-coded indices and objective indicators (Part 3) is unresolved, and it matters: the policy response to a world-historical slide is different from the response to a normal fluctuation.
• Where's the line? There is no agreed threshold at which aggressive but legal politics becomes executive aggrandizement. That's partly why the same events are described so differently by supporters and critics.
• Do the guardrails cut both ways? Some norms of restraint protected unjust arrangements, and some were broken in the name of extending democracy. Telling a dangerous norm violation from a necessary one is a judgment, not a measurement.

The takeaway

Democracies today rarely die by coup. More often they erode through executive aggrandizement: elected governments weakening courts, media and elections one legal step at a time (Bermeo 2016). Researchers measure this with expert-based indices such as V-Dem and Freedom House, which both report long global declines, but critics argue that objective indicators show much less change, and that expert judgment has its own biases. Levitsky and Ziblatt's guardrails, mutual toleration and institutional forbearance, help explain why written rules alone don't protect a democracy. Weimar, Chile and Venezuela show three different paths: legal dismantling after an elite deal, a breakdown of mutual trust ending in a coup, and gradual change through elections and referendums. The counter-evidence matters too: most democracies resist the start of decline, many declines are reversed, and independent courts are the strongest predictor of which ones hold.

Try this: pick any country you're interested in, your own included, and look it up in both V-Dem's latest Democracy Report and Freedom House's Freedom in the World. Do they agree? Then read the methodology pages and find one indicator where you think the score could reasonably be argued either way. You'll understand the debate better than most commentators.

Sources & Further Reading

Sources
- Nancy Bermeo, "On Democratic Backsliding," Journal of Democracy 27, no. 1 (2016): 5–19 (typology and conclusions from the published abstract).
- Anna Lührmann & Staffan I. Lindberg, "A third wave of autocratization is here: what is new about it?" Democratization 26, no. 7 (2019).
- Freedom House, Freedom in the World research methodology (2026 edition: 195 countries, 13 territories; 25 indicators; 134 analysts, 28 advisers); Freedom in the World 2026 (20th consecutive year of decline; 54 declines, 35 improvements); Hungary country reports, 2017, 2019 and 2025 editions; Nations in Transit 2020.
- V-Dem Institute, "The V-Dem Project" and methodology pages (600+ indicators, 1789–present, 4,000+ experts); Democracy Report 2026 (average citizen back to 1978 levels; 92 autocracies, 87 democracies; 44 autocratizing, 18 democratizing; Hungary episode and LDI change; ten U-turns).
- Kenneth Bollen & Pamela Paxton, "Subjective Measures of Liberal Democracy," Comparative Political Studies 33, no. 1 (2000); Gerardo Munck & Jay Verkuilen, "Conceptualizing and Measuring Democracy," Comparative Political Studies 35, no. 1 (2002); Nils D. Steiner, Journal of Comparative Policy Analysis 18 (2016); Sarah Sunn Bush, "The Politics of Rating Freedom," Perspectives on Politics 15, no. 3 (2017).
- Andrew Little & Anne Meng, "Measuring Democratic Backsliding," PS: Political Science & Politics 57, no. 2 (2024): 149–161; Carl Henrik Knutsen et al., "Conceptual and Measurement Issues in Assessing Democratic Backsliding," same issue, 162–177; Daniel Treisman, "How Great is the Current Danger to Democracy?" Comparative Political Studies 56, no. 12 (2023).
- Steven Levitsky & Daniel Ziblatt, How Democracies Die (Crown, 2018); summary of the four indicators and the norms from published reviews. Juan Linz, The Breakdown of Democratic Regimes (Johns Hopkins, 1978).
- Matthew Graham & Milan Svolik, "Democracy in America?" American Political Science Review 114, no. 2 (2020).
- Weimar: German Historical Museum (LeMO) on the 1920 general strike and the Enabling Act vote; US Holocaust Memorial Museum on Article 48, the Nazi rise to power (vote shares; Papen's role) and the Reichstag Fire Decree.
- Chile: 1970 election and congressional vote, the August 1973 Chamber resolution and the 1988–1990 transition as summarised in standard references; CIA, CIA Activities in Chile (2000), via the National Security Archive; Rettig (1991) and Valech (2004) commissions.
- Venezuela: Human Rights Watch, A Decade Under Chávez (2008); World Bank, national poverty headcount (SI.POV.NAHC: 54.0% in 2003, 27.5% in 2007, 33.1% in 2015) and Gini index (48.9 in 2002, 44.7 in 2006); 1998, 1999, 2004 and 2009 electoral results and the Carter Center and OAS assessments of 2004 as summarised in standard references.
- Marina Nord et al., "When Autocratization is Reversed: Episodes of U-Turns since 1900," Democratization 32, no. 5 (2025), as summarised in V-Dem's Democracy Report 2026.
- Vanessa A. Boese et al., "How democracies prevail: democratic resilience as a two-stage process," Democratization 28, no. 5 (2021); Adam Przeworski & Fernando Limongi, "Modernization: Theories and Facts," World Politics 49, no. 2 (1997).

Go deeper
- 🌐 V-Dem — Democracy Reports (the annual reports, free, with country rankings and charts; read the methodology boxes too)
- 🌐 Freedom House — Freedom in the World methodology (all 25 indicator questions, so you can judge the scoring yourself)
- 🌐 Nancy Bermeo — "On Democratic Backsliding," Journal of Democracy (the article that named executive aggrandizement)
- 🌐 Boese et al. — "How democracies prevail" (open access) (the best single study of resilience, free to read)
- 📘 How Democracies Die — Steven Levitsky & Daniel Ziblatt (the guardrails argument, with historical cases from Europe and Latin America)
- 📘 The Breakdown of Democratic Regimes — Juan Linz (the classic comparative study that most later work builds on; academic but foundational)
- 🌐 Companion pieces in this library: How Political Change Actually Happens and Rights, Constitutions & the Rule of Law