In July 2022 a prime minister in the United Kingdom was forced out in a matter of days, because his own party stopped supporting him. In the same year a president in Brazil, whose term was fixed by the constitution, could be removed only by impeachment or an election. A president in France could dissolve the National Assembly; a chancellor in Germany could not dissolve the Bundestag at will. All four countries are democracies. They are not governed the same way, and the difference decides what a crisis looks like.
Open any newspaper and the stories are about people: who won, who resigned, who is blocking whom. A political scientist sees the same stories as outputs of rules. Change who can remove the leader and the same leader behaves differently. This piece is about those rules, and about the older question underneath them: why do some countries manage to keep a democracy for generations while others lose it?
This describes how political systems work. It does not take sides in any country's current politics, and the regime labels below are tools for thinking, not verdicts.
Before asking how a country is governed, ask whether it is governed at all. Max Weber's definition, from his 1919 lecture "Politics as a Vocation", is the one most courses start with: a state is the organisation that successfully claims a monopoly on the legitimate use of physical force in a territory. Two words carry the weight. Successfully: a decree nobody can enforce does not make a state. Legitimate: a gang that holds the same streets by threat is not one, because people do not accept its right to.
That gives the first idea in the field, state capacity: the ability of a state to collect taxes, enforce laws, deliver services and keep order across its whole territory. It is a matter of degree. A state can be tyrannical and weak (it can jail critics in the capital but cannot collect taxes in the provinces), or mild and strong (Denmark). Charles Tilly's famous line, "war made the state and the state made war", sums up one influential story of where capacity came from in Europe: rulers who needed money to fight built tax collectors, armies and records, and those became the state. Scholars of Latin America (Miguel Centeno, for one) and of Africa point out that the path was different there, so this story does not explain every region.
Why does this come first? Because every other contrast, democracy against dictatorship, federal against unitary, is about how power is used, and a state that cannot reach its own territory has little power to organise. Political scientists also warn that a democracy needs a state to be a democracy: elections cannot be run, counted and enforced where nobody can guarantee safety at the polling station.
Every democracy has to answer one design question: how does the government get its power, and what can end it? There are three answers in common use.
The classic argument about which is safer was made by Juan Linz in a 1990 essay, "The Perils of Presidentialism". His case: two actors each claim to be the people's true agent ("dual legitimacy"), fixed terms leave no peaceful way out of a stalemate, and winner-take-all makes losing frightening. The failure he had in mind was Chile in 1973. Others reply that the evidence is mixed. José Antonio Cheibub's 2007 book argued that the apparent fragility owes much to presidential democracies being found where armies were already politically powerful, not to the constitution itself, and that presidential democracies with parties able to form coalitions (as in Brazil) can work. The honest summary is that presidentialism raises risks in some conditions and does not doom any country.
Pick a crisis, guess what happens in each system, then read the answer. Typical constitutional design only; real countries add their own twists.
Notice the pattern in the answers. Parliamentary systems have an exit built into the rulebook: lose the majority and the government ends. Presidential ones do not, so a crisis has to be resolved by bargaining, or, in the worst cases, outside the rules. That is the price of fixed terms, and the benefit is predictability.
A second design choice is where power sits geographically. In a unitary state (France, Japan, China, Chile) the national government holds sovereignty and may create or reshape regions. In a federation (the United States, Germany, India, Brazil, Mexico, Argentina, Nigeria, Switzerland) regions hold constitutionally protected powers that the centre cannot simply remove. The United Kingdom and Spain sit in between, devolving powers to Scotland, Wales, Northern Ireland and to Spain's autonomous communities.
George Tsebelis made the effect precise with the idea of veto players (2002): any actor whose agreement is required to change the rules, such as a second chamber, a constitutional court, a regional government or a coalition partner. Each additional veto player shrinks the set of policies that all of them would accept over the status quo. His prediction: the more veto players, and the farther apart they are, the more stable policy becomes, for better (it protects minorities and prevents lurches) or worse (it blocks needed reform). Try it. Players live on a left-right line; each will accept a change only if it moves policy closer to their ideal point than the status quo is.
Illustrative one-dimensional model. Player positions: 35, 55, 20, 80 (spread scales them around 50). Winset = intersection of each player's interval [x − |x − SQ|, x + |x − SQ|].
With the status quo at 50 and the players spread on both sides of it, the set of acceptable changes is empty: someone is worse off whichever way policy moves. Move the status quo to an edge and a window opens, because everyone wants to move away from an extreme. Squeeze the players together (a unified government) and almost any change that suits them passes. This is the reason unified one-party control produces sharp policy swings in some countries and gridlock in others: it is the veto players, not the party names, that matter.
"Democracy" and "dictatorship" are not two boxes. The Yale political scientist Robert Dahl broke democracy into two dimensions, contestation (can opposition compete and win?) and inclusion (who gets to take part?). Modern measures keep that idea: indices from V-Dem (Sweden), Freedom House (United States) and the Economist Intelligence Unit (UK) all rate the same countries on several scales, and they disagree at the margins, especially about where "democracy" ends.
Hybrid regimes live in that margin. Steven Levitsky and Lucan Way (2002) called them competitive authoritarian: elections are real enough that the opposition can sometimes win, but the playing field is tilted. The incumbent controls the media, the courts, the money, the electoral authority. Nobody is shot at the polls; the contest is just unfair. V-Dem's "Regimes of the World" scheme splits the world into closed autocracies, electoral autocracies, electoral democracies and liberal democracies, and counts a country as an electoral democracy only if its elections are clean and free enough that the opposition could win.
Illustrative thresholds, not V-Dem's actual rules: elections below 2 = closed autocracy, below 4.5 = electoral autocracy, 4.5 to 6.5 = grey zone (competitive authoritarian if liberties are below 5), 6.5 and above = electoral democracy, and liberal democracy if liberties reach 6.5.
The zone in the middle is where most of the scholarly argument happens, and it matters in practice: a country in the grey zone can drift either way, and it is often hard to tell from the inside which way it is going.
In 1959 Seymour Martin Lipset observed that rich countries tend to be democracies. The natural story ("modernization theory"): as countries get richer, they get more educated, urban and middle-class, and those people demand a say. For decades that story was treated as the answer.
Then came a sharper test. Adam Przeworski and Fernando Limongi (1997) asked which event wealth explains: democracies starting, or democracies surviving? Their finding, widely cited since: transitions to democracy happen at all income levels, roughly at random with respect to income, but once a country is a democracy, higher income makes it far more likely to last. Their headline version was that no democracy had ever fallen in a country richer than Argentina in 1975, about $6,000 per head in 1985 prices (as reported in summaries of the paper; later datasets have moved the threshold). Carles Boix and Susan Stokes (2003) replied that wealth does help cause transitions in the long run, and the argument is still alive.
Here is why the distinction is subtle. Both stories produce the same picture in a snapshot: rich countries are mostly democratic. A toy model shows it. Each of 300 countries starts with a random income and random regime and runs for 100 years. Income grows 2% a year. Each year an autocracy may turn democratic (base chance 2%), and a democracy may fall (base chance 3%). Use the sliders to let wealth raise the chance of becoming a democracy, to let wealth protect democracies, or both.
An invented model, not data. Chance of becoming a democracy per year = 2% × (income ÷ 10)a; chance of a democracy falling = 3% × (10 ÷ income)b; both capped between 0.1% and 60%. Income in thousands at start log-uniform from 1 to 30. Fixed random seed, 300 countries, 100 years.
With both sliders at zero the bars sit roughly level, apart from random noise. Now look at the three bars under "Wealth only protects" and under "Wealth causes transitions". They look alike. A correlation between income and democracy across countries cannot tell the two stories apart; you need to follow countries over time, which is exactly what Przeworski did. This is a recurring lesson in the field, and the reason Research Methods in Political Science (35/18) exists.
For most of the twentieth century a democracy ended with soldiers at the palace: Chile 1973, Argentina 1976, and the long run of coups in Africa. Coups still happen: Mali (2020 and 2021), Burkina Faso (2022) and Niger (2023) are recent examples. But a large part of today's decline works differently, and political scientists have names for it. Nancy Bermeo (2016) calls it executive aggrandizement: elected leaders weaken checks on their own power one legal-looking step at a time, by packing courts, rewriting electoral rules, tilting media regulation. Levitsky and Ziblatt's 2018 book How Democracies Die argues that the guardrails are informal norms (mutual toleration, forbearance) as much as formal rules. Peru's 1992 autogolpe, in which a sitting president closed congress with army support, is the dramatic version; most modern cases are quieter.
How common is the pattern? According to the V-Dem Democracy Report 2026, covering 2025, there were 87 democracies and 92 autocracies among 179 countries, with about 74% of the world's population living in autocracies, and 44 countries autocratizing (as reported in coverage of the report; the institute's counting rules differ from Freedom House's or the EIU's, so the totals shift between indices). Whatever the index, the contested question is not whether the rules matter but what keeps leaders and parties inside them.
Two things the evidence supports. First, democracies are much more likely to end where incomes are low and growth is poor (the survival result above). Second, the strongest barrier is rarely a clause in a constitution; it is parties and elites who accept losing an election and still expect to compete again. That second thing is hard to engineer, and it is why the same constitutional text works in one country and not in another.
Countries are governed differently because their rules answer three questions differently: can the state enforce, who can remove the leader, and how many actors must agree to change anything. Democracy and autocracy are the ends of a spectrum with a crowded grey zone in the middle, wealth seems to protect democracies more than it creates them, and today's backsliding comes more often through legal-looking steps than through tanks.
Try this: next time a country is in the news for a political crisis, find out two things before reading the commentary: who can remove the head of government, and by what vote? Then see how much of the story the rule already explains.
Sources. Weber, "Politics as a Vocation" (1919); Tilly, "War Making and State Making as Organized Crime" (1985) and Coercion, Capital and European States (1990); Duverger's term semi-presidentialism (1980); Linz, "The Perils of Presidentialism", Journal of Democracy 1(1), 1990; Cheibub, Presidentialism, Parliamentarism, and Democracy (Cambridge, 2007); Tsebelis, Veto Players (Princeton, 2002); Dahl, Polyarchy (1971); Levitsky and Way, "The Rise of Competitive Authoritarianism", Journal of Democracy 13(2), 2002; Lipset, "Some Social Requisites of Democracy", APSR 1959; Przeworski and Limongi, "Modernization: Theories and Facts", World Politics 1997 (the $6,055 threshold, 1985 dollars, comes from secondary summaries; later datasets put the threshold higher); Boix and Stokes, "Endogenous Democratization", World Politics 2003; Bermeo, "On Democratic Backsliding", Journal of Democracy 2016; Levitsky and Ziblatt, How Democracies Die (2018). Counts of democracies and autocracies: V-Dem Democracy Report 2026 (87 democracies, 92 autocracies, 179 countries, 74% of world population in autocracies, 44 autocratizing), as reported in coverage of the report. The interactives are invented models, computed live; the numbers they show were checked in node.