Governments

If the incumbent camp keeps power (in a democracy, when the equation of E15, draw included, gives it at least 50; in an electoral autocracy, according to E23), the party that leads it stays at its head. In parliament, the government in office stays as it is if it keeps more than half of the seats. Otherwise this party forms a new government. It takes in, one by one, the party closest to the centre of its coalition (the profiles of its parties, weighted by their seats), until it has more than half of the seats. If there is nobody left to take in, the government is a minority government. In the presidency, nothing changes.

If the incumbent camp loses, the government is formed without the party that led it. In parliament, the largest of the other parties forms it, by the same rule; its former partners can join it. In the presidency, the two largest opposition parties go to the runoff (unless one has more than half of the vote), where the votes of each other party, those of the outgoing government included, split between them according to their proximity, halved every h points of distance (vote_proximity_half_distance, E20). The winner governs alone. Where the data hold the presidential election in two rounds on two days (France), the two rounds are those of E30. The first says who goes to the runoff, the second decides, and its winner takes the presidency.

Where a ruling group governs, nothing changes.

A government that is being formed shares out its posts by a rule, and the player who leads it then reshapes it as they see fit. The government's parties then follow the posts (E41).

Where a president is elected separately from an assembly that the government comes from (E21, today only France), two powers coexist. One is the presidency, held by a party (at the start, the one that leads the government). The other is the government, which is formed in the assembly. Both elections judge the president's camp, and that is the camp the equation of E15 scores. The president is the leader of that party, and takes no post in the government (E41).

• At the presidential election, the camp that keeps the presidency changes nothing, not even the government. The camp that loses it leaves the presidency to the winner of the runoff, who governs alone and dissolves the assembly. The assembly is elected dissolution_days days later.
• At the assembly election, when the president's camp wins, its party forms the government by the rule above (the government in office stays as it is if it is already the president's and keeps the majority). When it loses, the government is formed without the president's party, by the same rule, and the president stays. This is cohabitation. It lasts as long as the president's party does not lead the government, and ends at an election that gives the lead back to it, or when another party takes both the presidency and the head of the government.
• Dissolution. The player who holds the presidency can dissolve the assembly at any time. It is elected dissolution_days days later, and cannot be dissolved again for dissolution_lock_days days after that election (Constitution, article 12). The dissolution that follows a lost presidential election sets the same lock. A non-player president never dissolves of their own accord.

After a fall (the default, E16), a new government is formed without the party that led the previous one: in parliament, on the seats of the last election, by the rule above; in the presidency, the party that had the most votes after it. A separately elected president does not fall with it. If their party led the government, the fall opens a cohabitation.

The player plays a party, the one that leads the government at the start, which also holds the presidency where there is one. They lose power when their party no longer holds either the presidency or the head of the government, through the ballot box or through a fall. They keep it if they now lead only a coalition that has lost some of its partners, and, as president, during a cohabitation. Their pending orders and their bills before parliament are then cancelled. They can take up the game again with the party that now leads the government, and from then on they play that party, under the rules by which it can lose. As president under cohabitation, they can also choose to play the government instead, and leave the presidency to the model.

Why. The equation of E15 is what is calibrated. It makes the outgoing camp lose nearly one election in two, as in the data. A first version let the seats decide. A government that started with a large majority almost never lost it, and parliamentary democracies lost only 12% of their elections, against 46% without parties. The world changed government 0.52 times per country per decade, whereas WhoGov counts 1.31 from 2000 to 2024 (the leader's party changing, all regimes). The parties then say who governs. In parliamentary democracies, the largest party gets the post of Prime Minister in 72.8% of cases in Western Europe and 77.5% in Eastern Europe (Glasgow, Golder and Golder 2011), and a coalition most often brings together neighbouring parties. The formation rule is the simplest one that respects these two facts. It is a choice, which no study calibrates.

What it simplifies. There is no negotiation. Portfolios, programmes, vetoes and grand coalitions against the extremes do not exist, and a party joins a coalition as soon as it is the closest, even if it is not needed for the count. The incumbent camp can lose while keeping a majority of the seats, or win while losing it, because E15 and the seats are not tied to each other. Between two elections, a non-player government falls only through a default. There is no motion of no confidence, no coalition breaking up and no coup. This is why the model changes government even less often than the world does (0.82 times per country per decade over fifty years, against 1.31). Only the government that the player puts together can lose a minister or a partner, or fall to a motion of no confidence (E42 to E44). Where the assembly is elected separately, a president who wins against the government's camp always dissolves it, and a non-player president never dissolves it otherwise; the date of the dissolution is the same everywhere. Cohabitation is played as the player experiences it, not as the Constitution describes it. The government “determines and conducts the policy of the nation” (article 20), but the player as president keeps their levers, and their measures go before an assembly where the opposing government votes like an opposition (E25). The model has only one popularity per country. Under cohabitation, both elections judge the president's camp on the economy that the player runs, and a player who has taken the government sees their good management serve the opposing president. There is no honeymoon. The legislative elections that follow a presidential election are no easier for the new president than the others.

The regime says which elections a country holds. A democracy holds competitive elections. E15 says whether the government loses, and E20 to E22 say the rest. An electoral autocracy holds controlled elections. Votes and seats move as elsewhere, but the government loses only if its draw of the day, read as a probability, falls below q, whatever E15 says:

Φ is the cumulative distribution function of the normal distribution, z^E the election draw (E15). The defeated government is replaced without its leading party, as after a fall (E22). A closed autocracy holds no elections, whatever its calendar.

Why. In an electoral autocracy, the government rarely loses, but it does lose: Senegal in 2000, Malaysia in 2018. Making defeat depend on the draw alone, and not on the votes, says that the result does not follow opinion. Since the draw is that of E15, it uses up no extra randomness.

What it simplifies. The probability is the same in all electoral autocracies, hegemonic or competitive, and depends neither on the economy nor on popularity. A regime never changes. There is no democratisation, no coup and no autocratisation.

A non-player government follows two rules: the Taylor rule for its central bank (E7), and the fiscal rule for its budget (E9), with the automatic stabilisers (E8). These are the rules of a perfectly rational government. Real governments depart from them, and five of these departures are measured; a sixth, which concerns who pays and who receives, is a choice. Each one is a behaviour: a function that reads the calendar, the country's group and the party that leads the government, and that adjusts its targets. The ai.pure_rationality setting switches them all off. Each one also has its own switch and its own strength.

J_t is the number of days before the next election, P_L the position, on the populism slider, of the party that leads the government. The first line is the electoral cycle. The targeted primary balance falls by c points of GDP during the D days before the vote. The second is procyclicality. Where spending follows output with elasticity e, the stabilisers (ε, E8) keep only 1 − e of their effect. The third is populist pressure on the central bank. Under a populist leader, the response of the Taylor rule to inflation loses the share κ.

The last two are what a populist leader does, and their cost comes from the economy itself (E36 and E37). The fourth is the tariff. Under a populist leader, the country's average tariff (τ, the customs_rate lever, E35) moves towards R points above its starting level, τ₀, reached in Y_τ years in regular steps, one step on the first of each month. Under any other leader, it returns towards τ₀ at the same pace. The fifth is the rule of law. Under a populist leader, the country's index (I, from 0 to 100, E37) moves towards Δ_I points below its starting level, I₀, covering half the distance every h_u days (rule_of_law_erosion_half_life_days). Under any other government, a player's included, it returns towards I₀, half the distance every h_r days (rule_of_law_recovery_half_life_days). Under any government, the level the index aims for is lowered by what the country's exit regime costs it, and the index comes down to it at the speed h_u (E89).

A behaviour at strength a (a setting) shifts by a times its gap (c, R, Δ_I), and multiplies by 1 − a (1 − x) what it reduces by a share x (e, κ). When switched off, the tariff stays where it is, and the rule of law wears away under nobody.

A fifth behaviour does not touch the rules but who pays and who receives (E29). It is its camp's distribution. A non-player government that comes to power sets the three levers of E29 according to the gap between its party and the one that led at the start of the game. The data describe the policy of that starting party, which is the neutral point of each lever:

E_L and E₀ are the positions, on economic_left_right, of the party that leads and of the starting one, h the half-scale, and a the strength of the behaviour. Further to the left (v > 0), the progressivity of the income tax rises by v, and the minimum pension rises by v. Further to the right, progressivity falls, and the minimum pension stays at zero. Each lever is held within what it can still change (E29). A government on the same side as the starting one puts the levers back to zero. The VAT rates by family of goods (E38), on the other hand, stay where the government finds them.

Why these five. The electoral cycle is one of the best-established facts of political economy, and its meta-analyses confirm it for the primary balance (Cazals and Mandon 2016), small in established democracies, clear elsewhere. The procyclicality of the budgets of developing countries is just as well documented. More than nine in ten were procyclical from 1960 to 2009 (Frankel, Végh and Vuletin 2013). Populist leaders put pressure on their central bank more often than others (Binder 2021), protect their market more and wear down institutions (Funke, Schularick and Trebesch 2023).

Why there is no longer a direct loss. The same study measures that a country loses about one point of growth per year under a populist leader, more than 10% of GDP per head in fifteen years. The model used to take this point away from potential growth for as long as the populist governed. It did so with no time limit, whereas the study stops at fifteen years; for 42 governments out of 240 from the start, whereas the study classes 51 leaders out of 1,482 as populists; and with no cause, whereas the study describes policies. Measured before the removal, over four hundred-year games without a player, France spent 71% of its time under a party beyond the threshold, and 37% of its years in negative growth, against 5% in the data from 2000 to 2019. The studio chose to count only what the policies that the model knows how to count actually do. Their effect is small, from 1.90 to 2.45% of output in fifteen years (below), a fifth at most of what the study measures. The study itself does not say that these policies explain the loss. It documents them as facts that accompany it.

The results. Without shocks and without elections, the output of a country that a populist party leads at the start, compared with the same country without the two behaviours.

What is not in it, and why. Two expected behaviours were set aside for lack of evidence. The left-right orientation does not move the budget balance in a robust way. The left has a slightly larger state, spending and revenue together, but the effect on the deficit is “ambiguous” (Potrafke 2017), and the two published figures have opposite signs (Alesina, Cohen and Roubini 1993; Müller, Storesletten and Zilibotti 2016). Coalitions do not consolidate measurably more slowly. The only published test of the speed of adjustment finds nothing (Roubini and Sachs 1989), and the effect of coalitions on the deficit disappears outside the OECD (Elgie and McMenamin 2008). The orientation of a government therefore counts here through its populism, and through what it does to elections (E20 to E22).

What it simplifies. The electoral cycle depends only on the country's group. Brender and Drazen find it mostly in new democracies, during their first four elections, and a strong fiscal rule dampens it (Gootjes, de Haan and Jong-A-Pin 2019). Neither the age of a democracy nor its fiscal rules are in the model yet. For low-income countries, the study measures the overall balance, interest included, and not the primary balance. Procyclicality is the same for a whole group, whereas it falls with the quality of institutions (Frankel, Végh and Vuletin 2013). Populist pressure is read from governors leaving for political reasons, in a working paper. No study directly measures the effect of a populist leader on the policy rate. A populist is recognised there by a threshold on a slider, whereas Funke et al. classify leaders one by one. The threshold picks out 19% of governments at the start, and they pick out 3.4% of leaders. The tariff rises and the erosion of institutions are read off the curves of their figures, to which the article gives no numbers, and the first is their least certain result. They are the same for all populists, of the left as of the right, and depend neither on the time spent in power before the start of the game nor on what the country can decide alone. A member of a customs union raises its tariff like any other, whereas the study leaves these countries out of its measurement. A country that starts under a populist sees its rule of law wear away by a further ten points, even though the starting index already carries part of that erosion. Debt, which the study finds ten points of GDP higher after fifteen years, and finance, which it finds less open, are not counted. The electoral cycle and procyclicality also act in a country without parties; the three populist behaviours act only where a party leads.

A seventh behaviour is described elsewhere. It is migration_laws, by which a non-player government shifts its migration laws when its arrivals depart from those it wants (E86). It reads no slider, because studies find no effect of a government's political colour on how restrictive its migration policy is.