Opinion and elections
What links the economy to the survival of the government. This page covers equations E12, E14 and E15.
In plain words
The government's popularity depends on unemployment, inflation and purchasing power, compared with what they were when it took office, and it wears down over time. Each population group departs from it according to what its own income has become, so the country's popularity is the average of its voters (E12, E28). Purchasing power, for its part, measures the growth over one year of household income after taxes and benefits (E14). On election day, popularity and growth give an expected score to which an element of chance is added, and it is this score that says whether the incumbent camp loses (E15). The parties then say who gains votes and who governs (E20 to E22).
In the game
The player loses power when their party holds neither the presidency nor the leadership of the government any more, whether it is beaten at an election or brought down after a default (E22), but the player can then carry on the game with the winning party. Each country votes on its own calendar. France elects its president on 2 May 2027 and its National Assembly in July 2029, while the United States votes on 7 November 2028. A stimulus placed eight months before the 2027 election leads to defeat less often than the same stimulus placed sixteen months before, because purchasing power is measured over twelve months (§7.4). In the three months before an election at which they are judged, the player campaigns, with debates and promises (E30 to E32).
The basics
Popularity is the share of people who approve of the government, as in opinion polls. Studies of these polls show that popularity mostly follows two figures, unemployment and inflation. The game adds purchasing power, as well as the wear and tear of power, which costs a government a little popularity every year, even when everything is going well.
Everything is measured from the day the government took office, because a government is judged on what it has done with the economy it inherited, and not on the absolute level of that economy.
Purchasing power compares household income with that of a year earlier. A tax cut makes it rise for twelve months, then the effect drops out of the comparison, so the cut must come a few months before the vote to count on polling day.
Not everyone judges the government in the same way, and a retiree whose pension falls judges it more harshly than the average. The game therefore follows the opinion of each population group, and the country's popularity is the average of its voters (E28).
On election day, a popular government in a growing economy starts as the favourite, but after two terms, voters tend to want a change. Either way, it is not sure to win, because part of the vote cannot be explained by the economy. The game represents that part by a random draw whose typical spread is three points of the vote, while a well-timed stimulus brings in about one. The studio keeps this element of chance because it is realistic, but it knows that a player may experience the result as bad luck.
E12. Popularity
In plain words
Every new government starts at 50%. It then loses 0.6 points for each extra point of unemployment since it took office and 0.6 points for each extra point of inflation, while it gains 0.5 points for each extra point of purchasing power. Whatever it does, it also loses 0.6 points per year in power. The result is a target, the popularity the government is heading towards, as a voter who judged only the country's economy would see it. Each group departs from it according to what its own income has done, and the country's target is the average of its voters (E28). Popularity follows this target slowly and covers half the distance in 70 days. The gap in inflation counts only up to 20 points, in either direction. The studio sets this bound because the studies that give these figures cover inflation rates of a few tens of per cent, and because without it, Argentine inflation falling back from 220% to its target would lift its government above 100% popularity.
A government takes office with 7% unemployment, 2% inflation and purchasing power rising by 1.5%. Two years later, unemployment has risen to 8% and inflation to 5%, while purchasing power is rising by only 0.5%. Its popularity then heads towards 50 − 0.6 × 1 − 0.6 × 3 − 0.5 × 1 − 0.6 × 2 = 45.9%.
In the formula:
| Symbol | What it is | In the example |
|---|---|---|
| \(P^{*}\) | the popularity the government is heading towards | 45.9% |
| \(P_0\) | the starting popularity of every government | 50% |
| \(u - u_0\) | the extra unemployment since taking office | +1 |
| \(\pi^{12} - \pi_0\) | the extra inflation since taking office, counted up to 20 points | +3 |
| \(g^{R} - g^{R}_0\) | the extra purchasing power since taking office | −1 |
| \(n\) | the years in power | 2 |
| \(a_u\), \(a_\pi\), \(a_y\), \(a_c\) | the weight of each term | 0.6; 0.6; 0.5; 0.6 |
Show the full formula and its sourcesThe exact form at the daily step, the population in groups, every parameter with its source, and what the equation leaves out.
u0, π0 and gR0 are the unemployment, published inflation and purchasing-power growth (E14) that the government found when it took office. For the one in office at the start, they are the starting values, with potential growth from the data standing in for purchasing power. n counts its years in power, from the last election before the start of the game, and is reset to zero when the government is replaced (E15, E16). A re-elected government keeps its years, and it keeps wearing out. At the start, popularity already equals P*, so the government in office has lost ac points per year elapsed since the last election.
Where the population is described in groups (E27, E28), P* is the target of a voter who would judge only the country's economy. Each group departs from it according to what its own income has done since the government arrived, the unemployed sit below the others, and the country's target is the average of its voters. The groups.enabled setting brings back P* alone.
| Variable | Meaning | Unit | Bounds |
|---|---|---|---|
popularity (P) | government popularity | % | 0 to 100 |
unemployment (u) | unemployment rate (E6) | % of the labour force | 0 to 50 |
inflation (π12) | published inflation over twelve months (E5) | % | −11 to 500 |
income_growth (gR) | purchasing power: growth of real disposable income over twelve months (E14) | % | not controlled; they follow from those of the output gap |
years_in_office (n) | years in power, internal | years | not controlled |
starting_unemployment, starting_inflation, starting_income_growth (u0, π0, gR0) | the economy the government inherited, internal | % | not controlled |
| Parameter | Value | Status | Source |
|---|---|---|---|
starting_popularity P0 | 50% | every new government starts at 50%; the one in office at the start has already lost cost_of_ruling per year elapsed since the last election | |
popularity_unemployment au | 0.6 points per point of unemployment | Paldam 2004: typical coefficient of a quarterly popularity function | |
popularity_inflation aπ | 0.6 points per point of inflation | Paldam 2004 | |
popularity_inflation_span S | 20 points of inflation | see “The bound S” | |
cost_of_ruling ac | 0.6 points per year in power | Paldam 2004: 0.15 points per quarter; Nannestad and Paldam 2002: 2¼ points per term on average | |
popularity_persistence ρP | 0.74 per month | Berlemann and Enkelmann 2014, Gallup approval 1953-2006 (half-life: 70 days) | |
popularity_income ay | 0.5 points per point of growth in real disposable income over a year | Frey and Schneider 1980, see E14 |
Why. Unemployment and inflation are the “big two” of popularity functions. Their effect is one of the best-established results of the literature (Nannestad and Paldam 1994; Lewis-Beck and Stegmaier 2013). And ruling wears a government out, whatever the economy does. Purchasing power comes on top of them (E14).
The bound S (choice). These functions are estimated where inflation stays in the order of a few tens of per cent. Without a bound, Argentine inflation falling from 220% to its target would carry its government above 100%.
What it simplifies. Popularity functions are unstable from one country and one period to another, and the government's responsibility is clear only in two-bloc systems (Nannestad and Paldam 1994). Voters compare their economy with their neighbours' (Kayser and Peress 2012), which the game does not do yet. The coefficients are symmetric, while Nannestad and Paldam find that a deterioration weighs three times as much as an improvement; Hibbs (2000) does not find this asymmetry for income. The purchasing-power coefficient comes from studies where unemployment and inflation have their own coefficients, estimated together. Paldam's were estimated without income, so adding them up may count part of the same effect twice.
E14. Purchasing power
In plain words
Household disposable income is what the economy produces, minus what the government takes, plus what it pays back in benefits such as pensions or allowances. Purchasing power is the growth of this income over twelve months, so a tax cut adds to it straight away and no longer adds anything a year later. To weigh on the vote, the cut must therefore come six to nine months before the election.
In an advanced country whose revenue is 51% of GDP and whose spending is 55%, 42% of spending goes on benefits, that is 23.1 points of GDP, and disposable income is 100 − 51 + 23.1 = 72.1% of GDP. Cutting revenue by 2 points raises it to 74.1%. Income then rises by 2.8%, and the purchasing power on the dashboard gains 2.8 points for twelve months before returning to its previous level.
In the formula:
| Symbol | What it is | In the example |
|---|---|---|
| \(R\) | real household disposable income | 72.1% of GDP |
| \(Y\) | real GDP | |
| \(T\) | government revenue | 51% |
| \(TR\) | the benefits paid to households | 23.1% |
| \(\tau\) | the share of spending that goes on benefits | 0.42 |
| \(INT\) | the interest on the debt received by households, not yet counted | 0 |
| \(g^{R}\) | purchasing power, the growth of \(R\) over twelve months |
Show the full formula and its sourcesEvery parameter with its source, and what the equation leaves out.
R is the real disposable income of households: output, minus what the state takes, plus what it gives back, namely the transfers TR and the interest on its debt INT, in points of GDP. gR is its growth over twelve months. This is purchasing power, which the player sees and which enters popularity (E12).
| Parameter | Value | Status | Source |
|---|---|---|---|
popularity_income a_y | 0.5 points of popularity per point of growth in real disposable income over one year | Frey and Schneider (1980): United States 0.52 (quarterly) and 1.08 (monthly), Germany 0.31 to 0.56 depending on the government, 0.50 measured as a gap to the previous parliament; United Kingdom (1978): 0.39 to 0.81 on the government's lead | |
social_transfers_share τ | 0.42 / 0.27 / 0.05 of primary spending | public social protection spending excluding health (ILO 2024, figure 3.11, weighted by GDP: 16.2% of GDP for high-income countries; 8.5 and 4.2% for the middle-income groups, 7.7% combined; 0.8% for low-income countries), divided by the group's primary spending weighted in the same way (38.3; 28.6; 14.7% of GDP) | |
| INT | 0 | see below |
Why. Real disposable income is a classic variable of vote and popularity functions. In Hibbs (2000), it alone absorbs the effect of the economy on the US presidential vote (4.12 points of vote per point of average growth in income per head over the term); in Frey and Schneider, it adds to unemployment and inflation. The term is written like the others in Paldam's function, as a gap from what the government found when it arrived. Growth in income per head and growth in total income differ only by population growth, which is constant here and cancels out in the gap. This is what links the player's budget to their political survival: a tax cut gives back purchasing power at once, a tax rise takes some away; and a rise in spending also gives some back, through the share that goes to households as benefits. For France, τ G is 23.3% of GDP, and the model's disposable income is 72% of GDP (49% without benefits).
What the player learns from it. Purchasing power is measured over twelve months, so a tax cut that is a year old no longer gives anything back. To count on election day, it must come six to nine months before (§7.4). This is the political business cycle, which appears in the model without anyone having written it in.
What it simplifies. Benefits are a fixed share of spending in each group of countries. One extra point of spending is split like the group's average budget, and the United States receives the share of high-income countries, which is higher than its own. The ILO's social protection “excluding health” includes some benefits in kind, which are not income (ILO 2024, note 29 of chapter 4). Interest on the debt does not count yet, because nobody knows who holds the debt (unknown interest_in_income, held at zero). Until 23 September 2026, transfers were also an unknown held at zero. Disposable income was underestimated (49% of GDP in France), a tax cut gave back too much, and a spending stimulus took purchasing power away.
E15. Elections
In plain words
On polling day, the incumbent camp starts from about 51.4% of the vote. It gains a little over a tenth of a point per point of net approval, that is, the gap between those who approve and those who disapprove, and 0.6 points per point of growth over the year. If it has governed for two terms or more, it loses 4.3 points, which is called the time-for-change penalty, but only at the election that decides who governs. Where an assembly is elected separately from a president, its election judges the president's camp by the same equation, without this penalty. A random draw is added for everything the economy does not explain, along with, for the player's camp, what its debates have brought it (E31). At 50% or more, the incumbent camp is re-elected, and below that, it is replaced.
With 45% popularity, 1.2% growth over the year and a first term, the expected score is 51.417 + 0.109 × (2 × 45 − 100) + 0.604 × 1.2 = 51.05%. That is above 50, but by so little that, with the draw on the day, the incumbent camp wins only about 64 times out of 100. After two terms, the 4.265 points of the time-for-change penalty bring it down to 46.8%, and it wins only 14 times out of 100.
In the formula:
| Symbol | What it is | In the example |
|---|---|---|
| \(V\) | the incumbent camp's share of the vote | 51.05% |
| \(P\) | popularity | 45% |
| \(2P - 100\) | net approval | −10 |
| \(g\) | real GDP growth over twelve months | 1.2% |
| \(\mathbb{1}[\ldots]\) | equals 1 after two terms or more, 0 otherwise | 0 |
| \(z^{E}\) | the draw on election day |
Show the full formula and its sourcesThe campaign, the time-for-change penalty, every parameter with its source, and what the equation leaves out.
On its country's election day, the incumbent camp obtains
points of the vote of the two main camps, that is, with the values in the table, V = 51.417 + 0.109 (2*P* − 100) + 0.604 g − 4.265 · 𝟙 + 3 zE. g is real GDP growth over twelve months (E14), and zE a draw from a standard normal distribution. At 50 or more, the incumbent camp is re-elected and counts one more term. Below that, it is replaced. The new government starts at starting_popularity, with no years in power and one term, and is judged on the economy it inherits (E12). If it was the player's government, the player has lost, their pending orders are cancelled and their country goes back to the rules of non-player countries, unless they take up the game again with the winner (E22). The next election takes place one term later. The news item on the election gives V, draw included, as the share that decided.
The campaign. Where the president is elected in two rounds (France), the day in the calendar is that of the runoff, and the first round takes place fourteen days earlier. In the first round, V gives each party's votes, and in the runoff, the duel when the incumbent camp is in it (E30). For the player's camp, V also counts what its debates brought it (E31), and its popularity counts what voters grant to its promises (E32).
The time-for-change penalty (𝟙) weighs only on the election that decides who governs. Abramowitz measured it on presidential elections, for a camp seeking a third term. Where an assembly is elected separately from a president (E21, E22), its election judges the president's camp by the same equation, without the penalty, as in a first term, and does not add a term to it.
At the start, each incumbent camp is counted as being in its first term. This is the unknown incumbent_terms_at_start, held at 1.
| Variable | Meaning | Unit | Bounds |
|---|---|---|---|
| V | share of the two main camps' vote won by the incumbent camp (the president's camp under cohabitation, E22) | % | not controlled |
popularity (P) | government popularity (E12) | % | 0 to 100 |
growth (g) | real GDP growth over twelve months, the day compared with the same day a year earlier; not the growth the game publishes | % | not controlled |
terms_in_power | terms of the incumbent camp, internal | terms | not controlled |
next_election | date of the next election, internal | date | none for a country that does not vote |
| Parameter | Value | Status | Source |
|---|---|---|---|
vote_intercept a0 | 51.417% of the vote | Abramowitz 2008, table 5: United States, 1948-2004, 15 elections | |
vote_net_approval aP | 0.109 points of the vote per point of net approval | same source; net approval is “approve” minus “disapprove” (Gallup, mid-year), written here as 2*P* − 100, as if nobody answered “no opinion” | |
vote_growth ag | 0.604 points of the vote per point of growth | same source: annualised real GDP growth in the 2nd quarter of the election year; here, over twelve months | |
vote_time_for_change at | 4.265 fewer points of the vote | same source: when the outgoing camp has governed for two terms or more; only at the election that decides who governs | |
vote_noise σV | 3 points (standard deviation) | 1.78 (Abramowitz 2008), 2.99 for Germany and 4.15 for the United Kingdom (Lewis-Beck and Dassonneville 2015), 3.0 (Fair 2022, economy only), 4.5 in parliamentary systems without popularity (Nannestad and Paldam 1999) |
| Quantity | Value | Status | Source |
|---|---|---|---|
| the calendar | per country: the last national election that decides who governs, the next one, the length of a term | Wikidata, CC0 |
Why. Abramowitz's equation is the vote function that goes through popularity. This is what makes the model vote on what the player has done, and not on the economy alone. The draw on the day stands for everything the economy does not explain. The calendar is each country's own. France votes on 2 May 2027, the United States on 7 November 2028, India in 2029, and governments are not all “elected” on the same day.
What it simplifies. The equation is estimated on US presidential elections. It applies to all countries, parliamentary systems included, where coalitions and seats decide otherwise. The number of terms the incumbent camp has already served at the start is unknown, whereas in France the camp in power since 2017 should suffer the time-for-change penalty from 2027. Where the data describe parties, the election is played party by party, and the party that leads the government changes what a non-player government does (E19 to E24). Elsewhere, a new government follows the same policy as the old one. A country with no national election in Wikidata has none (Saudi Arabia, but also, wrongly, Denmark); a country where the election does not decide who governs still has one (China, Russia). A well-timed stimulus wins about one point of the vote, while the standard deviation of the draw is worth three. This is realistic, since the economy explains only part of the vote, but a player may experience it as chance.