Variables and bounds
What the model tracks for each country, and the values that no country should ever reach.
In plain words
Every figure the model publishes has bounds, set a little beyond what has already been observed in the real world. A bound does not stop the calculation. Without shocks, a game that crosses one points to an error in the model or in its starting figures, whereas with shocks a run of bad luck can take a figure across it, in at most one country-month in a thousand. The model also remembers what has happened, such as past inflation or the budget of the year just ended. These internal memories are not displayed, but because they are part of the state of the game, a save keeps them.
The ten published variables
A bound is never a cap applied to the calculation. It is a line that a simulation must not cross; if it does, it has found a fault in the model or in its calibration. Each bound is set a little beyond what has already been observed.
| # | Variable | Meaning | Unit | Bounds | Where the bounds come from |
|---|---|---|---|---|---|
| 1 | real_gdp | real GDP: potential × (1 + gap) | index, 1 at the start | > 0 | by construction |
| 2 | output_gap | output gap | % of potential | −25 to +25 | largest annual changes in real GDP observed since 1960: −21.3% and +28.6% (0.5th and 99.5th percentiles, World Bank) |
| 3 | inflation | published inflation, over twelve months | % | −11 to 500 | deflation of −10.9% in the United States in 1921; above 500% a year, it is hyperinflation (Reinhart and Rogoff 2008), which the core model does not represent |
| 4 | unemployment | unemployment rate | % of the labour force | 0 to 50 | the highest ever measured is 38.8% (North Macedonia, 1996); the lowest is 0.10% (Qatar) |
| 5 | policy_rate | policy rate | % per year | −1 to 500 | the lowest is −0.75% (Switzerland, Denmark, 2015); the highest since 2015 is 133% (Argentina, 2023); beyond that, it follows inflation |
| 6 | gov_revenue | government revenue | % of GDP | 0 to 100 | by definition; observed in 2024, from 2.9% (Sudan) to 74% (Kuwait) |
| 7 | primary_spending | primary spending (excluding interest) | % of GDP | 0 to 100 | by definition; observed in 2024, from 5.2% (Haiti) to 57.8% (Finland) |
| 8 | gov_debt | public debt | % of GDP | −600 to 600 | the highest ratio on record is 600% (Liberia, 2003); below zero, the state is a net lender, like sovereign wealth funds |
| 9 | effective_rate | effective rate on the debt: interest paid / debt | % per year | −5 to 500 | negative when the neutral rate is negative; it is then the return on public savings |
| 10 | popularity | government popularity | % | 0 to 100 | by definition; observed: 22 to 90% (Gallup, United States), 13 to 75% (Ipsos, United Kingdom), 17% (Ifop, France, 2014) |
The primary balance is not a state variable. It follows from revenue, spending and the gap (E8). Its bounds are checked all the same, from −40 to +55% of GDP. The most extreme balances observed are −32.1% (Ireland, 2010, bank bailout) and +52.3% (Kuwait, 2000).
The stakes (E13 to E17) publish four more figures, which the player sees:
| Variable | Meaning | Unit | Bounds |
|---|---|---|---|
growth | growth of real GDP over twelve months | % | follow from those of the output gap |
income_growth | growth of real disposable income over twelve months, i.e. purchasing power | % | same |
credibility | credibility of the central bank | from 0 to 1 | by construction |
default_probability | probability, over one year, that the markets stop lending | from 0 to 1 | by construction |
With shocks, a bound can be crossed by bad luck rather than because of a fault. The data themselves hold a few examples (−16.9% inflation in Lesotho in 2009, Macao's output halved in 2020). The world without shocks must stay within its bounds. The world with shocks may leave them, but in at most one country-month in a thousand (choice).
Sources of the observed figures: the World Bank (WDI) for growth and unemployment; the IMF (WEO and Fiscal Monitor of April 2026) for debt, revenue, spending and balances; the BIS and the ECB for policy rates; Gallup, Ipsos and Ifop for popularity. The inflation bounds come from Reinhart and Rogoff (2008) and from the Minneapolis Fed (price index series). Details in §11.
The internal memories
They are not shown to the player, but they are part of the state and must be saved:
| Memory | What it retains | Why |
|---|---|---|
inflation_memory | past inflation, weighted less and less (half-life of 474 days) | the unanchored share of expectations (E3) |
price_history | the price level over the last 365 days | published inflation over twelve months (E5) |
remembered_spending, remembered_revenue | the budget of the past year | the fiscal impulse (E1) |
debt_anchor | the lowest debt reached since the start | the fiscal rule of non-player countries (E9) |
starting_balance_gap | the starting gap between the balance and the balance that stabilises the debt, which fades | same |
private_bonds | the debt as seen by those who hold it | the check of the accounts (E11) |
years_in_office | the government's years in power, counted at the start from the last election | the cost of ruling (E12) |
starting_unemployment, starting_inflation, starting_income_growth | the economy the government inherited | popularity is measured from the day it took office (E12) |
underlying_price_level, price_deviation | the price level driven by the Phillips curve, and the temporary deviation that price shocks add to it | the published price level (E5, E13) |
output_history, income_history | real GDP and disposable income over the last 365 days | growth and purchasing power over twelve months (E14, E15) |
inflation_quarter | published inflation over the last 91 days | the credibility signal (E17) |
credibility | the central bank's stock of credibility | expectations (E3, E17) |
next_election, terms_in_power | the date of the next vote, and the terms of the incumbent camp | elections (E15) |
market_reference, debt_limit, excluded_until, defaulted_on, last_haircut, written_off | the debt on which the markets lent, the country's limit, the end of the exclusion, the day and the haircut of the last default, what the haircuts have written off | the crisis and default (E16) |
| the state of the random generator | the seed of the game and how far the draws have gone | everything drawn at random (E13, E15, E16); without it, a saved game does not replay identically |