Demographics

The demographics system models voter populations in each state. It is the foundation of election mechanics: every vote in every election flows through demographic groups and their alignment with candidates. Demographics are a voter and election mechanic, not a corporation mechanic.

The census: who the electorate is#

Each state's electorate is described by its census. In the US that is four dimensions:

Every state carries its own real shares in each dimension: a rural state has more no-college voters, a coastal metro more graduates. Other countries use their own dimensions and their own buckets (most use ethnicity, age, education, income and urbanization), because a bucket list that fits the US does not fit the UK or Japan.

Each bucket has an economic and a social position on the −5 to +5 scale, and a baseline turnout rate. Positions come from an era position table plus per-state overrides, so white voters in Texas genuinely sit to the right of white voters in Massachusetts: regional political texture comes out of the census rather than one national table.

Cells: how the census becomes voters#

The dimensions are combined into cells: a cell is one combination across all of them, for example "white, mature, no college, middle income". The engine estimates how many voters sit in each cell by raking the state's own dimension shares together with association priors (no-college voters are more likely to be low income, graduate degrees predict high income, and so on), then prunes the cells too small to matter.

A cell's lean is the average of its buckets' positions, and its turnout is blended from its buckets' participation rates. Cells are what the vote engine counts. Everything you see on a region page, in a poll, or in a turnout action is either a cell or an aggregate of cells.

Eras. Every playable era (1953-2023) has its own census bundles and position tables: a 1979 world has a 1979 electorate, not a reweighted 2019 one. Derived leans are calibrated against real historical election results per era and country, and an automated calibration suite keeps them honest.

How demographics drive elections#

During elections, candidates compete for votes from each cell through a multi-factor formula:

1. Reach#

reach = sqrt(min(100, politicalInfluence) / 100)

How much of the turned-out voters a candidate can actually contact and persuade. The square-root curve gives diminishing returns: PI=25 gives 0.5 reach, PI=50 gives 0.707, and it caps at 1.0 once PI reaches 100.

2. Appeal#

positionScore = 25 × (positionRaw / 50)^1.5 + floor(0.5)
appeal = positionScore + directionalBonus + normalizeNPI(politicalInfluence) × 12.5

Where positionRaw is derived from the absolute economic and social distance between the candidate's policy positions and the cell's leans, and directionalBonus adds a small amount per axis when the candidate's lean matches the cell's lean direction (tribal-voter effect). Maximum appeal is 50. Candidates who match a cell's positions score high; candidates far from their positions score low. The exponent (1.5, not squared) keeps position from swamping every other lever in the formula.

3. Approval scalar#

approvalScalar = (favorability / 100) ^ 0.8

Voters won't support candidates they don't approve of, regardless of policy alignment. The exponent below 1.0 means the scalar falls off less steeply than favorability itself at low approval, and 0% approval still means 0 votes.

4. Party organization (general elections)#

In general elections, each party gets a multiplier equal to its normalized share of the state's total Org:

orgShare = party.organization / Σ(every party's organization in this state)
orgMultiplier = orgShare ^ 0.2

The share is raised to the 0.2 power, a diminishing-returns curve: a party with 60 Org in a state where the total is 100 gets roughly a 0.9× multiplier, not 0.6×, softening a dominant party's structural Org edge. A party with no presence in the state still gets 0 (no votes). Two more general-election factors layer on top:

• Reg resistance (1.0×-1.3×): own-Reg makes voters harder to peel away through persuasion
• Support mood (0.6×-1.4×): short-term candidate momentum from debates / endorsements / scandals; neutral 1.0× at Support=50

In primary elections (intra-party), Org is applied as a uniform neutral 1× instead: within-party normalization would cancel out since all primary candidates share the same Org.

5. Vote allocation#

Within each cell, candidates split the cell's votes proportional to their relative appeal × reach × approvalScalar × orgMultiplier × regResistance × supportMood × infamyMult in general elections. Primaries drop the three general-only factors (orgMultiplier / regResistance / supportMood): Org is a neutral 1× intra-party. Either way, the cell votes as a bloc, and each candidate gets a fraction of the cell's total turned-out voters.

Total votes are summed across all cells. This cell-level split is the base pipeline; live general-election vote accumulation runs it through the swing-flow model, which additionally applies coattails, a median-voter policy reference, persuasion drivers, and party-tenure fatigue. See Election Mechanics.

State political lean#

Each state has an overall political lean calculated from its demographics:

1. For each cell: multiply population share × turnout to get voter weight
2. Compute turnout-weighted average economic lean and social lean
3. Average the two leans and clamp to −5..+5

Labels use the shared −5..+5 position ruler, rounded to the nearest step: 0 = Centrist, ±1 = Center-Left / Center-Right, ±2 = Lean Left / Lean Right, ±3 = Left / Right, and so on. Post-calibration state leans typically span about −2.5 to +2.5.

This display label appears on the state page and the political map. It is for display only: elections use raw position differences, not the display lean.

Turnout manipulation#

Candidates and parties can increase (or decrease) how many voters actually vote. Turnout modifiers sit on top of baseline turnout and range from −20% to +20%. They are applied to a census bucket, and reach every cell that carries that bucket.

Party GOTV spending#

Party budgets target a single chosen demographic bucket and spend automatically each turn: gotvBudgetPercent (0-25% of hourly revenue, with a legacy flat gotvBudgetPerTurn amount as a fallback). There is no binary "within 2 points" eligibility gate; instead a continuous alignment multiplier scales the boost by how close the party's position is to the target bucket's lean, from 1.0× at perfect alignment down to a 0.1× floor.

Boost formula per turn:

boost = (spend / 5000) × alignmentMultiplier × diminishingReturns(currentModifier)

$5,000 of spend buys one full turnout point before alignment and diminishing returns are applied. National spend divides evenly across all states; state-scoped spend applies in full to one state.

Diminishing returns: at a +10% existing modifier, a 1% boost becomes only 0.5% effective.

All modifiers decay 2% per turn toward 0, requiring sustained spending to maintain boosts.

Player canvassing#

The canvassing action targets a specific census bucket in your active campaign state (home state by default; presidential candidates use their travel/primary campaign state):

• Cost: 100 funds + 1 action point
• Base boost: 0.05 percentage points
• Alignment multiplier: Perfect policy alignment = 1.0×, distance 10 = 0.1× minimum
• Campaign season bonus: 2× effective during the 4 turns before an election
• Effect: Immediate (not queued)

You are most effective canvassing groups that already align with your positions. Mismatched canvassing wastes actions.

Demographic alignment and character policy#

Candidates declare economic and social policy positions on a −5 to +5 scale. These are compared directly to each cell's leans in that state when computing appeal. The closer a candidate's positions are to a cell's leans, the higher their appeal with those voters.

This means:

• A candidate at (−4, −4) naturally appeals to young, graduate-educated, urban voters
• The same candidate poorly serves older no-college voters in conservative states
• A centrist candidate (0, 0) has moderate appeal everywhere but strong appeal nowhere
• Because leans are state-derived, the same platform lands differently state to state: poll locally rather than assuming a national table

Winning elections requires either mobilizing your natural coalition (GOTV + canvassing) or moderating your positions to pick up the middle-income and mid-age groups that sit closest to the centre.

Polling#

Players can use poll actions to check demographic standings before an election. A full demographic poll breaks the state's electorate down into its cells and shows which segments are strongly for, weakly for, or opposed to your candidacy.

See also: Election Mechanics, Government Approval, Campaign Strategy