Monetary System (as shipped)
Money in A House Divided runs on four connected systems: a central bank per currency that sets a prime rate through a voting committee, a forex market where every active currency floats against an internal anchor, savings that pay a real (inflation adjusted) rate, and a line of credit whose price is that same prime rate plus a credit spread. This page describes what is actually in the code today, with the real constants and the files they live in. One turn is one real hour (MS_PER_TURN = 60 * 60 * 1000) and a game year is 48 turns (TURNS_PER_YEAR = 48), both in . Every rate and accrual below is expressed against that clock.
Central banking: who holds the rate#
Each currency has exactly one bank behind it. The mapping is explicit, not derived: COUNTRY_CURRENCY_MAP gives every country its home currency, and CURRENCY_ANCHOR_COUNTRY names the single country whose exchangeRates document and central bank are authoritative for that currency (both in ). Scotland and Wales are sterlingized: they use GBP with the UK as anchor. The Soviet ruble (SUR) is anchored on RU and shared by BLR, UKR and BAL. EUR is anchored on DE, and IE keeps its own IEP so a 1953 Bretton Woods par with sterling can exist without colliding with the DM/EUR rate. The URL slug /centralbank/[currency] is resolved by resolveCentralBankCurrency in , which also routes to the intorg API when the anchor bank is treaty run (the ECB).
Who may move the rate depends on the bank's independence. encodes the historical case: the Bank of England had no operational independence until BOE_INDEPENDENCE_YEAR = 1997, so a UK world whose era START year is before 1997 opens with the Treasury setting Bank Rate. The default keys on the era start year, not the live year. A 1991 world that plays past 1997 does not flip on its own, because transferring monetary power is a statute: Parliament passes a central_bank_independence provision, and an explicit bank.governmentControlled written by legislation always wins. A country on a shared bank cannot legislate here at all (canLegislateBankIndependence returns false whenever the country config carries sharedBankId), so an ECB member cannot rewrite a treaty institution and a sterlingized Scotland cannot legislate over the Bank of England.
Prime rate changes and their limits#
is the single write path. It checks authority first: on a government controlled bank only the head of government or the finance seat (via isNationalIssuer) may act, and the chair is told plainly that the bank has no operational independence. On an independent bank only the seated chair may act, unless chairControlsLocked is set by an administrator. Admins bypass the limits; players do not. The limits, from :
MAX_RATE_CHANGE_DELTA = 0.75: hikes are capped at +0.75pp per adjustment.MAX_RATE_CUT_DELTA = 1.75: cuts are capped at -1.75pp per adjustment.AGGRESSIVE_CUT_SCRUTINY = 10: a cut deeper than 0.75pp adds 10 tochairInfamy, capped at 100. If the government set the rate, the chair's infamy is untouched, since they had no hand in it.RATE_CHANGE_COOLDOWN_TURNS = 6: one change every 6 turns.
Every change appends to rateHistory with the previous rate, the new rate, who changed it and the stated reason, trimmed to the last 50 entries.
The FOMC committee#
Banks are committees, not lone chairs. FOMC_BOARD_SIZE = 7 seats, FOMC_TERM_TURNS = 192 (four game years), a meeting every FOMC_MEETING_INTERVAL_TURNS = 8 turns (six meetings per game year), and player seats get FOMC_PLAYER_VOTE_WINDOW_MS = 24 * 60 * 60 * 1000 of wall clock plus a hard turn deadline FOMC_VOTE_WINDOW_TURNS = 24.
populates any bank without a board, one technocrat NPP per seat (role fomcMember), seat 0 the chair mirrored onto the bank's legacy single chair fields. Alignments alternate hawk, dove, hawk, dove across seats, and term expiries are staggered so roughly one seat opens per window instead of the whole board at once. A government controlled bank gets no committee: the MPC was created by the independence grant, so a pre-1997 Bank of England has no board until legislation seeds one.
The voting logic in is pure and deterministic. Each seat forms a view by running the same Taylor rule the autonomous chair uses (computeNppChairRateTarget and computeNppChairRateStep in ), tilted by its alignment. A step smaller than FOMC_MOVE_THRESHOLD = 0.125pp reads as a hold. The chair proposes a motion; when the chair cannot change the rate (per term cap, cooldown, command economy) the motion can only be a hold. tallyMeeting requires a strict majority of the FULL board, majorityThreshold(7) = 4. Seats that never cast a ballot abstain and count against the motion, so a divided or apathetic board holds. NPP seats auto vote their preference when the meeting opens; player seats vote live or abstain at resolution.
Alignment is temperament, not ideology. defines CHAIR_ALIGNMENT_POLICY: a hawk multiplies the Taylor rule's inflation coefficient (1.5 for hawks), a dove weights growth harder and tolerates more inflation, and each side scales its own hike and cut step sizes. When no alignment is stored, deriveChairAlignment reads the technocrat's personality: stubbornness at or above ambition leans hawk, otherwise dove. oppositeAlignment flips the temperament on term expiry replacement so policy character changes over time.
Nominations#
FomcNomination (, lifecycle in ) is a President's nomination of a named nominee to a specific seatId, confirmed by the Senate through the same lifecycle as cabinet nominations. The nomination carries makeChair, the occupantType (a live player seat or an autonomous NPP seat) and the hawk or dove alignment the President assigns. On confirmation the nominee is installed into that seat on fomcBoard; on rejection the seat is untouched. The older executive path (CentralBankNomination plus ChairSelectionPending) still exists for chair picks that need the nominee's acceptance, and tracks which nominees declined this cycle so re selection skips them.
The rate corridor readout#
answers one question at a glance: is the bank ahead of inflation? corridorVerdict(primeRate, inflation) takes the delta, and with NEUTRAL_BAND = 0.5 calls it restrictive above +0.5pp, accommodative below -0.5pp, neutral in between, and writes the signed copy line. inflationTrendLabel reads the last TREND_WINDOW = 12 observations and reports cooling, rising or steady on a TREND_THRESHOLD = 0.15pp move. Both are computed, never editorial.
Reserves and the reserve currency ranking#
A bank holds two pools: reserveBalance (home currency lending reserves) and spreadFeeReserveBalances (foreign currency accumulated from forex spread fees). builds the portfolio view, converting each foreign balance to the home currency through the live rate map (convertCurrencyAmount goes through the internal anchor: amount / fromRate * toRate) and reporting each currency's share of the FX reserve stack, plus home only, foreign only and total views.
ranks currencies by how much of them is held in reserves across every central bank. It deliberately counts only spreadFeeReserveBalances, not reserveBalance: the home lending reserve is a country stuffing its own currency into its own bank, and counting it would let one large domestic balance crown the leading exchange currency regardless of international demand. Totals are valued in the internal anchor (units divided by the rate) so currencies can be compared, and getLeadingReserveCurrency returns the number one, or null when no bank holds rankable FX reserves yet.
Rank pays. RESERVE_CURRENCY_VOLATILITY_REDUCTIONS = [0.5, 0.25, 0.125] in means the leading exchange currency trades 50% calmer, the second 25% calmer, the third 12.5% calmer, and everything else takes the full market jitter.
Chairs can move money between the two pools, with limits in : RESERVE_POOL_TRANSFER_COOLDOWN_TURNS = TURNS_PER_DAY (24 turns, one real day) and RESERVE_POOL_TRANSFER_MAX_FRACTION = 0.5 of the source pool per action. Moving reserves out of lending is additionally capped so the 70% LOC pool stays at or above outstanding loans.
Forex#
Eighteen currencies trade: FOREX_ACTIVE_CURRENCIES covers USD, GBP, JPY, EUR, IEP, CNY, BRL, NGN, SUR, DDM, FRF, ITL, ESP, SEK, TRL, GRD, ATS and FIM. Nigeria trades even while the country is still coming soon, because its sovereign bonds already settle in NGN and holders would otherwise be trapped; the same logic keeps the 1979 only countries exchangeable in every era. The NPP run planned economies (HU, PL and the rest of the bloc) are not forex active until they decommunise.
Rates are quoted as local currency per one internal unit (₳). Higher rate means a weaker currency, which is why computeStrengthVsReferencePercent in returns (referenceRate / currentRate - 1) * 100 for strength while computeRawRateVsReferencePercent returns the plain rate change. Mixing the two is the classic way to read a devaluation as a rally.
Trading costs#
From :
MARKET_MAKER_SPREAD = 0.01(1%) for an instant market trade.LIMIT_ORDER_SPREAD = 0.0064(about 0.64%) for a limit order.DIRECT_TRADE_SPREAD = 0.0036(about 0.36%) for a direct player to player trade, withDIRECT_TRADE_EXPIRY_TURNS = 24.SECTOR_FX_SPREAD = 0.005(0.5%, half the market maker rate) on cross currency sector flows, so multinational operations carry real FX friction without hitting core business income as hard as a market trade.
A chair can dial the fee charged when their own currency is sold, between FOREX_SPREAD_STRENGTH_MIN = 0.5 and FOREX_SPREAD_STRENGTH_MAX = 1.5 (default 1.0), once every FOREX_SPREAD_STRENGTH_COOLDOWN_TURNS = 48. Higher means more reserve income and more capital flow friction but less liquidity and less reserve appeal; lower means the opposite.
How the rate moves each turn#
combines three components. First, macro drift toward a fundamental target at DRIFT_SPEED = 0.05 per turn (about 48 turns to 90% convergence). The target uses PRIME_RATE_SENSITIVITY = 0.02, INFLATION_SENSITIVITY = 0.015, GDP_GROWTH_SENSITIVITY = 0.01 and TRADE_GROWTH_SENSITIVITY = 0.005, each measured against an era aware baseline resolved by resolveMonetaryBaseline from keyed on the current in game year, so a 1953 world at 1955 is not judged against late 1970s inflation targets.
Second, a separate absolute inflation term. Because the inflation sensitivity measures deviation from each currency's OWN target, a currency with a deliberately high baseline (Italy, Turkey, Sweden in the late 1970s profiles) would otherwise sit at par despite crippling inflation and pay a free real carry. ABSOLUTE_INFLATION_DEPRECIATION_THRESHOLD = 6.0pp sets a healthy inflation ceiling and ABSOLUTE_INFLATION_SENSITIVITY = 0.015 weakens the currency per excess point above it. Developed markets around 1 to 2% inflation are unaffected.
Third, player flow and noise. Volume pressure uses VOLUME_PRESSURE_SENSITIVITY = 0.0001, caps at VOLUME_PRESSURE_CAP = 0.05 (plus or minus 5%), looks back VOLUME_LOOKBACK_TURNS = 24, and accounts for VOLUME_DIRECTION_WEIGHT = 0.2 of the movement direction against 80% macro. Random jitter is RATE_NOISE_MAX = 0.004 (plus or minus 0.4%) per turn, scaled down by the reserve currency multiplier above. On top of that, each currency rolls a directional regime every CYCLE_PRESSURE_TURNS = 12 (about 12 hours) from CYCLE_PRESSURE_BY_REGIME: plus or minus 0.0015 per turn for moderate, plus or minus 0.0007 for slight, zero for neutral. A sustained regime settles only about pressure divided by DRIFT_SPEED off the macro target (roughly 3% moderate, 1.4% slight), so fundamentals generally win. Everything is finally clamped to RATE_FLOOR_MULTIPLIER = 0.5 and RATE_CEILING_MULTIPLIER = 1.5 of the base rate.
Era anchored starting rates#
Seed rates live in four authored tables in : INITIAL_RATES (2019, the STARTING_YEAR), INITIAL_RATES_1991, INITIAL_RATES_1979 and INITIAL_RATES_1953. The 1953 table is Bretton Woods par: GBP 0.357 (the $2.80 par), JPY 360.0 (the Dodge Line peg fixed 1949 to 1971). The 1991 table uses IMF IFS annual averages: GBP 0.57, JPY 134.5, CNY 5.32, NGN 9.9, and a DEM to EUR equivalent of 0.85 derived through the fixed 1.95583 DEM per EUR parity. Brazil is deliberately excluded from 1991 (cruzeiro hyperinflation makes a parity meaningless) and keeps the 2019 BRL 5.0 as a divide by zero guard. The 2019 table has USD 1.0, GBP 0.75, JPY 106.0, EUR 0.92, BRL 5.0, CNY 7.2, NGN 1550.
getInitialRatesForYear(year) covers everything between. It returns the authored table unchanged at an anchor year, so 1953, 1979, 1991 and 2019 worlds are byte identical, clamps below the first anchor and above the last rather than extrapolating, and otherwise interpolates GEOMETRICALLY between the two surrounding anchors (exp(log a + (log b - log a) * t)). Exchange rates compound: straight line interpolation put Nigeria at roughly 450 naira in 1999 against a real rate near 97, while the geometric mean gives about 42. A country present in only one anchor keeps that anchor's rate, since a missing entry means "not authored", not "no currency".
Savings and the real rate#
Savings pay half the REAL prime rate. savingsApyPercent in is max(SAVINGS_REAL_RATE_FLOOR_PERCENT, prime - inflation) / 2 with SAVINGS_REAL_RATE_FLOOR_PERCENT = 0.5, so the floor APY is 0.25%. Paying the real rate rather than the nominal prime is what closes the carry trade: a currency with a high nominal prime also has high inflation, so its real spread, and therefore its APY, ends up near everyone else's, and parking money in a high nominal currency no longer yields a free real return. The floor is uniform across every currency, so it confers no cross currency advantage and cannot be arbitraged.
Accrual is simple, per turn: balance * (apy / 100) / TURNS_PER_YEAR, rounded to whole yen for JPY and two decimals otherwise (roundSavingsAmount). Interest accrues every turn but is credited to the balance once every SAVINGS_CREDIT_INTERVAL_TURNS = TURNS_PER_YEAR / 4 = 12 turns (a game quarter); turnsUntilSavingsCredit drives the countdown players see.
One more guard: SAVINGS_POOL_SHARE_CAP = 0.25. Interest accrues only on the lesser of the account's balance and 25% of the whole national pool for that currency (interestEligibleBalance), so a single actor cannot become nearly all of a currency's savings and farm the entire pool's interest. It binds only on accounts above that share; ordinary savers never feel it. The national pool is summed by sumSavingsInCurrency in , which adds per currency currencyBalances.savings buckets plus the legacy savingsOnHand for pre migration characters in the anchor country. Every credit, deposit and withdrawal writes a savingsLedger row via with the balance after, so the player's statement is a real ledger, not a recomputation.
Lines of credit#
A LOC is priced off the same prime rate the central bank sets, plus a spread that depends on the borrower. spreadPercentPointsFromComposite in maps a 0 to 100 credit composite to 5 - (composite / 100) * 6: +5pp over prime at the bottom, -1pp under prime at the top. The composite blends income and wealth. incomeScoreFromPerTurnCurrency annualizes one turn of personal inflow (salary and dividends, not campaign funds) through a saturating 100 * (1 - exp(-annual / 2_500_000)); netWorthScoreFromInternal does the same on internal unit wealth with a 4,000,000 scale. userCreditComposite weights 75% corporate composite plus 25% income for a CEO, or 50/50 income and net worth otherwise. applyDebtLeveragePenalty subtracts up to 30 points as debt to assets rises (ratio * 38, capped), and applyPrimeEnvironmentToComposite pushes the composite down when prime sits above baseline, partly shielding wealthier borrowers.
Interest is simple per turn on principal plus arrears: computeLocInterestForTurn charges (obligation * (prime + spread) / 100) / TURNS_PER_YEAR and the caller adds it to arrears. Repayment is automatic at LOC_PER_TURN_PAYMENT_RATE = 0.01625 / 4 (about 0.40625% of the obligation per turn). The derivation is a real world rule of thumb: roughly 1.625% of principal per month on a modest fixed rate loan, and one game month is four turns. Over about 23 game years a zero interest loan is about 99% repaid. A borrower can flip a currency's account to interest only mode, which pays arrears alone and leaves principal flat, at the cost of LOC_IO_SURCHARGE_PERCENT_POINTS = 2.0 extra on the annual rate for that currency, and mode flips are gated by LOC_PAYMENT_MODE_COOLDOWN_TURNS = 24 (one real day).
Three caps bound borrowing, all in . System wide, one currency exchange may lend at most LOC_DEPOSIT_FRACTION = 0.7 of its savings deposits plus reserves, and availableForExchangeInternal subtracts every borrower's outstanding principal and arrears from that. Per player, LOC_DTI_MAX_FRACTION = 0.7 caps the scheduled per turn payment at 70% of per turn income, which inverts to a maximum principal of (0.7 * income_per_turn) / LOC_PER_TURN_PAYMENT_RATE; income of zero means a limit of zero, since borrowing requires verified income. Second, LOC_NET_WORTH_LIMIT_MULTIPLIER = 1 caps total exposure at current economic equity, and because net worth already excludes current LOC debt, debt funded bond or stock purchases cannot recursively inflate the ceiling. Payments allocate to arrears first, then principal, in a stable currency order.
Money supply aggregates#
assembles M1 from every liquid pot: household, campaign, NPP, corporate, party, government, fund and organization balances. M2 adds household savings, externalBroadMoney, and bank deposits. Bank reserves are tracked as a capacity measure and reported separately. Negative and non finite components are normalized to zero before summing.
Growth is annualized geometrically: (closing / opening) ** (TURNS_PER_YEAR / turnsElapsed) - 1. It returns null, not zero, when the base is missing or the window is shorter than MIN_MONEY_GROWTH_BASE_TURNS = 12. Annualizing a bootstrap rebase over two turns raises the ratio to the 24th power and produces meaningless percentages, and returning null (rather than 0) makes downstream consumers fall back to GDP growth instead of reading a false "money supply is frozen".
Open market operations live in . planOpenMarketOperation clamps requested units to what is actually available (public float for QE, central bank holdings for QT), prices the consideration at face value times market price, and reports a moneySupplyDelta positive for QE and negative for QT. Accumulated holdings feed applyQePriceSupport, a persistent demand support of up to 20% on the rate derived bond price (min(0.2, qeSupportRatio * 0.5)), with the final price clamped between 0.05 and 2.
Autonomous banks act through chooseNppMonetaryOperation in , which compares inflation to getInflationTarget, compares annualized M2 growth to GDP growth (only when the growth reading is reliable), and picks QE, QT, a treasury advance, a liquidity injection or a hold. The emergency paths are bounded: MONETARY_OPERATION_COOLDOWN_TURNS = 6, DIRECT_ADVANCE_GDP_CAP = 0.01 and LIQUIDITY_INJECTION_GDP_CAP = 0.03 in . A treasury advance requires a genuine slump: inflation at least 3pp under target, GDP growth at or below -3%, and a treasury balance worse than half of GDP in deficit.
How it fits together#
The prime rate a chair or committee sets is not a dial in isolation. It feeds the macro target that moves the currency's exchange rate, it sets the savings APY through the real rate formula, and it sets the base cost of every line of credit in that currency. Savings deposits in turn become the lending pool that limits how much credit exists, and the fees on forex trades become the reserve stack that decides which currency is the leading exchange currency and therefore trades calmest. A hike tightens credit, strengthens the currency, pays savers more in real terms and shrinks the borrowing pool through the DTI cap. Nothing in that chain is a special case: it is the same numbers read by different systems.