Behind the numbers

Game modes

The fixed game-over limits count Angry + Furious adults: Hard: 15 million; Democrat: 30 million; Republican: 50 million. Reaching the selected limit loses the game. Annoyed stays visible in the yellow bar but never counts toward game loss. The party names are mode labels, not models of either party's policies.

Balancing the budget without reaching the loss limit wins; you can keep adjusting afterward. Defaulting on the debt immediately loses and does not remove interest or create savings. Calculator mode has no win or loss, no backlash limit, and no debt-default action; all budget and reaction estimates still update. Switching modes keeps the current scenario and checks it against the new rules. Restart restores all baselines and keeps the selected mode. These are game rules, not forecasts of a fiscal crisis.

The victory comparison

The overall federal budget last ran an annual surplus in fiscal year 2001, which ended September 30, 2001: 25 years before this historical snapshot on September 30, 2026. Treasury budget history.

The 1,036 arrivals shown at victory are a calculated count: 877 distinct people entered or returned to the voting House and 159 entered or returned to the Senate between October 1, 2001 and September 30, 2026. Each person counts once per chamber, so someone who entered both chambers counts twice. This is not 1,036 distinct individuals across Congress or a claim that each personally attempted and failed to balance the budget. Reelections, nonvoting delegates and future announced retirements are excluded. The count uses the congress-legislators service-record dataset; Senate arrivals were cross-checked against the Senate's official list. It is our calculation, not a published official total.

The victory time runs from opening the game to the first valid balanced budget or surplus, including time spent reading or away from the tab. Restart begins a new clock; changing modes preserves the current clock and scenario. The congratulations appears once per run and never in Calculator mode. A win in this simplified simulation is not equivalent to passing or implementing a real federal budget.

Starting point

FY2025 federal receipts were $5.235 trillion, spending was $7.010 trillion, and the deficit was $1.775 trillion. Those figures come from the U.S. Treasury. Non-tax receipts stay fixed; tax and spending changes are applied to that baseline.

Rate changes

Each percentage-point change is applied to a fixed, estimated tax base. The income-bracket and investment bases are simplified assumptions informed by IRS tax-return statistics; they are not official revenue scores. Payroll and corporate changes are scaled from recent receipts.

Cumulative upset people

A game mechanic, not measured public opinion. All tax and spending sections feed one cumulative reaction score. Individual income adds positive percentage-point increases across a taxpayer's top ordinary bracket and every lower bracket. Investment adds occupied dividend and gains-band increases and applicable NIIT. Payroll, corporate and consumer costs add estimated burden relative to income. Reactions now vary within each group instead of changing everyone at the same threshold.

As combined pressure rises from 0 to 5, people gradually become Annoyed. From 5 to 15, they gradually move from Annoyed to Angry. From 15 to 25, they gradually move from Angry to Furious; at 25 or above, the exposed group is fully Furious. For example, at score 10, half the group is Annoyed and half Angry; at score 20, half is Angry and half Furious. These linear tolerance distributions are tunable game assumptions, not polling. They are fixed, not randomized on each slider movement. Each adult counts once, and only Angry + Furious count toward loss. All changes accumulate before the reaction is calculated; cuts do not cancel increases elsewhere.

The bar covers 269,763,509 adults, the Census July 2025 adult population estimate. The original 174,979,293-adult ordinary-income proxy uses IRS 2023 Table 3.4: joint/surviving-spouse returns count as two adults; other returns as one. The remaining 94,784,216 adults form separate synthetic groups, including non-filers and excluded IRS classifications. They are not additional dependents and are not counted twice.

Residual groups use assumed income weights of 60% at $15,000, 30% at $40,000 and 10% at $100,000, and assumed Social Security exposure of 35%, 20% and 10%, respectively, with $20,000 average benefits. These are sensitivity inputs, not measured demographics. They receive spending and consumer effects but no invented ordinary-income, investment, corporate or payroll tax bills. Children remain excluded. Extending those direct tax effects requires better data.

Investment exposure and overlap

Modeled cells separate joint filers from other filers within each ordinary bracket. Their average income selects an income-source bin in IRS Table 1.4; qualified-dividend frequencies and Table 1.4A net long-term-gain frequencies estimate direct exposure. Reported NIIT counts come from Table 3.4. Average taxable income and investment amounts estimate occupied 2025 bands, using joint thresholds for joint filers and single thresholds as a proxy for other statuses. Net long-term gains are a proxy: loss offsets and special-rate gains are not reconstructed. Within each cell, investment-recipient groups are nested to maximize overlap conservatively. Wage and positive business/farm profit groups are also nested; earned-income and investment groups are independent. These are synthetic weights, not observed taxpayer combinations. Tax-deferred-account income is not counted as direct dividend or gains-tax exposure.

Corporate-tax effects

The model allocates 75% of additional corporate tax to capital recipients and 25% to workers, following the CBO incidence assumption described by CRS. These are shares of dollars, not shares of people. Within the included population, wage income weights the worker burden; taxable interest, ordinary dividends and net long-term gains proxy capital income. Amounts are normalized across modeled recipients, then divided by average income to estimate a loss percentage. This is an illustrative long-run allocation, not a wage, market or price forecast. It does not separately model foreign owners, retirement-only holdings or non-filers, and uses unadjusted 2023 income proxies. No consumer-price pass-through is assumed.

Payroll-tax effects

IRS wage and positive business/farm profit frequencies estimate worker households. Business/farm profits proxy self-employment earnings, with 92.35% treated as taxable; partnership earnings, losses and coverage exceptions are not reconstructed. Regular payroll increases include both sides, assuming the employer side ultimately reduces worker compensation, consistent with the long-run incidence assumption described by CRS. Social Security uses the 2025 $176,100 cap per modeled earner; joint returns assume two equal earners. Medicare has no cap. Additional Medicare applies only to earnings above $250,000 jointly or $200,000 for other statuses as a proxy; actual married-separate thresholds differ. IRS threshold guidance. Burden is divided by the cohort's average income, and linked benefit-spending caps still apply.

Customs, excise and estate exposure

Customs changes collections; excise controls change named rates or rate schedules. Additional consumer-tax revenue creates a broad cost proxy: 95% for customs, informed by CBO's tariff-incidence assessment, and 100% for consumption excises, informed by CBO's excise-tax analysis. An assumed 80% goes to modeled adults. Per-adult weights grow with the square root of income. Purchasers of individual products are not identified, so broad exposure is not a claim that everyone drinks, smokes, flies or gambles.

Buybacks, foundation income, college endowment income and tax-exempt executive compensation instead allocate added costs by modeled capital income. For institutional taxes this is a deliberately rough proxy: it does not reconstruct donors, students, charity beneficiaries or executive households. It is not an official incidence estimate. All channels add pressure before reactions are calculated; adults remain deduplicated. Decreases do not cancel increases elsewhere.

Estate / gift exposure uses 250,000 potentially concerned wealthy adults, informed by the IRS 2019 wealth estimate, distributed among capital recipients in the top three ordinary brackets. Exemptions stay fixed.

Excise tax rates and revenue

Controls use a statutory percentage, cents per unit, dollars per unit, or an annual fee pool. A schedule control at 100% preserves the 2025 rates; 200% doubles its rates and 0% removes them. Multiple rates within wine, tobacco, aviation fuel, tires, heavy-vehicle use, chemicals, coal, foreign insurance, wagering and sporting goods scale together rather than pretending they share one statutory rate. Eligibility thresholds, exemptions and most credits stay fixed. Beer and spirits scale reduced rates alongside the displayed general rate. Coal scales both per-ton rates and its percentage cap; fixed fishing-rod caps do not scale.

Sources: IRS Publication 510 (2025), 2025 inflation-adjusted flight charges, petroleum and arrow shafts, TTB alcohol and tobacco schedules, foundation income tax, and 2025 institutional excise taxes. The endowment slider uses the 2025 1.4% rate, not later graduated rates.

Revenue change = allocated baseline receipts × (new rate / baseline rate − 1). Gasoline ($26.377B), diesel ($11.854B) and airfare ($11.936B) retain their DOT FY2025 Note 20 proxies. Airfare uses an assumed 75% of ticket collections; the remainder weights domestic segments. The remaining adjustable pool is $54.770B. It is distributed proportionally using IRS FY2024 Table 20 weights, with DOT FY2025 weights for trucks, flight facilities, cargo and aviation fuel. Buybacks use a $9B weight informed by CBO's FY2025 summary. Endowment ($0.5B), executive compensation ($0.6B), wagering ($0.05B) and ozone chemicals ($0.01B) use explicitly assumed weights because comparable category amounts are unavailable. These are weights, not official FY2025 receipts; every weight is normalized to that pool.

Named controls together allocate $104.937B of the existing $105.937B excise baseline. A fixed $1B reconciliation provision represents unclassified receipts and net accounting differences; it is a modeling assumption, not a tax policy. Refunds, penalties, obscure unclassified taxes and occupational fees have no invented adjustable rate. No receipts are added at startup or double-counted. Product volumes, taxable income and fares remain fixed even at extreme rates. The linear model does not forecast avoidance, demand changes or behavioral revenue responses.

Spending changes

Annual category spending ranges from zero to its FY2025 baseline. Each slider starts at the maximum; restoring spending cannot exceed that baseline. Social Security payroll receipts remain capped at benefits, and combined Medicare payroll receipts at Medicare spending. Zero linked benefits means zero linked taxes. SSI is a separate program and does not change the Social Security payroll cap. These are game constraints, not current tax law.

Major categories retain the CBO FY2025 Table 3 rounded figures. Expanded categories use the Treasury final September 2025 statement, Tables 3 and 5, in billions. Agency net outlays are not gross spending, appropriations or loan volume.

The former $868.988B catch-all is replaced by $1,050.754B of named net categories plus a fixed −$181.766B reconciliation. That negative amount combines FDIC net recoveries, undistributed offsets, Executive Office / GSA net balances and differences between the retained rounded figures and Treasury detail. It is not a cuttable program. Interest remains fixed at $1,028B. Baseline spending remains exactly $7,009.988B; changes subtract savings from that total. Holding net receipts and offsets fixed is a simplification: abolishing an agency would not necessarily save its displayed amount.

HHS's $1,884.277B net total less the existing $992B Medicare and $668B Medicaid baselines leaves $224.277B. This is split into NIH ($47.341B), CHIP ($23.067B), ACF family services ($75.676B), public health / community care ($49.832B) and remaining health programs ($28.361B). Public health combines FDA, HRSA, Indian Health Service, CDC, SAMHSA, AHRQ and Community Living. The remainder is calculated, not a separately published budget category.

SSA's $1,646.515B less Social Security benefits ($1,569B) and SSI ($69.472B) leaves $8.043B for administration and transfers. Treasury's $243.296B non-debt total less the existing $225B refundable-credit category leaves $18.296B. Independent agencies ($37.952B) exclude the FDIC's negative $31.164B; FDIC stays in fixed reconciliation. Military retirement / defense civil programs and Army Corps civil works are separate from DoD military spending. Refundable credits include premium tax credits and are not added to HHS a second time.

Expanded spending reactions

New categories use the same gradual cut curves and partial synthetic overlap as existing departments. These are tunable gameplay assumptions, not official estimates of angry people. Direct groups include recipients, household spillovers, workers and service users. Broad concern is added only for selected public services. Children and overseas beneficiaries are not added to the U.S. adult population; CHIP and family services proxy adult household effects, and international assistance proxies U.S. workers and concerned adults.

Potential direct exposure in millions of adults / full-cut score: SSI 9 / 70; federal employee benefits 8 / 60; military retirement 6 / 60; transportation 35 / 35; housing 12 / 65; labor 12 / 50; energy 7 / 30; Interior 10 / 30; Justice 10 / 35; Commerce 6 / 25; State 3 / 25; international assistance 2 / 25; Army Corps 8 / 35; NIH 12 / 35; CHIP 8 / 60; family services 18 / 55; public health 20 / 45; health administration 6 / 25; NASA 4 / 25; NSF 5 / 30; Treasury 5 / 25; SSA administration 12 / 35; legislative branch 2 / 25; judiciary 5 / 35; SBA 2 / 35; independent agencies 15 / 30.

Housing, SSI, CHIP, family services and public health use low-income weights; employment, industry, diplomacy, NASA and SBA use worker weights. Retirement and SSA administration use the existing age / household proxy; NSF uses education weights. Remaining categories use broad adult weights. These weights do not reconstruct actual eligibility or employment.

At a full cut, broader concern reaches this share of all modeled adults / score: transportation 30% / 12; energy 20% / 10; Justice 30% / 15; State 15% / 10; international assistance 10% / 8; NIH 20% / 10; public health 25% / 12; NASA 15% / 8; legislative branch 20% / 12; judiciary 25% / 12. Both concern exposure and score scale with the square of the cut fraction. All affected adults overlap with the same population; amounts cannot be added as independent headcounts.

Spending-cut reactions

Exposure grows with the cut. If the fraction cut is c, directly upset exposure is the potential direct group multiplied by c^0.7. Its score is the full-cut score multiplied by c × (0.5 + 0.5c). Deep cuts reach more people and intensify reactions. This is an illustrative sensitivity curve, not evidence that small cuts eliminate only waste.

Defense, Veterans Affairs and Homeland Security also generate broader concern. At a full cut, potential direct groups are 12M, 16M and 8M adults; broader concern reaches 60%, 70% and 50% of all adults. Concern exposure and score scale with c². Full-cut direct scores are 35, 60 and 25; concern scores are 20, 30 and 20. Direct and concern groups overlap and are deduplicated.

Defense includes military families, civilian staff and contractor households as a worker-weighted proxy. The GAO defense supply-chain report explains the wider industrial network. VA includes benefits, healthcare and household spillovers: FY2025 benefit recipients and planned patients provide context, but their counts cannot simply be added. Homeland Security's concern reflects disaster response, transportation security and cybersecurity. Group sizes and concern percentages are game assumptions, not measured opposition.

Social Security uses IRS benefit-reporting frequencies plus residual-group assumptions; its direct score is adjusted by benefits relative to income including untaxed benefits. Medicare uses the same recipient group as an age/disability proxy. Existing potential direct groups remain Medicaid 18M, agriculture 22M, education 20M, EPA 12M and refundable credits 20M. Their full-cut scores remain 60, 35, 35, 20 and 30; Medicare's is 40. The old 25M catch-all reaction group has been removed in favor of the named groups described above.

Each category and concern channel uses a different reproducible stratum offset within each income group. This gives partial synthetic overlap, not measured joint eligibility or exact independence. Active scores add together while the population remains conserved. Restoring a category removes its pressure; automatically reduced payroll taxes do not offset other harms. Debt interest and accounting reconciliation have no spending-cut reaction. All exposure and anger assumptions can be tuned.

What is excluded

The budget estimate does not predict changes in work, investment, prices, avoidance, deductions, credits, growth, debt interest, or future spending. The reaction bar uses illustrative wage, consumer-cost and benefit/service-loss assumptions, not a forecast of public opinion. Exact eligibility, product purchasers, geographic exposure, contractors, implementation choices and effects outside the included adult population are not reconstructed. Customs lacks a single federal rate, so its control changes collections. Multi-rate excise categories scale their 2025 schedules.