DRAFT 12.0 — Clean baseline release.
Consolidates the v11.x series (v11.0 → v11.9) plus the v11.10 review fixes into a single coherent baseline. Folds prior change-log entries (§9.x) into established design where they describe current behaviour. The companion agi-economic-model-v12_0.jsx drops historical "What's new in Draft 11.x" UI panels and contains two calibration improvements caught in v11.x review.
Headline result at v12.0 defaults (unchanged from v11.9): 2035 endpoint +$25B, 6-year cumulative −$1,399B, 88% debt-to-GDP, 7.81M employment, Gini 2026 0.222 → 2035 0.176 (~21% reduction).
The v11.9 → v12.0 Gini reduction shifts from 22% to 21% solely because of the child-count calibration (smaller equivalence factor for couple+kids and lone-parent types). All macro fiscal numbers are identical to the dollar.
This document describes a model exploring how Australia could fund a guaranteed daily income (GDI) and absorb the labour-market shock as artificial general intelligence reaches full economic penetration over 2030–2035. The accompanying JSX file (agi-economic-model-v12_0.jsx) is an interactive React artifact with sliders, charts, and tables that compute fiscal balances, household outcomes, and population-wide purchasing power distributions year-by-year.
The model is intended to be resumed and refined. Load this summary and the JSX file into a new conversation to continue.
The model treats GDI as a redistribution mechanism, not a fiscal stimulus. AGI raises productivity; the question is how the gains are shared. GDI plus the wage-ratio variable GST is a mechanism to redirect a share of the gains from automated profits and overseas AI providers to displaced workers, financed partly by new tax instruments and partly by deficit. In an equilibrium framework, household consumption equals consShare × realGDP regardless of which sources fund it (wages, GDI, fund returns, dissaving). The macro GST/RST taxes that consumption once. No fiscal multiplier is added — once equilibrium is reached, all induced flows are baked into observed GDP, wages, profits, and tax collections; isolating "multiplier revenue" would double-count the same dollar.
Timeline: AGI appears ~2030, full augmentation by 2035. All dollar amounts in 2026 Australian dollars.
Population: 28M (2026), 30M (2030), 31.5M (2035). Adult share drifts from 78.0% to 79.5% over the transition (reflecting Australia's ageing population). Macro POP_A = 23.4M, POP_C = 6.6M at 2030.
GDP: $3,200B baseline (2030, in 2026 dollars). Real GDP grows at a configurable rate (default 3% p.a. compounding) plus an AGI productivity bonus scaling with adoption (default 2pp/yr at full adoption). By 2035 at v12.0 defaults, real GDP reaches $3,912B. Nominal GDP deflates as AGI reduces production costs, but real output grows.
AGI adoption: S-curve: 0% → 10% → 30% → 55% → 80% → 100% over 2030–2035.
Australia's position: Negligible AI supplier. Net importer of AI services. Commodity exports (mining, agriculture) and tourism hold up. Education exports decline.
The entire superannuation system is replaced by a single government-managed investment fund:
New progressive brackets (all configurable via sliders), v12.0 defaults:
No Medicare levy in the new system. Capital gains taxed as income, CPI-adjusted only.
Flat 30% for all companies (configurable slider 20–40%).
Franking credits: At the macro level, the 20% reduction in effective non-wage tax rate accounts for franking. At the household level, investment income is taxed at full marginal rates without explicit franking offset — this makes the PP index marginally conservative (~$500–$3K per household).
Base RST/GST: Applied to ALL goods and services including currently-exempt categories (fresh food, health, education, residential rent, financial services). v12.0 default 12% (slider 10–25%).
Variable RST/GST (wage-ratio AGI levy): Each business's variable RST/GST rate is determined by its wage-to-revenue ratio relative to a pre-AGI sector benchmark:
Variable rate = Max levy × (1 − actual wage ratio ÷ benchmark wage ratio)
Equivalently, at industry level: rate = max levy × AGI augmentation fraction.
GST mode (default ON): Uses ABS IO tables (2022–23) aggregated to the 18 model sectors to compute gross GST, input credits, and net GST.
Per-firm break-even cap: At high max levy rates, an AGI-using firm's consumer-facing price would exceed what a fully-human-labour competitor charges. A profit-maximising firm wouldn't adopt AGI under those conditions. Per-industry break-even formula:
r_break = PASSTHROUGH × labInt × (1 + baseRate) / (1 − PASSTHROUGH × labInt × maxAGI)
When maxLevy × aug exceeds r_break, effective adoption is capped via soft cap controlled by the "Adoption price sensitivity" slider (default 60%).
Source | Mechanism | Approx 2035 |
Personal income tax | Progressive brackets on wages + GDI + investment + fund | $236B |
Company tax | 30% flat (post-profit-shifting loss) | $241B |
Base GST/RST | 12% on all consumption | $185B |
Variable GST/RST | Wage-ratio levy on AGI-using firms | $652B |
Fund returns tax | Investment fund distributions taxed at marginal | ~$80B |
Other revenue | 14 decomposed categories with AGI sensitivity | ~$430B |
Import levy | Variable GST applied to imports (~80% collection) | $130B |
Total | ~$1,934B |
The ~$296B "Other" revenue base is decomposed into 14 categories with per-category AGI sensitivity:
Total "Other" revenue at 2035 default scales by (1 − sensitivity × af × otherRevSens) per category, where otherRevSens (slider, default 50%) is a calibration uncertainty factor.
Aggregate annual fund returns are split into three notional tax slabs based on likely-recipient income brackets — younger working holders (high marginal rate), mid-life accumulators (mid bracket), retiree drawdown (low/zero bracket). Tax yield computed by stacking each slab on the assumed primary income tier.
Aggregate non-wage individual income (rents, dividends, business profits) computed as a representative dollar amount per non-displaced worker, taxed at marginal rates. Avoids cliff-edge errors from earlier marginal-increment formulations.
Annual interest = previous-year debt × interest rate (default 4.5%). Compounds into the debt stock unless the balance is positive.
Component | Approx 2035 |
GDI (adult + child) | ~$870B |
Government services (post-AGI savings) | ~$870B |
Debt interest | ~$155B |
Investment fund operations | ~$5B |
Total | ~$1,909B |
Government services labour savings: AGI displaces some government workers. Saved wages can be applied to deflation passthrough (lower service costs to households), profit (returned to fund), or AI imports (purchasing AGI services from foreign providers). Allocation is controlled by the govtEffCap slider (default 70% to deflation, 30% to profit/imports).
18 industries with calibrated GVA shares, employment, labour intensity, and AGI displacement maxima. Sources: ABS National Accounts, Labour Force Survey.
Industry | GVA share | maxAGI | rBreak (default) |
Professional svcs | 11% | 75% | 62% |
Financial svcs | 9% | 80% | 43% |
Health care | 8% | 35% | 56% |
Manufacturing | 6% | 45% | 28% |
Construction | 8% | 25% | 31% |
Retail | 4% | 45% | 38% |
Education | 5% | 55% | 68% |
Mining | 11% | 35% | 15% |
... (10 more) | ... | ... | ... |
Adoption-cap binding: At v12.0 defaults (maxLevy 100%), the per-firm break-even cap binds in 13 of 18 industries — particularly Mining (uncapped 35% → capped 23%), Utilities (40% → 28%), Financial svcs (80% → 58%). Construction, Education, Health, Hospitality, and Other services don't bind because their r_break exceeds their maxAGI.
Industry-output feedback (§industry feedback): When the variable rate raises industry i's consumer price, households substitute toward lower-rate industries. Each industry's output and remaining employment scale by its quintile-weighted demand multiplier. High-rate industries (Financial, IT, Education) shrink ~2-3%; low-rate industries (Hospitality, Wholesale, Utilities) grow ~1-2%. Macro effect is small but consistent.
24 household types derived from ABS Census 2021, SIH 2019-20, HES 2015-16, National Accounts 2021-22, ATO 2022-23 individual percentile data, and Services Australia 2024-25 rates. All income figures are medians, uplifted 4% from SIH+15% to 2026 dollars.
Spending rates by income quintile: RST/GST liability and PP index calculations use ABS HES 2015-16 expenditure-to-income ratios by quintile:
Each type has: population count (hhK, in thousands), adults (a), children (c), two-earner wage/workday data (w1/d1, w2/d2), fund returns (si, scales with pool), investment income (invInc), current-system breakdown (cw=wages, ct=transfers, cs=super pension income), AGI exposure (agiExp), renter fraction (renPct), and tenure type.
Couple + Children (4 types, 2,970K HH, c=1.85 per ABS Census 2021): | Type | hhK | Wages | Transfers | Fund | Inv | AGI Exp | |------|-----|-------|-----------|------|-----|---------| | 1FT+1PT | 1,600 | $130K (85+45) | $8K FTB-A | $17.5K | $5K | 40% | | Both FT | 800 | $195K (105+90) | $4K | $25K | $8K | 55% | | 1 income | 400 | $82K | $12K FTB-A+B | $10K | $2K | 35% | | Own-biz | 170 | $90K (biz) | $5K | $9K | $3K | 20% |
Couple No Children (4 types, 2,040K HH): | Type | hhK | Wages | Transfers | Fund | Inv | AGI Exp | |------|-----|-------|-----------|------|-----|---------| | Both FT | 800 | $180K (100+80) | $0 | $22.5K | $6K | 55% | | 1FT+1PT | 500 | $110K (70+40) | $2K | $15K | $4K | 42% | | Ret self-funded | 500 | $0 | $0 | $60K | $15K | 0% | | Ret part-pension | 400 | $0 | $42K | $30K | $10K | 0% | | Ret full pension | 340 | $0 | $56K | $5K | $3K | 0% |
Lone Parent (3 types, 1,050K HH, c=1.71 per ABS Census 2021): | Type | hhK | Wages | Transfers | Fund | Inv | AGI Exp | |------|-----|-------|-----------|------|-----|---------| | PT work | 550 | $38K | $25K PPP+FTB+RA | $3K | $0.5K | 40% | | Not working | 350 | $0 | $38K PPP+FTB+RA | $1.5K | $0.5K | 0% | | FT work | 150 | $72K | $8K FTB | $5K | $1.5K | 40% |
Lone Person (6 types, 2,640K HH): | Type | hhK | Wages | Transfers | Fund | Inv | AGI Exp | |------|-----|-------|-----------|------|-----|---------| | FT work | 850 | $82K | $0 | $7.5K | $3K | 45% | | PT/casual | 400 | $35K | $5K JSP | $2.5K | $1K | 42% | | Unemployed | 250 | $0 | $20K JSP | $1K | $0.5K | 0% | | Ret self-funded | 350 | $0 | $0 | $40K | $12K | 0% | | Ret pension | 600 | $0 | $30K | $4K | $3K | 0% | | Ret part-pension | 190 | $0 | $18K | $20K | $8K | 0% |
Other (2 types, 780K HH): | Type | hhK | Wages | Transfers | Fund | Inv | AGI Exp | |------|-----|-------|-----------|------|-----|---------| | Other family (multi-gen/sibling) | 390 | $95K (55+40) | $8K | $12.5K | $4K | 40% | | Group household | 390 | $75K (40+35) | $4K | $5K | $2K | 40% |
High Income (4 types, 500K HH, ATO 2022-23 percentile-calibrated): | Type | hhK | Wages | Fund | Inv | AGI Exp | Notes | |------|-----|-------|------|-----|---------|-------| | Couple+kids: senior prof | 200 | $279K (164+115) | $32K | $12K | 55% | 95th pctl couple | | Couple no kids: senior prof | 100 | $279K (164+115) | $32K | $12K | 55% | 95th pctl couple | | Specialist / senior partner | 150 | $385K (308+77) | $48K | $25K | 45% | 98th pctl earner | | Top exec / partner | 50 | $577K (481+96) | $75K | $60K | 35% | 99th pctl earner |
Total: 10,480K households (within ABS estimate of ~10.5M for 2026).
Renter variants use lower income than their Avg counterparts: wages ~15–17% lower, investment ~50% lower, fund returns ~45% lower — reflecting the ABS-documented income/wealth gap. Rents at current capital city medians.
Variant | Wages | Fund | Inv | Rent $/wk | Annual Rent |
Couple+kids 1FT+1PT renter | $110K (72+38) | $10K | $2K | $750 | $39,000 |
Lone parent PT renter | $32K | $1.5K | $0.3K | $680 | $35,360 |
Lone parent not working renter | $0 | $0.8K | $0.3K | $680 | $35,360 |
Single FT renter | $68K | $4K | $1.5K | $500 | $26,000 |
Renter variants have hhK=0 (not counted in population totals or PP distribution). They allow direct comparison of the RST/GST-on-rent impact for politically sensitive groups. Rent attracts base GST only — variable GST is excluded because landlords have no AGI wage-ratio benchmark.
For each household type at each transition year:
Key v11.1 correction: Savings are eventually spent, so all income faces the GST wedge: PP = newNet × ppGain / (1 + effRateHH). Earlier drafts only applied the GST wedge to spent income, understating GST drag on saving households.
Working household types (those with agiExp > 0 and wages > 0) show two rows in the Household Outcomes table:
Retirees, non-workers, and renter variants show only one row.
PP-change distribution computed across PP-change bands (<−20%, −20% to −10%, …,
+20%) for 2030–2035.
Per-earner Bernoulli decomposition: Within-type displacement modelled as each earner independently facing displacement probability p = af × agiExp. Two-earner households produce 4 binomial scenarios (both employed, w1 only, w2 only, both displaced) with weights (1−p)², (1−p)p, p(1−p), p². Mean wage retention 1−p matches macro totEmp exactly. One-earner: 2 scenarios.
This captures within-type variance: "Couple+kids: both FT" at 55% AGI exposure and 100% adoption produces 4 cohorts: 20% both employed (high PP), 50% one displaced (medium PP), 30% both displaced (low PP).
Population PP Distribution chart uses ABS Cat 5204.0.55.011 methodology: sort households by 2026 equivalised disposable income (after-tax / OECD-modified equivalence factor 1 + 0.5×(adults−1) + 0.3×children), bin into 5 equal-population quintiles of ~2.1M HH each. Chart displays ABS 2026-equivalent per-household disposable income anchors ($60K / $98K / $133K / $174K / $326K).
Internal calibration: model's per-household disposable income means run roughly Q1 $48K / Q2 $64K / Q3-Q4 ~$96K / Q5 $182K — within ~15% of ABS at Q1-Q3, ~40% short at Q4-Q5 due to under-representation of mid-to-upper income working-age households in the 24-type taxonomy.
Distributional pattern at v12.0 default 2035: Q1 (lowest) 90% positive; Q5 (highest) 67% loss-or-flat, 33% in <-20% loss tail. The Gini reduction visualised.
Interactive tool letting a user enter their household's current (2026) financial details and expected 2035 work pattern. Computes their PP change at the model's current slider settings.
Wage-rate convention: PSC takes the user's hourly rate at face value as their actual 2035 wage. The Household Outcomes table uses a different convention: it scales 2026 median wages up by wageScale (~10% at default settings) to model "remaining workers complement AGI and share productivity gains". So a household-table row labelled with the same 2026 income may show a slightly higher 2035 PP than the calculator gives. Both are defensible — they answer different questions. PSC tooltip documents this.
S-curve adoption: Given target maxAGI per industry, adoption ramps over 6 years following the global af curve [0, 0.10, 0.30, 0.55, 0.80, 1.0].
Augmentation fraction: At year y for industry i, the unconstrained augmentation is aug_uncapped = maxAGI × af.
Per-firm break-even cap: Computed as r_break_i = PASSTHROUGH × labInt_i × (1 + baseRate) / (1 − PASSTHROUGH × labInt_i × maxAGI_i). When maxLevy × aug_uncapped > r_break_i, the effective adoption is suppressed by an "adoption price sensitivity" factor (slider, default 60%).
Industry-output feedback: Households substitute away from high-rate industries (quintile-weighted basket). Each industry's output and remaining employment scale by demandMult_i. aug and varRSTRate don't depend on demandMult, so single-pass execution is sufficient.
AGI productivity bonus: Each year, real GDP grows by realGrowth + agiProdBonus × af[y] (default 3% + 2pp at full adoption).
Two channels capturing realistic tax-system response under high-stakes AGI levy:
importLevyRevenue = importedConsumption × effVarRate × importBasketFactor × importLevyCollection. Defaults: importLevyCollection=80%, importBasketFactor=75% (reflecting goods-bias of imports). Yields ~$130B/yr at 2035 default. Builds on the 2017 low-value-import GST regime.
profitShiftLoss = compTax × profitShiftRate × (1 + 0.5×af). Defaults: profitShiftRate=7% (calibrated to ATO transfer-pricing audit estimates). Grows with AGI adoption since AGI/IP services are easier to misprice. Yields −$28B/yr at 2035.
Net contribution: +$102B/yr at 2035 default. After this is applied, m.compTax becomes the post-shift value (intentional but documented).
When GDI provides an unconditional floor, some workers exit. UBI evidence (Finland, Stockton, Kenya): 2-5%. Default 3%, range 0-8%. Mechanism: exited workers lose income tax, gain full GDI, reduce consumption.
At 2035 default: 0.24M workers exit (concentrated in low-wage and secondary earners), ~$15B/yr fiscal cost. Bounded sensitivity: 8% max only costs ~$24B more than baseline.
Variable GST levied on foreign-owned firms is partly passed through to AU consumers. Two sliders: foreignFirmShare (default 30%, ABS calibration) × foreignIncidenceConsumer (default 40%, mixed competitive conditions) = 12% of variable GST = ~$78B/yr at peak. Distributionally regressive. Redistributed across households via spend rate (low-income HHs spend more, bear proportionally more).
Productivity-boom-driven capex (compute, robotics, integration). Default 2% of GDP × af. Generates direct jobs in Construction/Mfg/Prof svcs (~4500 jobs per $B GVA), income tax, company tax, GDI savings, GST. Net peak fiscal benefit ~$30-40B/yr.
Computed on per-equivalent-adult disposable income across 24 ABS+ATO-derived household types. Per-earner Bernoulli cohorts capture within-type variance. Not directly comparable to the official ABS Gini (~0.323 in 2019-20), which uses individual-record data.
Cross-year and cross-scenario comparisons are reliable; absolute level is conservative because top-tail dispersion is under-represented even with the 4 high-income types.
Gini values are slightly different from v11.9 (was 0.224 → 0.175) because the v12.0 child-count calibration (couple+kids c=1.85, lone-parent c=1.71) reduces the equivalence factor slightly for these types, marginally redistributing weight in the Gini calculation. Macro fiscal numbers identical.
Each of the 24 types uses MEDIAN income for its hhK households, while macro uses GDP-implied aggregate consumption (mean-based). The ~50% gap between micro × hhK summed and macro consumption is structural:
The macro figure is correct for revenue accounting; the micro figure is correct for typical-household PP analysis. The two answer different questions and need not reconcile in absolute terms.
HH_TYPES sums:
The adult shortfall is structural — group households, multi-adult arrangements, and institutional residents are under-represented in the 24-type taxonomy. The child overhang is smaller (and reduced from 9.0M in v11.9) by calibrating to ABS Census 2021 averages: couple+kids c=1.85, lone-parent c=1.71.
Macro fiscal numbers use POP_A and POP_C directly (not HH_TYPES) so this gap doesn't affect headline results. Per-household child-GDI lines in the household table are now accurate to ABS averages.
Slider | Default | Rationale |
Adult GDI start (2030) | $80/day | Below all major current welfare rates → all replaced cleanly |
Adult GDI end (2035) | $105/day | Less regressive than $110-130 range |
Child GDI start | $20/day | ~FTB-A equivalent |
Child GDI end | $45/day | Substantial child-poverty reduction |
Base GST | 12% | Less regressive on renters; broader base |
Max AGI levy | 100% | Within "no make-work distortion" range |
Company tax | 30% | Counter offshoring (vs 25% would invite more) |
Import leakage | 25% | Status quo; collection improvements modelled separately |
Real GDP growth | 3% | Trend AU growth |
AGI productivity bonus | 2pp/yr at full | Centre of academic estimates 0.5-3pp |
Investment response | 2% of GDP × af | ICT capex 1995-2002 ~1.5pp; default moderate |
Voluntary withdrawal | 3% | Mid-range UBI evidence 2-5% |
Foreign firm share | 30% | ABS-calibrated |
Foreign incidence (consumer) | 40% | Mixed competitive conditions |
Govt efficiency capture | 70% | Govt as AGI-leadership signal |
Adoption price sensitivity | 60% | Slight conservatism vs 50% |
Other revenue AGI sensitivity | 50% | Calibration uncertainty factor |
Demand elasticity | 40% | Deflation passthrough volume effect |
Levy demand elasticity | 50% | Substitution magnitude |
Import levy collection | 80% | 2017 low-value-import precedent |
Import basket factor | 75% | Goods-bias of imports |
Profit shifting rate | 7% | ATO transfer-pricing audit estimates |
Tax-free threshold | $45K | GDI-supplemented baseline |
Brackets | 30/35/40/45% | More progressive than 2026 |
Bracket ceilings | $90K/$120K/$150K/$150K+ | Compressed top |
Calibration — child counts updated to ABS Census 2021 averages.
Effect: HH-attributed children 9.0M → 8.25M (closer to macro POP_C of 6.6M). Reduces equivalence factor slightly for these types. Marginal effect on Gini (reduction headline shifts from 22% to 21%). Per-household child-GDI lines in the household table are now accurate.
No headline impact: Macro fiscal numbers identical to v11.9 — calcModel reads WAGE_GROUPS, totEmp, and POP_C directly (not HH_TYPES.c). 2035 endpoint +$25B, cumulative −$1,399B, 88% debt-to-GDP, 7.81M employment all unchanged to the dollar.
UI cleanup.
PSC tooltip. Added explicit wage-rate convention note distinguishing PSC ("your real wage at face value") from household table ("median wages share productivity gains via wageScale ~10%"). Both conventions are defensible; they answer different questions.
Caught in code review of v11.9; all are bug-class issues with no headline impact:
v | Section refs | Key change |
11.0 | §9.39 | Fiscal multiplier removed (~$400-450B was double-counting equilibrium consumption). |
11.1 | §9.44 | Purchasing power formula corrected — savings eventually spent so all income faces GST wedge: PP = newNet × ppGain / (1 + effRateHH). |
11.2 | §9.45-46 | Quintile-weighted consumption basket; levy demand response (substitution-only, total spending preserved). |
11.3 | §9.47 | AGI adoption price-sensitivity per-firm break-even cap. |
11.4 | §9.48-49 | Industry-output feedback; AGI productivity bonus added to GDP growth. |
11.5 | §9.51-56 | GDP-base bug discovered & fixed; investment response (§9.53); tax arbitrage (Channel A: import levy collection; Channel B: profit shifting). |
11.6 | §9.55b, §9.57-58 | WAGE_GROUPS recalibrated to ATO percentiles; voluntary labour withdrawal; foreign-firm tax incidence. |
11.7 | §9.59-62 | Foreign incidence redistributed into Gini; hardcoded 0.11 GST rate replaced with m.effTotalRST_HH; applyTaxArbitrage cleanup; agiProdBonus slider max raised 3% → 5%. |
11.8 | §9.63 | Within-type displacement re-modelled as per-earner Bernoulli; binomial Gini captures within-type variance. |
11.9 | §9.64-65 | 4 high-income household types (HI category) added per ATO 2022-23 percentiles; income-band chart upgraded to ABS-comparable equivalised disposable-income quintiles. |
11.10 | (review) | Defensive bracket coercion; latent retirees-with-investment bug; cosmetic improvements. (Not released as a separate JSX; folded into v12.0.) |
Detailed change-log for v1 → v10b is preserved in the v11.9 summary (agi-model-summary-v11_9.md) §8-9 if needed. Highlights: v4 introduced the mature macro engine, fund-returns 3-tier model, real GDP growth, debt dynamics. v5 added ABS-grounded household types and AGI displacement per type. v6 introduced PP Change Bands; v8 added employer super cost saving and wage-scaling with growth. v10 added displaced household rows; v10b added Personal Scenario Calculator and corrected 2026 tax function.