REPATRIATE WASHINGTON POLICY RESEARCH INSTITUTE
Washington FY2029 Fiscal Explorer
Methodology and Technical Documentation
October 2026
INTRODUCTION
What follows is a formal description of the methodology used to estimate, for each of Washington State's 49 legislative districts (LDs): (1) the likely geographic source of revenue collected under Washington's millionaires tax; and (2) the likely geographic allocation of the expenditures and benefits funded by that revenue under the programs established by the Legislature.
These quantities must be modeled rather than directly measured because, to our knowledge, no official source reports either millionaires-tax collections or the resulting expenditures by Washington legislative district. We do, however, have authoritative statewide data and substantial geographic information from public sources. Internal Revenue Service data provide the statewide number of Washington tax returns reporting adjusted gross income of at least $1 million and the aggregate adjusted gross income reported by those filers. IRS and U.S. Census Bureau data then provide a basis for estimating how this upper-income population is distributed among Washington's legislative districts.
When these data are examined geographically, an important feature emerges: Washington's potential millionaires-tax contributors are not evenly distributed among the state's 49 legislative districts. Upper-income taxpayers are substantially concentrated in portions of the Puget Sound region, while many other legislative districts—including much of eastern and southwestern Washington—contain considerably smaller estimated shares of the statewide millionaire population.
This geographic concentration has an important consequence for the fiscal incidence of the tax. Millionaires-tax collections are expected to be considerably more geographically concentrated than the expenditures financed by those collections. The programs funded by the tax serve populations distributed much more broadly across Washington. The tax therefore creates the potential for substantial geographic redistribution: some legislative districts are expected to contribute more in millionaires-tax revenue than they receive in modeled tax-funded benefits, while others are expected to receive more in modeled benefits than they contribute.
The statewide totals are known or officially estimated; what must be modeled is how those totals are distributed among Washington's 49 legislative districts.
The accompanying interactive map estimates this relationship for each Washington legislative district. For every district, it presents the modeled Tax Outflow attributable to the district, the modeled Tax Inflow allocated to the district, and the resulting Net Balance between the two.
These district-level figures should not be interpreted as direct measurements. No official dataset currently provides the information necessary to make such measurements. They are estimates derived from publicly available federal and state data, independently constructed geographic relationships, statistical modeling, and authoritative statewide fiscal controls. Where the available evidence does not support a single precise estimate, the methodology explicitly retains that uncertainty rather than concealing it.
The sections that follow describe the data, assumptions, statistical models, geographic allocation procedures, validation tests, fiscal controls, and uncertainty framework used to construct these estimates and the interactive map.
EXECUTIVE SUMMARY
The Washington FY2029 Fiscal Explorer estimates the geographic incidence of Washington's new millionaires tax and the expenditures financed by that tax across the state's 49 legislative districts.
The analysis addresses two separate questions:
Where is the tax likely to be collected?
and
Where are the expenditures financed by that tax likely to accrue?
For every legislative district, the central accounting relationship is:
Net Balance = Tax Inflow - Tax Outflow
A positive Net Balance indicates that a district is estimated to receive more in tax-funded benefits than is collected from its residents under the millionaires tax.
A negative Net Balance indicates that the district is estimated to generate more millionaires-tax revenue than it receives in modeled benefits.
Neither quantity is directly observed at the legislative-district level. The Fiscal Explorer is therefore a geographic fiscal-incidence model, not a database of observed district-level tax payments and government expenditures.
A Common FY2029 Accounting Period
The final model is expressed as a Washington State Fiscal Year 2029 snapshot.
Where authoritative FY2029 statewide fiscal controls exist, those controls determine the statewide dollar amounts. Where the best available geographic evidence originates in an earlier period, those data are used to estimate geographic shares and are normalized to the applicable FY2029 statewide control.
Historical observations therefore determine geographic distribution, not FY2029 statewide magnitude.
Tax Outflow
The statewide Tax Outflow model is anchored to the Washington Department of Revenue's official estimate of:
$2.698 billion in FY2029 millionaires-tax revenue.
The Fiscal Explorer does not independently forecast that statewide total. It estimates how the official statewide control is distributed geographically.
For Tax Year 2022, authoritative IRS Statistics of Income data report:
21,530 Washington tax returns with adjusted gross income of at least $1 million
reporting approximately:
$65.957 billion in aggregate adjusted gross income.
IRS ZIP-code Statistics of Income data are geographically allocated to legislative districts using Census population geography.
The geographic procedure was independently tested against official IRS Congressional District data. Candidate models for millionaire-taxpayer geography were subsequently evaluated against independent Washington Department of Revenue capital-gains-taxpayer geography.
The selected model uses the geographic distribution of upper-income adjusted gross income and dividend income.
The resulting 49 legislative-district estimates are normalized to reproduce exactly the authoritative IRS statewide total of 21,530 millionaire returns.
Estimating the geographic distribution of income above the $1 million threshold is less certain than estimating the distribution of millionaire taxpayers. Rather than conceal that uncertainty by selecting a single preferred geographic specification, the methodology retains seven tax-geography scenarios.
Every scenario reconciles exactly to the same official FY2029 statewide revenue control of $2.698 billion.
Tax Inflow
The benefit-allocation model is constructed independently from the Tax Outflow model.
Millionaire-taxpayer geography is not used to determine expenditure geography.
The FY2029 model contains the following expenditure controls:
Component Modeled FY2029 Amount
Working Families Tax Credit $250.0 million
K-12 education $1.000 billion
Health and human services $750.0 million
Higher education $400.0 million
Fair Start $134.9 million
Unallocated $163.1 million
Total $2.698 billion
The first five components produce $2.5349 billion of geographically allocated benefits.
The remaining $163.1 million is retained as unallocated rather than assigned to districts without a sufficiently defensible geographic basis.
Net Balance and Uncertainty
Net Balance is calculated for every district under all seven tax-incidence scenarios.
Districts are then classified according to whether their donor-or-recipient status survives the full range of modeled tax geography.
Of Washington's 49 legislative districts, 38 retain the same donor-or-recipient sign under all seven scenarios.
The purpose of the uncertainty framework is not to make uncertainty disappear. It is to identify which conclusions remain stable despite it.
TECHNICAL METHODOLOGY
1. Analytical Framework
For every legislative district:
Net Balance = Tax Inflow - Tax Outflow
Statewide modeled Tax Outflow is:
$2.698 billion
while geographically allocated Tax Inflow is:
$2.5349 billion.
Consequently:
Statewide District Net Balance = $2.5349 billion - $2.698 billion = -$163.1 million
The statewide negative balance is intentional. It represents the $163.1 million retained as geographically unallocated rather than artificially distributed among legislative districts.
2. FY2029 Snapshot Rule
The final reporting period is Washington State FY2029.
The governing rule is:
Where authoritative FY2029 controls exist, they are used directly. Where geographic allocation data originate in another period, those data are used to estimate geographic shares and are normalized to the applicable FY2029 statewide control. Source year and normalization method are retained in provenance.
This separates two analytical functions:
Statewide control = magnitude
Geographic evidence = distribution
TY2022 IRS observations, for example, provide information about the geography of upper-income taxpayers. They are not represented as FY2029 tax receipts.
3. Geographic Foundation
IRS ZIP geography and Washington legislative-district geography do not coincide.
The model therefore constructs population-weighted relationships between Census ZIP Code Tabulation Areas (ZCTAs) and legislative districts.
For every ZCTA that intersects more than one legislative district:
ZCTA-to-District Weight = Population of the ZCTA Within the District / Total ZCTA Population
The weights for each positive-population ZCTA sum to 100%.
An IRS ZIP-level quantity is then allocated according to:
Allocated District Value = ZIP Value x ZCTA-to-District Weight
District totals are obtained by adding all allocated ZIP values belonging to that district.
This procedure conserves the geographically resolved source quantity while translating ZIP-level economic data into legislative-district geography.
4. Independent Validation of the Geographic Method
Before the geographic procedure was used to estimate millionaire geography, it was tested against an independently published IRS district geography.
TY2022 IRS ZIP $200,000+ return counts were allocated through Census geography to Washington congressional districts and compared with official IRS Congressional District Statistics of Income.
The comparison produced:
Statistic Result
Pearson correlation 0.999641
Spearman rank correlation 1.000000
Mean absolute district-share error 0.116 percentage points
Maximum district-share error 0.334 percentage points
The congressional-district controls were not used to construct the ZIP allocation.
They therefore provide an independent test of the geographic procedure.
5. Statewide Millionaire Controls
For Tax Year 2022, IRS Statistics of Income report:
21,530 Washington returns with adjusted gross income of at least $1 million.
Those returns reported:
$65,956,796,000 in aggregate adjusted gross income.
These are treated as authoritative statewide controls.
They do not provide legislative-district geography.
6. Upper-Income IRS Predictors
IRS ZIP Statistics of Income do not directly publish the number of $1 million+ returns by ZIP.
They do, however, report characteristics of returns in the $200,000+ AGI category.
Candidate predictors included geographic shares of:
$200,000+ return count;
adjusted gross income;
interest income;
dividend income;
business income;
capital gains;
pension income; and
partnership/S-corporation income.
These quantities are not treated as equivalent to millionaire taxpayers. They provide observable characteristics of the upper-income population from which the still-higher-income tail can be estimated.
7. Independent Washington DOR Validation
Washington Department of Revenue TY2022 capital-gains-taxpayer geography provides an independent measure of upper-tail taxpayer concentration.
Capital-gains taxpayers are not treated as synonymous with taxpayers having AGI of at least $1 million.
Their geographic distribution is instead used as an external validation target.
DOR individually disclosed 46 legislative districts. Districts 14, 19, and 29 were disclosed only as a combined group containing:
7 taxpayers and $344,057 in payments.
The model did not invent individual observations for those districts. They were excluded from the individual-district model-validation tournament.
Capital-gains payment dollars were retained as diagnostic evidence rather than used as the primary count-model selection target.
8. Millionaire-Taxpayer Model
Six predeclared candidate specifications were evaluated.
The selected model was designated M5_AGI_DIVIDENDS.
Its raw district score is calculated as:
**Raw Score = -3.677917462427
(937.623967905126 x District AGI Share)
(2532.374382281683 x District Dividend-Income Share)**
where:
District AGI Share is the district's share of statewide AGI reported by $200,000+ returns; and
District Dividend-Income Share is the district's share of statewide dividend income reported by $200,000+ returns.
These coefficients were frozen before the downstream fiscal-incidence calculations were performed.
9. Model Selection by Leave-One-Out Cross-Validation
Candidate models were evaluated using leave-one-out cross-validation (LOOCV).
Under this procedure, each validation district is withheld in turn. The model is fitted using the remaining districts, and the withheld district is then predicted.
Performance is measured primarily using Mean Absolute Error (MAE):
MAE = Average Absolute Difference Between Observed and Predicted Values
The selected M5 model produced:
LOOCV MAE = 13.098632
compared with approximately:
35.866
for the simple $200,000+ return-count baseline.
The selected model therefore reduced validation MAE by approximately:
63.5%.
Additional M5 validation statistics were:
RMSE: approximately 18.912;
Pearson correlation: 0.982132;
Spearman correlation: 0.835822.
The model was selected on independent validation performance, not on the appearance of the final donor/recipient map.
10. Calibration to the Statewide Millionaire Control
Raw model scores are converted into each legislative district's share of the statewide total:
District Share = District Raw Score / Sum of All District Raw Scores
The estimated number of millionaire returns in each district is then calculated as:
Estimated Millionaire Returns = 21,530 x District Share
Because the 49 district shares sum to 100%, the district estimates reconcile exactly to the authoritative statewide IRS control:
Total Estimated Millionaire Returns = 21,530
The model therefore determines where Washington's statewide millionaire population is estimated to reside.
It does not determine how many statewide millionaire returns exist; that statewide total comes directly from the IRS.
11. Statewide Excess AGI
The statewide millionaire population reported:
$65.956796 billion
in aggregate adjusted gross income.
The first $1 million associated with each of the 21,530 millionaire returns equals:
21,530 x $1 million = $21.530 billion
The resulting statewide excess-AGI control is:
$65.956796 billion - $21.530 billion = $44.426796 billion
This is an analytical measure of aggregate AGI above the $1 million floor in the IRS source population.
It is not used as a substitute for the Washington Department of Revenue's official FY2029 revenue estimate.
12. Dollar-Geography Reliability Screen
Locating millionaire taxpayers and allocating millionaire income dollars are different statistical problems.
Potential IRS variables were therefore separately evaluated for geographic reliability.
Three were classified HIGH:
Variable Mean Absolute Geographic Error Pearson Correlation
Adjusted gross income 0.540360 pp 0.990939
Business income 0.226157 pp 0.997681
Pension income 0.267503 pp 0.991782
Interest income, dividend income, capital gains, and partnership/S-corporation income did not meet the same reliability standard for excess-AGI allocation.
This distinction is important.
Dividend geography contributes to the independently selected millionaire-count model, but dividend-dollar geography is not treated as a HIGH-reliability basis for allocating millionaire excess income.
13. Seven Tax-Geography Scenarios
Because district-level millionaire excess AGI is not directly observable, the methodology does not designate one geographic specification as the true answer.
Four single-factor scenarios are retained:
E0 - COUNT
Millionaire-taxpayer count share.
E1 - AGI
Upper-income adjusted-gross-income share.
E2 - BUSINESS
Upper-income business-income share.
E3 - PENSION
Upper-income pension-income share.
Three additional scenarios combine the reliable geographic signals using geometric means:
E4 - COUNT + AGI
E4 = (Count Share x AGI Share)^(1/2)
E5 - COUNT + AGI + BUSINESS
E5 = (Count Share x AGI Share x Business Share)^(1/3)
E6 - COUNT + AGI + BUSINESS + PENSION
E6 = (Count Share x AGI Share x Business Share x Pension Share)^(1/4)
Each scenario is then normalized so that the 49 district shares sum to 100%.
No scenario is designated the winner.
The spread among the seven scenarios is retained as structural uncertainty.
14. FY2029 Tax-Harvest Control
The Washington Department of Revenue FY2029 statewide millionaires-tax revenue control is:
$2.698 billion.
For each of the seven scenarios:
District Tax Outflow = $2.698 Billion x District Scenario Share
Therefore:
All 49 District Tax Outflows = $2.698 Billion
under every scenario.
The model consequently contains uncertainty about where FY2029 tax revenue originates, not about the statewide FY2029 revenue control used in the analysis.
15. Benefit-Allocation Model
The expenditure side is constructed independently of the Tax Outflow model.
No millionaire-taxpayer variable, millionaire-count estimate, excess-AGI scenario, or tax-harvest result is used to determine benefit geography.
For every expenditure program:
District Program Benefit = Statewide Program Control x District Program Allocation Share
The district allocation shares are constructed independently for each program and normalized so that the 49 districts sum to 100%.
Total district Tax Inflow is the sum of the district's five modeled program allocations.
16. Working Families Tax Credit Allocation
The modeled FY2029 Working Families Tax Credit control is:
$250 million.
The WFTC geographic model uses household income and family characteristics to estimate the relative geography of likely program recipients.
The underlying district model includes:
number of households;
median household income;
income propensity;
total families;
families-with-children proxy;
child-family share; and
a child adjustment.
These characteristics are combined into a district WFTC weight and converted to each district's share of the statewide modeled WFTC population.
The final calculation is:
District WFTC Benefit = $250 Million x District WFTC Share
This represents modeled recipient-incidence geography. It is not a claim that actual FY2029 WFTC payments by legislative district are presently known.
17. K-12 Education Allocation
The modeled FY2029 K-12 control is:
$1.000 billion.
The geographic driver is estimated public K-12 enrollment by legislative district.
The underlying demographic source is the 2024 American Community Survey 5-Year estimates, Table B14002, with Washington Office of Superintendent of Public Instruction administrative enrollment used as an external benchmark.
For each district:
District K-12 Benefit = $1.000 Billion x District Share of Statewide Public K-12 Enrollment
This allocates the education component according to the geographic distribution of the population most directly associated with the funded service rather than general population or millionaire-taxpayer geography.
18. Health and Human Services Allocation
The modeled FY2029 health and human-services control is:
$750 million.
The geographic driver is estimated Medicaid population by legislative district.
The underlying demographic source is the 2024 American Community Survey 5-Year estimates, Table C27007, with Washington Health Care Authority Apple Health enrollment used as an external benchmark.
For each district:
District Health/Human Services Benefit = $750 Million x District Medicaid Share
This represents modeled geographic benefit incidence.
It is not an assertion that every dollar categorized as health and human services will literally be distributed in direct proportion to Medicaid enrollment.
19. Higher-Education Allocation
The modeled FY2029 higher-education control is:
$400 million.
The geographic driver is estimated public higher-education enrollment by legislative district.
The underlying demographic source is the 2024 American Community Survey 5-Year estimates, Table B14004.
For each district:
District Higher-Education Benefit = $400 Million x District Share of Statewide Public Higher-Education Enrollment
20. Fair Start Allocation
The FY2029-revenue-attributable Fair Start control is:
$134.9 million.
The geographic driver is the population of children under age five by legislative district.
The underlying demographic source is the 2024 American Community Survey 5-Year estimates, Table B09001.
For each district:
District Fair Start Benefit = $134.9 Million x District Share of Washington Children Under Age Five
Fair Start Timing Qualification
Fair Start requires special temporal treatment.
Its statutory July 1, 2029 start falls on the first day of Washington State FY2030.
The $134.9 million displayed in the Fiscal Explorer is therefore treated as a:
FY2029-revenue-attributable Fair Start allocation
rather than literal FY2029 cash expenditure.
This preserves the analytical relationship between FY2029 tax generation and the spending commitment associated with that revenue without misrepresenting the timing of the actual expenditure.
21. Benefit Reconciliation
The frozen FY2029 benefit controls are:
Component Control
Working Families Tax Credit $250.0 million
K-12 education $1,000.0 million
Health and human services $750.0 million
Higher education $400.0 million
Fair Start $134.9 million
Geographically Allocated $2,534.9 million
Explicitly Unallocated $163.1 million
Statewide Control $2,698.0 million
Every geographically allocated component contains all 49 legislative districts.
Each component's district shares are normalized to 100%.
Each component reconciles to its adopted statewide dollar control.
The five geographically allocated components total:
$2.5349 billion
and:
$2.5349 billion + $163.1 million = $2.698 billion.
The $163.1 million residual is deliberately left unallocated rather than distributed merely to force district Tax Inflow to equal statewide Tax Outflow.
22. Net Fiscal Incidence
For each legislative district and each tax-geography scenario:
Net Balance = Tax Inflow - Tax Outflow
If Net Balance is less than zero, the district is a modeled net donor under that scenario.
If Net Balance is greater than zero, the district is a modeled net recipient under that scenario.
Tax Inflow is held fixed across the seven scenarios.
The uncertainty represented in the final classification therefore arises from alternative defensible estimates of Tax Outflow geography.
23. Robustness Classification
For every district, the model counts how many of the seven scenarios produce a positive Net Balance.
Classification follows the predeclared rule:
Positive Net-Balance Scenarios Classification
0 of 7 Robust Donor
1-2 of 7 Likely Donor
3-4 of 7 Scenario-Dependent
5-6 of 7 Likely Recipient
7 of 7 Robust Recipient
An exact zero is treated separately rather than counted as positive or negative.
The model also retains each district's complete uncertainty range:
Net Minimum = Lowest Net Balance Across the Seven Scenarios
Net Maximum = Highest Net Balance Across the Seven Scenarios
If a district's entire range lies below zero, it remains a donor under every modeled tax geography.
If its entire range lies above zero, it remains a recipient under every modeled tax geography.
24. Public-Facing Middle Estimate
The interactive map presents one Tax Outflow and one Net Balance figure rather than displaying seven scenario values simultaneously.
For presentation purposes:
Middle Tax Outflow = (Minimum Tax Outflow + Maximum Tax Outflow) / 2
and:
Middle Net Balance = Tax Inflow - Middle Tax Outflow
This midpoint is:
not an eighth statistical scenario;
not a probability-weighted expectation;
not a forecast of the most likely outcome; and
not the basis of donor/recipient classification.
It is simply a descriptive center of the certified scenario range.
25. Modeled Local Economic Benefit
The interactive map separately presents an illustrative Modeled FY2029 Local Economic Benefit.
The calculation is:
Modeled Local Economic Benefit = 1.5 x Tax Inflow
The underlying unmultiplied modeled allocation is displayed alongside it.
The 1.5 multiplier is applied to gross Tax Inflow for every district.
It is not included in Net Balance.
The fiscal accounting remains:
Net Balance = Tax Inflow - Tax Outflow
not:
Net Balance = Modeled Local Economic Benefit - Tax Outflow
The multiplier is therefore a separate economic-impact presentation layer rather than part of the tax-and-expenditure accounting model.
26. Forensic Validation and Unexpected Results
Unexpected results were treated as reasons for investigation, not reasons for automatic correction.
A district estimate was not changed merely because it appeared surprising.
Instead, two questions were asked:
Can the result be reproduced from the frozen inputs?
Does independent evidence reveal a material geographic or statistical error?
27. Legislative District 14 Forensic Review
LD14 received particular scrutiny because its modeled millionaire population appeared high relative to intuitive expectations and because DOR did not individually disclose its capital-gains-taxpayer count.
The frozen model estimated approximately:
241 millionaire returns
for LD14.
A six-model sensitivity analysis produced a substantially wider range:
approximately 30.5 to 241.4.
The geographic contribution from ZCTA 98908 was subsequently examined using independent 2022 ACS block-group household-income data.
A whole-block-group diagnostic allocated approximately:
87.8% of $200,000+ households to LD14;
11.7% to LD15; and
0.5% to LD13.
Several of the strongest affluent block groups were entirely within LD14.
This analysis does not prove that LD14 contains exactly 241 millionaire taxpayers.
It does materially reduce the narrower concern that the LD14 result was primarily an artifact of incorrectly assigning the affluent portion of ZCTA 98908 across the LD14/LD15 boundary.
No post-hoc coefficient adjustment was made.
28. Statewide Anomaly Screen
The forensic review was subsequently expanded statewide.
The audit independently reconstructed the frozen millionaire model and examined all 49 districts for:
upper-income return intensity;
AGI concentration;
dividend concentration;
six-model sensitivity;
ZIP concentration; and
standardized anomaly measures.
Fourteen districts were flagged for review.
LD14 was materially unusual but not uniquely anomalous.
The statewide audit was classified:
PASS WITH TARGETED FOLLOW-UP.
It did not alter the frozen model.
29. Provenance Firewall
The current model was constructed de novo.
Historical legislative-district estimates of millionaire counts or millionaire AGI from earlier spreadsheets were prohibited from influencing:
predictor selection;
model coefficients;
calibration;
geographic allocation;
validation;
model selection; or
acceptance or rejection of results.
Historical estimates may be inspected only after model freeze as external benchmarks.
This prevents the current analysis from reproducing earlier assumptions simply because they already existed.
30. Separation of Analytical Stages
The workflow deliberately separates:
source acquisition;
source preservation;
geographic transformation;
validation;
model selection;
model freeze;
statewide calibration;
fiscal normalization;
benefit allocation;
scenario analysis;
certification;
reporting; and
visualization.
A later analytical stage is not permitted to silently rewrite an earlier certified stage.
The interactive map consumes certified reporting outputs.
It does not refit the statistical model or recalculate fiscal incidence in the browser.
31. Exact Reconciliation
Where authoritative statewide controls exist, district estimates are required to reconcile to them.
Millionaire taxpayers
Total Estimated Millionaire Returns = 21,530
Tax Outflow
Total District Tax Outflow = $2.698 billion under every scenario
Geographically allocated Tax Inflow
Total District Tax Inflow = $2.5349 billion
Complete statewide accounting
$2.5349 billion allocated + $163.1 million unallocated = $2.698 billion
These identities function as both accounting requirements and validation tests.
32. Interpretation
The Fiscal Explorer is best understood as answering:
Given the official FY2029 statewide revenue estimate, independently validated evidence about the geography of upper-income taxpayers, explicitly retained uncertainty concerning millionaire-income geography, and independently modeled program-benefit geography, what legislative-district pattern of fiscal incidence is implied?
The map should not be interpreted as claiming that the displayed district dollar amounts were directly observed.
A donor classification does not imply that residents of the district receive no benefits from the funded programs.
A recipient classification does not imply that residents of the district pay no millionaires tax.
The classifications describe the estimated net geographic relationship between the two.
33. Principal Limitations
No observed legislative-district tax liability. Washington does not publish actual millionaires-tax liability by legislative district. Tax Outflow must therefore be modeled.
No IRS ZIP-level millionaire count. IRS ZIP SOI does not directly report $1 million+ returns. Their geography must be estimated from observable upper-income characteristics.
Within-ZIP heterogeneity. Population-weighted ZIP allocation assumes that economic characteristics within a ZIP follow population unless better sub-ZIP evidence is available. Income can be substantially more geographically concentrated than population.
Imperfect external validation proxy. Washington capital-gains taxpayers are not identical to IRS millionaire taxpayers. DOR geography provides independent evidence concerning upper-tail concentration rather than ground truth for millionaire counts.
Excess-income geography is underidentified. No available source uniquely identifies the legislative-district distribution of millionaire income above the threshold. The seven-scenario framework exists because this uncertainty cannot defensibly be eliminated.
Benefit geography is modeled. Program allocators estimate where benefits accrue. They do not constitute direct observations of FY2029 cash payments to individual district residents.
Temporal stability. The methodology assumes that geographic relationships observed in source periods such as TY2022 and ACS 2024 remain sufficiently informative to allocate FY2029 statewide controls.
Fair Start timing. The displayed Fair Start component is FY2029-revenue-attributable rather than FY2029 cash expenditure.
Economic multiplier. The 1.5 local-economic-benefit multiplier is illustrative and does not affect Net Balance or donor/recipient classification.
34. What the Model Deliberately Does Not Do
The model does not:
use historical legislative-district millionaire estimates to construct the current results;
equate $200,000+ taxpayers with millionaires;
equate capital-gains taxpayers with millionaire taxpayers;
treat TY2022 dollar amounts as FY2029 dollar amounts;
independently replace the official statewide FY2029 DOR revenue control with a model-generated forecast;
force the unallocated $163.1 million into legislative districts;
select whichever excess-income scenario produces the most intuitive geographic result;
alter coefficients after examining surprising districts;
incorporate the 1.5 multiplier into Net Balance;
infer political meaning from donor or recipient status; or
permit the interactive visualization to alter certified analytical results.
35. Source and Provenance Framework
The methodology gives preference to authoritative primary sources.
Internal Revenue Service Statistics of Income
Used for statewide millionaire controls, ZIP-level upper-income characteristics, and independent Congressional District validation.
U.S. Census Bureau
Used for population geography, ZCTA-to-district allocation, ACS demographic allocators, and independent sub-district forensic validation.
Washington State Department of Revenue
Used for the official FY2029 tax-revenue control, Working Families Tax Credit geographic evidence, and independent TY2022 capital-gains-taxpayer geography.
Washington Office of Superintendent of Public Instruction
Used as an external administrative benchmark for public K-12 enrollment geography.
Washington Health Care Authority
Used as an external administrative benchmark for Apple Health/Medicaid geography.
Washington fiscal and legislative sources
Used to establish statutory purposes, fiscal controls, and timing associated with millionaires-tax revenue.
For reproducibility, retained source artifacts and model outputs are accompanied where available by source identity, source period, local filename, transformation, analytical role, and cryptographic checksum.
36. Provenance Categories
Inputs are conceptually separated into four roles.
CONTROL
Establishes an authoritative or adopted statewide magnitude.
ALLOCATOR
Provides evidence for distributing a statewide control geographically.
VALIDATOR
Provides independent evidence against which a geographic transformation or statistical model can be tested.
DIAGNOSTIC
Provides contextual or forensic evidence but does not determine coefficients, model selection, or final calibration.
Maintaining these distinctions reduces the risk of circular validation.
CONCLUSION
The Washington FY2029 Fiscal Explorer is designed to make an otherwise unobserved geographic fiscal relationship measurable.
It begins with authoritative statewide controls.
It separates statewide magnitude from geographic allocation.
It validates geographic transformations against independent evidence where possible.
It constructs Tax Outflow and Tax Inflow geography independently.
It freezes analytical stages before downstream results are examined.
It reconciles district estimates to statewide controls.
And where the available evidence does not support a single precise answer, it preserves alternative specifications rather than selecting one for convenience.
The resulting district estimates should therefore not be interpreted as exact observations.
They are transparent, constrained estimates whose assumptions and uncertainty can be examined directly.
The purpose of the model is not to eliminate uncertainty. Its purpose is to measure that uncertainty, preserve it, and determine which geographic conclusions remain standing after it is taken into account.