Pathway 05 · Learn

The Tax Ladder: An Interactive History of Who Pays for America

Homelessness is a systems problem, and the tax code is the system that decides what the safety net can afford. This page is a three-part tool. Act 1 is a scrubbable history of federal tax policy from 1913 to today. Act 2 hands you the controls: change the rates, close the loopholes, swap the system out, and see what happens to revenue, to each income group, and to growth — with the mechanism explained in plain language every time. Act 3 is twelve things people actually propose, each one a tap away from the same controls. Every figure traces to a government agency, a peer-reviewed paper, or a named research shop, dated and linked.

Rates over time — the master chart

The number most people know is the top statutory rate — the headline percentage in the law. It has ranged from 7% (1913) to 94% (1944). But almost nobody at the top ever paid the statutory rate: deductions, exclusions, and a narrower tax base meant the effective rate was always lower — sometimes dramatically. Toggle the layers below to compare the top rate against the corporate rate, the payroll rate, capital gains, and the government's own receipts as a share of the economy.

Where the money comes from

A chart of the top marginal rate alone hides the biggest structural shift in a century of federal taxation: the corporate income tax used to fund the government the way the payroll tax does now, and the two roles have swapped.

The wage cap and the estate tax

Two of the quietest levers in the tax code, because neither shows up in a "top rate" headline: the dollar amount above which payroll tax stops applying, and the size of an estate that owes nothing at all.

Who holds the income, who pays

Two questions, two charts. First: how much of the country's income does the top 1% hold, and how has that changed? Economists don't fully agree on the level — see note 6 — but they agree on the direction. Second: once transfers and taxes are counted, how much does each income group actually pay the federal government, from the bottom fifth of households to the top 1%?

The bottom line: life over the same century

Tax policy doesn't exist in a vacuum — it interacts with wages, housing costs, and health spending to determine what a household actually keeps and what it actually needs. Three more series, on their own scales:

Eras, 1913–2026

Seventeen turning points, each with the verified numbers attached. The 1950s card and the OBBBA "you are here" card carry the heaviest caveats — read them alongside the honesty notes below.

Nine caveats to this history

Nine places where this history is commonly flattened into a slogan, and what the evidence supports instead.

1. 1

   “91%” and the 1950s boom

   Effective rates were far lower than the statutory 91%. The top 1% paid about 16.9% of income in federal income tax on average during the 1950s (Tax Foundation). The top 0.01%'s total effective federal rate — income, payroll, corporate and estate tax combined, with corporate and estate imputed to their owners — was just over 70% in 1960 (Piketty & Saez), of which the individual income tax alone was only about 31%. The 1950s boom had other drivers too: the postwar international position, demographics, cheap energy. Say “high rates did not prevent fast growth” — never “caused it.”

2. 2

   Reagan's cuts and the 1980s recovery

   The recovery that followed the 1981 rate cut coincided with Paul Volcker's disinflation and large federal deficits. Top-rate-specific growth evidence from this period is weak, and Congress raised taxes repeatedly after 1981 — including in 1982 and 1984.

3. 3

   Clinton 1993 and the 1990s boom

   This is the mirror image of note 1. The honest claim is narrower than “the tax increase caused the boom”: the 1993 top-rate increase demonstrably did not derail the growth that followed it.

4. 4

   “Tax cuts pay for themselves”

   No major U.S. tax cut has. Verified behavioral offsets in this tool's data run roughly 1–32% of a change's static cost — real, but nowhere near 100%. Kansas's 2012–2017 experiment and its bipartisan repeal, in the sidebar below, is the sharpest single exhibit.

5. 5

   “Hauser's law” — flat receipts, not flat stakes

   Federal receipts as a share of GDP have stayed within a narrow band since 1950 — as low as 13.2% (1950) and as high as 19.8% (1945, and again 2000). That flatness does not mean tax policy doesn't matter: the composition underneath it shifted enormously (see “Where the money comes from” below), and the range itself is fiscally enormous — receipts were 18.80% of GDP in 2022 and 15.97% in 2023, a one-year swing worth hundreds of billions of dollars.

6. 6

   Inequality is contested — the numbers disagree by 8 points

   Two credible tax-data series disagree sharply for the identical year. For 2022, the top 1%'s share of pre-tax national income is 23.6% by the WID/Piketty-Saez-successor measure and 15.5% by Auten & Splinter's competing estimate. Both show the same rising trend since the 1970s–80s; the level is genuinely disputed among tax-data economists, driven by how each study allocates underreported income, retirement income, and corporate profit. Show both, never average.

7. 7

   Productivity and pay — a note this page doesn't chart yet

   Popular framing shows productivity and typical worker pay diverging sharply after the 1970s. Any honest version of that chart has to disclose its price deflator, whether benefits are included, and whether it compares medians or averages — Stansbury & Summers find the productivity-pay link “largely intact” once measured consistently, with the apparent gap driven mostly by forces acting on pay independently of productivity. This series is verified and ready (1948–2026), but this page doesn't chart it yet — a later update will.

8. 8

   The one association that does hold up

   The clearest pattern in this dataset: periods with lower top tax rates coincide with a rising top-1% income share, while the same rate cuts do not show up as faster growth. Two studies built on different designs reach that split result. Piketty, Saez and Stantcheva, looking across 18 OECD countries, put it in one sentence: “Top tax rate cuts are associated with top one percent pretax income shares increases but not higher economic growth.” Hope and Limberg, identifying every major tax reduction on the rich in 18 OECD countries between 1965 and 2015, find that “tax cuts for the rich lead to higher income inequality in both the short- and medium-term. In contrast, such reforms do not have any significant effect on economic growth or unemployment.”

   A third source is usually cited alongside those two and comes with a fight attached. Congressional Research Service report R42729, by Thomas Hungerford, concluded that US “changes over the past 65 years in the top marginal tax rate and the top capital gains tax rate do not appear correlated with economic growth” while “the top tax rate reductions appear to be correlated with the increasing concentration of income at the top.” That report was withdrawn from the CRS website in 2012 after political objection and later reissued. Its method is disputed on the record: the Tax Foundation argues it compares one-year changes when the effects of capital-tax changes accrue over one to ten years — “looking only at the first year effect throws out about 95 percent of the outcome” — and that it holds nothing else constant. That critique lands on the CRS report. It does not reach the two panel studies above, which is why they carry the claim here and Hungerford does not.

   What remains genuinely debated is the mechanism, not the correlation: whether lower top rates change how much people work, how much pay executives can bargain for themselves, or simply when and in what form income gets reported. Piketty, Saez and Stantcheva’s whole framework exists because those three channels have different policy implications and the same statistical signature.

9. 9

   “The 1950s were better” needs its caveats in the same sentence

   Healthcare was 5.0% of GDP in 1960 and 18.0% in 2024 (CMS National Health Expenditure data) — partly because medicine could do far less. Access to housing and credit was formally and informally restricted along racial and gender lines throughout the era: redlining, FHA underwriting exclusions, and routine credit discrimination against women were standard practice, not exceptions.

   Two statistics that usually appear in this argument are missing above. Typical 1950s home size and the era’s Black–white household-income ratio have no source in this project’s data that meets its own bar, so they are not shown. And the union-density figure often paired with the era — “about a third of workers” — is computed on wage-and-salary employment, a smaller denominator than the total-employment basis behind the 1954 peak of 28.3% this project verified. Both figures are real. Quoting them against each other would be comparing two different measures.

What state and local taxes add

Everything above is federal. States and cities layer their own taxes on top — and for most states, that layer runs regressive even where the federal system runs progressive.

Act 2 — you hold the pen

Everything above is history. This part is yours. Move the rates, close the loopholes, fund the tax collector, swap the whole system out for a flat tax — and watch four things at once: how much revenue changes, who pays it, what it plausibly does to the economy, and, in plain language, why.

Every number below traces to a government agency, a peer-reviewed paper or a named research shop — the same standard as Act 1. Where nobody has published a figure, the space stays empty. And the assumptions are yours to move: the dials at the bottom of the panel are not a settings menu, and moving them will change your answer.

Six caveats to the sandbox

Act 1 has nine of these. Building a model adds six more, because a model can mislead in ways a chart cannot.

1. 10

   A published score already contains the behaviour

   When you close a loophole above, the figure that appears is a conventional score from CBO, JCT, CRS, the Tax Foundation or the Yale Budget Lab. Conventional does not mean "before people react" — it means before economy-wide growth effects. The scorer has already assumed taxpayers shift, time and restructure their income. So this tool does not apply a second behavioural haircut on top; that would be counting the same response twice. The two rate sliders are different: those are computed here from raw IRS return data, so the behavioural layer is applied once, by us, and shown as its own bar.

2. 11

   Ten-year totals here are not uprated for growth

   Where the engine computes a figure from IRS data, it does so on tax year 2023 returns and multiplies by ten. It does not project income growth across the window, which real scorekeepers do. That makes every self-computed number on this page conservative — smaller than a properly uprated score of the same policy. It is one reason a lever here can land below a published estimate of the same idea; when that happens, the panel says so.

3. 12

   Most published scores predate the law you are living under

   The One Big Beautiful Bill Act became law in July 2025 and made the 2017 individual rates permanent. Almost every option score in circulation — including most of CBO's — was written in December 2024, assuming those rates would expire. Those scores are stale in direction, not just size: against a permanent-rates baseline, most revenue-raisers would score higher. Every pre-scored lever here carries a "pre-OBBBA baseline" or "post-OBBBA baseline" label. Only three sources in this project's dataset are post-OBBBA. Read the label before you read the number.

4. 13

   The class chart shows the shape, not a year-matched ledger

   Revenue is computed on IRS tax returns for 2023, ranked by adjusted gross income. The income groups are CBO's, for 2022, ranked by household income before transfers and taxes and adjusted for household size. Those are two different populations sorted two different ways, and mapping one onto the other — which this tool does, using CBO's own published group boundaries — is an approximation, not an identity. Read the chart for direction and relative magnitude between groups. Do not read the third decimal place.

5. 14

   Tax breaks do not add up

   Turn on three loopholes and the tool adds their three scores together, because that is the only arithmetic the published figures support. It is also wrong, and knowably so: each was scored on its own, against a world where the other two still existed. Close the pass-through deduction and some of that income reappears elsewhere; cap the health exclusion and wages rise, which changes what every other provision is worth. Interaction effects run in both directions and no published estimate in this dataset quantifies them. Treat a multi-lever total as an order of magnitude, not a sum.

6. 15

   Growth effects are shown, and deliberately kept out of the revenue figure

   The GDP band is real output from the model. It is never added back into the revenue number, and that is a structural decision rather than a modelling detail: it is what makes it impossible for any combination of settings on this page to show a tax cut financing itself. The evidence supports the choice. Verified growth feedback in this project's data runs from about −5% to +32% of a change's static cost — real, occasionally negative, and nowhere near 100%. CRFB's arithmetic explains why 100% is out of reach: because taxes capture only a fraction of income and some spending grows with income, "a policy would need to produce $5 to $6 of economic activity for every $1 of cost to be self-financing." No major U.S. tax cut has ever come close.

Act 3 — what people propose

Twelve things people actually propose. Tap one and it loads into the sandbox above: the same engine, the same guardrails, the same why-panel. Nothing here is scored a second way.

Each card carries what changes, what published estimates say happens, why it happens, how strong the evidence is, what its critics say, and where every figure came from. Where this tool's own arithmetic disagrees with a published score, both numbers are shown. That disagreement usually has a reason, and the reason is on the card.

One more thing the scenario library won't let you oversimplify

1. 16

   A card is not a bill

   Every card above compresses a policy into a handful of settings. Real legislation is where the money actually moves: an exclusion for family farms, a five-year phase-in, a threshold indexed to one price index rather than another. Those choices routinely matter more to a score than the headline number does — the payroll-cap card is the clearest case, where one design decision nobody argues about changes the answer by roughly a third. When a card and a real proposal disagree, assume the proposal has details this tool cannot see, and go and read it.