Did anything
trickle down?

The argument in one paragraph

The supply-side mechanism is real, measurable, and roughly a fifth the size it was sold as. Capital released at the top did some work — the best-identified study finds the 2017 corporate cut raised tangible business investment about 11% — but it delivered under $1,000 per employee in wages against a promise of $4,000 to $9,000, and corporate revenue fell 40% with growth offsetting only a sliver of the loss. Meanwhile the single largest measured source of American growth in the postwar record was not tax policy at all. It was the removal of discrimination against women and Black men, worth 20 to 40 percent of the growth in output per person between 1960 and 2010. America ran two experiments in those decades. The one it argued about produced modest results. The one it barely noticed produced the growth.

Part I

Did the capital go to work?

Start with the strongest evidence for the proposition, not against it. The honest answer is partly yes, and the magnitudes are the whole story.

The 2017 Tax Cuts and Jobs Act (TCJA) is the cleanest natural experiment available, and the definitive study — Chodorow-Reich, Smith, Zidar and Zwick, using confidential Internal Revenue Service microdata — found a real investment response. Firms experiencing the average tax reduction raised domestic investment roughly 20% relative to firms with no change. Aggregate tangible corporate investment rose about 11%. Long-run domestic capital ended around 7% higher.

That is a genuine finding and supply-side advocates are entitled to it. The mechanism they described exists. Then the magnitudes.

Gross investment held up. Net investment did not.

Two claims circulate about post-1980 business investment and they appear to contradict each other. Investment as a share of GDP looks roughly flat. Investment net of depreciation looks much weaker. Both are true, and the reconciliation is the interesting part.

The composition of American investment changed fundamentally. Federal Reserve Bank of New York research documents the shift: gross private investment in intellectual property products rose from below 1% of GDP to above 5%, while investment in structures fell from roughly 9–10% to near 7%. Software and intellectual property depreciate far faster than factories and bridges. A dollar of investment that used to buy forty years of service life now often buys five. So the same gross spending sustains a much smaller net addition to the capital stock — which is precisely the quantity NBS-1 tracks, and precisely the quantity that determines what the country actually owns.

This is not an argument that intangible investment is inferior. It is an accounting observation with a real consequence: a country can appear to be investing at historic rates while its net capital stock stagnates, because more of each year’s spending is replacing what wore out.

Hope and Limberg constructed an indicator of taxes on the rich across 18 wealthy countries of the OECD (the Organisation for Economic Co-operation and Development) from 1965 to 2015, identified every major tax reduction, and estimated the macroeconomic effects. Tax cuts for the rich raised the top 1% income share in the short and medium term. They found no significant effect on economic growth or unemployment.

Note what this does and does not say. It is a cross-country average with the identification problems that implies. It does not say tax rates never matter. It says that across the whole natural experiment, the growth dividend is not detectable against the noise, while the distributional effect is.

Part II

Where the capital went instead

If released capital did not mostly become new productive capacity, it went somewhere. Three destinations are documented.

It financed debt, not investment

Mian, Straub and Sufi traced top-1% saving through the financial system using tax records from 1963 to 2019. After 1982 the annual flow of saving from the top 1% rose by 2.9 percentage points of national income — over $680 billion a year in 2024 dollars. Their finding is unambiguous: this surge did not boost investment. It financed middle-class borrowing before 2008 and federal deficits after.

That is the mechanism behind the intuition about scarce assets. Capital accumulating faster than the supply of productive vehicles does not vanish. It buys claims on other people’s debt, and it bids up whatever is already scarce.

It bought back its own shares

Share repurchases were treated as illegal market manipulation until 1982, when the Securities and Exchange Commission adopted the safe harbor that made them routine. Between 2003 and 2012, S&P 500 companies spent about 54% of profits on buybacks. A buyback is not investment. It is a transfer to exiting shareholders that raises earnings per share by shrinking the denominator.

This deserves a caveat that is usually omitted: capital returned to shareholders can be redeployed elsewhere, and a firm without good projects arguably should return it. The critique is not that buybacks are inherently wasteful. It is that a regime which released capital on the premise that firms would invest it simultaneously legalized the most efficient available means of not investing it.

Why non-reproducible assets specifically

The pattern is not random. Assets with no carry cost and no cash flow have a specific tax advantage under current law, and naming it is more useful than describing the symptom.

Unrealized gains are not taxed. An asset held until death receives a stepped-up basis, and the accumulated gain is erased rather than deferred. In between, the holder can borrow against the position at low rates without triggering realization. The sequence — hold, borrow, die — converts appreciation into spendable cash while the gain never becomes taxable income.

That structure rewards exactly the asset profile in question: unique, non-reproducible, zero carry, no dividend, no realization event. It is not that wealthy holders have unusual taste in stores of value. It is that the tax code prices those stores of value more favorably than assets that throw off taxable income. This is the mechanism the Accord’s wealth-transfer stack is built to address — death as a realization event with step-up eliminated, and the closure of the deferral doors — and it belongs in the diagnosis rather than only in the remedy.

It bid up things that already existed

The logic from NBS-1 applies exactly. Bidding for an asset that already exists is a transfer to the seller, not capital formation. The $200 million painting is the clean case — the artist is dead, the canvas exists, and national productive capacity is unchanged. Elevated multiples on already-issued equity, land, and purely monetary assets are the same transaction at scale. The Accord takes no position on whether any particular asset is correctly priced. It observes only that a dollar spent acquiring an existing claim adds nothing to the produced-capital account in NBS-1.

Part III

The measured gap

RAND’s Price and Edwards asked a narrow, answerable question: what if incomes below the 90th percentile had kept pace with per capita GDP, as they did for the three decades after the war? The gap in 2018 alone was $2.5 trillion, about 12% of GDP. Cumulatively from 1975 to 2018: $47 trillion. The share of income going to the bottom 90% fell 17 percentage points over the period.

Two things must be said about that number, and the Accord says both. It is a counterfactual accounting exercise, not a causal attribution. It measures the size of a divergence. It does not demonstrate that tax policy caused it, and honest treatment has to concede that globalization, skill-biased technical change, the collapse of union density, and winner-take-all market structures are all live candidates alongside tax policy. Anyone presenting $47 trillion as the price tag of Reaganomics is overclaiming.

The sharpest single measure of the divergence is the productivity–pay gap. Between 1979 and 2017, economywide productivity rose 68.1% while median hourly compensation rose 13.0% — a 55.2 percentage-point divergence. Using the same price index to deflate both series, which is the more conservative construction, the divergence is 43 points.

The serious objection, which belongs here rather than in a footnote. Stansbury and Summers examined this gap and did not dispute the data, but disputed what it means: they find that productivity growth still passes through to typical pay, and that the divergence is driven by rising inequality and price-deflator effects rather than by a broken link between productivity and wages. That distinction matters for policy. If the link were broken, raising productivity would not help workers. If the link holds and the problem is distribution, then productivity growth remains worth pursuing and the fight is over who captures it.

And there is a complication that cuts against the simplest version of the story. Labor’s share of gross domestic income fell from roughly two-thirds in 1950 to about 58% today — but corporate profits captured only about 3 of the 8 percentage points labor lost. The remainder went to depreciation, interest, proprietors’ income and taxes. “Profits took it from wages” is too simple. Something broader happened to how output gets divided, and it is not fully explained by any single villain.

Part IV

The subsidy nobody booked

Here is the entry that belongs on the American balance sheet and has never appeared on one.

In 1960, 94 percent of American doctors and lawyers were white men. By 2010 it was about 62 percent. Innate talent did not redistribute itself across groups in fifty years. What changed is that a very large pool of capable people had previously been prevented from doing the work they were best at.

Hsieh, Hurst, Jones and Klenow measured what removing those barriers was worth. Between 20 and 40 percent of the growth in aggregate market output per person from 1960 to 2010 is explained by the improved allocation of talent.

Across 18 countries and fifty years, cutting taxes at the top produced no detectable growth effect. Across the same period in the United States, ending discrimination against women and Black men produced somewhere between a fifth and two-fifths of all growth in output per person. One of these was the central economic argument of American politics for forty-five years. The other was treated as a civil rights question with an economic footnote. The measured results run the other way round.

And the bill that came with it

The gain had a cost, and the cost landed in exactly one place. Before the barriers fell, American public education was staffed by women who were capable of far more demanding professional work and were barred from it. The schools got PhD-caliber labor at elementary-teacher wages. That was an enormous, invisible subsidy flowing from discriminated-against women directly into the nation’s human capital account.

When the barriers came down, the subsidy ended. Hoxby and Leigh documented the result: measured teacher aptitude declined as women’s outside options opened. But their finding contains the part that matters, and it is not the part usually quoted. Compression of teacher pay explains more of the decline than improved pay parity outside teaching does. The problem was not that talented women left. The problem was that once they could leave, American schools declined to pay what talent costs — and the pay scales were structured so that the most capable teachers gained least by staying.

America ran its education system for decades on an off-book subsidy extracted from women who had no other option. When justice removed the subsidy, the country pocketed the liberation gain — a fifth to two-fifths of all growth in output per person — and never replaced the subsidy with an appropriation.

This is the NBS-1 thesis in a different register. An asset was being funded by a hidden transfer. The transfer ended, correctly and permanently. The obligation it had been covering did not end. Nobody booked the difference, and the schools have been running on the shortfall ever since.

What the finding does and does not license

This module has identified a problem it cannot responsibly solve from a federal position, and it says so. Teacher pay is set by states, districts and bargaining agreements; the federal share of school funding is under a tenth; and the evidence points at pay structure, not pay level — a dimension no federal appropriation reaches without a measure of teaching effectiveness the country has no consensus on. Teacher salaries are a state and marketplace matter, and the Accord leaves them there.

What the Accord does own is the floor beneath the schools: the childcare guarantee and the Child Allowance get children into care and to the schoolhouse door ready to learn — the part of the ledger a federal program can actually deliver. The return evidence settles the case for investing in children — it clears the public-investment hurdle several times over — while leaving the instrument question where it belongs. A strong return finding is never authorization for whatever instrument is nearest; keeping those separate is exactly what the Accord’s investment gate exists to enforce.

Other unbooked subsidies of the same shape

The pattern generalizes. Each of these is a real input to national capability that appears in no account and is therefore free to erode without anyone recording a loss.

• Talent allocation of Black men — measured in the same study, part of the same 20–40%, and equally absent from any growth accounting used in political argument.
• Unpaid care work — still off-book, still disproportionately female, and the binding constraint on labor force participation for a large share of prime-age adults.
• Foreign-trained skilled labor — the United States receives the human capital and pays none of the formation cost. It is the single cheapest input on the national balance sheet and the most politically exposed.
• Public pension systems — by preventing elders from becoming a direct claim on their children’s income, they free prime-age earnings for investment and consumption. The counterfactual cost is never scored.

The common structure: an input the economy depends on, supplied by a group with limited bargaining power or by an arrangement nobody itemizes, which does not appear as an asset and therefore cannot be seen depreciating.

Part V

The four cushions, and their condition

Median wages stagnated for four decades, and yet living standards did not collapse. Four things absorbed the difference. Their current state is the reason the next forty years cannot resemble the last.

The point is not that any of these was illegitimate. It is that each was a one-time transition, and all four have run their course. An economy that has been living on the difference between stagnant wages and expanding cushions has no cushion left.

The first cushion, stated precisely

The claim that women’s labor has been “used up” as a growth reserve is half right and the half that is wrong matters for policy.

What is right: the level gain is complete. Aggregate female participation cannot go from 33% to 60% twice. That one-time reallocation, together with the occupational convergence measured by Hsieh and colleagues, is the growth engine that has already been spent.

What is wrong: prime-age women are not in decline. Participation among women aged 25–54 reached an all-time high above 77% in 2023, with mothers of children under five leading the rebound. The aggregate figure fell after 2000 substantially because the population aged and more young women stayed in school — composition, not retreat.

The precise finding is better than the loose one. Between 2000 and 2016, participation among prime-age women with a high school diploma or less fell from about 71% to 62%, while for women with bachelor’s or graduate degrees it barely moved. The constraint is not gendered in general. It is concentrated where childcare costs consume the largest share of potential earnings. The United States remains the only industrialized country without national paid family leave.

That makes care infrastructure a targeted labor-supply instrument rather than a general one, and it makes the case on capability rather than on a growth reserve that no longer exists in the form claimed. The Accord’s Childcare Plan is built to that finding: supply where the shortage indicators show it, demand support through the Child Allowance — aimed precisely at the households where the constraint binds.

Part VI

Believers or cynics?

The honest answer is documented, and it is both — in identifiable proportions, because one of the principals said so at the time.

The originators had a real argument

The intellectual core was not a fraud. Robert Mundell, whose work underpinned the policy mix, won the Nobel Prize. The Laffer curve is trivially true at its endpoints — a 0% rate and a 100% rate both raise nothing — so the only real question was ever empirical: where is the peak? That is a legitimate question, and NBS-1’s marginal-dollar analysis answers it with the same tool supply-siders use: at every credible elasticity the revenue-maximizing top rate sits far above where America stands. Martin Feldstein was a serious empiricist who later became one of the sharpest critics of the deficits the policy produced.

Treating the entire intellectual project as bad faith is both wrong and strategically foolish. It concedes the argument that critics are unserious.

The political operators were explicit

David Stockman ran the Office of Management and Budget and was the chief architect of the 1981 cut. In December 1981, in The Atlantic, he told William Greider that Kemp-Roth was always a Trojan horse to bring down the top rate, and that since trickle-down was hard to sell, the supply-side formula was the only way to get a tax policy that was really trickle-down. Confronted publicly, Stockman confirmed it: those were words he spoke. This is not an inference about anyone’s motives. It is the budget director’s own contemporaneous account of the policy’s purpose, published while the doctrine was still being sold to Congress.

And the skepticism was internal to the coalition. George H. W. Bush called it voodoo economics during the 1980 Republican primary, before becoming Reagan’s running mate.

Supply-side economics was a genuine academic argument that was knowingly used as packaging for a distributional objective the packaging concealed. Both halves of that sentence are supported, and the second half is supported by the person who did it. The inference stops there: it covers the 1981 Act’s construction, not the sincerity of anyone arguing the position today.

Part VII

The strongest case for the other side

Stated as its most capable defenders would state it, because an argument that has not survived its best opposition has not been tested.

The mechanism is confirmed, not refuted. The best-identified study on the largest corporate cut in US history found investment rose. Critics who claim tax rates do not affect investment are contradicted by the strongest microdata available.

The counterfactual is unobservable. The 1970s were genuinely bad — stagflation, a collapsing productivity trend, a top marginal rate structure with extensive shelter activity that created real deadweight. What the American economy would have done from 1981 without the reforms cannot be observed, and cross-country matching studies like Hope and Limberg carry identification assumptions that reasonable economists dispute.

Distribution and growth are separate questions. Demonstrating that inequality rose does not demonstrate that growth was forgone. A policy can raise both. The RAND gap measures divergence, not causation, and its authors say so.

Capital mobility is real and increasing. The world of 1965 had capital controls. Corporate tax competition since has been genuine, and a country that ignores it is choosing where its tax base is domiciled.

The comparison set matters. The United States generated most of the world’s frontier technology firms over this period. Any account of the last forty-five years that treats the American economy as a simple failure has to explain that.

The Accord’s response is about magnitude rather than direction. Every one of these points is partly right. None of them establishes that the returns justified the fiscal cost, and the best evidence — from researchers with no stake in the answer — puts the delivered result at roughly a fifth of what was promised while corporate revenue fell 40%.

Part VIII

What this analysis does not establish

It does not establish that tax cuts caused the $47 trillion income divergence. That figure is a counterfactual accounting gap with multiple plausible causes.

It does not establish that tax rates are irrelevant to investment. The microdata says the opposite, and this page reports it.

It does not price the asset-bidding channel. The claim that capital flowed into existing assets rather than new capacity rests on Mian, Straub and Sufi’s aggregate correlation, which is not a causal estimate of the marginal dollar.

The talent-allocation estimate is a Roy-model calibration with a wide reported range, 20 to 40 percent, and the range should always be quoted rather than the upper bound.

On motive: the Stockman evidence is dispositive about Stockman and about the 1981 Act’s political construction. It is not evidence about anyone else’s sincerity, and the page does not use it that way.

Sources

Supporting literature

1. Hope, D. & Limberg, J., “The Economic Consequences of Major Tax Cuts for the Rich,” Socio-Economic Review 20(2), 2022, 539–559.
2. Mian, A., Straub, L. & Sufi, A., “The Saving Glut of the Rich,” NBER Working Paper 26941, 2020, revised 2025.
3. Price, C. C. & Edwards, K. A., “Trends in Income From 1975 to 2018,” RAND Working Paper WRA516-1, 2020; updated through 2023 in WRA516-2.
4. Bureau of Economic Analysis, NIPA gross domestic income decomposition, 1950–2025; labor-share and profit-share series via FRED.
5. Chodorow-Reich, G., Zidar, O. & Zwick, E., “Lessons from the Biggest Business Tax Cut in US History,” Journal of Economic Perspectives 38(3), 2024, 61–88; underlying estimates in Chodorow-Reich, Smith, Zidar & Zwick, NBER Working Paper 32180, 2024. Council of Economic Advisers projections, 2017.
6. Joint Committee on Taxation and Federal Reserve Board-affiliated analysis of TCJA corporate revenue effects, as summarized by the Center for American Progress, 2024.
7. SEC Rule 10b-18 (1982). Buyback share-of-profits figure from Sakinç / Academic-Industry Research Network data as reported in PBS NewsHour, 2015 — primary-source substitution is an open item.
8. Hsieh, C.-T., Hurst, E., Jones, C. I. & Klenow, P. J., “The Allocation of Talent and U.S. Economic Growth,” Econometrica 87(5), 2019, 1439–1474.
9. Hoxby, C. M. & Leigh, A., “Pulled Away or Pushed Out? Explaining the Decline of Teacher Aptitude in the United States,” American Economic Review 94(2), 2004, 236–240; Corcoran, Evans & Schwab, JPAM, 2004.
10. Greider, W., “The Education of David Stockman,” The Atlantic, December 1981; Stockman’s public confirmation reported by UPI, 12 November 1981.
11. Federal Reserve Bank of New York, Staff Report 1190, on the shift of gross private investment toward intellectual property products.
12. Mishel, L. & Bivens, J., “Identifying the Policy Levers Generating Wage Suppression and Wage Inequality,” Economic Policy Institute, 2021 (1979–2017 figures); EPI productivity–pay series. “Net productivity” is output less depreciation per hour worked.
13. Stansbury, A. & Summers, L. H., “Productivity and Pay: Is the Link Broken?”, NBER Working Paper 24165 / PIIE Working Paper 18-5, 2017–2018.
14. Hamilton Project / Brookings, “Prime-Age Women Are Going Above and Beyond in the Labor Market Recovery,” 2023, updated 2024; The State of Women in the American Workforce; Council of Economic Advisers, 2016.

Companion to NBS-1 — What America owns. NBS-2 is evidence adjudication, not simulation: no new model, every contested figure reported with its range and its identification problem. Two external AI drafts were adjudicated during preparation; what was accepted, corrected, and rejected is recorded line-by-line in the module’s technical notes.