The rich used to pay the most tax. Now they pay the least.

Correct — it compares a stock (accumulated wealth) to a flow (one year of output), so the 42% is not a “share of the economy” in an accounting sense. But the chart's point is the trend at constant methodology: the same ranking, divided by the same GDP series, moved from 6% to 42% in 28 years. Critics of the ratio (IREF, Fondapol) dispute how to interpret the level; none of them produce a series where the concentration isn't exploding. Any alternative denominator — total household wealth, national income — shows the same direction.

From the tax returns themselves. The Institut des politiques publiques worked inside the French tax administration, on exhaustive administrative microdata, and measured all taxes paid against economic income — including the profits of companies the household controls. That definition matters because that's exactly where the money goes: dividends accumulate in holding companies and never appear on a personal return. Judge the system only on taxable income and you've defined the loophole out of existence. On taxable income alone, French billionaires' income tax comes to a fraction of a percent of what they actually earn. In the US, ProPublica's leaked IRS files showed the 25 richest Americans paid a “true tax rate” of 3.4% on $401bn of wealth growth (2014–2018).

The anchor curves (1950, 1960, 1980, 2018) are sourced from Saez & Zucman's published series; the in-between frames are linear interpolation, and the chart says so on its face. The book's underlying data is annual and tells the same story — we interpolate only to animate, never to create a data point we cite. Every number quoted in text comes from an anchor year.

They are — economists at AEI and elsewhere argue different assumptions (who ultimately bears the corporate tax, how to allocate unreported income) that would raise the top-400 rate by several points and lower the bottom's. Two things survive every version of that debate: top effective rates have fallen by tens of points since mid-century, and taxation stops being progressive at the very top. The 2018 crossover is the sharpest framing, not the whole argument — France (IPP), the Netherlands (CPB) and Italy (Guzzardi et al.) each found the same regressive kink with their own national data, using different teams and methods.

Run the arithmetic. The floor applies above €100 million — 1,800 households, 0.005% of taxpayers. A household with €1M of wealth and €100k of income pays roughly €35–45k in taxes and contributions each year: that's 3.5–4.5% of its wealth, already far above the 2% floor. The floor only binds when wealth is enormous relative to taxed income — the holding-company configuration — which is mathematically out of reach below tens of millions. Here is the part that should actually worry you: on the curve above, you are the peak. Rates rise all the way to P99 and fall only after — the current system taxes your salary at full freight while extreme wealth compounds untaxed, then meets you again in housing prices and the inheritance divide. The status quo is the thing eating the upper middle class; the floor is the correction.

France ran this experiment for 30 years: under the ISF wealth tax, departures peaked at about 0.2% of liable households per year (parliamentary reports) — and wealth kept concentrating anyway. The current bill keeps leavers liable for several years after departure, and the G20's Rio declaration (2024) exists precisely to close the exit at the global level, the same way the 15% corporate minimum did for companies in 140+ countries.

Large fortunes have grown about 7% per year for three decades; a 2% floor slows accumulation, it doesn't reverse it — nothing is “confiscated”. And it can't double-tax by construction: every euro of income tax, social contribution, IFI and surcharge already paid is deducted first. Whoever already pays 2% of wealth pays zero extra. The floor exists only for those whose current total is near zero.

“Paper” wealth is bankable wealth: the standard playbook — buy, borrow, die — pledges those very shares as collateral to finance untaxed lifetimes (that's what the ProPublica files documented). If shares can secure a nine-figure loan, they can pay a 2% assessment: the bill allows payment in kind (shares) and staggering, so no forced sales and no loss of control — 2% dilution against ~7% average growth.

The broadest study on record — Hope & Limberg (LSE), covering 50 years of major tax cuts for the rich across 18 OECD countries — found no significant effect on growth or unemployment, in either direction. The only robust effect was higher top-1% income shares. The US's strongest growth decades (1950s–60s) coincided with top rates above 70%. Four decades into the experiment, the trickle has not come down.

The estimate range is €15–25bn/year; even the low end exceeds France's entire Justice budget, from 1,800 households. Behavioral erosion is real (that's why the range is wide) and shrinks further with G20 coordination. And revenue is only half the point: the other half is that a tax system where cleaners out-pay billionaires in percentage terms doesn't survive contact with its own citizens — that's an institutional cost no budget line captures.

Not because the public is divided — Ifop (Oct. 2025) finds majority support in every electorate, left and right. It failed in the Senate and in budget arbitrage, under heavy lobbying — and the debate itself runs through concentrated megaphones: a handful of billionaires own the majority of France's national media (see RSF/Le Monde diplomatique's ownership mapping). When 0.005% of taxpayers outweigh 86% of citizens, that is not an argument against the tax; it's a live demonstration of what wealth concentration buys.