Two institutions, eight days apart, asked Italy for different things

The IMF asked Italy to eliminate the flat tax on self-employment income. Eight days later the European Commission published 85,857 words about the Italian economy without once naming the regime everyone assumed it meant.

By Jan Wangrat · 25 July 2026

On 27 May 2026 the International Monetary Fund closed its annual Article IV mission to Italy with a sentence that made the Italian trade press: eliminating the preferential flat-tax rate on self-employment income, it said, “would expand the tax base, improve equity, and support consolidation efforts”.1

Eight days later the European Commission published its own annual assessment of the Italian economy — the Country Report that accompanies the European Semester recommendations. By the following week, “l’Europa mette nel mirino il forfettario” was circulating on Italian social media, and weekly search interest in regime forfettario was climbing towards its highest level since January.

So we read the Commission’s document. Then we counted it — and the two before it.

What is not in 199,509 words

We took the English editions of the Country Reports for Italy published in 2024, 2025 and 2026, extracted their text in reading order, and counted a fixed list of tax terms decided before we opened the 2026 edition.23 Together they run to 199,509 words. The word forfettario appears in none of them. Neither does presumptive, the term the international literature uses for regimes of this kind.

Absence of a name is not absence of a subject, and the 2026 report is the first of the three to discuss flat taxation at all. It does so six times — though one of those, in the housing section, is about a flat someone lives in rather than a rate someone pays, which is what you get when you let a rule count for you and then check its work by hand. Of the five that are about tax, three concern the self-employed. One describes the regime with precision:

“since 2019, most self-employed people have been subject to flat-tax regime, which was extended in 2023 to include those with revenues below EUR 85 000”2

That is the forfettario — the €85,000 ceiling, the 2023 extension, the whole thing5 — described accurately and left unnamed. The only earlier appearance of a flat rate anywhere in the three reports is from 2025, and it is about overtime pay for nurses.

What the Commission did ask for

The Country Report is analysis. The instrument that actually asks Italy to do something is the accompanying recommendation, adopted the same day. Its fiscal paragraph is specific about targets, and the forfettario is not among them:

“make the tax system more conducive to sustainable growth, while ensuring fairness, including by further fighting tax evasion and reducing remaining tax expenditures, including those related to value added tax and environmentally harmful subsidies. Update cadastral values…”4

Evasion, VAT reliefs, fuel subsidies, the property cadastre. The criticism of “special regimes for self-employed” is real, and it sits in the recitals — the part that explains the reasoning rather than the part that sets the ask.4

The census makes the same point from the other direction. Adjusted for length, only two of the terms we counted rise in every one of the three years: self-employed (1.06 → 2.10 → 2.34 per 10,000 words) and tax expenditure (0.63 → 1.20 → 1.75). Attention to Italy’s self-employed is growing. So is attention to the reliefs the Commission wants cut. They are not yet the same sentence.

Meanwhile, the search box

Italian search interest in regime forfettario averaged 46.8 on Google’s relative index over the eight weeks to 23 May. Over the five weeks from 31 May it averaged 65.4 — 40% higher — peaking at 79 in the week of 21 June, the highest reading since the January turn of the tax year.6

We are not claiming the Commission caused that. The IMF spoke first, the Italian budget cycle starts talking about the forfettario every summer, and a relative index cannot separate the two. What the series does show is that the anxiety is real and datable, and that it attached itself to the document that never mentioned the regime rather than the one that did.

Is the criticism right, though?

The recital’s charge is that regimes like this “erode the tax base”. That is a testable claim, and it has been tested. Francesco Alosa, using Italian administrative data, measured what happened at the regime’s earlier €65,000 cut-off and found excess mass just below the threshold of about 337% of the counterfactual frequency — a very large behavioural response by the standards of this literature.7 Turnover elasticities came out at 0.071 for professionals and 0.058 for business intermediaries, and lower compliance costs explained less than half of the response: people were reacting to the tax, not just the paperwork.

So the diagnosis has evidence behind it — more evidence than either institution cites. The Commission asserts erosion without a number; the IMF asks for elimination without one. The estimate exists, in a working paper, and neither document points to it.

What this does not show

Counting words measures attention, not intent. A government can be pushed on something a document never names, and “reducing remaining tax expenditures” is a phrase with room in it. The Commission’s recitals plainly dislike the proliferation of flat regimes; our census only establishes that the operative ask, three years running, has pointed elsewhere. Nor does it forecast: the 2027 budget could do anything, and this piece is not a prediction that the forfettario survives it.

What it does establish is narrower and, we think, still worth knowing. If you are a freelancer in Italy who spent June worrying about a European document, it is reasonable to want to know what that document actually says. It says tax expenditures, VAT, evasion, cadastre. Not forfettario. Not once, in three years.

Method: PDFs downloaded from the Commission’s country page, text extracted with pdftotext in reading order (not layout mode — these reports are two-column, and layout mode interleaves them, which breaks any rule that depends on words being adjacent). Terms are case-insensitive regular expressions fixed in advance; “flat” counts only where “tax” or “income” follows within 40 characters in the same sentence, and every hit was checked by hand. The scripts that produce both datasets ship with the site.