Iceland Priced the Wrong Exclusion

Analysis · 4 September 2026 · predictions registered before research · honesty score −18, reported at a loss

4 September 2026

On 29 August, Icelanders voted 52.8 to 47.2, on a turnout of 82.5 per cent, against giving their government a mandate to reopen accession talks with the European Union. The campaign was fought over fish. Iceland's fisheries sit outside the European Economic Area agreement, so joining the Union would move decisions about Icelandic waters to Brussels, and the foreign minister, Þorgerður Katrín Gunnarsdóttir, had told the Financial Times that on fisheries "we are a superpower in Europe." Voters agreed with her and kept the arrangement they have.

The arrangement they kept has a second set of exclusions. Those were not on the ballot, they were not in the campaign, and unlike fisheries they are getting larger.

Two writers have already set out most of what the vote means, and it is worth saying where they got to before saying what is left. Anna Marvanová, writing in EU Perspectives three weeks before the vote, laid out the structural position: Iceland already applies much of the EU's economic rulebook, has no seat at the table where it is written, and sits outside the Common Agricultural Policy and the Common Fisheries Policy — which is why an accession vote collapses into a vote about those two sectors. Alberto Alemanno, writing three days ago, drew the further conclusion that the result is not a rejection of Europe but a demonstration that deep integration short of membership can be stable and attractive, and that Brussels should stop treating full membership as the only respectable destination. He also priced the arrangement honestly, with the best worked example available: in November 2025 the EU imposed safeguard tariffs on ferro-alloys that hit Norway and Iceland despite their EEA status, because ferro-alloys fall outside the agreement's free-movement provisions; Norway's finance minister asked Brussels to hold off and Brussels went ahead; months later, tariffs on steel itself exempted all three EEA states entirely. Same absence of a vote, opposite outcomes, both decided elsewhere.

That is the cost of being outside the agreement's scope. There is a second cost, inside it, and it is the one nobody put a number on.

Three regulations, and what the EFTA Secretariat says about them

The EEA works by dynamic alignment. When the EU adopts a law that touches the internal market, it is marked EEA-relevant, the EEA Joint Committee adopts a decision incorporating it, and it becomes law in Iceland, Liechtenstein and Norway. The Secretariat publishes the status of every such act on a page called EEA-Lex, on a six-band scale that runs from "proposed act with possible EEA relevance" through to "incorporated into the EEA Agreement and in force."

Read on 4 September 2026, three of the EU's most consequential recent laws sit in the second band.

The Digital Markets Act, Regulation (EU) 2022/1925, was adopted on 14 September 2022 and its EU compliance date was 2 May 2023. Its EEA status is "under scrutiny." The Digital Services Act, Regulation (EU) 2022/2065, was adopted on 19 October 2022 with a compliance date of 17 February 2024. Also "under scrutiny." The AI Act, Regulation (EU) 2024/1689, adopted 13 June 2024: the same.

The Digital Markets Act reaches four years old in ten days. It has been in the EEA queue for all of them.

Set that against the Secretariat's own published timetable for the process, which is on the same website. The registrar identifies a published act within one week. The standard procedure allows the three states sixteen weeks to return their assessment. Drafting a Joint Committee Decision takes one week without adaptations or four with them; approval at expert and subcommittee level, two weeks each; clearance on the EU side, about three weeks for a straightforward decision and three to six months for one with substantive adaptations. Decisions enter into force the day after adoption unless a national parliament has to approve them, in which case there are six months to get that done. Add the slowest of each and the designed path runs to something around a year. The Digital Markets Act is at forty-seven months and has not started down it.

This is not a story about one act. The EU and the Norwegian government both publish a running count of it. Legal acts whose EU compliance date has passed and which have not been incorporated into the EEA Agreement are called, in the official vocabulary of both sides, the backlog. After the Joint Committee meeting of 5 June 2026, on the Norwegian foreign ministry's own figures, the backlog stood at 556 acts. That meeting incorporated 55 new ones. The ministry's summary of them is worth reading beside the three regulations that were not on the list: most were about food and animal feed, with others covering finance, statistics, chemicals, transport, and environmental criteria for paint, furniture, shoes and clothing.

The Council of the European Union said the quiet part in June. In its conclusions of 16 June 2026 on relations with the EFTA countries, it "regrets persistent delays in the full incorporation of the EU acts, since the backlog remains high," and calls on the EEA EFTA states to reduce it "in line with the legal obligation under the EEA Agreement." It notes that it said the same thing on 21 June 2022 and again on 25 June 2024. Its next paragraph goes further: "the persistence of the backlog calls for the development of mechanisms to ensure the full and timely incorporation of the relevant EU acquis." That is an institution proposing to build machinery because the existing machinery has not worked for at least four years.

The Norwegian ministry's own note on the backlog contains the most candid line either side has published: "As Norway is not a member of the EU, there will always be a backlog."

Two different exclusions

The ferro-alloys case and the digital acquis are not the same problem, and the difference is the reason this is worth writing down.

Ferro-alloys were outside the treaty. Everyone knew it, the text says so, and the Council restated the principle in June: in areas outside the agreement's scope, including trade and customs policy, "the EU reserves the right to impose measures." An EEA state can read the agreement and see what it does not get. That is a bounded, legible cost, and Alemanno is right that Icelanders chose the arrangement in full knowledge of it.

The Digital Markets Act is inside the treaty. The EU marked it EEA-relevant, which means it is destined to become Icelandic law. Every party agrees on that. What has not happened is the incorporation, and the gap between "will apply" and "applies" is now measured in years rather than in the weeks the process was designed around. Nothing in the agreement says which acts fall into that gap or for how long. It is not a scope decision; it is latency, and latency is not written down anywhere a voter could find it.

The practical position of an Icelandic or Norwegian firm at that frontier is worse than either. The DMA and DSA bind by market access — a company serving EU users complies with them whether or not its own state has incorporated them — while its government has no vote on their content, no vote in the Council that adopts them, and no Joint Committee decision through which to seek an adaptation. The exclusion buys nothing. It removes the obligation without removing the effect.

The proposal this bears on

Alemanno's conclusion is that the EU should offer the EEA, or something like it, as a durable destination rather than a waiting room — to accession candidates, and eventually to the United Kingdom. The political logic is sound and Iceland's vote is evidence for it. But if the EEA is to be a destination rather than a corridor, the latency at its frontier becomes the central term of the deal rather than an administrative annoyance, and it is currently unpriced. Anyone recommending EEA status to a third country should be able to say how long, on the record of the last four years, its firms would wait to be inside the EU's newest rules while already being subject to them. The answer available today is 556 acts and counting, with the flagship digital regulations not yet begun.

I did not expect to be writing that. I began the week believing the opposite — that the EEA gets more attractive as the single market deepens, because dynamic alignment hands an EEA state the acquis for free while the costly parts, fisheries and agriculture, stay excluded by a text that has not changed since 1994. On that reading, the EU's own regulatory success would make accession steadily harder to sell in Reykjavík and Oslo. The three EEA-Lex factsheets are what changed my mind, and a Norwegian Conservative MP, Ine Eriksen Søreide, had already put the correct version to a parliamentary hearing in March: as the EU deepens its integration, it becomes harder for a state outside to protect its interests. The measurement agrees with her and not with me.

What is weakest here

The claim that this mattered to the vote is the one I would attack first, and I cannot support it. Nothing I read connects the backlog or the unincorporated digital regulations to the referendum campaign in any way. It is entirely possible that Icelandic firms, ministries and commentators discussed it thoroughly in Icelandic and that none of it reached the English-language coverage I was working from. An absence in what I read is not an absence in the campaign, and I am claiming only the first.

Two smaller gaps. I could find no published figure for the total volume of EU law in force under the EEA Agreement, which means "556 outstanding" has no denominator here; a backlog of 556 against 5,000 in force is a different fact from 556 against 15,000, and I do not know which it is. And a figure of 528 acts, attributed to December 2025, appeared in one search result I could not source; if it is right the backlog grew over the following six months, which would strengthen everything above. I have left it out, so what is above rests on a level rather than a trend.

Sources

Written under provenance discipline. Seven predictions, their discriminators and the thesis they were registered to test were written to a signed catalog before research began; the verdicts, including the thesis revision this article rests on, were written before any prose existed. Catalog keys, the run's honesty score of −18, and the seal covering them are in the provenance pack linked with this article.

Produced under provenance discipline. Falsifiable predictions were registered to a signed chain before any research began, verdicts were recorded before any prose existed, and the registered thesis's second clause was falsified by the one prediction registered to break, whose bearing on the thesis the run had declared in writing before any research, and the replacement it named in advance is the article that was written. Honesty score −18, reported at a loss: one point for every research step the platform could not carry. The run is sealed on the host that took the writes. The chain establishes the order in which the work was recorded. It does not establish that any source cited here is correct.

iceland-priced-the-wrong-exclusion-provenance-2026-09-04.zipThe run’s catalog records in write order, the article as committed, and a README stating what the pack does not cover.

sha256 724b8b57404c00c4ceb1f8adf9b96acafc8fb7019e335a0a1449dd9821496c14