Most people can name the date the Weimar Republic ended. Almost nobody can name the date it stopped working.

That absence is not an accident of memory. The end has a scene attached to it, a chancellery, a signature, a torchlit procession, and the scene is what survives. The date that mattered has no scene at all. It has a committee room, a dispute over an insurance contribution rate, and a resignation letter. On 27 March 1930, the cabinet of Hermann Müller resigned. It was the last government of the Weimar Republic that rested on a parliamentary majority. For the next thirty-four months, emergency decree progressively displaced legislation as the working mechanism of government.

The dispute that ended it was about half a percentage point.

On 6 September 2026 the German state of Saxony-Anhalt recorded the highest turnout in its history and gave forty-three point eight per cent of the vote to a party its own state intelligence service classifies as a confirmed right-wing extremist organisation. Both facts were reported. Only one of them was treated as the news.

The German unemployment insurance scheme, created by the Reichstag in July 1927 and in force from October that year, was funded by a contribution levied equally on workers and employers. In December 1929 the rate had been lifted from three per cent of wages to three and a half. By March 1930 the fund was failing and the free trade unions demanded four per cent, in order to avoid cutting benefits. The German People's Party, close to the export industries, refused, and proposed cutting benefits instead. Heinrich Brüning offered a compromise that postponed the decision to the autumn. The Social Democratic parliamentary group, under pressure from the unions, rejected it. Three days later Paul von Hindenburg appointed Brüning to lead a minority cabinet that had announced in advance that it would govern by decree if the Reichstag defeated it.

Here is the part that is almost never told. On 26 July 1930, four months after the coalition died rather than concede four per cent, Brüning raised the contribution by decree to four and a half. By October it stood at six and a half. The number that the last democratic majority in Germany could not agree on was overtaken within a season, by a government that did not need to agree with anyone.

The dispute was never about the money. It was about who would still be in the room when the decision was made.

The Vote Nobody Remembers

What the March 1930 resignation actually removed was not a coalition. It was a procedure.

The arithmetic of the following three years is a matter of record. In 1930 the Reichstag sat for ninety-four days and passed ninety-eight laws, against five emergency decrees issued under Article 48. In 1931 it sat for forty-one days, and the ratio inverted: thirty-four laws, forty-four decrees. In 1932 it sat for thirteen days. Five laws. Sixty decrees.

This is a familiar story told with the wrong emphasis. The usual reading treats the presidential cabinets as a symptom of Nazi pressure, a democracy buckling under an assault from outside the chamber. The sequence says otherwise. The decrees began in July 1930, before the September election in which the National Socialists took eighteen point three per cent. In May 1928 that party had polled two point six. The mechanism of emergency government was in place, tested and normalised before the electoral surge it is usually blamed on. It was not built to contain the radicals. It was built because a parliamentary majority could not resolve a technical question about social insurance, and the executive found it easier to proceed without one.

In Weimar it did not begin when the fringe became strong. It began when the centre stopped being believed.

The centre in Weimar was not disbelieved because it was corrupt or because its speeches were poor. It was disbelieved because it had visibly, publicly, and by its own choice removed itself from the decision. The voters of 1930 were not being asked to choose between democratic and anti-democratic government. They were being asked to choose a parliament that had already demonstrated that it would not be the body that decided.

The Republic did not fall in 1933. It stopped governing in 1930, and the three years that followed were spent finding out what fills the space.

The Fund Was in Surplus

The decrees that followed were not chaotic. They were methodical, and their method is the reason this story is worth retelling now.

Brüning issued five major fiscal decrees between July 1930 and December 1931. The first tightened eligibility criteria for social benefits and raised the insurance contribution. The second, in December 1930, cut civil service salaries and pensions by six per cent and reduced unemployment and health insurance benefits. The third, in June 1931, cut unemployment benefit and crisis relief by a further five per cent, lengthened the waiting period before entitlement began, and reduced the child allowance. The fourth, in October 1931, shortened the entitlement period and raised the qualifying age. The fifth, in December 1931, rolled all wages set by collective agreement back to their January 1927 level and fixed the maximum insurance entitlement at twenty weeks. By the beginning of 1932 civil service pay had fallen cumulatively by between thirteen and twenty-three per cent depending on grade.

Registered unemployment rose from one point nine million on annual average in 1929 to five point three million in 1932, peaking above six point one million in February of that year. Set against that, here is what the insurance system actually did.

In 1928 and 1929, sixty-seven and a half per cent of the unemployed drew the insurance benefit. By 1932 the figure was nineteen point four per cent. Another twenty-five point nine per cent had fallen through to time-limited crisis relief. Thirty-six point six per cent had fallen further, to municipal poor relief. And eighteen point two per cent received nothing at all.

Now the fact that ought to be better known than it is. In the fiscal year 1932 to 1933, with more than five and a half million people out of work and roughly one in five of them receiving no support of any kind, the Reich Institution for Employment Placement and Unemployment Insurance recorded a surplus of three hundred and seventy-two million Reichsmark. The figure sits in the Federal Employment Agency's own published history of its predecessor institution.

The fund was solvent.

The insurance did not run out of money. It ran out of people it was permitted to pay.

This is the shape that matters, and it is the shape that recurs. The institution was not abolished. Its budget was not exhausted. Its name, its staff, its statutory basis and its accounts all continued. What was removed, by a sequence of individually defensible administrative decisions taken in conditions of genuine fiscal emergency, was its capacity to reach the people it had been built for. The form outlived the function by three years, and the three years were the ones that counted.

The Descent Was Administered

Those are categories, and the descent between them was administered. A worker who lost his job in the spring of 1932 drew insurance for twenty weeks and crisis relief for sixteen more. What came after was not a benefit but a means test, applied not to him but to every person living under his roof, including relatives who owed him nothing in law. What it paid was repayable. Work could be assigned in exchange for it under paragraph nineteen of the 1924 relief ordinance, six to twenty-nine hours a week depending on the size of his family, formally voluntary and refusable only at the price of the payment itself. From June 1932, if he lived within five kilometres of the labour exchange, he reported daily to have a card stamped, and that same month the relief was cut by a further fifteen per cent. A schedule published in one Schleswig-Holstein district the previous August had already taken a family of six from twenty-nine Reichsmark a week to twenty, put a single man in lodgings on five, and removed anyone still living with their parents from the rolls altogether.

One thing the system did not take. Under the Kaiserreich, drawing poor relief had cost a man his vote. The Weimar constitution ended that: Article 22 gave the franchise to every adult, and Article 17 extended the same rule to municipal elections. So by the autumn of 1932 the state had tested his household, priced his relatives, assigned his labour, stamped his card daily and cut his payment twice, and had left him exactly one instrument through which to register an opinion about any of it.

He used it. Turnout that July was the highest in the Republic's history.

Mobilisation, Not Apathy

The second thing the record refuses to support is the story of exhaustion.

There is a comfortable version of the interwar collapse in which a demoralised, atomised population stops participating, drifts away from politics, and is eventually collected by whoever is still shouting. It is comfortable because it is passive, and because it flatters the assumption that democratic failure looks like silence.

Turnout at the Reichstag election of May 1928 was seventy-five point six per cent. In September 1930 it was eighty-two. In July 1932 it was eighty-four point one, the highest of any free election in the Republic's history. Across the same period the National Socialist vote went from eight hundred and ten thousand to six point four million to thirteen point seven million. In 1930 alone, roughly four point two million more Germans voted than had voted two years earlier.

The surge was not built out of people who had given up. It was built out of people arriving.

Honesty requires the qualification. At the election of November 1932 turnout fell to eighty point six per cent and the National Socialist vote fell with it, from thirty-seven point three per cent to thirty-three point one. The party was past its electoral peak when Hindenburg appointed Hitler on 30 January 1933, at the head of a cabinet containing two other National Socialists, on the basis of a presidential appointment rather than a majority. The March 1933 turnout of eighty-eight point seven per cent was recorded under the Reichstag Fire Decree with the Communist leadership already arrested. Neither of those figures is evidence of anything except the collapse of the conditions under which figures mean something.

But the shape of it is not in doubt. The Nazi surge was a participation event.

Saxony-Anhalt, 6 September 2026. Turnout seventy-seven point eight per cent, against sixty point three in 2021, an increase of seventeen and a half points and the highest at any free state election the Land has held. The Alternative für Deutschland took forty-three point eight per cent and thirty-nine of eighty-three seats, thirty-eight of them won directly. The Christian Democrats fell from thirty-seven point one per cent to seventeen point two and did not win a single constituency; the minister-president entered parliament from the list after losing his own seat.

According to the Infratest dimap voter migration analysis, one hundred and seventy thousand of the AfD's five hundred and seventy-six thousand second votes came from people who did not vote in 2021. That is close to thirty per cent of its entire vote, and it is more than every other party on the ballot gained from the non-voter pool combined.

Democracies in difficulty do not go quiet. They get loud, in a direction nobody planned for, and the loudness is the first measurement worth taking.

The Control Group

British consumer price inflation reached twenty-four point two per cent in 1975 and eighteen per cent again in 1980. Italian inflation hit nineteen point two per cent in 1974 and twenty-one point one in 1980. French inflation peaked at thirteen point seven per cent. The spot price of crude rose from three dollars and fifty-six cents a barrel in July 1973 to ten dollars and eleven cents by January 1974, and from fourteen dollars and eighty-five cents in January 1979 to thirty-nine dollars and fifty by the summer of 1980. 1975 delivered the first synchronised contraction in postwar Western European history. British unemployment nearly tripled between 1970 and 1980; West German unemployment rose fivefold.

The violence was worse than anything Europe has seen since. The Red Army Faction killed thirty-four people. In the autumn of 1977 alone West Germany lost a federal prosecutor and his escort, a bank chairman, the president of its employers' federation and his four-man security detail, and an airline captain. Italy's Years of Lead produced something over four hundred deaths, ninety-one armed organisations by 1977 and two hundred and sixty-nine by 1979, the murder of a five-times prime minister and his five bodyguards in 1978, and the Bologna station bombing that killed eighty-five people in 1980. ETA killed close to a hundred people in 1980 alone.

Against all of that, the electoral record of the European far right in that decade reads as follows.

The German National Democratic Party peaked at four point three per cent in 1969, before the oil shock, before stagflation, before the German Autumn. It took nought point six per cent in 1972, nought point three two in 1976, and nought point one eight in 1980. In the year after the second oil shock it polled less than a tenth of what the newly founded Greens polled. Jean-Marie Le Pen took nought point seven five per cent in the French presidential election of 1974 and could not assemble the five hundred sponsorships required to stand in 1981. The British National Front fielded three hundred and three candidates in the general election of 1979, more than any insurgent party since Labour in 1918, held immediately after the Winter of Discontent with inflation at thirteen per cent, and won nought point six per cent. The Vlaams Blok sat at between one point four and one point nine per cent from 1978 to 1987, then quadrupled in 1991, in the middle of a boom. The Netherlands did not elect a single far-right member of parliament in the entire decade; the first arrived in 1982 on nought point eight three per cent.

Two exceptions must be stated plainly rather than managed. Italy had a genuine neo-fascist party, descended directly from the Republic of Salò, which took eight point seven per cent in 1972 and was the fourth largest party in the country throughout the decade. And the Austrian Freedom Party, sitting at five to six per cent, was led until 1978 by a man who had volunteered for the Waffen-SS at seventeen and served in a unit attached to Einsatzgruppe C, and it propped up Bruno Kreisky's Social Democratic government while doing so.

But the Italian trend through the shock years ran downward, from eight point seven per cent in 1972 to six point one in 1976 to five point three in 1979, and the party was excluded from government for the entire forty-six years of its existence by an informal agreement that never once broke. The Austrian surge came under Jörg Haider from 1986, a decade after the shocks ended.

The 1970s were not a calmer decade. They were a far more violent and more inflationary one, and the difference is not in the size of the shock.

What the Seventies Had

The difference is in what stood between the shock and the ballot box, and it can be measured.

Public social expenditure in West Germany rose from sixteen point six per cent of GDP in 1970 to twenty-three point one per cent in 1975. In the Netherlands it rose from nineteen point three to twenty-five point four over the same five years. Those welfare states grew by roughly forty per cent relative to national income in half a decade, in direct and visible response to the crisis. Since the mid-1990s the equivalent ratios have been broadly static, and the sharpest single cut in the set, Dutch disability provision, took that budget from six point eight per cent of GDP to three point nine. The contrast is not between a generous welfare state and a mean one. It is between one that was visibly expanding in answer to a shock and one that is visibly defending a perimeter.

Trade union density across the OECD averaged thirty-eight and a half per cent in 1975 and stands at fifteen point two per cent today. German density peaked at thirty-four point three per cent in 1978 and is now fourteen point one. Collective bargaining coverage in West Germany stood at around eighty-five per cent from 1970 to 1990; in 2023 Germany fell below half of all employees for the first time.

Party membership tells the same story from a different angle. In 1980, just under four per cent of the eligible population of the Federal Republic belonged to one of its parties, close to two million citizens. At the end of 2024, with three additional parties in the count including the AfD, the figure was just over one point six per cent. The Social Democratic Party reached its all-time membership peak of one million and twenty-two thousand in 1976, at the exact centre of the shock decade. It now has three hundred and fifty-seven thousand members. The Christian Democrats have fallen from seven hundred and eighty-nine thousand in 1990 to three hundred and sixty-four thousand.

And the conflict itself was not suppressed. It was conducted. The United Kingdom lost twenty-nine and a half million working days to industrial disputes in 1979. In 2018 it lost two hundred and seventy-three thousand, a difference of roughly a hundredfold. The 1970s were not a decade in which people were calm. They were a decade in which fury had an address, a meeting, a subscription, a shop steward and a negotiating table, and in which the institutions receiving it were visibly growing rather than visibly retreating.

Anger needs an address.

There is quantitative support for exactly this reading. Zack Grant's analysis of three hundred and ninety-three general elections across twenty-two democracies between 1950 and 2016 finds that economic hardship and the ideological convergence of mainstream parties interact, and that their combined effect on anti-system voting is greater than the sum of the parts. In a recession with a converged mainstream, anti-system parties take around fourteen per cent of the vote. In the same recession with a polarised mainstream, they take around nine. Convergence raises anti-system support by more than half.

The 1970s were the high-water mark of mainstream polarisation in Western Europe. A communist party at thirty-four per cent in Italy, a social democratic party at forty-six per cent in West Germany, the Programme Commun in France, Labour against the ascendant Thatcherite right in Britain. In 1972, at ninety-one point one per cent turnout, two West German parties took ninety point seven per cent of the vote between them. In February 2025, at eighty-two point five per cent turnout, the same two parties took forty-four point nine.

The absorbers were not overwhelmed by a larger shock. Their mountings had loosened between shocks, and the next shock arrived before anyone had checked them.

The Instrument With No Voter

The unbolting has a paper trail, and it is the least read paper trail in European government.

The European Semester was agreed by finance ministers in September 2010 and ran its first cycle in 2011, anchored in EU law by the Six-Pack the following year and resting on Articles 121 and 148 of the Treaty. Each year the Commission analyses every member state's economy and proposes country-specific recommendations. The drafts pass through Council preparatory committees, are discussed by finance and employment ministers, endorsed by the European Council in June and formally adopted by the Council in July.

The European Parliament, the Union's only directly elected institution, cannot amend a recommendation, delay one, or block one. What it has instead is what the Six-Pack calls an Economic Dialogue: its committees may invite the presidents of the Council, the Commission, the European Council and the Eurogroup to come and discuss their decisions. The Commission that drafts is appointed. The ministers who adopt hold national mandates and answer to national parliaments, not to any European electorate, and those national parliaments have no formal role in the procedure at European level at all. Nothing in the chain is elected to do the thing the chain does.

What that machinery recommended is documented, counted and peer-reviewed. Between 2011 and 2019 it issued eighty-nine wage-related recommendations to sixteen member states, coded by Mattia Guidi and Igor Guardiancich into one hundred and ten distinct prescriptions, with a consistent direction that the authors describe as unambiguously preferring reforms that reduce workers' protection. Between 2011 and 2023 it issued four hundred and thirty-eight pension recommendations, of which seventy explicitly proposed raising the retirement age and roughly seventy per cent of the general ones were framed in terms of financial sustainability rather than adequacy. Health system recommendations went from eight member states in 2011 to twenty-one in 2014, framed throughout in the language of fiscal sustainability rather than of health.

The texts are public and they are not ambiguous. Belgium, 2012: reform the system of wage bargaining and wage indexation, ensure that wage growth better reflects productivity and competitiveness, and, in the operative phrase, facilitate the use of opt-out clauses from sectoral collective agreements. Slovenia, 2012: adjust employment protection legislation as regards permanent contracts. Italy, 2015: promote an effective framework for second-level contractual bargaining. Malta, 2015: accelerate the already enacted increase in the statutory retirement age and link it to changes in life expectancy.

The objection is on the record too, inside the machine. A Council document of June 2016 notes that the trade unions found the recommendations on wages and collective bargaining, particularly calls for decentralisation of collective bargaining, rather worrying, and rejected what they saw as an overarching Commission view that employment protection legislation was too strong. That is not a press release from a union. That is a Council document recording the objection and proceeding anyway.

There is a party on the other side of this and it is not the Commission. Where German coverage actually fell, from seventy-two per cent in 1996 to fifty-one by 2017, the mechanism was domestic and it had a name. Employers' associations began offering a second class of membership, ohne Tarifbindung, which gave a firm the association's services without the obligation to apply its collective agreement. A firm could belong and not be bound. Meanwhile the employers' side holds parity on the body that decides whether an agreement is extended to firms outside it, and extensions fell for two decades against the European trend. Gunther and Hopner, who traced that fall, call it a German institutional singularity rather than externally driven liberalisation. The recommendations named a direction. The instrument that moved in it was held at home.

This is not a policy imposed on Europe from somewhere outside it. This is an address removed from inside it.

Who Asked For It

Here the honest version of the argument diverges sharply from the convenient one, and the honest version is more damaging.

The recommendations were almost never implemented. The European Court of Auditors, examining the years 2011 to 2017, found that one point six per cent of country-specific recommendations were fully implemented in the year they were issued, with fifty point nine per cent recording some progress and thirty-five point five per cent limited progress. On a cumulative multiannual basis over 2011 to 2018 the full implementation rate reaches nine per cent. Bruegel's index of implementation fell every year from 2013 to 2017. Pension and unemployment benefit recommendations were not implemented more than average; they were implemented less. The Court also observed that there is no publicly available consolidated database from which the recommendations and their implementation status can be retrieved.

An instrument fully implemented in fewer than two cases in a hundred cannot have dismantled a continent's institutions by force. It did something more interesting. It supplied an external justification, renewed annually, for domestic choices that could then be presented as responses to a constraint rather than as preferences originating at home.

Kenneth Dyson and Kevin Featherstone gave this mechanism its name three decades ago, studying Italy's road into the single currency: the vincolo esterno, the external constraint, sought by domestic elites precisely because it empowers technocrats and disarms veto players at home. The direction of agency runs the opposite way from the one most critics assume. It is not principally Brussels overriding unwilling nations. It is national executives outsourcing unpopular decisions to a body that cannot be voted out, and then standing for election as though the decision had been made elsewhere.

Three further corrections belong here, because a reader will find them otherwise.

Where collective bargaining actually collapsed fastest, the instrument was harder than a recommendation. Portuguese sectoral agreements fell from two hundred in 2008 to forty-six in 2013, and coverage from one point nine million workers to two hundred and forty-two thousand, under commitments made to the Troika in a memorandum of understanding with legal force. Romanian coverage fell from around ninety-seven per cent to thirty-five under a 2011 law passed alongside balance-of-payments conditionality. Those were loan conditions, not recommendations. The Semester extended a softer version of that logic into the routine governance of states that were never in a programme, which is a different and more troubling claim than the one usually made.

German bargaining erosion is substantially domestic. It runs from 1996, predates the Semester by fifteen years, and is driven by employers' associations offering membership without collective agreement obligations and by the withdrawal of statutory extension. Germany received comparatively few wage recommendations and declined anyway.

And the direction has since reversed. Directive 2022/2041 obliges every member state with collective bargaining coverage below eighty per cent to publish an action plan to raise it. Only ten of twenty-seven are above that threshold; Germany is not one of them. In November 2025 the Court of Justice annulled the Directive's provisions on setting statutory minimum wage levels, and left Article 4 standing. The obligation to promote collective bargaining survived judicial review.

Nobody in the chain had to face an election for the decision the chain had just made. That was not merely a flaw in the design. For governments looking for an external constraint, it was part of the attraction.

Saxony-Anhalt as an Instrument of Measurement

Which returns us to a state of two point one million people that has lost eighteen per cent of its population since 2000 and eighty-two thousand of its registered voters in the five years since the last election.

Eastern Germany is where the removal of the absorbers can be read directly, because those institutions never reached the depth there that they reached in the west. Fifteen per cent of eastern German establishments are bound by a sectoral collective agreement and a further three per cent by a company agreement, which leaves eighty-three per cent bound by nothing. Thirty-one per cent of eastern employees are covered by a sectoral agreement, against forty-three per cent in the west. Works councils reach thirty-one per cent of eastern private sector employees. The median eastern household holds thirty-five thousand nine hundred euro in net wealth against one hundred and forty-three thousand in the west, a ratio of one to four. Among the top leadership positions of Germany's most important companies, the share held by east Germans in 2024 was zero point zero per cent, down from two point nine in 2018, against an east German population share of around a fifth.

And then the finding that should end any comfortable reading of this election. In Saxony-Anhalt, according to Forschungsgruppe Wahlen, trade union members voted for the AfD at forty-four per cent. Non-members voted for the AfD at forty-four per cent.

In this election, union membership sorted nothing. That is not a measurement of how many members are left. It is a measurement of what being one still does.

This is the same shape as the Reich insurance fund running a surplus while a fifth of the unemployed received nothing. The institution is present. Its accounts are in order. The function does not register.

France demonstrates the point from the opposite direction and must be allowed to. French collective bargaining coverage is ninety-eight per cent, the highest in Europe, achieved through administrative extension. French union density is ten per cent, the lowest in the OECD. France has the strongest far right in Western Europe by vote share. Coverage without membership delivers a wage floor. What it does not deliver is the meeting, the steward, the argument, the political education, the experience of being angry among others. It is a payment, and a payment is not a membership.

Not a German Story

Saxony-Anhalt is legible because the data there are unusually clean, not because eastern Germany is unusually susceptible. The same reading runs across the continent, and in one country it runs sharper than in Germany.

Dutch spending on disability and incapacity benefits fell from six point eight per cent of GDP in 1980 to three point nine by 2021, counted across public and mandatory private schemes together so that the fall is not an accounting effect. On the headline public measure the Dutch decline looks larger still, but most of that gap is reclassification: when the 2006 reform moved health insurance to private carriers under a public mandate, more than four points of GDP crossed from the public column to the mandatory private one, and total Dutch social spending rose that year. The disability retrenchment is the real one. Dutch union density fell from thirty-one point seven per cent in 1980 to thirteen point eight. Collective bargaining coverage fell from eighty-six per cent in 2010 to seventy-two. At the general election of October 2025 the parties classified as far right took twenty-seven per cent of the vote between them.

That figure needs its complication stated, because the headline ran the other way. The Party for Freedom lost eleven seats and fell from twenty-three and a half per cent to sixteen point seven, edged out of first place by D66 by the narrowest margin since 1956. Read as a party, that is a defeat. Read as a bloc, JA21 gained eight seats and Forum for Democracy four, and the vote did not leave. It moved sideways.

Elsewhere the arithmetic is blunter. The Austrian Freedom Party took twenty-eight point eight per cent in September 2024, first place and the best result in its history, having sat at five to six per cent throughout the 1970s under a leader who had served in the Waffen-SS. Chega took twenty-two point eight per cent in Portugal in May 2025 and became the second largest party in the parliament, ahead of the Socialists. The National Rally and its allies took thirty-three point two per cent in the first round of the 2024 French legislative election, against Jean-Marie Le Pen's nought point seven five per cent fifty years earlier. Brothers of Italy, the organisational descendant of a party excluded from government for the entire forty-six years of its existence, now leads one.

The PopuList project, which classifies parties across thirty-one European countries by the combination of nativism and authoritarianism, puts the far-right share at five point two per cent in 1995 and twenty-three point two per cent as of 2026, measured on the most recent election in each country and weighted by population, so that Germany carries a hundred and forty-seven times the weight of Malta. The entire increase sits in one column. The non-populist far right, the parties descending directly from the interwar movements, has stayed below two per cent throughout and sits roughly where it started.

What has grown is not the interwar right recovering its audience. It is something else occupying the space where representation used to be.

The Loop

One word carries the load in all of this, and it is worth asking what it measures. Sustainability, in the language of the recommendations, is a property of a budget line over time. It can be computed from a pension liability, a wage bill, a health expenditure path. It cannot be computed from whether a wage-setting system also produces meetings, stewards and arguments, because that is not a cost and appears in no column. A term meaning fiscal durability has been doing duty as a term meaning institutional durability. They are not the same word.

A crisis arrives that no national government can resolve alone. The response is coordinated at a level where electoral accountability for the resulting package is indirect, because that is where coordination is possible. The coordination recommends reforms that thin the institutions through which economic grievance is collectively negotiated, because fiscal sustainability is legible to the machinery and political absorption is not. National executives adopt as much of it as they can carry, and campaign at home as though the choice had been made abroad. The next crisis arrives to find fewer institutions standing between the shock and the household. The household, finding fewer organised places able to carry the grievance, increasingly puts it on a ballot paper. The centre responds by adopting the vocabulary of the party that received it, while retaining the machinery that produced it. That is not a policy failure. That is a loop, and since 2008 it has run through a banking crisis, a sovereign debt crisis, a migration crisis, a pandemic and an energy shock.

Fiscal cost was visible in the model. Political absorption was not. It never entered the assessment as a cost at all.

When the Ballot Does Not Close the Loop

The loop has one further stage, and it runs backwards.

Elections do not only count preferences. They tell people whether political action does anything, and that can now be measured. Caner Simsek, working with European Social Survey data covering twenty-two countries and thirty-three thousand respondents, finds that what populist voters tend to share is not an ideology but a sense of political inefficacy, the belief that acting politically does not translate into influence. The kind of grievance then leans the direction: cultural grievance toward the populist right, economic grievance toward the populist left. The finding has a limit, and the limit is the interesting part. In Bulgaria, Czechia, Hungary, Italy and Poland the pattern inverts, and it is the more efficacious who vote populist. Those are the countries where populist parties have held power, which suggests the mechanism describes challengers and stops describing incumbents.

The other half of the circuit was closed in August 2026. Nicholas Charron and Jana Schwenk exploited elections that fell in the middle of European survey fieldwork, comparing people interviewed in the days before a result with people interviewed in the days after, across nineteen thousand respondents in Poland, Slovakia, Romania, Luxembourg and two German states. Elections change political trust, causally. The effects are asymmetric: winners gain roughly twice what losers lose. And the losing half is not general. It comes, in their words, entirely from supporters of populist parties, who also gain more than twice as much as anyone else when their party advances.

A vote is not only an instrument for changing a government. It is the instrument through which a citizen finds out whether the instrument works.

The Netherlands ran that experiment in public. In the spring of 2023 the Social and Cultural Planning Office found fifty-nine per cent of Dutch respondents dissatisfied with politics in The Hague, and fifty-eight per cent giving the government a failing grade for trust. Asked why, they did not name a crisis. The Office's own word was opeenstapeling, an accumulation, and it recorded that opinion was dominated not by one or two concerns but by a whole range: a political system seen as failing to function, income differences and poverty, immigration, the shortage of affordable housing, the way people live alongside each other.

Then they voted, and the party promising to break the accumulation became the largest in parliament. Political trust among its voters rose. The National Election Study measured the same people before and after and found the increase among the party's long-standing core as well as among its new arrivals.

Four qualifications, which matter more than the finding. The rise was small, and the Planning Office said so. It came off a floor so low that a majority still failed both parliament and government afterwards. It was not confined to one party but appeared across the winners, while voters of the losing left did not move at all. And it had happened before, in 2010, after the same party's previous breakthrough.

What the ballot restored was not confidence in government. It was the belief that voting registers.

What Keeps the Loop Turning

Work covering Western Europe from 2002 to 2016 found that where mainstream parties fail to respond to a changing conflict structure, openings appear for challengers of both the radical left and the radical right, and that political discontent raises their support only while those parties sit in opposition. Discontent is not a general engine. It runs while the party is outside the government.

That is not a case for admitting them to it, and nothing here should be read as one. A study of a hundred and nine elections across twenty-six member states found exactly what the argument in this article would predict, that mainstream parties are less responsive to the losers of globalisation, and then found no systematic relationship between that gap and right-populist electoral success. The mechanism is real and it is not sufficient.

What survives is narrower and harder to dismiss. Governments do not have to persuade anyone toward the edges. They only have to keep producing the experience of unresolved grievance alongside the sense that ordinary political correction changes nothing.

Europe's governments are not observers of the movement now challenging them. They are one of its conditions.

The Strongest Case Against This Reading

Nor was the direction uniform even at the time. Coding of the social and employment recommendations finds the share framed as social investment rather than retrenchment rising across the decade, and the European Pillar of Social Rights was proclaimed in 2017, inside the window described here. The claim is about what was recommended persistently, to identifiable states, in identifiable domains. It is not that one doctrine was applied everywhere.

The most serious counterargument does not dispute a single figure above. It accepts the decline of the absorbers and offers a cleaner explanation of the outcome.

Manuel Funke, Moritz Schularick and Christoph Trebesch examined more than eight hundred elections and over a hundred crises across twenty advanced democracies since 1870, and found that far-right vote shares rise by about a third in the five years following a systemic banking crisis, while producing no comparable effect after ordinary recessions or severe non-financial shocks. On that evidence the 1970s null result requires no institutional explanation at all. The 1970s were an inflation and supply shock. 2008 was a banking crisis. The far right should have risen after one and not the other, and it did. A second objection is at least as strong: the defining issue of the modern radical right was not available in the 1970s. West German asylum applications stood at thirty-three thousand in 1978 and around four hundred and thirty-eight thousand in 1992, and every European far-right breakthrough, in France in 1984, the Netherlands in 1982, Austria from 1986, Belgium in 1991, tracks that curve far more closely than it tracks unemployment. David Art's interview work adds a third: whether a country acquired a viable radical right depended on the activists available to staff one, and on whether the initial official reaction was permissive or repressive.

These are serious and they are partly right. The claim made here is narrower than they assume. It is not that strong intermediary institutions prevent far-right voting. It is that they help determine what a grievance can become before it reaches a ballot, and whether the voter believes the ballot changed anything after it was cast. The banking-crisis finding explains why demand intensified after 2008. It does not by itself explain the political form that grievance took once it had intensified, or why an earlier era of hardship produced organised strikes, mass party recruitment and a rapidly expanding welfare state instead. Something between the grievance and the ballot changed, and it is measurable, and it was written down.

What Remains

The vote of 27 March 1930 is worth returning to one last time, because of what it was not.

It was not a vote about democracy. Nobody in that room was choosing between parliamentary government and rule by decree. They were arguing about whether an insurance contribution should be three and a half per cent or four, and both sides had defensible fiscal positions, and both believed the argument would continue the following autumn. The decision that ended the Republic's parliamentary life was taken by people who did not know they were taking it, in a dispute whose subject was so technical that no one has bothered to remember it.

Nothing in the European Semester was decided in a room where anyone knew they were doing this either. Each recommendation was reasonable on its own terms, addressed to a real fiscal problem, drafted by competent officials, adopted by ministers who had won elections at home. There is no memorandum anywhere proposing that Europe should dismantle the machinery through which its citizens convert economic anger into political demands, because nobody proposed it. It was the aggregate of a hundred separate decisions about sustainability, each of which had a named author and none of which had an author responsible for the sum.

That is not an excuse. A system that can price a pension liability to four decimal places and cannot tell whether anyone still belongs to anything is not a neutral instrument. It has decided, without ever deciding, which costs are real. It could price what it could not protect. It could not protect what it did not price.

That is what the comparison with the 1930s is actually about, and it has nothing to do with the resemblance between one politician and another. Weimar's absorbers were stripped by emergency decree in three years by men who thought they were saving the currency. Europe's were left to erode across fifteen years, reviewed each year by a machinery that recommended reforms running in the same direction, by people who thought they were saving the euro. The mechanisms are not comparable and nothing here claims they are. What is comparable is the residue. In both cases the institutions kept their names, kept their staff and kept their accounts, while losing part of the representative function that had made people believe in them. In both cases the electorate noticed before the institutions did, and voted accordingly, in record numbers, from a position of maximum participation rather than minimum.

Sixty-eight per cent of Europeans surveyed for the European Council on Foreign Relations this spring described the continent's condition as general deterioration or a loss of normality. Twelve per cent described it as renewal. Nearly three-quarters said Europe is coping badly with its greatest challenges. That is not a measurement of ideology. Read through this structure, it is what a continent sounds like when conditions are felt everywhere and responsibility can be located nowhere.

Weimar's loop had an ending. It did not stop because the mechanism corrected itself. It stopped because the constitutional system that contained it was destroyed, and what replaced it was worse than the loop. That is not a forecast about Europe and nothing here should be read as one. It is an observation about the mechanism. Nothing in it guarantees its own correction. It does not exhaust itself and it does not self-correct, and there is no scheduled review at which anybody is required to ask whether the institutions being reformed were doing anything else.

The architecture accounts for every actor. The Commission that proposes, the Council that adopts, the government that implements, the party that inherits the anger. Every one of them can point to the next.

There is only one position in the entire structure that cannot point anywhere, and it is the one from which the ballot is cast.

Evidence Map

Core claim: Between 1930 and 1932 in Germany, and between 2011 and 2026 in the European Union, the institutions that convert economic grievance into organised political demand were reduced in function while being preserved in form, in both cases through administratively defensible decisions whose electoral accountability was either displaced from parliament or indirect across levels of government, and in both cases the electorate responded by mobilising rather than abstaining.

Observed conditions, high confidence, directly documented: The Müller cabinet resigned on 27 March 1930 in a dispute over raising the unemployment insurance contribution from 3.5 to 4 per cent, and was the last Weimar government resting on a parliamentary majority; the contribution was raised by decree to 4.5 per cent on 26 July 1930 and to 6.5 per cent by October. Emergency decrees against Reichstag laws ran 5 against 98 in 1930, 44 against 34 in 1931 and 60 against 5 in 1932, while sitting days fell from 94 to 41 to 13. Insurance coverage of the unemployed fell from 67.5 per cent in 1928/29 to 19.4 per cent in 1932, with 25.9 per cent on crisis relief, 36.6 per cent on municipal poor relief and 18.2 per cent receiving nothing, while the Reich Institution recorded a surplus of 372 million Reichsmark in fiscal 1932/33, a figure carried in the Federal Employment Agency's own history of its predecessor and given as 372.8 million in the Bayerische Staatsbibliothek's commentary on the 1927 Act. Turnout rose from 75.6 to 82.0 to 84.1 per cent between 1928 and July 1932, and fell to 80.6 in November 1932 as the NSDAP fell from 37.3 to 33.1 per cent. Saxony-Anhalt on 6 September 2026: AfD 43.8 per cent and 39 of 83 seats, CDU 17.2 per cent and no constituency won, turnout 77.8 against 60.3 in 2021. Union members and non-members both voted AfD at 44 per cent. Eighty-three per cent of eastern German establishments are bound by no collective agreement.

Documented structural dependencies, medium-high confidence: The European Parliament, the Union's only directly elected institution, has no power to amend, delay or block a country-specific recommendation; adoption rests with nationally mandated ministers in the Council. Eighty-nine wage-related recommendations were issued to sixteen member states between 2011 and 2019, and 438 pension recommendations between 2011 and 2023, of which 70 explicitly proposed a higher retirement age. Council document 9684/16 records the trade union objection to recommendations for decentralisation of collective bargaining. The European Court of Auditors found 1.6 per cent annual full implementation for 2011 to 2017, and no publicly available consolidated database of recommendations.

Analytical inferences, medium confidence: That the withdrawal of intermediary institutions, rather than the magnitude of the economic shock, governs whether grievance becomes organised demand or anti-system voting. Supported by Grant (2021) on mainstream convergence, by the 1970s divergence and by the Saxony-Anhalt union-membership null result, but observational throughout. The direction of the recommendations was not uniform across the window: the social-investment share of social and employment recommendations rose through the decade and the European Pillar of Social Rights arrived in 2017, so the claim is about persistence in identifiable domains and states, not about a single doctrine applied everywhere.

Open hypothesis, stated as inference: That the European Semester functioned as a vincolo esterno through which national executives escaped domestic veto players, and that voters did not attribute the resulting reforms to the European level. The first limb is established for Italy and monetary union by Dyson and Featherstone. No study measures the second.

What would confirm this: A study measuring attribution, showing that electorates in heavily recommended states did not assign responsibility for wage and pension reforms to the European level. Panel evidence that intermediary institutional strength predicts the conversion of economic shock into organised demand rather than anti-system voting, after controlling for financial-crisis incidence.

What would disprove this: A study finding that voters did attribute those reforms to the European level and voted accordingly, since the argument depends on the absence of attribution. Evidence that bargaining coverage and union density declined at equivalent rates in states that received no wage recommendations, which would remove the Semester from the chain. A specification in which far-right vote share is fully predicted by financial-crisis incidence with no residual explained by institutional strength. France weakens the reading without destroying it: near-universal bargaining coverage coexists there with the strongest far right in Western Europe, which is why the argument is made about density and organisation rather than about coverage.

Instrument boundary and watchlist: The collapse of collective bargaining in Portugal, Greece and Romania occurred under Troika memoranda and balance-of-payments conditionality, which are binding loan conditions and not country-specific recommendations. Nothing here attributes those cases to the European Semester. Under watch: the action plans required under Article 4(2) of Directive 2022/2041 from the seventeen member states below 80 per cent coverage; the main proceedings on the federal AfD classification before the Verwaltungsgericht Köln, the state branch in Saxony-Anhalt having been classified a confirmed right-wing extremist endeavour since 7 November 2023; and whether the net-expenditure paths binding from 2024 reopen the channel that Directive 2022/2041 was meant to close.

The withdrawal of correction has two documented layers in the archive. The Architecture Cannot Move traced the same absence through energy, defence and currency, three systems locked into a tension none of them can release alone. Palantir and the Price of Exit recorded the earlier half of the mechanism, a procurement decision that removed the option of leaving before anyone had asked whether leaving might be needed.