The majority that does not exist — what the 1 September vote revealed about Romania's next prime minister

1. One hundred and nineteen days, not one hundred and thirty

The motion of censure filed jointly by PSD and AUR was carried on 5 May 2026. From that day, the cabinet led by Ilie Bolojan has been dismissed and exercises only caretaker powers. By 1 September 2026, 119 days had passed.

The figure circulating in the press — "more than 130 days" — does not match the date of the vote. The discrepancy is not cosmetic: over a stretch this long, every week means a budget revision postponed and a European milestone missed. We count from the vote, because the vote is the legal act that produced the dismissal.

It is, however one counts, the longest caretaker government of Romania's post-1989 era. And its chronology reveals a pattern, not a run of accidents:

• 5 May — the PSD-AUR motion passes. The Bolojan government is brought down with 281 votes.
• 4 June — Eugen Tomac is nominated. PNL objects that his list of ministers favours PSD.
• 14 June — Tomac hands back his mandate without ever calling a vote in Parliament.
• 22 June — Adrian Veștea's government seeks investiture and is rejected: 189 votes in favour, 23 against, out of the 233 required. The opposition did not vote against; it refused to supply the political quorum.
• 31 August — the recovery plan (PNRR) comes to an end. The public pay law, a milestone Romania had signed up to, is not adopted.
• 1 September — Parliament opens its ordinary session. The leadership of the Chamber of Deputies changes through a vote we return to below.

2. The arithmetic of investiture

A government is invested in a joint sitting of the two chambers, by a majority of deputies and senators. Parliament has 464 seats, so the threshold is 233.

The breakdown by parliamentary group at the opening of the autumn session:

Group	Seats
PSD	127
AUR	90
PNL	76
USR	59
UDMR	31
Unaffiliated MPs	23
National minorities	17
SOS România	15
Uniți pentru România	15
PACE	11
Total	464

The table checks itself. PSD, AUR, PACE, SOS, "Uniți pentru România" and the unaffiliated MPs add up to exactly 281 — the figure of the 5 May motion. The remainder, 183, is the combined strength of the PNL, USR, UDMR and national-minorities groups.

This is where the standard reading of the past four months began: there is a majority that can tear a government down and none that can build one. The reading was right as description and wrong as prediction, and the difference showed on 1 September.

3. The 1 September vote: the majority showed itself

At the first plenary sitting of the ordinary session, the Chamber of Deputies altered the composition of the Standing Bureau. Two secretary posts held by PNL and USR passed to AUR and SOS România: Raluca Turcan and Oana Murariu were replaced by Ramona Bruynseels and Virginia Vedinaș. PNL and USR walked out of the chamber before the vote, then filed a challenge with the Constitutional Court, arguing that the change breaches the rules on proportional representation of parliamentary groups.

The detail that matters most is not who won the posts. It is who voted. The USR vice-president of the Chamber, Cătălin Drulă, noted publicly that the new line-up also passed with the votes of UDMR and of the national minorities group — precisely the two groups that the "281 to 183" reading had placed on the other side of the line.

The causal chain runs as follows. Cause: Parliament is fragmented in such a way that the only stable arithmetic majority includes AUR. Mechanism: every pro-European party that joins that majority pays a reputational cost greater than the gain from governing, and therefore prefers deadlock — for as long as deadlock is free. Effect: 119 days of caretaker government, a budget run through emergency ordinances, missed milestones. What changed on 1 September: deadlock stopped being free, because the leadership posts in the Chamber are shared out at the start of the session, once and once only. The moment the cost of abstaining became concrete, the majority materialised in a single sitting.

The counter-hypothesis, which deserves to be taken seriously. A vote on the composition of the Standing Bureau is cheap; an investiture vote is expensive. UDMR can hand AUR a seat in the Chamber's leadership on Tuesday and refuse on Wednesday to invest a government with AUR behind it — the leader of UDMR's deputies said precisely that: the party does not rule out a minority Grindeanu government, but does not want to "end up in the same boat as AUR". The 1 September vote demonstrates the majority's capacity, not its willingness. That distinction is the whole stake of September.

There remains, even so, a piece of arithmetic that no statement can undo. PSD, AUR and SOS România together hold 232 of the 464 seats. The investiture threshold is 233. Exactly one vote is missing.

And from there upwards, every group that joins produces a comfortable majority: with "Uniți pentru România" the total reaches 247, with UDMR 278, and with the national minorities group 295. This is the real stake of September, and the reason the negotiation is being conducted over small groups rather than large parties. A government does not need UDMR to enter the cabinet; it needs UDMR not to leave the chamber.

4. Why dissolving Parliament is not on the table

AUR is calling for early elections, and part of the public commentary treats dissolution as a valve available at any moment. It is not.

Article 89 of the Constitution allows the President to dissolve Parliament only if two conditions are met cumulatively: at least 60 days must have passed since the first request for investiture, and at least two requests to invest a government must have been rejected.

The first condition was met long ago. The second has not been. There has been a single rejection — the Veștea government, on 22 June. Eugen Tomac withdrew his mandate on 14 June without putting it to a vote in Parliament, and a withdrawn request is not a rejected one.

The practical consequence: the road to early elections necessarily runs through a third government proposed and voted down in plenary. And even then, the President may, not must, dissolve. This is the institutional trap that explains why AUR has simultaneously pushed a resolution to amend the Constitution, with binding deadlines for nomination: under the present architecture, a party that wants elections has no way of triggering them on its own.

5. NRRP: the big percentage and the denominator cut three times

On 31 August the National Recovery and Resilience Plan (NRRP) closed, and on 1 September the tallies appeared. The interim prime minister announced take-up of more than 90% on the grant component and more than 95% on loans. The minister for European funds set out the final estimate: 90.58% of grants and more than 97% of loans, or around 93% overall, "with possible fluctuations". The Ministry of Development announced that it had met all its milestones.

All of these figures are real. None of them answers the question that matters.

The absorption rate is calculated against the plan in force at the end. Romania's plan was revised downwards three times, from roughly 29.18 billion euros in its original form to about 20.11 billion (13.57 billion in grants and 6.54 billion in loans). The sums actually drawn down come close to 13 billion euros.

Measured against the reduced plan, those 13 billion work out at around 64% — precisely the absorption rate the business press put in its headlines when the plan closed, alongside Romania's 25th place out of 27 on milestone delivery. Measured against the original allocation, it amounts to roughly 45%. A figure above 90% on one particular component is true and, at the same time, the least informative way of describing what happened: a plan stripped of the projects that could not be delivered has, by construction, a good execution rate.

What was left outside can be counted: more than 2,000 unfinished local projects, the A7 Bacău–Pașcani and A1 Margina–Holdea motorways dropped from the financing and moved onto the state budget or onto conventional programmes, and the milestone on the public pay law, missed because the parties failed to reach agreement before the deadline.

That last milestone is the one that ties the European chapter to the political one. Its failure did not only cost money. It cost the signal.

6. What the markets did — and why the leu is not telling the truth

Market reactions to a political crisis are misread if you look where everyone else is looking.

The exchange rate did not break. The euro/leu spot rate stood at 5.2547 lei on 1 September. The 2026 high for the same series was 5.2639 lei, reached on 6 May — the day after the government fell. In other words, after 119 days of crisis, four months of a caretaker administration and two failed nominations, the leu is marginally stronger than it was the day immediately after the no-confidence motion. The official reference rate of the National Bank of Romania says the same thing over the period it covers: its 2026 high, 5.2688 lei, also dates from 6 May.

That does not mean the market cannot see the risk. It means that the exchange rate is a policy variable, not a thermometer. The central bank has reserves, instruments and a long record of using them. A stable exchange rate through a drawn-out political crisis is evidence that someone is holding it stable, not evidence that there is no crisis.

Equities did break. The BET index hit its high for the year on 19 August, at 36,487.21 points. On 31 August it closed at 33,743.73 — a fall of 7.52% over eight trading sessions, 3.61% of it in a single day, on 31 August itself. On 1 September it recovered 1.03%, to 34,090.29 points, still 6.57% below the peak.

An honest caveat is required here. 31 August was also a risk-off day on external markets, on the back of a jump in the oil price. The attribution cannot be clean: it cannot be claimed that the entire 3.61% decline is Romanian. What does hold is the internal comparison — the managed instrument did not move, the free instrument fell. Political risk shows up in equities and in government bond yields, not in the reference rate.

The macroeconomic backdrop on which all of this rests:

• GDP, quarterly change: 0.0% in Q2 2026, after −0.1% in Q1 and −1.9% in Q4 2025.
• Annual inflation: 8.16% in July 2026, down from 10.42% in June — an improvement that owes a great deal to the 2025 price rises dropping out of the base, rather than to fresh disinflation.
• Harmonised unemployment: 6.4% in July 2026.
• Cash budget deficit: 2.34% of GDP in the first seven months, down from 3.99% in the same period of 2025 — 48.08 billion lei against 76.44 billion, 37% less. But with interest payments of more than 40 billion lei over those same seven months.

Romania's sovereign rating is BBB−, the lowest rung of investment grade, at all three major agencies — and, at each of them, with a negative outlook. Fitch confirmed the level in August 2026, after the government challenged the initial assessment and submitted additional data. The margin to speculative grade is, literally, one notch. After the public pay bill failed, market analysts flagged that downgrade risk was rising, and the interim finance minister publicly pointed to the S&P review as the next test. This is the real external constraint: not Brussels, which has just closed the NRRP and no longer holds the leverage of milestones, but three agencies watching to see who signs the 2027 budget.

7. The debate: 14 virtual personas, three rounds

We put the question to an olivLaw panel of 14 virtual personas, each with its own objectives and constraints: the parliamentary parties and their internal factions, the presidency, the central bank, the Fiscal Council, the European Commission, a sovereign rating analyst, a fund manager and two voters — one on the left, one sovereigntist. Three rounds of deliberation, with each persona seeing the others' arguments before revising its position.

The question explicitly asked for 100 probability points to be distributed across five scenarios, plus a nomination of the main risk to the rating and to the exchange rate.

The aggregate result: a weighted average probability of 50.1% that a government with full powers is formed within eight weeks. Median 55%. Range 0–85%. Level of agreement 0.64. Distribution of positions: nine cautious, three optimistic, two pessimistic.

Convergence was unexpectedly strong, and it formed around the same fact: the vote of 1 September. The personas representing PSD, PNL, USR, the central bank, the Commission and the investment fund independently reached the same conclusion — that taking the leadership posts in the Chamber with AUR and SOS support is the tactical gain that makes forcing a majority possible. Their estimates for "a PSD prime minister" settled between 40% and 65%. The second shared argument: the closing of the NRRP has removed both an external pressure and a bargaining instrument, which leaves the parties freer to calculate in purely political terms.

Divergence was just as instructive. The political personas gave high probabilities, the financial ones gave low probabilities. It is precisely the gap you see in real life between a backroom calculation and a sovereign risk note.

A measurement flaw, reported as such. For two of the fourteen personas, the numerical value extracted automatically contradicts their own text: the rating analyst comes out at 0%, although his argument states 40% for the main scenario, and the Fiscal Council comes out at 6%, although it states 65%. This is a parsing error, not a position. Corrected against the personas' own text, the simple average rises from 47.6% to 54.6%. We mention it because a panel that reports its own instrument errors is more useful than one that does not — and because, in this case, the error pushes the conclusion in exactly the direction in which we disagree with the panel anyway.

The asymmetric risks identified by the panel, in order of severity: a caretaker period that drags on with no clear prospect, with a direct effect on fiscal credibility; and an unstable majority that resolves the crisis in the short term but generates a permanent risk of legislative deadlock. And the main invalidator, as the panel framed it: a public statement by the AUR leader that he will not vote for any PSD government would destroy the basis of the entire prediction.

Fact-checking the panel. Two factual claims made in the deliberation are wrong and we correct them here: one persona used an inflation rate of 3.8% (the real figure is 8.16% in July), and another put the budget deficit at over 6% of GDP (the seven-month cash deficit is 2.34%). Their arguments did not depend on these figures, but the reader needs to know that a panel of this kind produces reasoning, not data.

8. The name: Sorin Grindeanu

The question "who will be prime minister" only has an honest answer if it comes with a probability attached. Ours is Sorin Grindeanu, at roughly 25% as an individual name — which, on the calibrated scale we use, reads as "unlikely". That the best-placed name in Romania is unlikely is the finding, not an evasion.

The arguments in his favour have almost all accumulated in the past few days:

• At the PSD leadership meeting in Sinaia on 31 August, the party's position was stated without ambiguity: PSD's only candidate for prime minister is Sorin Grindeanu.
• Grindeanu publicly claimed the job, arguing from experience — he has been prime minister before.
• Asked about a PSD government passed with AUR votes, he did not rule it out: "If we get the vote, all the better".
• PSD has put forward a governing agreement with ten priorities and wants a government in place by 15 September.
• UDMR does not rule him out and has noted that its own candidate, Siegfried Mureșan, "has vanished from public life".
• The 1 September vote showed that the bloc that would install him can act together in a real division.
• The olivLaw panel of 14 virtual personas independently arrived at the same scenario.

The arguments against him, however, are structural rather than circumstantial, which is why the probability stays below one in three:

• It is not parliament that makes the designation, but the president. Nicușor Dan has twice avoided naming the PSD leader, picking first a candidate from a small party and then a Liberal. A third designation is not automatic.
• PNL has shut the door explicitly. The Senate president said on 1 September that the party is ready to move into opposition and will not back any government that includes PSD — "whether with an independent prime minister or any other kind of prime minister". He also confirmed that Liberals voting otherwise would be expelled.
• The cost to AUR is real. A PSD government propped up by AUR takes from AUR the very asset it sells: its standing as anti-system opposition. Voices inside the party warned as early as June that such support would "bring the party down".
• The referral to the Constitutional Court over the composition of the Permanent Bureau could legally unpick the show of strength of 1 September.

The second option also has a name: Alexandru Nazare, the interim finance minister. His is the name that surfaces every time the formula "a prime minister who is not a PSD member but governs with PSD" comes up — he was rated above Adrian Veștea as early as June, and PSD leaders have described him as "a very good option". He has the unique advantage of being the public face of the relationship with the rating agencies, which is precisely the constraint that is not up for negotiation. We give him roughly 12%, "very unlikely", for one specific reason: the PNL position announced on 1 September hits him directly. A Liberal minister leading a government with PSD would, by his own party leadership's declaration, be expelled from the party on the day of his investiture. His scenario becomes the main one only if PNL changes its position — and at that point the probabilities in the next section have to be rewritten.

9. Scenarios, with probabilities summing to 100

Scenario	Probability	What triggers it
(a) A PSD prime minister — Grindeanu or another party politician	35%	AUR accepts a price payable in portfolios; UDMR and the minorities repeat the 1 September vote
(b) An independent or technocrat prime minister, with PSD in government	20%	PNL revises its position under the pressure of a negative rating review
(c) A joint PNL-USR-UDMR candidate	5%	The numbers are not there: the three groups plus the minorities come to 183 of the 233 required
(d) Early elections called	10%	A third designation rejected on the floor, followed by the president's decision to dissolve parliament
(e) A prolonged caretaker period, with no government invested by 31 October	30%	PSD-AUR negotiations collapse over portfolios and the president postpones the third designation

We are more optimistic than the panel: 60% for a government invested by 31 October, against the panel's aggregate of 50.1% (54.6% once the parsing error is corrected). The difference has a single, verifiable cause: we give the 1 September vote more weight than the virtual personas did, because it is the only behavioural evidence in the entire crisis. Everything else is talk.

10. What would invalidate this analysis

Three things would dismantle the reasoning above, and each of them is publicly observable.

First: if George Simion states explicitly that AUR will not vote for any PSD government, under any formula. The whole construction rests on the assumption that the bloc of 1 September has a price. Such a statement would say that it does not, and the probabilities would shift heavily towards (d) and (e).

Second: if a third designation is again withdrawn before the vote, as in the Tomac case. That would show the deadlock runs deeper than a calculation over portfolios, and that the president has no candidate who can get through.

Third: if PNL votes for a government with PSD ministers after announcing expulsions. That would mean the public statements of party leaderships carry no predictive value in this system, in which case our prediction on Nazare has to be rewritten — upwards, not downwards.

We also flag the limits of the factual base. The breakdown by parliamentary group is the one recorded at the opening of the autumn session and can change through individual defections, which have been frequent in this parliament. We were unable to verify against primary sources the exact composition of the votes in the Chamber's sitting of 1 September, only the statements of those taking part. The exact date of the next S&P review has not been confirmed from an official calendar. The NRRP figures on the initial plan and the successive revisions come from secondary reporting rather than from the Commission's consolidated document.

11. Falsifiable predictions

Ten predictions with probabilities on the calibrated ICD-203 scale, an explicit horizon and a public verification criterion. They will be scored at resolution using the Brier score, in a separate record.

#	Prediction	Horizon	Probability	How to verify	Verification source
F1	A government with full powers will be invested by Parliament	31 Oct. 2026	more likely than not (60%) — the majority was demonstrated on 1 Sept., but its price has not yet been paid	Published decision granting confidence	Monitorul Oficial (the Official Gazette); Chamber of Deputies
F2	Sorin Grindeanu will be designated as candidate for prime minister	31 Oct. 2026	even chance (45%) — the only PSD proposal, but the designation belongs to the President	Official designation statement	Presidential Administration
F3	Sorin Grindeanu will be invested as prime minister	31 Dec. 2026	less likely than not (35%) — designation does not guarantee investiture	Published appointment decree	Monitorul Oficial
F4	Alexandru Nazare will be designated as candidate for prime minister	31 Dec. 2026	very unlikely (12%) — the PNL position of 1 Sept. rules him out directly	Official designation statement	Presidential Administration
F5	The PNL groups will vote to invest a government that includes PSD ministers	31 Dec. 2026	unlikely (25%) — a position publicly reaffirmed, with the threat of expulsion	Transcript of the investiture vote	Chamber of Deputies; Senate
F6	Parliament will be dissolved and early elections called	31 Dec. 2026	very unlikely (10%) — Art. 89 requires one more rejection, and dissolution remains discretionary	Published dissolution decree	Monitorul Oficial
F7	Romania will keep its sovereign rating in the investment grade category at all three major agencies	31 Mar. 2027	likely (70%) — the cash deficit is falling, but all three ratings already sit on the last notch, with a negative outlook	Rating action statements	S&P Global Ratings; Moody's; Fitch Ratings
F8	At least one of the three agencies moves Romania's outlook back from negative to stable	31 Mar. 2027	unlikely (25%) — the outlook is already negative at all three, so the next move is binary: either it is repaired or it is downgraded; repairing it requires a government and a budget voted through	Rating action statements	S&P Global Ratings; Moody's; Fitch Ratings
F9	The BNR reference rate for the euro will exceed 5.30 lei on at least one day	31 Dec. 2026	unlikely (30%) — spot at 5.2547 on 1 Sept., below the 6 May high; the rate is managed	The series of daily reference rates	National Bank of Romania
F10	The BET index will close the year below 36,487.21 points, the high reached on 19 August	31 Dec. 2026	likely (70%) — 34,090.29 on 1 Sept., that is 6.57% below the peak	The closing quotation of the year's final session	Bucharest Stock Exchange

12. Methodology and sources

Anchor data (verified on 1 September 2026): the euro/leu spot rate series, which gives 5.2547 lei on 1 September and the annual high of that same series, 5.2639 lei, on 6 May, each series being compared only against itself so as not to mix two sources; the series of official reference rates of the National Bank of Romania, whose 2026 high is 5.2688 lei, also on 6 May; the closing series of the BET index, which gives the high of 36,487.21 points on 19 August, the recent low of 33,743.73 on 31 August and 34,090.29 on 1 September; the quarterly series of gross domestic product change (0.0% in Q2 2026, −0.1% in Q1 2026, −1.9% in Q4 2025); the annual inflation rate reported by the National Institute of Statistics for July 2026 (8.16%, against 10.42% in June); the harmonised unemployment rate for July 2026 (6.4%); the interim finance minister's statements of 31 August on the cash deficit of 2.34% of GDP over seven months and on interest payments exceeding 40 billion lei; the National Recovery and Resilience Plan (NRRP) progress reports presented publicly on 1 September by the interim prime minister and by the European funds minister; the reports on the Chamber of Deputies sitting of 1 September and on the referral to the Constitutional Court; the statements of 31 August and 1 September by the leaders of PSD, PNL and UDMR; the rating actions published by the three agencies, including the Fitch affirmation of August 2026.

Method. The parliamentary arithmetic was verified by an independent tally: the six groups that voted for the motion of 5 May add up to exactly 281, and the remainder to 183, which confirms the breakdown used. The conditions of Article 89 were read as cumulative, in line with the constitutional text, and the number of rejections was counted by formal acts, not by intentions — a withdrawn request is not a rejected request. The NRRP absorption figures are reported in both forms, against the revised plan and against the initial allocation, because only the second form allows comparison over time. The debate was run with an olivLaw panel of 14 virtual personas over three rounds of deliberation, with aggregation of probabilities; its outcome is reported in full, including a numerical extraction error identified in two of the personas and two factual errors in their arguments. The final probabilities in this analysis are the newsroom's, not the panel's, and where they differ we have said in which direction and for what reason. All probabilities are expressed on the standardised ICD-203 scale and will be scored by Brier score at the resolution date, through a separate record, not by amending this article.

Disclaimer: This material is intended for information and analysis. It does not constitute investment advice, a political recommendation or a legal assessment. The probabilities are explicit estimates, framed so that events can contradict them, not certainties. The composition of the parliamentary groups and the parties' stated positions can change from one day to the next; the analysis reflects the public information available on 1 September 2026.