A government in a parliamentary system falls when it loses its majority. That is the rule, and it is close to the only one that matters. That is what the majority is for: it is the mechanism by which a prime minister can pass laws, survive confidence motions and remain in office.
The exception is what happened in London on Monday. Keir Starmer announced his resignation as prime minister and Labour leader with the parliamentary majority he won in 2024 entirely intact. He had not lost a confidence vote. He had not lost a division that required him to go. Nobody outside his own party removed him.
What everybody assumes
That a leader with a large majority is secure, and that the size of a mandate is a measure of how long it will last.
It is a reasonable assumption. It is how the system is described in every account of it, and it is how it worked for most of the twentieth century. A prime minister with a workable majority governed until an election, and elections removed them.
What the record now shows
Britain is about to have its sixth prime minister in seven years. It has held two general elections in that period.
Do the subtraction. Most of those changes were not made by voters. They were made inside a parliamentary party, by members of it, on the basis of a judgment about what would happen at an election that had not yet been called.
Which reverses the assumption completely. The majority is not what keeps a prime minister in office in this system any more. It is the thing that makes them removable, because a party with a large majority knows it has seats to lose and a mechanism for changing the person it will lose them under.
A leader clinging on with a majority of three is difficult to replace, because the replacement is risky and the arithmetic is fragile. A leader with a hundred and fifty seat majority can be replaced comfortably, and the successor inherits a working government rather than a crisis. Size makes it safe.
What actually did for him
Polling, principally, and the pressure that follows it.
Local election results in May were poor. The rise of Reform UK has been reorganizing the electoral map in a way that threatens Labour seats from a direction the party is not organized to defend. Economic conditions have been stagnant throughout, which is the condition under which incumbents lose everywhere.
Then a set of specific difficulties: a record of reversing announced positions, an awkward relationship with the American administration through a period when it mattered a great deal, and the damage done by revelations about his ambassador to Washington, Peter Mandelson, and his connections to Jeffrey Epstein. That last was not a policy failure and was arguably not his fault, and it did as much damage as anything that was.
None of those is a confidence motion. None of them would have removed a prime minister in a system with a higher threshold. All of them together produced the same result faster, and the speed is supplied by the mechanism rather than by the gravity of the charges.
The mechanism nobody voted for
It is worth being precise about what has replaced the election as the removal mechanism, because it is not written down anywhere.
A prime minister goes when enough of their own MPs and cabinet conclude they are an electoral liability and say so, privately and then publicly, until the position is untenable. There is no threshold, no ballot, no defined process, and no moment at which anybody can point to a rule being satisfied. The party’s formal machinery then arranges a succession afterwards.
So the effective electorate for a British prime minister, between general elections, is a few hundred members of one party, applying a forecast rather than a verdict.
Why Britain does this and comparable systems do not
Other parliamentary democracies remove leaders too, and most do it considerably less often, which suggests the mechanism rather than the era is doing the work.
Germany has a constructive vote of no confidence: the Bundestag cannot remove a chancellor without simultaneously electing a successor by majority. That single requirement makes removal a deliberate act with a named alternative rather than an accumulation of pressure, and it has been used successfully once since 1949.
Spain adopted the same device. Several others have fixed term provisions that raise the cost of an early change.
Britain has none of that. A party leader can be removed by that party’s own internal procedure, which is a matter for the party rather than for the constitution, and the prime ministership follows the leadership automatically. So the threshold is set by a private organization and can be changed by it.
That arrangement was tolerable when parties changed leaders rarely. It becomes the governing mechanism when they change them often, and nobody designed it to be one.
The part that does not resolve
What I cannot work out is whether this is a pathology or an adaptation.
The unattractive reading is obvious: governments become unable to do anything difficult, because anything difficult costs polling, and polling is now the removal mechanism. A four year program requires surviving four years, and nobody is surviving four years.
The other reading is that a parliamentary system is supposed to do this. It is the distinguishing feature of the type: the executive sits inside the legislature and is removable by it at any moment, without the fixed terms and elaborate procedures a presidential system requires. Removing a leader who has lost the country is the system working, and doing it without an election is the efficiency the arrangement was designed to provide.
Both are true, which is unhelpful. What is not in doubt is the frequency, and frequency is its own fact regardless of which reading you prefer: a country that changes government leader every fourteen months on average is a country where nothing that takes longer than fourteen months gets done by the person who started it.
There is a second order effect that a country notices later. Officials plan around ministers, and ministers plan around a leader, and a system that replaces the leader every year or so teaches everybody inside it to discount anything announced. A civil servant handed a ten year strategy by the fourth prime minister in five years is not being cynical in privately assigning it a low probability; they are being accurate, and the accuracy is corrosive.
The nominations open in a fortnight and there will be a new prime minister before Parliament returns. Whoever it is will begin, as the last five did, by being ahead in the polls.
