Analysis · Democracy & Accountability

The System in Which Parties Lose

On the June morning a British prime minister with a landslide majority prepared to surrender power, Ethiopia certified a contest that almost no one lost. A study in what turnout figures conceal — and what democracy actually demands.

By E. Frashie

On the morning these words were set down, Sir Keir Starmer — barely two years into a Labour government returned with a Commons majority of around 172 seats, among the largest in modern British history — was widely reported to be arranging the choreography of his own departure. British newspapers said he had concluded, after consultations with cabinet colleagues, party donors and union leaders, that his position was no longer tenable, and would set out a timetable for an orderly exit. Across the Atlantic, the President of the United States had pre-empted him, announcing on his own platform that the British leader “will resign” — a breach of protocol that commentators in London called a final humiliation.¹

What undid Starmer was not the loss of a general election. He still holds his majority in the Commons; no one has out-voted him at the national ballot. What undid him was a cascade of lesser verdicts. Catastrophic local-election results in which Labour shed more than eleven hundred council seats. The surge of Nigel Farage’s Reform UK, now a fixture at the head of the polls. A by-election won decisively by a rival from his own benches, Andy Burnham. And, finally, the cold arithmetic of more than ninety Labour members of parliament concluding that he could no longer hold together the coalition that had carried them to power.

It was his authority that drained away, not his honesty. He stands accused of no personal wrongdoing; the scandal that shadowed his final months — the appointment, since reversed, of an envoy with ties to a convicted sex offender — bruised his judgement rather than his conduct. But a leader’s mandate is a perishable thing, and in a working democracy a spent mandate is fatal to the office-holder even when it is survivable for the office. Hold that image — a prime minister with a parliamentary supermajority forced out by the judgement of voters and the nerve of his own party — and turn to Addis Ababa.

Democracy, the political scientist Adam Przeworski wrote, is a system in which parties lose elections. By that test, what Ethiopia declared this month was not a democratic event but its photographic negative.

The test Przeworski set

Three traditions in political theory tell us what an election is for, and the distance between them is where this month’s result lives. The minimalist school, descending from Joseph Schumpeter, reduces democracy to a method: a genuine competitive struggle between rival elites for the people’s vote. The procedural school, built on Robert Dahl’s idea of polyarchy, sets out the institutional guarantees that make such a struggle real — free and fair and recurring elections, freedom of expression, alternative sources of information, the freedom to organise, an inclusive franchise, and officials who actually wield the power they are elected to.

But it is Adam Przeworski who gives the sharpest single line. Democracy, he argued, is distinguished by one feature above all others: the institutionalised uncertainty of its outcomes. It is the system in which incumbents can lose, and sometimes do. The test is not whether a vote is held, nor even how many turn out for it, but whether the result was ever genuinely in doubt — and whether, having lost, the powerful actually surrender what they held. By that measure Westminster, for all its present squalor, was staging a democratic event this week. Addis Ababa was not.

The arithmetic of unanimity

The National Election Board of Ethiopia certified the 7th General Election with an average turnout of 96.2 per cent of registered voters, ranging from 94 per cent in the Somali and Afar regions to 99 per cent in Harari.² Against that near-universal participation sat a near-universal verdict. The Prosperity Party took 523 of the 528 declared regional-council seats in Oromia, 257 of 277 in Amhara, and clean or near-clean sweeps almost everywhere else. In the federal House of Peoples’ Representatives the pattern repeated: 167 of 173 declared seats in Oromia, the great majority in every other region, with opposition parties and independents reduced to a scattering of single mandates.

Set that pair of figures — almost everyone voting, almost everyone endorsing one party — against the most recent national elections in the established democracies, and the anomaly announces itself.

Turnout and its meaning, by regime type

Latest national elections. Turnout uses differing denominators across systems and is indicative, not exact.³

The competitive democracies cluster in the low-to-mid sixties: the United Kingdom at 59.7 per cent in 2024, the United States at roughly 63.7, India — the largest electoral exercise on earth — at 65.8. Only an extraordinary mobilisation lifts a free system into the eighties, as Germany reached 82.5 per cent this year. The sole established democracies that routinely clear ninety are those where the law compels attendance, as in Australia and Belgium; and there the high turnout still yields fragmented parliaments and alternating governments. Ethiopia compels no one to vote. The company its figure keeps is elsewhere: Rwanda, where President Kagame was returned last year with 99 per cent of the vote on a turnout of 98 per cent.

What turnout conceals

The reason the comparison cuts is that turnout means opposite things in different systems. In a competitive democracy, participation is explained by the economists’ paradox of voting — a single ballot is almost never decisive, so people vote out of duty, habit and the sense that the contest matters — and by institutional design: compulsory voting, proportional representation, automatic registration, all of which lift the figure. In a dominant-party state, very high turnout measures something different again: the mobilising reach of the administration, the party and the public payroll, the social cost of being seen not to vote, and, in some cases, the elasticity of the figure itself. The same number, read in the two settings, carries contrary meanings.

Ethiopia’s own electoral history makes the point more forcefully than any foreign comparison can. The freest contest in the country’s modern record — the election of 2005, when the opposition made real gains before the crackdown that followed — produced the lowest turnout of the era, around 82.6 per cent. The least competitive — the 2010 election, which returned the ruling front and its allies to all but two of 547 federal seats — was held on a turnout above 93 per cent.⁴ The more genuine the choice on offer, the fewer came; the more foregone the conclusion, the larger the crowd. Read in that order, the figures stop flattering and begin to confess.

The freest election in Ethiopia’s modern history produced its lowest turnout. The least free produced near-unanimity. Enthusiasm does not explain numbers that move in that direction.

The universal criteria

There is no single global registrar of democracies, but three bodies function as the working consensus, and on Ethiopia they converge. The V-Dem Institute, whose framework is built directly on Dahl’s polyarchy, classifies Ethiopia as an electoral autocracy — a state that holds multiparty elections but lacks the freedoms of expression and association that make them meaningful — with an Electoral Democracy Index score of 0.263 on a scale to one. Freedom House rates the country Not Free, and names it explicitly among a small group of states that have fallen from Partly Free to Not Free on the strength of manipulated elections since 2005. The Economist Intelligence Unit places it in the lowest of its four bands, the authoritarian regimes.

These are not the verdicts of Ethiopia’s domestic opposition, who might be discounted, but of three independent international monitors applying published criteria to the same facts. They reached their classifications before this month’s vote. The 96.2 per cent does not unsettle that picture; it illustrates it.

Elections without democracy

The scholarship has a name for the species. Andreas Schedler called it “electoral authoritarianism” — the regime that no longer cancels the vote but instead rigs the conditions around it, deploying what he called a menu of manipulation: disqualifying inconvenient candidates, controlling the media and the count, harnessing the resources of the state, and manufacturing the appearance of consent. Steven Levitsky and Lucan Way mapped the same terrain as “competitive authoritarianism,” where elections are real enough to matter yet tilted enough that the opposition cannot win. In this literature a landslide on a near-total turnout is not the refutation of the diagnosis. It is the diagnosis.

The tell is never the polling day alone. Electoral integrity, as Pippa Norris insists, must hold across the whole cycle — the drawing of boundaries, the registration of voters, the eligibility of candidates, the freedom of the campaign and the press, the casting, the counting and the resolution of disputes. A flawless count behind a curated ballot is not a free election; it is a well-run formality. The question to ask of Ethiopia’s 96.2 per cent is not whether the sums add up, but whether anyone was ever permitted to make them come out differently.

The price of accountability

Which returns us to London, and to the spectacle that prompted these reflections. By one reading it is a portrait of dysfunction: a governing party devouring its own leader; a prime minister with a historic majority driven out within two years; the sixth occupant of Downing Street to depart in a single decade; a foreign president crowing over the wreckage. Stability it is not.

But the churn is the accountability. The reason a British prime minister can be felled by council results and a restless backbench is that the mandate is real, and therefore revocable. Power in that system is held on terms, and the terms can be enforced — messily, humiliatingly, but enforced. The Ethiopian prime minister faces no such hazard, not because he is more loved than his British counterpart, but because the mechanism that could translate disaffection into consequence has been hollowed of its content. The 96.2 per cent is the measure of that hollowing. A leader who cannot lose is a leader who need not listen.

So the two events of this June week belong together. In one country, an election that produced near-perfect agreement and changed nothing. In another, results that produced no clear winner of anything and brought down a government. Unanimity is not the higher form of democracy; it is, more often, the sign of its absence. Where parties cannot lose, the people have not yet been given the one power that makes the rest worth counting. By Przeworski’s plain test, the messy capital had the democratic week. The orderly one merely held a vote.

A note on sources

1. Reporting on the British leadership crisis follows The Observer, the BBC, Reuters and the Associated Press, 20–22 June 2026. At the time of writing a formal statement was widely expected but not officially confirmed; Downing Street pointed to the prime minister’s earlier insistence that he remained focused on the job.
2. Official figures from the National Election Board of Ethiopia’s declaration of the 7th General Election. Turnout is reported against registered voters, the NEBE denominator.
3. Comparative turnout: UK House of Commons Library (2024); University of Florida Election Lab and US Census Bureau (2024); Election Commission of India (2024); the Federal Returning Officer, Germany (2025). Denominators differ across jurisdictions; the bands, not the decimals, carry the argument.
4. Classifications: V-Dem Democracy Report 2025/2026 (Electoral Democracy Index 0.263; ‘electoral autocracy’); Freedom House, Freedom in the World 2025/2026 (‘Not Free’); Economist Intelligence Unit Democracy Index (‘authoritarian regime’). Ethiopian historical turnout: IFES Election Guide. Theory: Schumpeter (1942); Dahl (1971); Przeworski (1991); Schedler (2002); Levitsky & Way (2010); Norris (2014).