13.6 — Elections

In the 2019 Indian general election, the Bharatiya Janata Party won 37.4 percent of the votes and 55.8 percent of the seats.

In the 1997 British general election, Labour won 43.2 percent of the votes and 63.4 percent of the seats.

Neither result was fraudulent. Both are what the voting system used is designed to produce, and the gap between vote share and seat share is not a defect being tolerated — it is the system working as intended.

Which is the point of this chapter. How votes are counted changes the result more reliably than how people vote.

The systems

First past the post. The country is divided into constituencies; each elects one member; whoever gets the most votes wins, even without a majority. Used in India, Britain, the United States, Canada and much of the former British empire.

What it produces.

A tendency toward two large parties. Duverger's law: because a vote for a third party is usually wasted, voters and donors consolidate. India's exception is that its third parties are regional rather than national — a party concentrated in one state can win seats there while a party with the same national vote spread thinly wins none.

Manufactured majorities. A party with a plurality of votes gets a majority of seats, which produces stable single-party government and misrepresents the electorate.

A geographic bias. Votes are wasted in two ways: every vote for a losing candidate, and every vote for the winner beyond the number needed. A party whose support is efficiently distributed wins more seats than one with the same vote share concentrated in fewer places.

Proportional representation. Seats allocated in proportion to votes, by party list or by single transferable vote. Used in most of Europe, Israel, New Zealand and South Africa.

What it produces. Seat share close to vote share. Multi-party systems and coalition government, which requires negotiation and can be slow — Belgium once went 589 days without a government. And small parties can hold disproportionate leverage as coalition partners.

Thresholds — typically 3 to 5 percent — are used to keep very small parties out, and Germany's 5 percent threshold exists specifically because of Weimar's fragmentation (Chapter 11.6).

Mixed systems. Germany and New Zealand elect some members from constituencies and top up with list seats to make the overall result proportional. It is the arrangement most often recommended by electoral commissions that study the question from scratch.

Ranked choice. Voters rank candidates; the lowest is eliminated and their votes transferred until someone has a majority. Used in Australia and Ireland. It removes the wasted-vote problem within a constituency and does not produce proportionality overall.

And there is a mathematical result worth knowing. Arrow's impossibility theorem, from 1951, proves that no voting system converting individual rankings into a group ranking can satisfy a short list of reasonable fairness conditions simultaneously. There is no perfect system. The choice is between different sets of trade-offs, and pretending otherwise is the most common error in electoral reform argument.

Manipulation without fraud

The most effective ways of altering an election result do not involve miscounting.

Gerrymandering — drawing constituency boundaries to advantage a party. Two techniques: packing, concentrating opponents into a few seats they win overwhelmingly, and cracking, splitting them across many seats so they win none. The word comes from an 1812 Massachusetts district shaped like a salamander, drawn under Governor Elbridge Gerry.

Malapportionment — constituencies of unequal population, so a vote in one is worth more than in another.

India's position on both is unusual and worth knowing. Boundary delimitation is done by an independent Delimitation Commission whose orders cannot be challenged in court, which removes the partisan drawing that afflicts the United States. And the number of Lok Sabha seats per state has been frozen since the 1970s, based on the 1971 census, specifically so that states that reduced their population growth would not lose representation. The freeze is due to expire, and the resulting redistribution — which would shift seats from southern states with lower fertility to northern states with higher — is one of the largest political questions India faces this decade.

Voter suppression — registration requirements, identification rules, polling station distribution and hours — which shifts turnout among particular groups without altering any count.

Media control and campaign finance (Chapter 13.2).

And disinformation, which is the newest and least understood, discussed below.

What makes an election free and fair

The standard criteria.

Universal adult suffrage with no property, literacy, sex or race qualification.

Secret ballot, which matters enormously: an observable vote is a vote that can be bought or coerced.

Genuine choice — an opposition permitted to organise, campaign and access media.

An independent administration running the process.

Accurate counting with observation permitted.

And acceptance of the result, which is the norm rather than the rule and which is the one currently under most strain (Chapter 13.2).

The Indian election machinery

The largest exercise of its kind and the details are worth stating.

Around 970 million registered voters in 2024. Over a million polling stations. Around 15 million polling personnel and security staff. Conducted in phases over several weeks so that security forces can be moved.

The Election Commission's rules. No voter should have to travel more than about two kilometres to a polling station — which is why booths are set up in forests, on islands and in a Gir forest location for a single voter. The Model Code of Conduct takes effect on announcement and restricts government announcements, transfers and inducements.

Electronic voting machines since the late 1990s. They are standalone devices, not networked, with a paper audit trail — VVPAT — added from 2013 and audited by sampling. The security debate is genuine: the machines are not connected to anything, which removes the remote attack surface, and physical chain of custody and the sample size of the audit are the live questions.

Reservations. Seats are reserved for Scheduled Castes and Scheduled Tribes in proportion to population, in a joint electorate rather than a separate one — the Poona Pact arrangement (Chapter 6.21). The consequence is that a reserved-seat candidate is elected mostly by non-Dalit voters, which is Ambedkar's objection, and the alternative — separate electorates — is what Chapter 6.20 showed producing permanent communal division.

Women's representation. The 73rd and 74th Amendments reserved a third of local government seats for women, which put over a million women into elected office and has been studied extensively — with findings that women sarpanch invest more in drinking water and roads and that exposure to women leaders reduces bias against them. The Women's Reservation Act of 2023 provides for a third of Lok Sabha and state assembly seats, contingent on a census and delimitation.

And campaign finance. Electoral bonds, introduced in 2018, permitted anonymous corporate donations to parties. The Supreme Court struck the scheme down in February 2024 as unconstitutional, holding that voters have a right to information about political funding. The disclosed data showed a very heavy concentration of donations to the governing party of the day.

Why turnout and information matter

Two findings that apply everywhere.

Turnout is not neutral. Where turnout is low it is systematically higher among the older, richer and better educated, which shifts what governments do. Compulsory voting, as in Australia, removes the effect; automatic registration reduces it.

And Indian turnout has an unusual pattern. Rural turnout frequently exceeds urban, and turnout among the poor frequently exceeds turnout among the well-off — the reverse of the pattern in most democracies. The best explanation offered is that for a poor Indian voter, the state's decisions matter more directly and the vote is more evidently valuable.

Information. The classic argument is that voters are rationally ignorant — the effort of becoming informed exceeds the chance of affecting the outcome — and that they use shortcuts: party label, caste or community identity, incumbency performance, and the opinion of trusted people.

Which is why the information environment matters more than the ballot. India's Association for Democratic Reforms compelled the disclosure of candidates' criminal cases, assets and education through litigation, and the data are published — and a substantial share of elected members have criminal cases pending, which the disclosure made visible and has not changed.

And disinformation is the new problem. Cheap distribution through encrypted messaging, which cannot be moderated, at a scale no fact-checking operation can match. Chapter 8.2's finding applies exactly: a technology that lowers the cost of spreading an idea lowers it for every idea.

Where this shows up in your life

Your vote's weight depends on where you live, under any constituency-based system.

Your state's share of national representation depends on decisions about delimitation that will be made in the next few years.

And the practical point. When a result looks disproportionate, the first question is what system produced it. A party with 37 percent of the vote holding 56 percent of the seats is not evidence of fraud; it is evidence of first past the post. Understanding that distinction is the difference between criticising a system and disbelieving a result, and Chapter 13.2 showed what happens when a large number of people stop distinguishing them.

What the next page covers

Chapter 13.7 covers international law and the United Nations — what international law actually is when there is no world police, how the UN is structured and what it can and cannot do, what its agencies actually deliver, the record of peacekeeping, and the honest assessment of an organisation that is simultaneously indispensable and unable to stop the things it was founded to stop.