The Options on the Table, With Their Trade-Offs

The options listed here are not the only ones; electoral systems are a vast design space. Here are the ones most often raised in the Sri Lankan debate, and together they span the main trade-offs Horowitz lays out.

FPTP (First Past the Post, The Pre-1978 System)

The country is divided into single-member constituencies. Whoever gets the most votes in each constituency wins the seat. This is the system Sri Lanka used from independence in 1948 to 1977. It is also the system used in the United Kingdom, India, and many former British colonies.

The strengths are real. Every voter has one MP, who lives in or campaigns in their constituency, and whose name they know. Accountability is high. Governments tend to be durable because the system inflates the seat share of the winning party. A party with 40% of the vote often gets a comfortable majority of seats. Counting is also simple.

The weaknesses are also real, and Sri Lanka has lived through them. FPTP wipes out dispersed minorities, especially Muslims, who are spread across many districts without a concentrated geographic base. It produces alarming seat bonuses.

For example, the 1970 election gave the United Front a two-thirds parliamentary majority on 49% of the vote, and the 1977 election gave the UNP a five-sixths majority on 51%. It wastes the votes of everyone who did not vote for the winner in each constituency.

Returning to FPTP would solve the accountability problem and recreate the representation problem that drove the 1978 switch in the first place.

List PR (Proportional Representation) Without the Preferential Vote

This is what is often meant by “pure list PR” or list PR “simplified.”

Voters mark only the party, not individual candidates. Seats are allocated to parties in proportion to vote share, and party leaders decide which candidates fill those seats from a pre-published list. Used in Israel, Spain, and many European countries.

The case for it is that it removes the worst pathology of the current system – the intra-party preferential vote contest. No more poster wars between candidates of the same party. No more candidates spending tens of millions of rupees to climb their own party’s list. Proportionality is preserved. Small parties continue to get fair representation.

The case against it is that it hands even more power to party leaders, who already have too much. Voters lose what little control they currently have over which individuals represent them. Accountability is essentially gone — voters can only reward or punish parties as wholes, never specific MPs.

For a country whose central complaint about its political class is that it is a closed and unaccountable cartel, doubling down on party-list control is a strange remedy.

Mixed Member Proportional (MMP)

Voters cast two ballots. One for a local candidate in a single-member constituency, decided by FPTP. One for a party. A proportion of the seats (say 60%) are filled from the local constituency winners. The remaining seats (say 40%) are allocated from party lists to “top up” each party’s seat count until the overall result is proportional to the party-vote share. Used in Germany, New Zealand, Scotland, and Wales.

MMP is genuinely attractive because it tries to combine the best of two worlds. Every voter has a local MP they can identify, reach, and hold responsible. At the same time, the overall composition of Parliament reflects the national party vote, so small parties and dispersed minorities are not wiped out. New Zealand switched to MMP in 1996 after its own dissatisfaction with FPTP, and the result has been broadly judged a success.

The complications are also significant. The ballot is more complex. Voters must understand they are casting two votes that do different things; a common source of confusion in early MMP elections elsewhere.

The threshold design is delicate (the minimum vote share a party needs to get any list seats). Too low and Parliament fragments into many tiny parties; too high and votes are wasted. The total size of Parliament can vary from election to election as “overhang” seats are added (to maintain proportionality).

MMP also says nothing at all about ethnic conciliation. It optimises for proportionality and accountability, not for cross-community vote-seeking.

Alternative Vote (AV)

Used in single-member constituencies. Voters rank candidates in order of preference: 1, 2, 3, and so on. To win, a candidate must get more than 50% of the votes. If no one does on first preferences, the candidate with the fewest first preferences is eliminated. Their ballots are redistributed to the candidate marked next on each ballot. The count continues, eliminating the lowest each round, until someone passes 50%. Used in Australia and Papua New Guinea, and previously in Fiji.

AV is the system Horowitz spent decades arguing for in divided societies, and it is worth understanding why.

The 50% threshold means that in any ethnically mixed constituency (where no community is a majority), a candidate cannot win on their own community’s votes alone. They must seek second and third preferences from voters of other communities. A Muslim candidate in Trincomalee cannot win without some Tamil and Sinhala second preferences. This creates direct electoral pressure to moderate, to campaign in other communities, to avoid inflammatory rhetoric.

Horowitz calls this “vote pooling.” It is also strong on the Condorcet goal, the system tends to elect the candidate most voters could live with, not just the candidate with the largest hard core of support. Accountability is also high, since constituencies are single-member.

The problems are equally specific. AV is inert in ethnically homogeneous constituencies. For example, it does nothing in the Hambantota Electoral District, where almost everyone is Sinhalese (97%) and Buddhist (97%) and the second preferences just flow among candidates of the same community.

The system is not proportional, so dispersed minorities, especially Muslims, can lose representation entirely if they are not the largest group anywhere.

Implementing AV in Sri Lanka would require redrawing the country into roughly 160 single-member constituencies, which is a politically explosive boundary exercise.

There is also a practical counting problem. AV requires multiple rounds of elimination and redistribution. In a five-candidate constituency, the count could take up to four rounds; in a Colombo-style contest with ten or more candidates, nine or more. Australia’s count often takes days. Sri Lanka currently counts report within hours.

The fix could be electronic counting of paper ballots (preserving a re-countable paper trail), but this requires significant Election Commission investment and is not a small undertaking.

Fiji’s experiment with AV, adopted in 1997 partly on Horowitz’s advice, ended with the 2006 coup. In fairness, the collapse was driven by factors largely outside the electoral system; military intervention and constitutional design choices (notably retaining communal voter rolls alongside AV) that diluted the vote-pooling incentives Horowitz had argued for.

AV is not a vaccine against a coup, but the case that it failed as an electoral system is harder to make as some headlines suggests.

Interlude: Sri Lankan Presidential Elections

Sri Lankans might find the AV mechanism familiar. Presidential Elections since 1978 use a near-identical system, with one simplification: voters mark up to three preferences rather than ranking all candidates, and only the top two survive to a single redistribution round.

The 2024 election was the first to actually trigger a second count, where Anura Kumara Dissanayake won after second and third preferences were redistributed. But the numbers also show how rarely Sri Lankans use their preferences.

Of the roughly 3,321,666 voters whose first-choice candidate was eliminated (those who didn’t vote for Dissanayake or Sajith Premadasa), only 273,131 or 8% had marked a usable preference for one of the top two.

AV’s conciliation argument depends on voters expressing preferences across parties and communities; if the vast majority of voters eliminated have no transferable preference, the system collapses back toward FPTP.

Single Transferable Vote (STV)

Used in multi-member districts of typically 3 to 5 seats each. Voters rank candidates in order of preference, across parties. A candidate is elected once they reach a quota — total valid votes divided by (seats + 1), plus one.

For example, in a 4-seat district with 100,000 valid votes, the quota is (100,000 ÷ 5) + 1 = 20,001 votes, or roughly one-fifth of the vote. Surplus votes above the quota are transferred to the next preference on each ballot. The lowest candidates are eliminated and their votes transferred too. The process continues until all four seats are filled. Used in Ireland, Malta, and the Australian Senate.

STV is the system that most explicitly combines proportionality with voter choice of individuals. Voters can rank a less-favoured party’s candidate above a favoured party’s hack, which weakens party machines in a real way.

Irish politicians work their constituencies obsessively because they know voters can punish them individually, not just their party. The result tends to be proportional but personal, the opposite of list PR’s faceless party-determined slates.

The complications are practical. Counting is genuinely complex; Ireland’s elections take days. The system requires medium-sized multi-member districts, which means yet another boundary exercise.

Sri Lanka would have to be redrawn into perhaps 40–50 districts of three to five seats each. The incentives for cross-ethnic cooperation are weaker than AV because the quota in a five-seat district is only about 17%, so candidates can win on relatively narrow support without needing to court other communities.

STV solves the accountability and party-machine problems better than any other option, but does little for conciliation.

Lebanese-Style Reserved Seats With a Common Roll

Seats in each multi-member constituency are reserved for candidates of specified communities, so many Sinhalese, so many Tamil, so many Muslim per district, proportional to local demography. But all voters in the district vote on all seats, regardless of community. A Sinhala candidate for the “Sinhala seat” needs the votes of Tamil and Muslim voters in that district to win.

The conciliation incentives are unusually strong. Candidates from each community must seek votes from voters of every other community, because they cannot win on their own group’s votes alone.

Lebanon has held together (barely, and with horrific interruptions) partly because of this system. The variant in Mauritius works reasonably well in a calmer context.

The objection is one Sri Lanka has been making to itself for forty years. The system entrenches ethnic categories formally in the constitution and on the ballot. Every voter is registered by community. Every seat is labeled by community.

The post-war Sri Lankan consensus, to whatever extent there is one, has been to move away from formal ethnic categorisation, not toward it. Adopting a Lebanese model would be an explicit reversal of that direction, and would need to be defended on those terms.

The Details (where Reform Succeeds or Dies)

The headline choice of system matters less than people often think. The parameters that sit underneath it matter more, and they matter a great deal. A well-designed system with badly chosen parameters will fail. A mediocre system with well-chosen parameters can still work tolerably.

Thresholds. The minimum share of the vote a party needs to get any seats. Too low (1–2%) lets tiny parties hold coalitions hostage. Too high (10%, as in Turkey in 2002) wastes huge shares of the vote (46% of Turkish votes in that election produced no representation at all).

Constituency boundaries. Partisan bodies produce partisan boundaries. Delimitation must be done by independent, professional, neutral bodies. Sri Lanka’s Delimitation Commission has a mixed history.

Apportionment. How many seats each region gets; a separate question from boundaries, and equally consequential. Sri Lanka already has significant malapportionment that nobody is talking about.

Parameters alone are also not enough.

Voter education has to run in parallel: every new system asks voters to do something different in the polling booth, and the 2024 presidential election showed that even after fifty years of a preferential ballot, most Sri Lankans don’t use their preferences effectively.

Politician adaptation matters too. AV rewards cross-community campaigning, STV rewards constituency service, MMP rewards local presence; the full effects of any new system take two or three election cycles to surface as candidates learn the new rules.

And there is a hard political constraint behind all of this. The current Parliament voted in on the old system has to vote any new system into existence.

Sitting MPs were elected under the old rules and are unlikely to approve a new system that threatens their re-election. This is the iron law of electoral reform everywhere, and it is why most reforms get watered down in drafting or blocked at the vote.

Designing a system that is good for the country and survivable for the people who must approve it, is its own design problem.

Concluding Questions

The discipline Sri Lanka has avoided is the discipline of naming the goal first. Every serious reform proposal has begun with a system (MMP, AV, STV, some hybrid) and worked backwards to justify it. To me, Horowitz’s primer, and his presentation are a polite but firm instruction to reverse the order. Three questions, asked publicly and answered publicly, would change the conversation:

• Of Horowitz’s six goals, which two does Sri Lanka most need its next electoral system to deliver?
• What is the country willing to give up to get those two?
• Are the boundary and threshold institutions strong enough to make any system work, or does that work need to come first?

These are political questions, not technical ones. They cannot be answered by importing another country’s textbook. They cannot be settled by a panel of senior commentators nodding at each other. They have to be argued out, in Parliament, in the press, in the polling divisions that the system will actually govern.

Until Sri Lanka does that work, every reform proposal will fail the same way the last one did. Not because the system was wrong, but because nobody could say what it was supposed to do.