Let’s start a conversation about Singapore’s electoral system

TL;DR? Singapore’s parliament is really unrepresentative of the way people have voted compared to other countries’ parliaments. Instead of (or in addition to) speculating about which GRCs will shrink or disappear, let’s start thinking about switching to proportional representation. The PAP might improve their vote share; the opposition will increase their seat share; we get a more representative parliament. Everybody wins, right?

For some time I’ve been thinking about how Singapore’s politics might change in the coming decades. Here, my focus is electoral politics—the formal, institutionalised politics of political parties, voting, and elections. (I’ll treat things like inter-election parliamentary politics, and socio-political contestation from civil society/social movement organisations as a separate issue.) It’s topical, given that elections are just around the corner. But I’d like to take a step back from elections and examine our electoral system itself. Here, I’ve thrown together a bit of evidence which indicates that at least according to one simple measure, Singapore’s electoral system—the set of rules which transform voters’ preferences into representation in parliament—is fairly unrepresentative compared to other similar systems around the world. Moreover, the evidence suggests that Group Representation Constituencies (GRCs)—the mega-constituencies that send four to six party-list members to parliament—themselves are not to blame for the opposition’s sorry state. Because of that, I think we shouldn’t just look at having smaller GRCs; rather, more radical alternatives should be considered—namely, some form of proportional representation.

The next general election has been all over the news over the last two weeks—ironic since the date hasn’t even been announced yet. The Straits Times‘ Insight section last Sunday (19 July) covered speculation on possible changes in constituency boundaries, and today’s Straits Times covered the boundary changes in-depth. Here’s a timeline of official election-related announcements:

  • 13 July (CNA; Business Times): MPs filed parliamentary questions on when the Electoral Boundaries Review Committee was formed (answer: two months ago)
  • 21 July: New polling district boundaries were announced for eight Group Representation Constituencies
  • 24 July: Electoral boundaries—the demarcation of each constituency—for the next election were released

There’s also been some commentary from alternative media on the fairness of the electoral process. Bertha Henson, a longtime journalist and commentator, posted an article on the new website The Middle Ground that delved into the history of the GRC system, and argued that after a half-century of nation-building, race should not be a factor in voters’ preferences—the electorate should be more mature than that. Others such as this writer have called for an independent Electoral Boundaries Review Committee. Mothership.sg just went straight for it and called it gerrymandering.

Nevertheless, though these writers have questioned aspects of the GRC set-up and the way boundaries are drawn, nothing I’ve read recently in Singapore has questioned the first-past-the-post (FPTP) electoral system itself. FPTP is the simple rule that whichever candidate gets the largest share of votes wins. In our last general election in 2011, the PAP won Potong Pasir by getting 50.36% of the vote, just over a hundred votes more than the challenger. Now, we have gotten used to thinking of FPTP as the ‘usual’ system by which elections function. For instance, referring to FPTP before GRCs were introduced, Henson writes: “I rather like the old way of one-man-one-vote-one-representative. It seems fairer, doesn’t it?” Well, one-man-one-vote-one-representative is one way to frame it, but the 7,878 voters who didn’t vote for the winning candidate in Potong Pasir are, in practice, not represented. They were one-man-one-vote-no-representative; it’s like their votes didn’t count.

Of course, defenders of FPTP would argue this isn’t true—those votes did count; that’s just what happens when you vote on the side that loses an election! And on a micro scale, with 17,327 voters picking one person to represent them, that seems fair—since there was only one spot to be filled, let it be filled by the person the majority picked. Next question though: is it also fair to have that same system operate nationwide, with 87—now 89—spots to fill? Or should some representatives also be assigned to give voice to the people who voted for the “losing” side? (After all, nationwide they did constitute 39.9% of the vote altogether in 2011.) That in essence is the case for proportional representation (PR) electoral systems—electoral rules in which votes are translated proportionally into seats, so that all votes are reflected in the makeup of Parliament.

The argument for PR, as I’ve made it so far, therefore rests on how proportional or otherwise the FPTP system is. If 60.1% of votes gave the PAP 60.1% of seats under FPTP, there would be no difference switching to PR. However, that’s not quite the case. I’ve assembled a few macro-level statistics (shabby, but short of somehow acquiring precinct-level data, the best we’ve got). This will help make the case that Singapore’s parliament, even by FPTP standards, is particularly unrepresentative compared to other FPTP systems in the world. Surprisingly, there is also some evidence that the usual suspect, GRCs, isn’t wholly to blame.

Disproportionality is actually a problem in most FPTP electoral systems, even those without GRCs. But just how unrepresentative is our parliament? I’ve created a table which compares the discrepancy between seat share and vote share[1] for twelve FPTP countries. These countries are all British-influenced and include (I think) every FPTP country with a population over 20 million this year, except the US.[2] The average discrepancy for UK elections since 1929 was 11.4 percentage points, meaning that on average, the party in power won 11.37 percentage points more seats than their vote share. For Malaysia since the first elections in 1955, this was 21.3. In Singapore, the all-time average was 28 percentage points. Within this dataset, obviously there are within-country quirks (temporally and geographically). For instance, Canada’s 1984 election result is remarkable, because the Liberal party leader was demolished during a TV debate resulting in a nationwide swing the party didn’t recover from. Yet what is striking is the result recorded on sheet 2 (click the second tab): the all-time top eight most disproportionate elections, out of the 134[3] I recorded, were all Singaporean. The mean disproportionality is 16 percentage points, and the median is about 15.39 points; Singapore’s most recent elections have been double that. Compared to most other countries, FPTP serves us poorly when it comes to giving our voters voice in parliament.

Are GRCs to blame? The conventional wisdom is that GRCs make it harder for the opposition to win for two reasons: they force opposition parties to spread out already-thin resources (money, ground teams) and field more candidates in an electoral campaign, and they even out local deviations in the vote. The first makes sense, but opposition parties have struggled to overcome that. The second theoretically makes sense but in practice the opposition has also had a hard time winning single-member constituencies since 1988, which suggests that despite local variations, most seats would be won by the PAP in any case. There are other structural reasons why FPTP is poorly suited for Singapore. It’s a small place, with a homogeneously-distributed population. Almost every constituency has a mix of public and private housing (i.e., a wide income distribution), and an ethnic mix that mirrors the national average. We are also entirely urban, which means that there are no parties with agrarian or upper-class landed interests (again, these are associated with geographical quirks). This is reflected in the disproportionality before and after GRCs were introduced: even pre-1988, the discrepancy between government vote share and seat share was 25.6 percentage points; after that, it was 31.4 percentage points, which is a lot more but hardly a revolution. This suggests that GRCs did not cement the PAP’s grip on power; rather, there were other structural factors at play.

This also suggests that if you think elections in Singapore are structurally unfair, your target shouldn’t just be GRCs. Besides the fact that the Election Boundaries Review Committee’s work is secret (hence no information, hence no use speculating until it’s out), the vote share minus seat share statistic implies that this ink has been mis-spilt. Constituency boundaries don’t matter; GRCs don’t matter; it’s the rules themselves that ought to change—specifically, the first-past-the-post rule.

Here’s where proportional representation comes in. The number of ways PR could be implemented is too many to summarise here (here’s a map), but some version of it is used in Germany and New Zealand (mixed-member proportional), Japan, South Korea and Taiwan (mixed-member majoritarian), and many European and Latin American countries (party-list proportional representation). I’ll explain the New Zealand model, because New Zealand has a similar population size and also inherited the British FPTP model, but changed it in 1994. New Zealanders get two votes in every election: one for their local MP, and one for which party they support (which doesn’t have to be the same)—the second is the “party vote,” and helps ensure proportionality. New Zealand has around 70 electorates (63 geographical, 7 Maori—the number of geographical electorates is revised according to population). Those are awarded through FPTP. Then the other 50 or so seats in the 120-seat parliament are shared out among the parties that have achieved at least 5% of the party vote and at least one geographical constituency, in proportion to how the party vote was won but subtracting the geographical and Maori seats. (There’re formulas to determine how this allocation is done.) The parties fill those places with their list candidates, in an order that’s secret to them.

Here’s the impact of PR: it would reduce the amount of wasted votes that aren’t represented in Parliament, and would give a truer indication of the level of support of each party. Currently the PAP vote is biased down by the people who vote opposition at all costs, as well as biased up by the people who would vote opposition—but not that opposition running in their ward. My guess is that the former effect outweighs the latter (more on that later). As for the opposition parties, the level of support they muster, as measured by % of electorate captured, is dependent on the number of constituencies they run in, and is therefore not a true measure of their nationwide support. Compare that to a PR system. On the party vote, the electorate votes for who they really want to vote for—intelligently, not tactically (there is still tactical voting in the geographical contests, but that’s been the case even for FPTP). Minority representation can be achieved on party lists (and parties whose lists were grossly unrepresentative of national demographics would come under scrutiny; there could even be official requirements for this). Opposition faces increased pressure to compete based on platform rather than personality. PAP knows how much support it really has vis-a-vis all the other parties. Everyone wins!

Of course, political palatability is also important. But I’d argue that leaving FPTP would also benefit the PAP in three ways: insuring against a sudden opposition tsunami, increased vote share, and spreading the opposition even more thinly. A few scholars (Linzer 2012, citing Powell 2000) have made the point that first-past-the-post sacrifices proportionality for responsiveness. That is, proportional systems convert vote shifts to seat changes in a fairly consistent manner. But in FPTP, small vote swings can make large changes in the allocation of seats. This was illustrated in the 1984 Canadian election, where a 17.6 point change (54% swing, admittedly a big one) was manifested in a knockout 104.9% increase in the Conservatives’ seat share. In the last UK election, a 0.8 percentage point increase in the Conservative vote yielded a 3.7 percentage point increase in their seats, giving them a majority that no one predicted. Closer to home, in the last Malaysian election one website released an online tool which showed how a uniform vote swing would affect the result; at a crucial point around 60% share of the vote, a sea of seats would flip to Pakatan.[4] This is exactly the “freak election result” that Lee Kuan Yew warned about. The FPTP system might one day prove too responsive, and the party with the most to lose is clearly the PAP. It should change FPTP well before it reaches that tsunami point.

I also make the claim that proportional representation is likely to increase the PAP’s vote share. Here’s why. I think most people’s voting intentions fall into five types: (1) will never vote Opposition, (2) will never vote PAP (Cherian George’s famous “would vote for a cow if this were the only creature challenging the PAP”), (3) voted Opposition reluctantly but for the sake of having alternative voices, (4) will consider voting Opposition but only the right Opposition, and (5) voted Opposition because it was the right Opposition that contested their ward. Let’s call these blocs 1–5. 2, 3, and 5 voted opposition; 1 and 4 voted PAP; but 1 & 2 are “diehards,” 3 is those voters who value contestation in parliament, and 4 & 5 are those who evaluate the parties in some way, perhaps based on platform or personality. For instance, 4 and 5 could be those who would vote WP but think SDP is “too extreme,” or vote SDP and think WP is “too tame,” or Reform Party but not WP because they like RP’s policies.

Blocs 1, 2, and 5 will stay the same from the PAP’s point of view, although there will be some intra-opposition movement. Bloc 3 and 4 are the interesting ones. Given PR, they will have no incentive to vote Opposition on the party vote, which increases the PAP’s tally. Bloc 4 (for instance, those WP supporters who nevertheless voted PAP because SDP was standing in their ward) will defect to the particular opposition party that they like better. I suspect that those who evaluate parties based on platform (4 & 5), are few in number; if they were many, opposition parties would have more of an incentive to run on platform rather than promises like a “First World Parliament.” I could be wrong on that, but for that reason, I suspect that “reluctant opposition voters” who would switch to PAP on a PR system (Bloc 3) outnumber the “evaluative” voters. (Would welcome comments on this argument, particularly if there are opinion polls!)

Lastly, I made the claim that the PAP would benefit from PR because it would thin the opposition’s resources out even further. Here’s why: imagine the difficulties that opposition parties had mustering field teams, distributing flyers, and organising walkabouts to cover the geographical scope of a GRC. Now tell them that the whole of Singapore is just going to be one mega-constituency, returning 89 members.[5] Pengsan!

One objection is that PR is unfamiliar to Singaporeans, and will be difficult to explain. The full details of any new PR system ought to be worked out in a public debate, and it will take some time for the electorate to understand. Nevertheless, there are precedents for introducing proportionality in previously FPTP electorates, and the public debate (and subsequent public education) on that could be studied. Australia, which runs a “single transferable vote” system, distributes “how to vote” cards for every election. Many countries that run PR systems have lower levels of education and literacy than Singapore, or—especially in the case of European and Latin American PR electorates—introduced their systems at a time when much of their population was uneducated and illiterate. Moreover, in practice we already have a very crude proportionality-like measure in the form of Non-Constituency Members of Parliament (NCMPs). Up to 9 opposition members may be admitted, which means that the real size of the elected house could have been as high as 96 last election. The NCMP rule ensures that the proportion of opposition MPs (under the current arrangement of 87 MPs and at least 9 opposition members) will not fall below 9.375% (9 out of 96, if all are NCMPs). The NCMP scheme falls away once there are 9 elected opposition members (10.3% of the House). While in absolute terms this keeps at least 9 opposition members in Parliament, the proportion of opposition votes has not since 1968 fallen anywhere close to 10.3%. Tagging on this to introduce full proportionality over the course of perhaps two or three elections (perhaps by gradually increasing the number of balancing seats) will help familiarise the electorate and ease into the new rules.

At the end of the day, implementing proportional representation will be a political decision. But it won’t be done unless voices begin to be heard on electoral reform. Instead of wasting our breath wishing for smaller GRCs or the abolition of GRCs altogether, it’s time to talk about a more meaningful measure that could transform the relationship between our electorate and our Parliament.

Linzer, Drew A. 2012. “The relationship between seats and votes in multiparty systems.” Political Analysis 20:400-416.

[1] There are many much more sophisticated ways of calculating disproportionality: for instance, the Gallagher Index takes the sum of squared differences. I’ve used the simplest possible metric (1) to aid understanding and (2) because not every constituency has been contested by the opposition, which also has an effect on proportionality (we’ll never know how those votes would have gone, during the “by-election strategy” era). It would not be fair to hide behind a mathematically more sophisticated metric and claim disproportionality on the basis of incomprehensibility.

[2] Yes, there’s not much logic to the choice of these 11 comparisons, except that all of them are Commonwealth. Pakistan’s and Kenya’s systems are technically mixed, but I’ve removed the proportionally-allocated seats; likewise for Botswana and its small number of appointed seats. Many large African countries lacked data (e.g., Ethiopia, Nigeria, Kenya, Uganda). Myanmar has not held a true multiparty FPTP election. The US is the major and glaring omission; I simply didn’t have time to find a database with just the information I needed. Aside from that I think I’ve covered every country with a population above 20 million this year. The full list of countries using FPTP is here.

[3] Excluded ones in Bangladesh and Ghana where large opposition parties boycotted. (Because of this, I should also exclude SG 1968 but it doesn’t really swing the results.)

[4] I can’t find it, but you might remember playing with it. If you know where on the internet it lives, drop me a comment!

[5] This is not the way many PR systems work: New Zealand, for instance. Technically, they would also have geographical constituencies (by the New Zealand model), so perhaps the fallout from this effect, in the form of increased demands on opposition resources, could be limited. It would certainly increase the amount of information and communication that the parties would have to release, probably online, and it could contribute to intra-opposition consolidation.

One comment

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s