Hong Kong 2015-a tale of two incentives

Hong Kong has a long political history of semi-democracy. Until 1997, it was a British colony, governed by a bewildering mix of appointed and elected councils and officials. The main elected body was the Municipal Council, which had only four elected members elected with strict qualifications (the number increased over time, and from 1956 half of the Council was elected). However, this council had very limited roles in serious governance, and elections to it had poor turnout anyway. Most power resided with the Governor, who was appointed by the Queen on the advice of the British Prime Minister, and the Legislative Council, half of which was appointed by the Governor and half of which was comprised of government officials.

As Hong Kong (or, at least parts of it) had been leased from China for 99 years, it was due for return in 1997. The details of this were complicated by the Communist takeover of China after the Civil War, but in 1983 an agreement was signed between the two countries. It stated that Hong Kong would be handed back to China in 1997, but that China would give it substantial autonomy.

The first elections to the Legislative Council took place in 1985. 24 seats were actually elected. 12 of these were elected by professional associations; chambers of commerce, trade unions and the like. The other 12 were indirectly elected, in twelve single-member districts, by local councils (looking at the 1988 White Paper, it appears that an exhaustive ballot system was used, where, if one candidate did not have a majority, the lowest polling candidate would be excluded, and the voters would vote again. This process would be repeated until one candidate had a majority). These local councils were partially elected through universal suffrage and partially appointed.

Similar arrangements were in place for the next election (held in 1988), despite plans being put out for direct elections that year. The number of appointed seats was reduced by two, though, and the British made it clear that direct elections would be on the cards for future elections. A number of changes were made to the electoral system, including the introduction of preferential voting for the seats elected by local councils.

The first direct elections took place, as promised, in 1991. 18 seats were directly elected, 21 were 'functional constituencies', 18 were directly appointed by the Governor, and three seats were reserved for officials (again, British appointed). The directly elected seats were elected in two-member districts, with voters having two votes.

The result was a landslide for the liberal pro-democratic parties. Off a low turnout, broadly pro-democratic parties won 16 out of 18 seats, and 58% of the vote. Two other candidates were elected. One of these was Tai Chin-wah, a pro-Beijing candidate, who was later discovered to have falsified his legal credentials. The by-election to replace him was won by a pro-democracy candidate. The other independent was Andrew Wong, and, while the paper on the election lists him as independent, he seems to have been recognised as a pro-democracy candidate later in his career.

However, this win was balanced out by a strong result for conservative and pro-British and Beijing candidates in the functional constituencies. The pro-democrats won only four seats out of 21, and the remaining seats were apparently mostly won by pro-establishment conservative businessmen.

As a result, direct elections did not change the majority in the legislative council. However, they did draw the dividing lines for the future of Hong Kong's political competition; pro-democracy versus pro-Beijing parties.

The next elections, in 1995, saw a change in the electoral system. Instead of two-member districts, single-member plurality districts were used. Twenty members were elected in these districts. Thirty were elected in functional constituencies (28 single-member districts, and ten were elected by members of an Election Committee, elected from local councils. Election Committee seats used the single-transferable vote (electronically counted).

As in the last election, pro-democratic candidates made a near sweep of the directly elected seats. They won 17 of the 20 seats, with the Democratic Party, the largest party in the pro-democratic group, winning 12 of these seats and 42% of the vote. The other five seats were won by independents and by the Association for Democracy and People's Livelihoods (ADPL), a slightly more economically leftist group. The pro-Beijing parties won three seats; two by the more pro-Beijing Democratic Alliance for the Betterment and Progress of Hong Kong (DAB) and one by the Liberal Party (a more pro-business group).

The DAB victories in their districts were all very narrow. The DAB won New Territories North by only 48 votes, while they won Kowloon North-East by a more comfortable margin of 5.5%. The Liberal was elected in New Territories North-East with 35% to 28% for a DAB opponent, and 26% and 11% for two democrats. The competition patterns were quite interesting. Most seats saw competition between one pro-democrat and one pro-Beijing candidate. However, in some seats, there was more than one candidate from one of these groups. These seats were mostly safe ones, and there were no cases where strategic nomination would have made a difference.

In the functional constituencies, the Liberal Party dominated. Their pro-business stance allowed them to win many of the seats. Most of the others were won by Beijing-aligned independents, although pan-democrats were elected in the Construction district (a case where strategic nomination might have made the difference; the Democrat was elected with 38% against a field of three pro-Beijing candidates). Most other pro-democrat victories in the functional constituencies were more clear-cut, and it does not appear that nomination was a very big factor.

The Election Committee seats split 6-4 in favour of the pro-Beijing candidates, thus meaning that the overall Legislative Council was controlled 31-29 by pro-democrats. Beijing was obviously not happy with this, and before the handover, decided to handpick a Selection Committee to elect a new Legislative Council, that would sit until elections could be held. As you can imagine, this committee was dominated by pro-Beijing members, and most of the pro-democrats did not contest. The ADPL won two seats. The remaining seats were won by pro-Beijing parties and independents.

Following the handover, a new electoral system was put in place. Twenty seats would be directly elected, using party-list proportional representation and the Hare quota, with an average district magnitude of four. The goal of this appeared to be to prevent the pro-democrats sweeping all the directly elected seats, as they had done under the British.

Thirty seats would be elected in functional constituencies, which, as under the British, were elected by various businesses and trade unions. A further ten seats were elected by the Election Committee (despite the commitment to proportionality demonstrated in the direct constituencies, these seats were elected by the multiple non-transferable vote). Unlike under the British, this was a body elected in a similar manner to the functional constituencies; from various sectors of business and society. Members of the Legislative Council were eligible to sit in the Election Committee, too. The Committee also elects the Chief Executive of Hong Kong, who sits as the equivalent of a British colonial Governor; he signs legislation from the Council, appoints judges, and can issue decrees.

The result of the election was a win for the pro-Beijing group. In the direct constituenices, however, pro-democracy parties won 66% of the votes and 15 seats, with the pro-Beijing parties winning 5 seats and 30% of the vote. The pro-Beijing parties won a landslide victory in the functional constituency, with 25 seats, and they won all the Election Committee seats. This meant that the pro-Beijing parties controlled the Legislative Council 40-20.

A similar pattern was followed in 2000, although thanks to the Electoral Committee losing four seats which were transferred into the directly elected constituencies (thus increasing the average district magnitude to 4.8), the pro-democrats gained an extra seat overall despite losing about 6% of the vote.

The 2000 election was notable for the first use of a tactic not uncommon in party-list systems which use quotas. It involves running more than one list in a district, and then dividing your vote amongst the lists. Why is this an advantage? Let's look at the below table for an example, from the 2012 election in the seven member district of Hong Kong Island.

As you can see, in this heavily fragmented election the Civic Party was the only one to win two seats, very narrowly. However, this is not how the election worked. Instead, the DAB ran two lists, each headed by a different politician (one by Legislative Council president Jasper Tsang, and one by Christopher Chung), and split their votes roughly between the two lists. This was the result.

So, the DAB comfortably won more seats than Civic, despite winning slightly fewer votes. How did they manage that? Well, when you run one list if you have support levels like the DAB, one quota is tied up, thus meaning a smaller share of the vote is left over to compete after the quota is used to win a seat. Therefore, it makes sense for a party with substantially more than a quota's worth of to spread its votes amongst multiple lists.

A similar system was used in Colombia up until 2006. While it was nominally party-list with the Hare quota, the ability for parties to nominate multiple candidates meant that it was effectively single non-transferable vote. The lower house had an average magnitude of 4.8 (lower than Hong Kong's current average of 6.7), but the upper house used a single national district with a magnitude of 100. The highest polling list in this election got 2.2% of the vote (enough to secure two seats, though) and the lowest polling list to win a seat got 0.4% (full results here).

Hong Kong took some time to adapt to this system, with early elections being more conventional party-list. The Democrats were the first to try the system, running some multiple lists in 2000, with mixed results (in some cases, it would have made a difference, in some cases not). Indeed, they were the only party to do it up to 2012. At this point, the Democrats had lost support dramatically, as a result of their support for a controversial electoral reform package. This dramatically fragmented the pro-democrat vote, as a large number of Democrat voters switched to existing more radical democratic parties and new radical democratic parties. At this election, the DAB also tried running multiple lists.

The result of this was that, for the first time ever, no list elected more than one member. For most of the parties, this was not a deliberate strategic move. Very few of them had the support for more than one seat in each district, and so the running of one list made perfect sense. However, as the above example shows, the DAB's running of multiple lists meant that they were able to win more seats.

The election also demonstrate another factor in the multiple-list strategy; it requires tactical calculation. The Democrats ran two lists in New Territories West, for example, and very effectively divided the vote between them. However, they only won 11.8% of the vote in the district, and this meant that both lists went down to defeat, while parties with much smaller overall support that had not run multiple lists won seats.

In the same election, as demonstrated above, the Civic Party's failure to use multiple lists cost them seats. The Civic Party tried running single lists in districts where they had multiple incumbents, but this backfired on them, as described above, and they lost a few seats due to their failure to run multiple lists.

The main message here is that under an electoral system such as this, it is vitally important that parties learn to evaluate their electoral support, and the main way of doing this is through opinion polling. Hard, certainly, in certain countries where few people have phones, or where people are geographically spread. However, this is surely not the case in Hong Kong. While the accuracy of some polling is questionable (although this is not too great a sin in an area with many parties, and where small numbers of votes can have a dramatic impact on the result), it seems rather odd that the parties did not better evaluate their support.

District Councils, and why they matter

Like most countries, Hong Kong has multiple tiers of governance. There are eighteen districts of Hong Kong, and they all have local councils called District Councils. These councils provide certain public services, and spend grants from the Hong Kong-wide government. As far as I can tell, they do not levy taxes. Elections to these bodies are partisan.

The interesting thing about these council is their electoral system. The councils use single-member plurality/first-past-the-post (voters have one vote, the candidate with the most votes wins). This is a very different system from the single non-transferable vote, and has a practically opposite incentive. Under single-member plurality, having large numbers of parties is usually rare. As Duverger's Law states, plurality systems of any sort will lead to a two-party system, as minor parties with weak support will not be able to win seats. Given that that will lead to their votes being 'wasted', so to speak, their votes will head to larger parties with similar ideologies, in order to avoid opposing parties being elected.

Now, obviously there are exceptions to Duverger's Law. Nations like India, for example, have large numbers of parties despite using single-member plurality. However, in most of these countries, there is great regional variation in terms of party systems, and individual districts mostly have two-party competition. Hong Kong does not have this.

This means that there are two very different electoral systems in Hong Kong. One, in the Legislative Council, that encourages parties to split their vote between different candidates and encourages the formation of new parties. The other encourages small numbers of large and broad-based political parties. This creates an interesting conflict.

There are a number of ways parties and broader political groups could manage this conflict. They could hold internal primaries, they could agree on running a candidate from a party that had high support in the area, or they could decouple the LegCo party system from the District Council party system.

Now, I have had trouble finding the data available for District Council elections. However, a cursory look at the data available suggests that independents are a key fixture of these elections. 22% of pro-democrat votes were cast for independent candidates and 30% of pro-Beijing votes were cast for independents. This suggests a degree of decoupling between the LegCo and District Council electoral systems, as very few independents are normally elected in the geographic constituencies of the Legislative Council.

The other ideas for how is something that needs to be looked into further. However, a vaguely cursory look suggests that seats have a relatively low number of candidates. Overall statistics suggest this too; apparently, the pro-Beijing group runs 1.13 candidates to each seat, while the pro-democracy group only runs 0.77 candidates to each seat. This suggests a highly restrained nomination strategy by the pro-democrats (as well as some seats going uncontested), and a fairly restrained strategy by the pro-Beijing group.

While I don't have enough data to provide a proper analysis of these elections, it is very interesting to see these two electoral systems operating in parallel. The elections, however, seem to have sufficiently different vote shares, which makes any comparison rather hard and perhaps flawed.

Scotland 2016-One weird trick to improve a party's seat count

With elections to Scotland's devolved parliament coming up next year, it looks almost certain that the left-wing nationalist Scottish National Party, which has been governing since 2007, will win a comfortable victory. Polling done in recent months suggests that, off their near sweep of Scotland's Westminster seats earlier this year, the SNP will win a comfortable victory, under the leadership of First Minister Nicola Sturgeon.

However, unlike in Westminster, their gains will be limited somewhat, because of Scotland's usage of a mixed-member proportional representation system. Even if the SNP wins all of Scotland's constituencies, or makes a near sweep, the mechanism of the mixed-member system (described below) will keep their vote shares aligned with their seat shares. There is an option, though, for the SNP to make use of a tactic that has been used before in MMP and compensatory mixed-member systems to avoid this, called decoy lists.

How does Scotland's electoral system work?

Initial plans for a Scottish parliament were put forward by the Wilson-Callaghan governments in 1978, following an inquiry triggered by a strong SNP performance in the 1974 election (the October one). Exact details of the electoral system are unclear, as I don't seem to be able to find the specific legislation. It seems that some variation of the plurality system was proposed. This would have likely led to a Labour-dominated assembly, and this may have been one of the reasons that the proposal was rejected in a 1979 referendum (while 51.62% of voters voted Yes, this only represented 32.8% of the electorate, substantially short of the 40% of the electorate requirement).

Shortly after this referendum, a general election resulted in a Conservative government, led by Margaret Thatcher, take office. The Conservatives had little interest in devolution, and as a result it stayed off the agenda until 1997. In this year, a Labour government, led by Tony Blair, took office. Labour were committed to devolution. A referendum took place in 1997, which resulted in a 74-26 vote in favour of a Scottish assembly, and a 63-37 vote in favour of said assembly having tax-varying powers.

In order to get broad support for an assembly from the SNP and Liberal Democrats, the Blair government proposed a proportional system. Much discussion was had on the issue of this system, but the eventual result was a mixed-member proportional system that was later adopted by London and Wales.

The system works in this manner. Voters have two votes; one for single-member districts and one for regional party-lists. The single-member districts are simple: the candidate with the most votes wins. In the regions (of which there are currently eight, each electing seven members), all the party-list votes are counted up. The D'Hondt system is applied in these regions; however, the first divisor in the sequence is increased by the number of single-member districts won by the party.

For example, if Labour wins 5 district seats and 50,000 votes, and the SNP wins 1 district seat and 40,000, Labour's vote is divided by 6, meaning that Labour's vote in the party list contest is 8,333. The SNP's vote in the party list contest is 20,000. Party list seats are then distributed using D'Hondt.

This system serves to reduce the disproportionality that happens with single-member plurality. For example, at the first election to the parliament, Labour won 53 constituency seats off 38.9% of the vote, but this strong result in the districts meant that they only won three seats in the lists, thus making the overall result relatively proportional (off 33% of the list vote, Labour won 43% of the seats).

What is a decoy list? 

A decoy list is a tactic used under a mixed-member proportional (or compensatory) system in order to turn it into a mixed-member majoritarian system. The basic idea is that you run your single-member district candidates as members of a different party to your district candidates. This means that you can avoid the penalties that these sort of electoral systems levy on parties that win big in the districts.

An example of this took place in Italy's 2001 election. At this time, the lower house of Italy used a mixed-member system called the scorporo. The voting process was similar to Scotland, except the compensatory mechanism was somewhat different; the list the winning candidate district was affilated to had its vote reduced by the votes cast for the second place candidate in the district plus one. If a candidate affiliated to the list of Party X wins 10,000 votes, and the runner-up gets 5,000 votes, Party X's list votes are deducted by 5,001.

A key element of this system is that parties are permitted to form coalitions, as parties need to win 4% of the nationwide vote to get party-list seats. As a result, a district candidate can display themselves as a candidate of a national coalition on the ballot paper, while being affiliated to a list within that coalition. All party lists were on the national ballot paper.

In the 2001 election, the centre-right coalition, Casa della Liberta (House of Freedoms), led by Silvio Berlusconi, affiliated many of their district candidates to a list called Abolizone Scorporo (Abolish Scorporo). This list used a dull symbol and was not publicised. Single-member district candidates affiliated to it used the Casa della Liberta symbol. Its only purpose was to soak up the vote reductions caused by the scorporo*. The centre-left list retaliated by running their own similar list, called Paese Nuovo (New Country). The two lists actually received some votes; 0.09% for Paese Nuovo and 0.07% for Abolizone Scorporo.

Given that statistics are vague and calculations are difficult, it's a bit hard to measure the exact impact of this tactic. However, the Gallagher index of disproportionality (measured out of 100, based on list votes) spiked from 6.91 in 1996 to 10.22. It's also notable that the Communist Refoundation Party got only 11 seats out of 630, for 5% of the vote. They relied entirely on list seats for their representation in parliament.

How would it work in Scotland?

Let's look at an example. The below table shows the result of a hypothetical election in a region. The SNP has won all 9 single-member district seats, off just 55,000 (33.5% of the vote). With what is clearly an unlucky vote distribution, Labour wins no seats and 27.4% of the vote. The Conservatives (they have the cross logo) win 21.3%, the Lib Dems win 12.2%, and the Greens win 5.5%.

If the SNP ran with their single-member district candidates affiliated to the list, they would win no list seats. Three seats would go to Labour and to the Conservatives, and one to the Liberal Democrats. This would mean that the severe disproportionality in the single-member districts would be compensated for as much as possible.

However, let's say the SNP did not run a party list attached to their district candidates. Instead, they ran a nominally independent list, called, say, the National Scottish Party, that all SNP voters voted for. As you can see, this would mean that the SNP's landslide in the district seats would not reduce their share of the list seats. They would win three of these, bringing their total up to 75% of the total seats in the region.

What if the SNP had tried this strategy in the last election? If we assume that all SNP voters would vote for the decoy list, the results would be as below.

As you can see, it would be a quite dramatic gain for the SNP. They would go from 69 to 96 MSPs, three-quarters of the parliament for 44% of the vote. This is not necessarily an accurate reflection, though. Some voters might not recognise the SNP's decoy, or, more importantly, the other parties could run decoy lists of their own as a defensive measure. This would effectively turn Scotland's electoral system into a pure mixed-member majoritarian system, with complete separation of both tiers. Given the small size of the regional districts, this would lead to a relatively disproportional electoral system (average magnitude would be 1.6).

How would it be done?

So, if you are Scottish First Minister Nicola Sturgeon, and you are reading this blog, you are probably thinking 'Great! But how do I put up a decoy list? What do I call it?'. Well, Ms Sturgeon, this is where you have to use your imagination.

Now, there are a wide variety of ways in which this particular nut can be cracked. However, my best idea is one that takes advantage of Nicola Sturgeon's personal popularity and recognition. First, Ms Sturgeon should resign her membership of the SNP, and form a new party, called, say, 'Nicola Sturgeon for Scotland' or 'Nicola Sturgeon Team'. As there are no formal membership requirements, and it could not be argued that the label was intended to confuse the SNP and the new party, as Sturgeon would not be an SNP member. This party could run in the party-list section, and it could be mentioned in SNP literature as an encouraged second vote.

Now, this article is not to present a strategy only for the SNP, and this strategy could also be used by the Scottish Conservatives, for example, who have strong local appeal in the South of Scotland. I'm not sure what sort of name recognition the Scottish Conservatives have, but had they registered their candidates in the South as independents, they would have won an extra three seats (assuming no SNP decoy and all Conservative voters voting for the independents).

What would be the consequences?

The SNP trying a trick like this would not go unnoticed in Westminster, which formally controls the election legislation for Scotland. If the Conservative government had significant doubts about the SNP's actions, and if the Scottish Conservatives were substantially hurt by decoys, there is the possibility of Westminster legislating to make decoy lists useless.

One way which they could do this would be by introducing one-vote MMP. Under this system, voters would vote only in the single-member districts, and these votes would be used to calculate the list totals, replacing the list votes in the previous system. If SNP candidates recieve 5000 votes in a region, the SNP list is credited with 5000 votes. As it is impossible to separate single-member district and list votes, this would blunt the SNP's tactic.

It's unclear what impact such a system would have on support for the parties. With one exception, no party has a marked difference between their list and district votes. The key exception, of course, is the Scottish Greens. They have adopted a policy of running lists only, in the hope of picking up list votes from left-leaning voters who, while recognising that the Greens cannot win a district seat and therefore not voting Green in the districts, are willing to vote Green in the lists. Under a single-vote system, the Greens would have a much harder time attracting voters.

Another option would be dumping the single-member districts entirely, and having a party-list system only. This option would likely prove controversial, as this has only been used in the European elections, and if it were introduced, it would probably have to have some form of mechanism to open up the lists, and give voters personal control over the elected members. Perhaps, too, the single transferable vote could be used. Scotland has experience with the single transferable vote, as it has been used in local elections since 2007. While this sounds like a good idea, the issue remains that Scotland has lots of land that is virtually empty (and for good reason). This would mean that some STV constituencies would be quite large.

This post is not intended to endorse the Scottish National Party. Indeed, this is a strategy that could be used by all of Scotland's parties. The SNP, however, is in an especially strong position to exploit this particular loophole. I would suggest that it would be advisable for Westminster to close it, in some way, to ensure Scotland's parliament continues with its original intention of proportional representation.

*The Abolizone Scorporo list was used again in the 2004 European election. In an attempt to reduce the vote for the Green party, the Abolizone Scorporo symbol was placed in the middle of the symbol for the 'Green-Greens', a right-wing Green party. Under Italian electoral law, this means the list is entitled to get on the ballot automatically.