Wednesday, September 28, 2016

Hong Kong 2016-Electoral systems against political parties

The small city-state of Hong Kong held general elections for their Legislative Council on September 4. Hong Kong, though technically a part of China (PRC), was granted substantial autonomy upon its return to China in 1997 by the British (it had previously been a British colony). The citizens of Hong Kong enjoy civil liberties not held by other Chinese, and, to some extent, they are able to choose how they are governed.

It is this extent that has been one of the key areas of debate in Hong Kong politics ever since the handover, and explains why the state's politics have traditionally been between pro-democratic parties (which support rapid transition towards democratic governance) and pro-Beijing parties (which support either no change towards democratic governance, or more gradual change than the pro-democrats).

Executive government in the region is controlled by the 'Chief Executive', and a cabinet of Secretaries appointed by him. The exact method of selection of this Chief Executive is, at present, by a 1,200 member 'Election Committee', which is indirectly elected and comprised of representatives mostly chosen by business interests. This has led to Chief Executives coming from the pro-Beijing faction of Hong Kong politics, which is usually in an electoral minority.

The Basic Law states that "the ultimate aim is the selection of the Chief Executive by universal suffrage upon nomination by a broadly representative nominating committee in accordance with democratic procedures", though no specific timeline is laid down for this process. Last year, I wrote about the proposal for a new way to elect the Chief Executive; direct elections, but with candidates chosen by a somewhat unrepresentative nominating committee, and with the possibility of no candidates from the pro-democratic camp being permitted to contest the election. The change would have required amending Hong Kong's Basic Law, which needs support from two-thirds of the Legislative Council. As the pro-democrats have over a third of the seats, and as they refused to acquiesce to the proposals, the changes were not made.

The Legislative Council's composition is also a bone of contention. Out of the seventy members of the council, thirty are elected from what are known as 'functional constituencies'. These members are elected by people registered as having an interest in certain areas of business, like law, fisheries and banking. Five additional members are elected in a 'super' functional constituency, in which all electors ineligible to vote in one of the other functional constituencies are permitted to vote; only members of local councils may run as candidates in this functional constituency.

On average, each of the Functional Constituencies that went up for election this year had approx. 8,500 voters. This is compared to approx. 62,000 voters per seat elected in a 'geographic' constituency, in which anyone over 18 may vote and stand as a candidate (with the exception of people who refuse to acknowledge Chinese sovereignty over Hong Kong). There is huge variation between the seats, with the largest district (Education) having 482 times the number of voters as the smallest (Agriculture and Fisheries); both elect one member. 

This over-representation of business interests, which tend to support pro-Beijing parties, have ensured a majority for the pro-Beijing faction has prevailed since Hong Kong was returned to China, despite these parties never receiving a majority of the votes in the geographic constituencies. Pro-democratic parties forcefully argue against this, claiming that it is undemocratic, while pro-Beijing parties would argue that given that business makes a large financial contribution to Hong Kong, they should have a large say in governance.

The electoral system for universal suffrage elections

As stated above, however, Hong Kong does allow all voters to choose some members of the Legislative Council. Thirty-five members are chosen in the geographic constituencies; as the name implies, these members are elected from five specific geographic areas. Another five are chosen in the special functional constituency.

The first universal suffrage elections to the Legislative Council took place under British rule in 1991. This was for only eighteen seats, with twenty-one being appointed by the colonial Governor, and another twenty-one being chosen by functional constituencies. These elections took place in nine two-member districts, with voters having two votes for individual candidates. The two candidates with the highest votes would win. 

In general, this system would be expected to result in the party or faction with the most votes winning both seats, given that voters for a party would vote for both of that party's candidates. The results seemed to bear this out, with the pro-democracy parties winning sixteen seats (later increased to seventeen after a by-election) of the eighteen, despite only receiving 58.2% of the vote. This clearly had little impact on the government of the city, since the members were outnumbered by the appointed, generally pro-establishment members. 

At the next election, held in 1995, a number of changes were made to the composition of the Council. Twenty members would be directly elected, and they would be elected in single-member districts, though the electoral system would still be plurality (one vote, candidate with the most votes wins). Nine new functional constituencies were added.

These results were a landslide win for the pro-democratic parties in the directly elected seats; they won 17 with 63% of the vote. This landslide, along with relatively strong support in the functional seats, allowed the pro-democrats to win a narrow majority in the Council. This obviously displeased Beijing, which was scheduled to take over the city in 1997.

For the immediate time, Beijing destroyed the pro-democrat majority Legislative Council by appointing their own Elections Committee to choose a Provisional Legislative Council, which took over the role of the provisional Legislative Council.  However, upon the resumption of elections post-Beijing takeover, an important change was made to the electoral law; for the twenty directly elected members of the LegCo, the single-member plurality system was replaced by party-list proportional representation, in districts with an average magnitude of four members.

The results of the 1998 election, the first held after the handover, resulted in the three main pan-democratic parties securing 57% of the vote. This figure does not include independent candidates, some of whom may have been sympathetic to the pan-democrat cause. These three parties were able to secure thirteen out of twenty seats in the direct seats. However, once the results from the thirty seats elected by the functional constituencies and the ten elected by the Elections Committee came in, the narrow pro-democratic win in the directly elected seats were more than balanced out by the massive pro-Beijing majority on these other seats, allowing the pro-Beijing parties to win a majority.

For the most part, the statement made by the Hong Kong Electoral Commission that results with PR would "more accurately reflect the wishes of the electorate" was correct. At the same time, the Legislative Council was not representative as a whole of the electorate of Hong Kong. PR also had its clear advantages to the pro-Beijing political parties. But the advantages for Beijing do not stop with preventing pan-democrat landslides in the elected seats.


The specific party-list electoral system that was used for the geographical constituency elections was the largest-remainder method with the Hare quota. This particular method is notable for, compared to other electoral systems, advantaging small parties. It sounds rather counterintuitive, but for largest-remainder proportional representation systems, higher quotas advantage small parties. The reason for this is that distributing seats by remainder is very advantageous to small parties, given that fewer votes are needed for a seat won by remainder than a seat won by quota. Higher quotas also reduce the remainders for parties that win seats with a quota, but do not do that for parties that reach a quota. If you're more interested in this, I would encourage reading Michael Gallagher's paper on the matter.

Under an electoral system where small parties are advantaged, the rational thing for a large party to do is to transform itself into several small parties. The advantages given to parties that win seats by remainder means that these small parties should aim to win seats by remainder only, thus they should aim for each of these lists to win one seat. If elections are between lists, each trying to elect one member, with the lists with the highest M votes being elected (where M is the number of seats to be elected), the electoral system becomes roughly equivalent to the single non-transferable vote.

This process did not happen immediately after PR was introduced, as can be seen below. However, it is in the last two elections that it has become completely pervasive; all lists elected only one member.
For small parties, with enough support to win one seat in a district, SNTV is an easy electoral system to manage. They run a single candidate, and concentrate all their votes on that candidate. However, for large parties with enough support to win more than one seat, there is substantial difficulty involved in maximising that potential. Run too many candidates, or spread your vote too thinly between those you run, and you run the risk of all these candidates receiving too few votes; run too few, or concentrate your votes on only one, and that candidate could win far more votes than other elected candidates, thus depriving your party of seats.

The advantages that the system gives small parties has led to a more fragmented legislature as the number of lists electing only one member has increased, as can be seen below (the 'effective number of political parties' figure measures, roughly, the number of equal-sized political parties equivalent to the actual makeup of the legislature).
The effect of the electoral system, therefore, appears to have been to stunt the creation of a large party on either side of the major political divide. It would not be unreasonable to say that the possibility of a large, unified anti-Beijing party could be concerning to Beijing's interests. This is why, even as the electoral system appears to hurt the pro-Beijing parties, they may be wary about changing it to reward party consolidation.

Did the electoral system hurt the pro-Beijing parties?

The issue of whether the SNTV system hurt the pro-Beijing political parties is somewhat important, given that these parties have authority to amend the electoral law if they so desire. At the 2012 election, there were a number of cases where the pro-democracy parties made nomination errors. This was especially substantial for the Civic Party, which decided to run one list in the New Territories West and Hong Kong Island seats with two incumbents, and encourage all their voters to cast a vote for this particular list.

The strategy backfired quite substantially. In New Territories West, the Civic list received 14.5% of the vote. The quota was 11.1% of the votes, as there were nine seats. As a result, the Civic Party was allocated one seat for their full quota, and then had 3.4% of the votes left as a remainder. This was not enough to beat the other parties; the last seat went to a pro-Beijing list with 6.8% of the vote. Had Civic successfully divided their vote into two, they could have beaten the bottom list and won two seats. Nomination errors along similar lines occurred in other Civic districts.

The usage of functional seats means that the contest in the geographic seats cannot be meaningfully considered important for overall control of the Council. However, given that a two-thirds majority is needed to amend the Basic Law (which entrenches many of the civil liberties given to Hong Kongers), the pro-democrats need to do relatively well in the elected seats in order to ensure that this law cannot be amended without consultation with them.

At this election, however, a substantial fragmenting factor appeared on the pro-democratic side. Localist candidates, supportive of increased autonomy and in some cases independence for Hong Kong, ran for election. These candidates appear to have taken most of their support from existing pan-democrats, which fragmented the anti-Beijing vote. Fragmentation, however, can be an advantage under this electoral system; as stated above, vote division that is a consequence of fragmentation can be useful to parties under SNTV. On the other hand, there were only three large pro-Beijing parties running, and they would have to make difficult internal tactical decisions about how to win the most seats.

So, did the SNTV system advantage the anti-Beijing parties? In order to calculate this, I compared the SNTV results to results using the D'Hondt system and party vote figures. The D'Hondt system with party vote figures is equivalent to what the SNTV system would look like if parties nominated the exact number of candidates as seats they can win, and then divided these votes perfectly between these candidates. This represents what the result would look like if vote distribution were not a factor.

The results of the analysis were that the pro-Beijing parties would gain two extra seats out of thirty-five elected seats. In both of these seats, the seats were lost due to an overly conservative nomination strategy; the party nominated too few lists to win the number of seats that they are entitled to. For example, in the Kowloon West seat, the pro-Beijing Democratic Alliance's one list received 18.8% of the vote, above the 16.7% quota. Under largest-remainder, this list's remainder of 2.1% was not nearly enough to win a seat. However, under D'Hondt, the party's vote was divided by two, meaning that the 9.4% left after the first seat was enough to beat the localist 'Youngspiration' candidate, with 7.4% to the last seat.

If the results of this election are anything to go by, then, it may be in the political interests of Hong Kong's pro-Beijing parties to introduce a system, potentially like D'Hondt, that rewards consolidated parties. However, as stated above, this could conflict with broader political goals in Beijing to encourage the pro-democratic parties to be fragmented and to have internal arguments.

A pattern? Macau's electoral system

Macau is another one of the territories returned to China by Portugal where China's political system does not fully operate. Like Hong Kong, Macau has a Legislative Council which is partially elected directly and partially elected through 'functional constituencies', though there are also seven members appointed by the Chief Executive. The Chief Executive, who performs a similar function to the identically named leader of Hong Kong, and is chosen by a similar, not directly elected, Elections Committee.

In Macau, only 14 of 33 members of the Legislative Council are directly elected, so there is a much smaller risk to the government of the pro-democratic parties winning a large number of seats. However, the electoral system used for those elected seats is uniquely designed to produce party fragmentation.

The electoral system for the elected seats is a variant of the highest averages method, like the D'Hondt method described above. However, when seats are calculated, the list of numbers used starts with one, and then are multiplied by the next number; the list goes 1, 2, 4, 8, 16.... A formula for a party's divisor would be 2^n, where n is the number of seats the party has already won. 

The effect of this is to very heavily penalise large parties, since their vote is divided into much smaller groups as they win more seats. The below chart shows how a party with 1000 votes would have their vote go down to 2 under the Macanese system after ten seats, compared to 100 under standard D'Hondt.

The effects of an electoral system that disadvantages large parties is, obviously, to encourage small parties, or to encourage large political factions to form small parties. This is certainly the case in Macau, where the largest party at the last election had merely 18% of the vote.

While there are only two regions to sample from, it would certainly not seem inconceivable that Beijing would have a political interest in keeping political organisations in the areas that are democratic divided. Electoral systems can have a substantial impact on political party systems, and the two chosen for these regions, by pro-Beijing political parties, encourage small, divided political parties.

Monday, September 12, 2016

Optional Preferences and Pauline Hanson

At the recent 2016 Australian federal election, the right-wing anti-immigration and anti-Islam One Nation Party performed remarkably strongly, winning four Senators and coming fourth in the nationwide Senate vote, with 4.3%. The party was formed in 1997 by Pauline Hanson, a Member of the House of Representatives from Queensland elected as an independent, after she was dis-endorsed by the Liberal Party for controversial comments about Indigenous Australians and government benefits.

Ms Hanson was defeated in the seat of Blair at the 1999 federal election, but unsuccessfully contested elections for a variety of offices after that, sometimes as an Independent and sometimes as a One Nation candidate. In some cases, she came very close to election, but it is only now that she has won a seat in the Senate, with 9.2% of the vote for One Nation in Queensland (Malcolm Roberts, Hanson's running mate was also elected).

With this strong a result, attention has turned towards the next Queensland state election, which is scheduled to take place in 2018. Polls show the minority Labor government of Premier Annnastacia Palaszczuk is currently in a tight race against the opposition Liberal National Party. One Nation's recent revival in support in the state, however, could potentially have a substantial impact on this election.

The other remarkable thing about the next Queensland election will be that it will take place under a somewhat different electoral system. Earlier this year, the Palaszczuk government changed the electoral system from optional to compulsory preferential voting. This means that voters must number every box on the ballot paper to cast a formal vote. From 1992 to now, voters were permitted to number as many or as few boxes as they liked. In the event that the count reached a stage where a voter had not marked preferences between the remaining candidates, that ballot was placed in the 'exhausted' pile, and exited the count.

The question now is; what impact will the change to compulsory preferential voting have on One Nation's chances at this election? One Nation has performed strongly in the state in the past; they won 11 seats, and 22% of the votes, at the 1998 election under optional preferences.

Do preferences help or hurt Pauline Hanson?

The strongest ever performance for One Nation in any sort of election was the party's result in the Queensland election 1998. The party won 22% of the first preference vote, and eleven seats, coming close to the balance of power.

What would this election have looked like under compulsory preferential voting? In order to do this, I assumed that exhausted votes would have flowed in the same way as those cast with preferences (for example, if at a certain count the preferences of Candidate C 40% of preferences went to Candidate A, 40% to Candidate B and 20% exhaust, I distributed 50% of Candidate C's votes to A and the same to B).

I looked at twelve seats in the course of conducting this analysis; the six most marginal seats won by Labor, and the six most marginal won by the National Party, where the final preference count was between that party and One Nation. The results of this analysis are below.

Seat Party held by OPV margin CPV margin Seat Party held by OPV margin CPV margin
Crows Nest NAT +0.9 +1.5 Bundaberg ALP +2.0 +0.6
Gympie NAT +1.7 +3.7 Cairns ALP +2.3 +1.5
Callide NAT +2.3 +3.3 Ipswich ALP +3.4 +3.3
Burnett NAT +8.8 +2.3 Kallangur ALP +3.9 +2.0
Hinchinbrook NAT +8.6 +9.8 Murrumba ALP +5.0 +3.8
Cunningham NAT +8.9 +10.1 Waterford ALP +5.2 +4.0

As can be seen, CPV tightens margins in seats eventually won by Labor, and increases margins in seats eventually won by the National Party. This intuitively makes sense; preferences from centre-right parties would be expected to flow to the right-wing candidate ahead of the centre-left candidate, and thus the simulated stronger flow of these preferences would be expected to strengthen the right-wing candidate.

This is from an election where One Nation were very prominent, though. In the 2015 election, Ms Hanson ran only for the seat of Lockyer. In this seat, Ms Hanson received 26.7% of the primary vote, to 33.7% for Liberal National incumbent Ian Rickuss. The final count resulted in Mr Rickuss narrowly holding the seat, winning 50.2% to 49.8% for Ms Hanson. Oddly enough, Ms Hanson received the preferences of 41% of Labor voters, compared to 29% that went to Mr Rickuss and 29% that exhausted. Compare this to the result for the same electorate in 1998. One Nation candidate Peter Prenzler received just 16.5% of Labor preferences, against 45.4% for the National candidate.

The difference may well have had something to do with the lower statewide profile of One Nation in 2015. The party ran only eleven candidates and they had not won any seats at the state or national level for a long period of time. Opponents of the incumbent LNP government decided upon a campaign message of 'number every box, and put the LNP last'. In Lockyer, such a vote would have gone  to Ms Hanson. Something similar may have happened in the Maranoa House of Representatives seat, where One Nation received a roughly even split of preferences from the votes for lower candidates, despite 65% of these votes coming from Labor and the Greens (the AEC does not have exact figures on where these preferences came from).

It may be that the increased profile of One Nation will hurt them in some way; increased awareness of the party's support by left-leaning voters, combined with the LNP not being in government (and thus not so antagonistic to the left), might lead to them preferencing the LNP as a lesser of two evils.

Of course, the situation would be somewhat different based on the One Nation votes cast in 2016. The party received a much lower share of the vote than in their 1998 result, though admittedly there were many more parties on the Senate ballot paper than in the average Queensland state electorate. Nonetheless, even in the House districts that the party contested, the highest vote share that they managed was 17.8% in Maranoa with seven other candidates. No One Nation candidate receiving this vote was elected in 1996. It may well just be that, on these support levels, One Nation would be unable to win any seats, compulsory preferences or no compulsory preferences. This would be made even more likely given that One Nation's most prominent member, Pauline Hanson, has six years in the Senate to go.

It is rather hard to predict precisely what the impact of the change to compulsory preferential voting will be. The LNP were able to get away with telling voters to 'just vote 1' under optional preferences. Under the new system, their voters will have to choose between supporting a partisan enemy (Labor) or an radical party which they could feel close to in a partisan sense; evidence from 2001 suggests that most voters opted for the latter. On the other hand, in seats where Labor preferences are distributed, compulsory preferences could mean a stronger flow of Labor preferences to the more moderate LNP candidate; whether Labor voters will do that is debatable, though, given that they do not appear to have done that in Lockyer, and perhaps not in Maranoa.