Posts Tagged ‘SNTV’

How about MMP?

October 16, 2011

I haven’t been inspired to write much about this year’s election.  This is partly because I’ve been busy on other projects and partly because I’m incredibly bored by the election so far.  However, reader Mike recently asked a question that hit me in exactly the right way; it’s impossible for me to resist a question about electoral systems.  So I thought I’d give a lengthy answer in a full post.  Mike’s question is:


Hi Frozen Galic. Was recently been reading through Tsai Ing-Wen’s 10-Year Platform, which she is using as part of her campaign. What struck me with more interest was her section on democracy, which proposed to make referendums more reasonable *gasp*, ideas to make the financial activity of political parties transparent to the public, and the adoption of the Mixed-member proportional voting system (MMP) for the Legislative Yuan (the current system that Germany and New Zealand adopts for their parlimentary elections).

I certainly like these ideas, but was wondering what you thought about the MMP system, particularly in the context of Taiwan? I do understand that the political climate in Taiwan is so largely different from that of Germany or New Zealand, but MMP seems to be quite a fair and well thought out system. How else do you think the political climate of Taiwan would change, should it be that MMP is adopted?


To be clear, I haven’t read Tsai’s platform, so I’m just talking about MMP as I understand it, which may be different than Tsai’s vision.


First, the executive summary.  I think Taiwan’s current electoral system is a disaster.  The current MMM system is prone to too much disproportionality, which leads the party in power to have too much power relative to its popularity.  This is dangerous in a country like Taiwan which faces existential questions that should only be addressed when there is a real consensus, not an apparent consensus that is actually manufactured by the electoral system.  The current system also fails to balance the potential disproportionality of the nominal tier because the (proportional) list tier only elects 22% of the seats.  Technically, the system is MMM, but because the nominal tier is so dominant, it is effectively a modified single-member plurality (SMP) system.  The legislature is also too small.  With only 113 members, there are not enough legislators to encourage specialization.  Moreover, with such a small legislature and a tiny list tier, small parties have very little chance of survival.  Finally, the current system combined with the dominance of the KMT in the north and the DPP in the south leaves too many citizens without a representative from their preferred party in their area.

Would MMP be better?  Almost anything, including MMP would be preferable to the current system.  However, the list tier still needs to be enlarged, and the size of the legislature still needs to be increased.  If MMP would be better for Taiwan, perhaps it still is not quite the best choice.  Party votes dominate the results in MMP, and Taiwan has a long tradition of nominal voting.  A system with a higher degree of proportionality and a ballot that allows voters to choose specific candidates would be more in line with Taiwan’s electoral culture.  This could mean simply returning to the old SNTV system.  Alternatively, an open list PR system or STV system would be quite appropriate for Taiwan.

And now the extended answer:

First, we’re going to need some definitions.  Electoral systems can seem like a bowl of alphabet soup to people looking at them for the first time, so don’t get overwhelmed.  This is just a handy reference guide.

SMD (single member district): a district that elects only one winner.  Currently, Taiwan has 73 SMDs.

SMP (single member plurality): an SMD in which the winner is elected by the plurality rule.  That is, whoever gets the most votes, whether or not they pass 50%, is the winner.  Also called first past the post.  Taiwan currently uses SMP to elect 73 seats in its national legislature.  SMP is also used in the USA, UK, India, and many others.

MMD (multi-member district): a district elects multiple winners.  Ex: In 1995, Taipei County elected 17 seats to the legislature.  The county was not cut into separate districts; all voters had exactly the same ballot.  There are many different ways to determine who wins the seats in an MMD system.

SNTV (single non-transferable vote).  SNTV is an MMD system.  If there are k seats in the district, each voter votes for a single candidate, and the k candidates with the most votes win.  This is the system used by Taiwan before 2008 and by Japan before 1994.

List PR (list proportional representation): MMD system. Each party presents a list of candidates.  Voters vote for a party.  If a party gets X% of the votes, it should get X% of the seats.  Assume party A gets 30% of the votes in a 10 seat district.  It is awarded 3 seats.  These seats go to the first three people on the party list.

Closed list PR.  Each party presents a list of candidates.  The party determines the ranking of the candidates on the list, and the voters can only accept or reject the list.  Taiwan currently uses closed list PR to elect the list tier.  Examples: Spain, Israel.

Open list PR.  Each party presents an unordered list of candidates.  The voters can vote for a specific candidate within the list, and their votes determine the ranking of the candidates.  However, the number of seats each party wins is determined by the overall vote share of all the candidates in the party, not the popularity of any single candidate.  Examples: Brazil, Poland, Finland.

STV (single transferable vote).  MMD system.  Each voter ranks the candidates in order of his or her preference.  First preferences are counted.  If a candidate has enough votes to win, any excess votes are given to the other candidates according to the second preferences.  If no one has enough votes to win, the last place candidate is eliminated and his/her votes are given to the other candidates according to second preferences.  This continues until all the seats have been filled.  Examples: Ireland, Malta.

Mixed member (MM) system.  There are two types (tiers) of seats.  In the nominal tier, voters vote for particular candidates.  In the list tier, voters vote for party lists.  In the prototypical MM system, the voter has two ballots, one for the nominal tier and one for the list tier.  Half the seats are elected by SMP and half the seats are elected by closed list PR.

MMM (mixed-member majoritarian).  The two tiers are independent.  Suppose there are 200 seats, 100 in the nominal tier and 100 in the list tier.  Party A wins 70 seats in the nominal tier, and gets 50% of the vote in the party list tier.  Party A wins 120 (70+50) total seats.  Taiwan currently uses MMM.  Other examples: Japan, South Korea, Thailand.  Italy and Russia both switched to MMM in the mid-1990s but switched away in the past several years.

MMP (mixed-member proportional).  The two tiers are linked.  The number of seats a party wins is determined by the party list vote.  Using the same example, Party A gets 50% of the party list vote, so it should win 50% of the overall seats, or 100 of 200 total seats.  Since it won 70 seats in the nominal tier, it gets 30 more seats from the party list.  If it had only won 5 seats in the nominal tier, it would get 95 seats from the party list.  Either way, it gets 100 total seats.  This is sometimes called a compensatory system.  Examples: Germany, New Zealand, Bolivia.


Still with me?  It’s actually more complicated than this, since each country puts in a few unique wrinkles.  So the MMM system in Japan is actually quite different from the MMM system in Taiwan which is different again from the MMM system formerly used in Italy.  There are also other electoral systems that we could consider.  But for the sake of parsimony, we’ll ignore all that wonderful complexity.


One thing you should have figured out by now is that there are lots and lots of ways to elect a legislature.  None of them are intrinsically right or wrong (or more or less democratic).  They each have their advantages and disadvantages, so when you think about which system is best for a particular country, you have to think about how each system fits the particular needs of that country.  Eventually, it is simply a judgment call.


Taiwan has a long history of SNTV, dating back to the colonial era.  For most of Taiwan’s electoral history, representatives were chosen by SNTV.  Even now, it is still the most common system.  All county assemblies and city councils are still elected by SNTV.  In fact, SNTV is still used in legislative elections to determine the six aboriginal seats.  (We will now continue in the grand tradition of ignoring aboriginal politics by pretending that these six seats don’t exist.  Otherwise, it might get complicated.)

Political scientists generally have a dismal view of SNTV.  It is said to encourage money politics, factionalism, localism, and an extreme focus on constituency service.  Moreover, SNTV places a heavy burden on parties and voters since they have to nominate the right number of candidates and then spread their votes out among those candidates appropriately.  Mistakes are often made, and parties end up with too few seats.  A few years ago, a couple of political scientists surveyed specialists in electoral systems to see which systems their colleagues liked best.  SNTV came in last place.[1]

Personally, I think these criticisms are somewhat overblown.  Localism, factionalism, money politics, and so on exist in a wide array of electoral systems, but for some reason, SNTV gets condemned more enthusiastically than, say, STV.  Also, I rather like the fact that SNTV pays a premium to well-organized and highly disciplined political parties.  Organization is the very essence of politics and should be rewarded.  And even if some parties win a few less seats than they might expect in a PR system, SNTV is fairly proportional.  It’s not perfectly proportional, but it’s close enough.  Nevertheless, even I don’t think SNTV is a great system.  It may not be a disaster, but there are other systems that engender preferable incentives.  I’m fond of SNTV because it is wonderfully complex and endlessly fascinating, not because I think it encourages desirable political behavior.

There had been calls to change the electoral system during the 1990s, but these never seemed to gain much political traction.  This is not surprising.  Electoral reform is generally difficult because the people who must approve it are precisely the people who have been successful under the current system.  The actual reform passed in 2005, and I was quite stunned by it.  Few of the incumbent legislators wanted it, and I doubt the leaderships of either party really strongly supported it either.  However, everyone found themselves in a position where they did not want to oppose it.

Electoral reform brought together several currents.  First, there were the traditional DPP calls for electoral reform going back to the Lee Teng-hui era.  These calls were mostly political theater.  They were designed to highlight KMT corruption, factionalism, and ties with organized crime without really proposing any serious remedy.  To give an analogy, they were a lot like the proposals by the legislative candidates in my district to move the Taipei airport to the north coast and build a new MRT line to service it.  People in my area don’t like the noise from planes taking off at the current airport, so they are happy to hear politicians addressing those complaints and validating their unhappiness, but this is just not a realistic proposal.  Second, President Chen found himself continually stymied by the blue camp majority in the legislature.  As presidents in divided government often do, he wanted to delegitimize and/or cause pain to that opposition, and electoral reform is one way of doing this.  Third, the KMT leadership (as opposed to the blue camp legislators) probably saw that they would do quite well under a plurality system.  Individual members might suffer, but the blue camp could probably count on winning 51% in more areas than the green camp.  At some point, when the pressure from society to pass reform became heavy, the KMT leadership (Lien Chan) made the decision to sell out his members (and all the local factions) and stop resisting the pressure.  Fourth, Lin Yi-hsiung led a moral crusade for reform.  Lin appears to envision himself as something of a Taiwanese Gandhi, and his campaigns for referendum and electoral reform were presented with a sort of moral fervor.  His ideas were “correct” in an absolute sense, and anyone opposing them didn’t understand democracy.  Because of Lin’s personal history, few in the green camp dared to openly oppose him.

I’m not sure about this, but I think it was also Lin who proposed cutting the legislature in half.  The main justification for this was to save money.  The logic ran as follows: Legislators are corrupt, and so if we only have half as many, we can save on half the corruption.  At the very least, we can cut salaries and legislative aides by half, and this will save money as well.  This was a punitive logic.  The president, in particular, was receptive to the idea.  After all, he was fighting the legislature, and this discourse both delegitimized it and threatened to invoke a harsh penalty upon it.

However, I believe that cutting the legislature in half was a terrible idea.  First, in the grand scheme of things, the salaries of 112 legislators and a thousand or so legislative assistants don’t add up to that much money.  Moreover, there might be half as many legislators who can be corrupt, but each legislator now carries double his or her previous weight.  Presumably, each legislator can now extract bigger bribes than before.  I really doubt cutting the legislature in half is saving much money at all.  Second, democracy, like almost everything else, entails certain costs.  Skimping on those costs is a good way to ensure problems that eventually cost you more than you saved.  One thing a legislature is supposed to do is review laws, and a healthy committee system is crucial to this institutional capacity.  Committees are the institutional repositories of expertise.  Of course, there were many flaws in the old system, and committees often did not develop much expertise.  But things were slowly improving.  At any rate, there was a possibility of developing institutional capacity.  In the old system, there were 12 committees with up to 21 legislators serving on each committee.  Now there are only 8 committees with up to 15 legislators serving on each one.  Now each committee is spread much more thinly, and there are fewer junior members to do any of the hard work.  With such a small legislature, it simply isn’t feasible that the Legislative Yuan can ever become a powerful legislature, replete with the institutional capacity to formulate its own laws and effectively oversee the government’s exercise of power.  When the legislature has no internal expertise, it has to rely on the cabinet’s expertise or do without expert advice.  Currently, it mostly opts for the latter course.  Most decisions are made by a handful of party leaders who usually base their decisions on political expediency.  Is this any way to save money?

So in 2005, the DPP leadership wanted reform, the KMT leadership decided to stop blocking reform, and the rank and file legislators from both parties were left out in the cold.  There was a little debate about whether Taiwan should opt for MMP or MMM.  The DPP generally wanted MMP while the KMT generally wanted MMM.  As far as I can tell, these preferences were based entirely on political calculations.  If you expect to win more votes than the other party, MMM is the way to go.  Over the past two decades, the KMT (or blue camp) has usually outpolled the DPP (or green camp).  So the KMT wanted MMM.  The DPP was strategically boxed in.  It wanted electoral reform more (or at least more publicly), and so the KMT was able to insist on its preferred system.


So what kind of system do we have with MMM?  First of all, let’s recognize that Taiwan’s MMM system is heavily weighted toward the first vote (the district vote).  73 of the 113 seats are elected from SMDs.  Only 34 are elected from the second vote, the party list.  What does such a small list tier mean?  Well, the list tier is supposed to do a few things.  First, it is supposed to produce a group of legislators whose primary loyalty is to the party and to the national issues championed by nationally oriented parties.  To put it another way, you don’t want everyone thinking about constituency service or the narrow interests of a few townships in some obscure corner of the country.  For the most part, list legislators do a reasonable job of representing national and party interests.  However, one effect of the party list being so small is that competition for a spot on the list is ferocious.  Many list legislators fear they will not be able to win re-election on the party list and cultivate a district as Plan B.  In other words, the list legislators sometimes act a little like district legislators.

Second, the list is supposed to be a survival mechanism for small parties.  Everyone understands that small parties have very little chance of winning seats in the SMDs unless a big party yields that district to them.  However, small parties are supposed to be able to survive by winning seats in the list tier.  Now, there are two things that currently stop this from actually happening.  On the one hand, there is a 5% threshold for winning seats.  This threshold is designed to keep the miniscule fringe parties out of the system.  We really don’t want to have 20 fringe parties with one seat each.  We want all significant interests in society to be represented, but we also want at least a moderate level of interest aggregation.  Most political scientists (and most politicians) around the world would agree that 5% is a fair threshold, not too high and not too low.  In 2008, both the New Party and the TSU failed to exceed this threshold, so they didn’t win any seats.  I have very little sympathy for them in this regard.  On the other hand, what if they had gotten 6%?  What would they have won?  With 34 list seats, 3% gets you one seat.  So 6% would get you two seats.  A party caucus with only two seats isn’t much of a force in the legislature.  You can only sit on two committees, and if one member is sick or out of the country, half your party is missing.  Really, what can you do with a mere 2 people?  You can barely hold a press conference.  In contrast, the German legislature has 598 members, and when a party wins 5% it gets 30 seats.  They can sit on every committee, develop some expertise in different areas of government, and spread out the burdens of political communication, party organization, policy research, and so on among 30 people.  They can even give a few seats to younger members in hopes of cultivating talent for the future.  With two seats, forget it.  Small parties are simply screwed in the current system.

Third, the party list tier is supposed to compensate for the disproportionality produced in the district tier.  SMDs can give one party a very high percentage of seats on a much lower percentage of votes.  For example, in 2008 the blue camp won 82% of the district seats on about 58% of the votes.[2]  The party list tier is apportioned proportionally to each party’s list tier vote share, so the KMT and DPP split those seats 20 to 14 (as opposed to 60-13 in the district tier).  As a result, the blue camp’s overall seat share was pulled back from 82% toward their 58% vote share.  Unfortunately, because the list tier is so small, it wasn’t pulled back very far.  The list tier simply isn’t big enough to adequately compensate for the disproportionality produced in the district tier.

Why am I so concerned about disproportionality?  I’m saying that disproportionality is perhaps the most problematic feature of SMP elections, but others might reply doesn’t seem to be much of a problem in the USA or UK.  That is a fair objection.  The USA doesn’t suffer much disproportionality in congressional elections.  There are various reasons for this, including the regional strongholds of each party, the fact that the two national parties are roughly equal in size, the fact that there are only two major national parties, dumb luck, and several other things that I don’t understand.  For the most part, parties that win 52% of the national vote seem to win somewhere around 55% of the seats, give or take a few percentage points.  But other countries have not had such good luck.  In the UK, disproportionality might be THE dominant fact of politics.  Britain usually has single party government because a single party usually holds a majority of seats in the House of Commons.  However, these majorities are almost always artificially created.  There is almost never a party that wins a majority of votes.  To give a dramatic example, in 1997 Tony Blair and New Labour swept to power in an electoral tsunami, winning 63% of the seats in Commons.  How many votes did they win?  43%.  In a proportional system, New Labour wouldn’t even have been able to form a cabinet, much less had such an overwhelming majority that it didn’t have to worry about dissenters within its own ranks.  This majority was entirely manufactured, and that is not unusual.  The last party to win a majority of votes was the Conservative Party – in 1931.  Margaret Thatcher never won a majority; Tony Blair never won a majority; Clement Atlee never won a majority.  All governed with a level of power that was not commensurate with their support in society.  Somehow, that hasn’t caused a catastrophe in Britain.

I don’t think that Taiwan would be so lucky.  Taiwan faces big, existential questions.  There are important decisions to be made that will affect everyone’s core beliefs about who they are.  These decisions should not be made lightly.  Any decision about Taiwan’s future HAS to have a political consensus underneath it.  Imagine if a party with 40% of the votes but 60% of the seats decided to act as if it had a 60% coalition supporting it and made some decision about Taiwan’s future.[3]  What kind of backlash would that create?  Street demonstrations, riots, a coup, or civil war?  The point is that political parties should not be empowered to make momentous decisions unless they have extremely high levels of support in society.  SMP elections can (and sometimes do) give parties far too much power.

This brings me to a fourth point about the current system.  Most people think that the current system is a disaster for the DPP because it got nearly wiped out in 2008.  There are people who think the DPP was just stupid, and there are people who think that it had reasonable logic but things just didn’t work out as expected.  (For example, one analysis assumed that the KMT and PFP wouldn’t be able to cooperate, and the DPP would win a majority of seats with a plurality, not a majority, of the votes.  The problem is that the KMT and PFP did, in fact, figure out how to cooperate.)  There is also a train of thought that says that the DPP’s primary goal was to consolidate its place as the primary party on the independence side of the spectrum, and it accomplished that by reducing the TSU to zero seats.  I think this misses another point.  The DPP may have lost the 2008 election, but the MMM system (which is really a SMP systems with a few party list seats) might give them a better chance of winning power than a more proportional system.  In SMP systems, small swings in vote shares can cause big swings in seat shares.  When you get close to 50% of votes, you can manufacture a majority of seats.  If there are significant 3rd parties, such as the Liberal Democrats in Britain, you can win a majority with less than 50% of votes.  If you assume the DPP has less than 50% support, this may be their best shot at eventually winning the legislature.  Alternatively, when the choice is between only two parties, the governing party sometimes becomes unpopular.  So the LDP in Japan went half a century without losing an election.  It seemed impossible that any other party could ever win power.  However, in 2009, the DJP won 48% of the votes and 64% of the seats.  It is not a coincidence that Japan had changed from SNTV to MMM.  The DPP might have been calculating that MMM made both ugly losses and big wins more probable.  If you just want to win power once or twice so that you can make a few big changes, that is probably a good tradeoff.  However, these big swings in popularity often don’t reflect a solid consensus in society.  They come and go based on less consequential factors.  For example, I believe that the KMT won 58% in 2008 because of the personal failings of Chen Shui-bian rather than because of a 58% consensus on attitudes toward increased integration with the China market.  It was probably too easy for the KMT to push through ECFA.  In a more proportional system with viable smaller parties, unhappiness with Chen Shui-bian would have been channeled into other parties on the same side of the political divide.  With only two big parties, unhappy voters could only turn to the KMT regardless of whether they agreed with the KMT’s policy choices.  However, once Chen was removed from the scene, that 58% receded back closer to 50%.

Fifth, SMDs have to be drawn.  As readers of this blog know, I have thought a good deal about the redistricting process in Taiwan in 2005-7.  Overall, I think the end product was reasonably fair.  Each camp got something they liked and something they didn’t like.  However, there is little guarantee that this will be the case in the future.  Redistricting presents a tremendous temptation to politicians to bend ideas of fairness as far as they can.  Sooner or later, one side will go too far.  The danger is both that this will create an unfair outcome and that it will create a deep distrust of party politics that could ultimately undermine Taiwan’s democracy.  Why risk this?  In MMD systems, the district boundaries don’t matter very much, so redistricting is mostly uncontroversial.


Before I go on to discuss MMP, it should be obvious by now that I don’t care much for Taiwan’s current system.  There are two things that I believe would be clear improvements.  First, the overall size of the legislature should be increased.  How much is a question with no right answer, but I thought the previous size of 225 was just about right.  Second, if Taiwan retains its current MMM system, the size of the list tier should be increased from the current 22% of all seats.  One reasonable option would be to keep the current nominal tier at 79 (73 SMD and 6 aboriginal seats) and simply increase the party list tier to 79.  This would create a legislature with 158 seats, which is not too far from the 161 seats Taiwan had in 1992.


What if Taiwan changed from MMM to MMP?  Remember, the big difference is that in MMM the two tiers are independent while in MMP they are linked.  In MMP, the total number of seats a party gets depends entirely on its vote share in the party list ballot.  This change has several repercussions.

Small parties would do much better.  Under the current 113 seat system, a 6% party would win 6 or 7 seats rather than only 2.  That’s a big difference.  A party with 7 seats has a viable future, and we could probably expect to see at least two smaller, more extreme parties along with the two bigger parties (who have to remain somewhat moderate in order to contest the presidency).  Legislators from these small parties would almost all be elected from the party list.

Disproportionality would cease to be a problem.  To give an example of how things would be different, let’s consider the 2008 election.  Only the KMT and DPP exceeded the 5% threshold, so only those two would have gotten party list seats.  (In MMP systems, small parties and independents that win district seats but do not reach the threshold are allowed to keep their district seats, but let’s ignore this to keep things simpler.)  The blue camp won 66 district seats (60 SMD plus all 6 aboriginal seats) and we’ll pretend that those were all KMT seats, while the DPP won 13 district seats.  (Yes, I know there were some allied independents, but let’s keep it simple.)  The KMT won 58% of the two-party vote in the list tier, so they should get 58% of the overall seats, or 66.  The DPP won 42%, so they get 47 total seats.  Under MMP, all 34 list seats would have gone to the DPP.  Final result: KMT 66 (66+0); DPP 47 (13+34).

Note that the small list tier is almost a serious problem in this example.  With only 34 seats, it is barely big enough to compensate for the disproportionality created in the district tier.  If the KMT had won one more seat, the list tier would have been too small to restore proportionality.  Also, note that the KMT caucus is entirely made of district legislators, while the DPP caucus would be heavily weighted toward list legislators.  There is some evidence that the two types of legislators behave differently, with district legislators favoring more parochial interests.  This research is ongoing, and it is not clear when these differences are larger.  However, whether or not it is a good idea to create district-oriented majority parties and list-oriented opposition parties is something to wonder about.

In MMP, redistricting would be much less important.  If you draw funny lines, your party might win one more district seat.  However, your party’s total number of seats is fixed by the party list vote, so you simply win one fewer list seat.  There simply isn’t much of an incentive to manipulate the redistricting process.

On the other hand, in MMP the district vote doesn’t matter a whole lot, and that bothers me.  Why should a voter care who gets elected in her district.  If she spends a lot of time and energy stumping for a particular candidate and that candidate wins, what is the reward?  Her favorite party wins one fewer list seat.  I don’t like the idea of asking voters to cast irrelevant ballots.  Elections should have consequences.

Double candidacy exacerbates this problem.  Double candidacy refers to candidates running in both the district and list tier.  For example, a candidate might be the party nominee in Taipei City District 7 and also ranked #8 on the party list.  If he wins in Taipei City 7, he wins a seat (and his name is erased from the party list, with all the lower ranked candidates moving up one spot).  If he loses in Taipei City 7, he can still win a seat if his party gets 8 or more party list seats.  Double candidacy does not have to go with MMP, and sometimes it goes with MMM.  For example, Japan has double candidacy in its MMM system.  However, as far as I know, all MMP systems have double candidacy.  In Japan, double candidacy has created quite a bit of controversy and dissatisfaction.  With a history of SNTV, Japanese voters were used to the idea that they could vote someone out of office.  However, with double candidacy, it can be impossible to kill a bad politician.  Even if you don’t vote for them in the district, they can still win a seat on the party list.  The Japanese call these legislators “zombies,” and most people think there is something wrong with having zombies.  If you know that both candidates in your district will be elected, one as the district representative and the other as a zombie, why should you care about the district election at all?

(Note: In countries with more of a tradition of party-oriented elections, voters do not seem to care about zombies.  For example, Germans don’t seem to think there is anything illegitimate about it, and they certainly don’t call politicians who lose a district but win a list seat anything negative sounding, such as a “zombie.”  In fact, Helmut Kohl once won a seat this way and then he was elected Prime Minister, so it couldn’t have been that bad.)

So to answer the original question, I guess MMP would be better than the current MMM system.  But it’s not what I would choose if I had my druthers.


Taiwan has a tradition of candidate-oriented elections, and the electoral system probably should not ignore that tradition.  MMP is perhaps a bit too party-oriented for this context.  I think the best system for Taiwan would be Open List PR.  STV might also work, though I don’t know if Taiwanese voters would be up to the task of ranking the candidates.  Going back to the old SNTV system wouldn’t be a terrible idea either, though I don’t think Taiwan is going to go back to something it already rejected (and that most political scientists say is a disaster).


So that’s my answer.  I’m sure there’s more to say, but that’s it for now.

[1] I attended a conference presentation of this survey.  A member of the audience suggested that the two people who had ranked SNTV as their most preferred system should have their PhDs revoked.

[2] Sorry, I don’t have the exact figure for blue camp district votes in front of me right now.

[3] Let’s be clear.  This applies to both sides.  Any move toward either unification with China or Taiwan Independence without a consensus in society would be a disaster.

KMT links mayoral and city council campaigns

April 26, 2010

Several days ago the KMT announced that it would require all its candidates for city councils to coordinate campaign materials with the mayoral campaigns.  The most noticeable result should be that there will be two people in every photo in direct mail, billboards, and other campaign advertisements.  I’m not sure if the KMT will require them to coordinate slogans, policy positions, campaign events, TV commercials, and so on.

From my standpoint as a SNTV nerd, this might be the most interesting development in Taiwanese elections in quite a long time.  (For those of you who aren’t electoral systems specialists, SNTV (single non-transferable vote) is the electoral system used in city council elections.  It was used in legislative elections until 2004.  SNTV features multi-member districts — more than one person is elected from each district — and each voter only casts one vote for a single individual.)  When I was first exposed to Taiwanese elections, I used to always wonder why DPP candidates always prominently displayed their party logo while KMT candidates rarely did.  I assumed that being identified as a KMT nominee must be a bad thing, probably because the party’s image had been sullied by years of corruption, the authoritarian era,  and all the baggage that comes from several decades in power.    I also assumed that this was an indication that the KMT was incapable of operating as a unified team.  Since the KMT was ridden with factions both at the national and local levels, the KMT party logo was simply not a useful device to ask voters for support.  Then, when the DPP and New Party started running unified campaigns in the mid- and late-1990s, I was rather confirmed in these thoughts.  Now those were unified parties.

As you might infer from the tone of the previous paragraph, I have exchanged those ideas for a set almost diametrically opposed.  The fact that KMT candidates in SNTV elections do not usually display their party identity is a triumph for the party.  DPP logos, likewise, are a sign of weakness.

Getting people who are already predisposed to your party to actually vote for your party is not much of an accomplishment.  The real challenge is to convince people who don’t care much for your party to vote for your party’s candidates.  Going back to the beginning of the Dangwai era over 30 years ago, the KMT has consistently been able to get more votes in SNTV elections than in one-on-one executive elections.  There are lots of possible reasons for this.  The KMT has been able to recruit more quality candidates.  The KMT candidates are better at doing constituency service.  KMT candidates have talked more about local matters and other things unrelated to national politics.  The KMT buys votes.  And so on.  However, in multi-member SNTV districts, the KMT has usually gotten a higher vote share than in straight party to party contests.  Likewise, the DPP has done worse.  This gap narrowed in Legislative Elections through the 1990s, and in the 2001 and 2004 elections was basically gone.  However, in lower level elections this gap has been enormous.  In the early 1990s, it was not unusual for the DPP to have a credible shot at winning a county executive seat in a county that it could only win 3 or 4 county assembly seats.  While the DPP has gradually raised its share of these local seats, it still wins far too few, given its vote share in national level party to party contests.  In short, while the DPP has not even been able to win all the votes of people who would generally support it, the KMT has not only been able to defend its vote share, it has also won the votes of people who don’t generally support it.

So now the KMT has decided to link its mayor and city council candidates together as tightly as possible.  They are hoping that this will pull up their Tainan and Kaohsiung mayoral candidates.  One of the Tainan mayoral candidates even said that he thinks this strategy will raise his vote share by 6-8%.  I think this is mistaken.  I do not think that people will vote for a mayoral candidate (identified very closely with the national party and national party cleavages) just because they like that party’s local city council candidate.  Just because you think your local KMT council member is good at getting local pork, because he is your mother’s brother’s friend’s boss’s golfing partner, because you know him from school or whatever does not make you likely to change your opinions on the broad national questions critical to support for a political party.  If the candidate who you know and like in spite of his party is packaged with the mayoral candidate who you don’t know personally (and you still don’t like his party), you are probably more likely to reject both of them than to accept both of them.  After all, what you are doing is emphasizing the part of the city council candidate’s background (his party affiliation) that many voters do not like.  The KMT would be better off trying to hide this, which is exactly why KMT candidates have not put their party logo on their campaign materials.  In other words, instead up lifting the KMT’s unpopular mayoral candidates up, I think the KMT will pull down its popular city council candidates.

The DPP is in exactly the opposite situation.  It would reap major gains in city councils if only it could convince all the voters who lean Green to actually vote for DPP city council candidates.  The DPP city council candidates would love to turn their election into a straight party contest.

(By the way, when I talk about people supporting or not supporting a party here, I am talking in a very loose sense.  I do not mean party identifiers or hard-core supporters.  Rather, I’m talking about the preferences revealed when you force voters to choose between the parties, as in a presidential election or the party list section of the new legislative electoral system.  Voters may not prefer one side over the other by very much, but they end up making a choice.  Even if the preference for one party is not very strong, that means they have (slightly) less reason to support candidates from the other party.)

Let’s look at how this might turn out in Greater Tainan.  In 2005, the city and county assembly election results were as follows:

votes Vote % seats 2008 party list vote
Tainan County
Total 530931
Blue Camp 173323 33 19 42
Green Camp 143934 27 10 54
Independents 211428 40 21 4
Tainan City
Total 323905
Blue Camp 116041 36 18 48
Green Camp 119190 37 14 50
Independents 88674 27 9 2

Ok, so I forgot that Independents have been eating into the KMT’s dominance in several counties.  It’s less important than it looks, because the overwhelming majority of independents end up cooperating much more closely with the KMT than with the DPP.  (This is why one of Frozen Garlic’s most important rules is always look at the DPP/Green camp first.  The KMT/Blue Camp totals can be misleading if there are strong independents in the race.)

By the way, the Blue and Green Camps in Tainan are almost entirely the two big parties.  There were a total of six PFP and TSU candidates.

Look at these results from the DPP point of view.  These elections are a disaster.  In Tainan, the DPP is the majority party.  Even in 2008, when DPP popularity was at its nadir, the DPP still won a majority of the party list votes in Tainan.  However, they lost half this support in Tainan County and a quarter of their support in Tainan City.  (Historically, this is fantastic for the DPP – they used to do far worse!)  (The DPP was also lousy at turning their votes into seats.  Their seat share is less than their vote share in both places.  However, that is a different essay.)  Would the DPP candidates love to turn these elections into a straight party fight?  You betcha!

What about the KMT?  It is tempting to conclude that the KMT candidates would do better with a party-dominated election.  They could take back all those votes that the independents have stolen away.  To some extent, this may be their calculus.  However, remember that the KMT is highly unlikely to be as popular in 2010 as it was in 2008.  Right now, I think the possible KMT nominees all look like 35% candidates.  I’d be very impressed if any of them broke 40%.  So even if the KMT grabbed a huge chunk of its support back from the independents, it would likely still be splitting a smaller pie among its own candidates if it had to divide the mayoral votes among it and (most of) the independent candidates.

The KMT would also be messing up years of carefully constructed support constituencies.  Lots of KMT candidates have spent years cultivating voters who are ambivalent about the KMT.  However, perhaps the worst affected would be the solid, true-blue KMT city council members.  In the past, they could show off their position and appeal to the hard-core KMT supporters by running party-dominated campaign ads.  Now everyone will have party-dominated campaign materials.

DPP nomination strategy in Taichung County

April 10, 2010

The DPP’s Taichung County branch has announced its nomination strategy for the upcoming city council elections.  If the national party goes along with this plan, the DPP will be nominating a very high number of candidates in Taichung County.

In these lower-level elections, the DPP never does very well.  In both the party list section of the 2008 legislative elections and the 2008 presidential election, the DPP got roughly 40% all over Taichung County.    In the table, I listed the share of the vote the DPP plus TSU got in each district.  It isn’t close to 40%.  In grassroots elections, they don’t come close to that, especially in the more rural townships.  The denser social networks there tend to work in the KMT’s favor, and lots of independent candidates also get considerable numbers of votes.

I imagine the DPP will raise these vote shares slightly in 2010 since it has slowly closed the gap between what it gets in national elections and local elections over the past two decades.  However, if they raise all these vote shares by 5%, they should be thrilled.  Even if they could get a 10% increase across the board, it would be difficult to support the number of nominees they are planning to put forward.

On the other hand, I think every party will be overnominating this year.  There are simply too many incumbents and hopefuls chasing too many seats.  As a result, a party that rations its votes very well could do spectacularly well at turning votes into seats this year.  This year is a tightrope act, with certain death facing most of the participants.

district Total



plans to


2005 CC DPP+


1. 中西區 3
2. 東南區 4
3. 北屯區 6
4. 北區 3
5. 西屯區 5
6. 南屯區 4
7. 后豐區 5 3 30.3 4
8. 潭雅神區 6 3 31.9 3
9. 東勢石岡新社和平區 2 1 17.6 1
10. 大甲大安外埔區 3 1 16.7 0
11. 清水沙鹿梧棲區 5 2 17.8 0
12. 烏日大肚龍井區 5 2 15.8 1
13. 大里霧峰區 6 2-3 26.2 2
14. 太平區 4 2 29.5 2
15. 平地原住民Plains Aborigines 1
16. 山地原住民Mountain Aborigines 1
total 63 16-17

(Note: The first six districts are in the current Taichung City, so the Taichung County office has no opinion about nomination strategy there.)

New municipal council sizes

February 20, 2010

With the passing of the amended Local Government Act 地方制度法 last month, we know the sizes of the new city councils for the five direct municipalities.  Since these are the highest level SNTV election left in Taiwan, I’m enthralled by them and will be paying far more attention to them than anyone else of the course of the next year.

But first, a bit of background info.  Taiwan used to have a four, maybe five tier system of government.  Under the central government, all territory was separated into provinces.  After the ROC lost mainland China, all that was left was Taiwan Province 台灣省 and a teeny slice of Fujian Province 福建省.  Under the provinces, territory was separated into counties and county-level cities.  So Taichung City 台中市 and Taichung County 台中縣 were of equal status underneath Taiwan Province.  Fujian Province had two counties, Jinmen County 金門縣 and Lianjiang County 連江縣.  Actually, most of Lianjiang County is still on the mainland and the ROC governs only a couple of islands with about 5000 people, but it still counts as a full county in ROC organizational terms.  Under the counties, territory was separated into township-level governments.  Depending on how urbanized it was in the 1950s, it was classified as either a city 市, town 鎮, or village 鄉.  So Nantou County 南投縣 has 13 townships, including Nantou City 南投市, Caotun Town 草屯鎮, and Mingjian Village 名間鄉.  The county level cities also have township-level entities, called districts 區.   However, unlike cities, towns, and villages, which elect their own mayors, district heads are appointed civil servants.  Finally, underneath the cities, towns, villages, and districts, you have cun 村 and li 里 (which are often translated as wards, boroughs, and/or villages).  Cun and li are usually have 1000-5000 people, and don’t have any real independent powers.  Rather, they are organizational units that the higher level governments use to implement policy.  Their heads are elected, however.  Villages have cun, and the other three have li.  (English simply doesn’t have enough words to translate all these terms well, so in the previous sentence I might as well have written, “villages have villages underneath them.” Sigh.)  Oh, and the provincial government was gutted in 1998.

Are you still awake?  Well, in 1967, the government decided to elevate Taipei City 台北市 to a direct municipality, with status equal to a province.  In 1976, Kaohsiung City 高雄市 was also elevated.  Direct municipalities are different from county-level cities in several respects.  First, under the old non-democratic regime, municipality mayors were, like the Provincial Governor, appointed rather than elected.  That, however, changed in 1994, so now all are elected.  Second, direct municipalities have much better funding.  They get a different set of tax revenues and end up with a much larger budget than county governments.  As such, direct municipalities have much better public infrastructure.  Third, mayors of county governments can only appoint about four people.  Everyone else is a civil servant.  In contrast, mayors of direct municipalities can appoint lots of people, including the heads of all the city government departments and districts.  So life is much better as mayor of Taipei City than as mayor of Hsinchu City 新竹市 (county-level city).

This year, after years and years of political wrangling, three more direct municipalities will be created.  Taipei County 台北縣 (which has a larger population than Taipei City) will be elevated.  It is also changing its name to 新北市, which the Taipei Times (my favorite source for translations) is rendering as Sinbei City.  I would have preferred New Taipei (a la New Delhi).  The new Taichung City is being created combining the old Taichung City with the old Taichung County.  Similarly, the new Tainan City combines Tainan City and Tainan County.  Finally, the new Kaohsiung City will expand the old Kaohsiung City (which was already a direct municipality) to include the old Kaohsiung County.

New entity Old entities Eligible voters

(2008 LY list tier)

% of electorate
Taipei City Taipei City 2037601 11.8
Sinbei City Taipei County 2870678 16.6
Taichung City 1900142 11.0
Taichung City 767549
Taichung County 1132593
Tainan City 1430245 8.3
Tainan City 576785
Tainan County 853460
Kaohsiung City 2114221 12.2
Kaohsiung City 1161599
Kaohsiung County 952622
Rest of Taiwan 6935664 40.1

Maybe now I’m finally ready to get to the actual post.

The Local Government Act also set the guidelines for determining the sizes of the direct municipalities.  (All this is preliminary, since the final determinations will be based on populations in June 2010 (I think), but we don’t expect too much change between now and then.)  So here’s what I think we have (updated June 17):

current new
台北市 Taipei City 52 63
北投士林 10 12
南港內湖 7 9
信義松山 9 10
大同中山 7 8
萬華中正 7 8
大安文山 11 13
平地原住民 Plains Aborigines 1 1
山地原住民Mountain Aborigines 0 1
新北市 Sinbei City 65 66
板橋 9 9
中和 7 7
永和 4 4
土城樹林鶯歌三峽 9 10
三重蘆洲 10 9
新莊五股泰山林口 10 10
淡水三芝石門八里 3 3
汐止金山萬里 4 4
瑞芳平溪雙溪貢寮 1 1
新店深坑坪林石碇烏來 5 5
平地原住民Plains Aborigines 2 3
山地原住民Mountain Aborigines 1 1
台中市 Taichung City 103 63
中西區 6 3
東南區 8 4
北屯區 10 6
北區 7 3
西屯區 8 5
南屯區 6 4
后豐區 8 5
潭雅神區 9 6
東勢石岡新社和平區 4 2
大甲大安外埔區 5 3
清水沙鹿梧棲區 7 5
烏日大肚龍井區 7 5
大里霧峰區 9 6
太平區 6 4
平地原住民Plains Aborigines 2 1
山地原住民Mountain Aborigines 1 1
台南市 Tainan City 91 57
東區 10 6
南區 7 4
中西區 5 2
北區 7 4
安南區 9 5
安平區 3 2
新營鹽水柳營 6 4
白河後壁東山 4 2
麻豆下營六甲官田 5 4
佳里西港七股 5 3
學甲將軍北門 3 2
大內新化山上 3 2
善化安定 3 2
玉井楠西南化佐鎮 2 1
仁德歸仁關廟龍崎 8 5
永康新市 11 7
平地原住民Plains Aborigines 0 1
山地原住民Mountain Aborigines 0 1
高雄市 Kaohsiung City 98 66
鹽埕鼓山旗津 5 4
左營楠梓 10 8
三民 10 8
新興前金苓雅 8 6
前鎮小港 10 8
鳳山 14 8
大樹大社仁武鳥松 7 4
林園大寮 7 4
岡山橋頭燕巢梓官彌陀永安 10 5
田寮阿蓮路竹湖內茄萣 6 4
旗山美濃六龜甲仙杉林內門茂林三民桃源 6 3
平地原住民Plains Aborigines 2 1
山地原住民Mountain Aborigines 3 3

So the new law means that Taipei County essentially retains the same number of seats, and everyone else adjusts to that standard.  I haven’t heard how Sinbei City will be apportioning its seats to individual districts, but I don’t see any reason for changes.  The only difference is that they get one more seat for Plains Aborigines.  (I don’t know if this is because the population has increase or because the apportioning rules are different.)

The Taipei City Council will actually be increasing in size from 52 to 63.  This means that each district is going to get one or two extra seats.  Hooray!  It’s a great time to be a candidate there.  Incumbents have a little more breathing room, and new candidates will be able to get nominations.

Taichung, Tainan, and Kaohsiung are exactly the opposite.  They will be bloody.  The number of seats will be reduced by anywhere from a third to a half.  A newspaper article gave the plan for the new districts in Taichung City.  They were debating on whether to use the legislative districts, each of which would have around eight seats or keep the old districts and simply cut the number of seats in each.  As of now, they have decided to keep the old districts.  As you can see, a lot of incumbents are going to become unemployed at the end of the year.  For new candidates, it will be next to impossible to get a nomination.  It will be hard enough for the incumbents to get nominated.  I don’t have the district by district breakdown for Tainan and Kaohsiung yet.  Tainan will almost certainly have to redraw its districts, since the current ones are fairly small.

This certainly was a long-winded way to say Taipei—easy; Sinbei—normal, Taichung, Tainan, and Kaohsiung—meat grinders.