ABSTRACT
To achieve stable democracy in
ethnically divided countries, pan-ethnic parties are needed--ones responsive
to, and obtaining votes from, each ethnic group. Therefore, seventeen new
electoral systems are put forward, designed to provide strong electoral
incentives for parties to become pan-ethnic. Each system uses, to provide this
pan-ethnic incentive, one or both of two basic principles: (1) Binary
Competition--two parties compete for the votes of the whole electorate,
especially for votes from electors of other ethnic groups who give neither of
them first preference; (2) Distributed Vote--rewarding a party with more seats
for getting an even distribution of its votes between the different ethnic
groups.
CONTENTS
1: Introduction
2: Binary Single Transferable Vote
3: Binary Proportional Representation
4: The Territorial List Scheme
5: The Distributed Vote
5.1: Tract Distributed Vote
5.2: Set Distributed Vote
5.3: Ethnic-Roll Distributed Vote
6: Binary Distributed Vote
7: Distributed STV
8: Ethnic-Roll Alternative Vote
9: Ethnic-Area Alternative Vote
10: Coalition government
11: Direct Election of Government (DEG)
11.1: Incumbency Condorcet
11.2: All-Preference Representation
11.3: Distributed-Vote DEG
11.4: Overview of DEG
12: Discussion
Endnotes
1: INTRODUCTION
In ethnically divided countries, democracy rarely functions
well. Under the usual types of electoral system, parties tend to divide on
ethnic lines, and each tends to become more or less mono-ethnic, that is, it
appeals to and draws votes from only one ethnic group. This makes it difficult
to create a stable government which is responsive to all ethnic groups. If one
party gets a majority of seats and forms a single-party government, as in
pre-1972 Northern Ireland,
this government will tend to be unresponsive to the ethnic minority. If no
party gets a majority of seats, and a majority coalition is formed in the
normal manner, this is likely not to include the parties of the ethnic
minority, so that again the government will tend to be unresponsive to the
minority. Alternatively, the "consociationalist" solution might be
attempted, which seeks to achieve power-sharing by means of a grand coalition
containing parties from each ethnic group, each group having a veto, at least
on certain important issues. However, this is likely to be unworkable and prone
to deadlock, especially when the ethnic conflict is severe, and power-sharing
and compromise are most needed.
What therefore seems to be required is an electoral system which
discourages ethnic nationalism, and gives each party the incentive to become
pan-ethnic, that is, to draw its candidates from each ethnic group, and to
respond to and seek votes from each ethnic group. With such pan-ethnic parties,
the ethnic power-sharing would take place not between the parties, as under
consociationalism, but instead within each party. Instead of the
consociationalist scheme where mono-ethnic parties negotiate after the
elections in the attempt to reach a compromise, the system would give each
party the incentive to draw up before the elections its own compromise policy
which responded to the concerns of each ethnic group.
If such a system could be achieved, the ethnic minorities would
be well protected: whatever parties won the election and formed the government,
they would always be pan-ethnic parties, and the government would be responsive
to the needs of the ethnic minorities, as well as to those of the majority. A
normal majority coalition could then be formed, and there would be no need
either to form a "grand coalition" of all major parties, or to give
each ethnic group the power of veto.
This paper therefore puts forward a range of new electoral
systems, in fact seventeen of them, designed to provide each party with strong
incentives to seek the votes of, and respond to, each ethnic group.
In order to provide this pan-ethnic incentive, all the new
systems use one or other, or in some cases both, of the following two basic
principles: (1) Binary Competition; (2) the Distributed Vote.
Systems using the Binary Competition principle set up a number
of contests between two parties (or in some cases between two candidates) in
each of which the two contestants compete for the votes of the whole electorate
of the country (or of the constituency). A party gets more seats, or a
candidate gets an increased chance of election, not only by getting more
first-preference votes, but also by getting a higher non-first ranking from
electors who give their first preferences to neither of the two contestants.
This gives the contestant the incentive to appeal more widely, to electors of
ethnic groups and other groups to which it had no reason to appeal before, in
order to obtain a higher non-first ranking from them.
This Binary Competition principle is used by Binary STV
(two-candidate competition: see section 2), by Binary PR (two-party
competition: section 3), by the Binary Distributed Vote system (two-party
competition: section 6), and by the Direct Election of Government system
(two-party competition: section 11).
Systems using the Distributed Vote principle operate by
rewarding a party with more seats or increased chance of government office, or
by rewarding a candidate with increased chance of election, for getting an even
distribution of its votes between the different ethnic groups. For example, in Northern Ireland,
a party which gets its votes from both the Catholics and the Protestants, will
be rewarded with more seats than a party which gets the same number of votes,
but only from the Protestants.
The Distributed Vote principle can be implemented in various
ways. In Set DV (section 5.2) and Ethnic-Roll DV (section 5.3), the electors
are registered on separate ethnic electoral rolls, each roll containing one
ethnic group, so that a party's votes in each ethnic group can be directly
measured.
Under Tract DV (section 5.1), the electors are registered on one
common roll, but the country is divided into small "tracts", the
boundaries of which are drawn so that each tract contains as high as possible a
percentage of electors of one or other of the ethnic groups. For example, in Northern Ireland,
each tract should be drawn up so as to consist of as high a percentage as
possible, either of Protestants, or of Catholics. A party then receives seats
according to its votes in those tracts in which it has obtained its lowest
percentages of votes. Thus if the party gets very few votes from let us say the
Catholics, its lowest-percentage tracts will be those which are mainly or
entirely Catholic, it will get seats according to its votes in these tracts,
and so it will lose seats. In future elections, it will thus have the incentive
to get more votes from the Catholics.
The three forms of Distributed Vote described above apply the
Distributed-Vote principle to the party-list form of proportional
representation. But it can also be applied to the other form of PR, that of the
Single Transferable Vote. Thus Distributed STV (section 7) is produced, which
can be expected to give very much better protection for the ethnic minorities
than would normal STV.
Binary PR (section 3) uses the Binary Competition principle
alone to provide its pan-ethnic incentive, and the government is formed by a
coalition of parties. The Binary Distributed Vote system (section 6) combines
both the Binary Competition principle and the Distributed Vote principle, in
order to achieve a stronger and more reliable pan-ethnic incentive than Binary
PR would be likely to achieve alone.
In the Direct Election of Government system (section 11),
single-party government is guaranteed, since one party is elected as the
government party by direct vote of the whole electorate. DEG uses the Binary
Competition principle, both in the election of a party to government office,
and in the determination of each party's number of seats, thus providing a
double pan-ethnic incentive for each party. Distributed-Vote DEG (section 11.3)
uses the Distributed Vote principle along with the Binary Competition
principle, in order to achieve a stronger and more reliable pan-ethnic
incentive than DEG would be likely to achieve alone.
All the above-mentioned systems are modifications of some form
of proportional representation. But non-proportional systems using
single-member constituencies, also can be modified to give a stronger
pan-ethnic incentive. Such systems are Ethnic-Roll Alternative Vote (section 8)
and Ethnic-Area Alternative Vote (section 9), which apply the Distributed Vote
principle to the Alternative Vote system, in order to provide much better
protection for the ethnic minorities than would be obtained by unmodified AV.
2: BINARY SINGLE TRANSFERABLE VOTE
Let us first apply the Binary Competition principle to the
individual candidates. This can be done by modifying the conventional system of
the Single Transferable Vote, so as to make the competition it engenders
between candidates, provide a pan-ethnic incentive. The rules of this new
system of Binary STV are as follows.
Multi-member constituencies are used, as with ordinary STV, let
us say ones of five members. Smaller constituencies could be used, especially
in sparsely populated areas, but not ones of less than two members.
Each constituency must be ethnically balanced. That is, each
constituency is drawn up so that its percentage of each ethnic group is
approximately the same as in the country as a whole. For example, in Northern Ireland,
each constituency would have about 55% of Protestants and 45% of Catholics, or
whatever are the current percentages in the whole province. (To achieve this
balance, it may sometimes be necessary to create divided constituencies, ones
consisting of two or more geographically separated areas, for example, one area
containing mainly Protestants and one containing mainly Catholics.)
The main difference from ordinary STV is that candidates are
allowed to stand in a constituency only if they register in pairs. Candidates
in the same pair are placed next to each other on the ballot paper. The pairs
are placed in order of the first-preference votes which they obtained in the
previous election, those with more votes being placed higher. However, within
any one pair, the weaker candidate, that is, the one who in the previous
election was preferred to the other by fewer voters, is placed higher. (This is
likely to assist the weaker candidate, thus strengthening the competition
between the two candidates.) The electors vote by putting the candidates in order
of preference, as under normal STV, writing "1" against their first
preference, "2" against their second preference, and so on. An
optional provision of the system is that electors should be required to put all
the candidates in order of preference, as is done in Australia, where a vote with an
incomplete ordering is rejected outright as invalid. (A less severe penalty
might be more acceptable, such as giving a one-half value to any vote in which
no preference has been expressed for more than one of the candidates. This
provision is not essential, but if adopted, it would strengthen the pan-ethnic
incentive given by the system.)
In any five-seat constituency, the counting is carried out as
follows. If there are more than five pairs of candidates, then both candidates
are excluded in that pair which contains the candidate who has fewest
first-preference votes. If there are still more than five pairs remaining, each
vote of the excluded pair is transferred to that candidate it indicates as the
next preference, and that pair is excluded which contains the candidate who
then has fewest votes. And so on, if necessary, until there are five pairs
remaining. Then in each of the five pairs, that candidate is elected who is
preferred to the other by a majority of the voters in the constituency.
How then will this system operate in practice? In any
five-member constituency, a party will have the incentive to register at least
one pair of candidates for each one-fifth of first-preference votes it expects
to get, and it will probably run an extra pair so as to be able to gain another
seat in the event that enough of the other parties' candidates get low first
preferences and get excluded. Thus it is likely that in total more than five
pairs of candidates will run. The party will try to ensure that each of its
pairs consists of two strong candidates who are about equally matched, so that
they get about the same number of first preferences, and thus minimise the risk
that they will be excluded. Thus each candidate in a pair is likely to be able
to compete strongly with the other, not only for first preferences, but for
lower-preference votes as well.
Let us consider a Northern Ireland example, in which
there is a pair of candidates of the Democratic Unionist Party, which might be
regarded as an extreme Protestant party. Which one of them gets elected, will
depend not only on the first-preference votes of the extreme Protestant
electors, but also, and mainly, on the lower-preference votes cast between them
by moderate Protestants and by Catholics. Thus each of these originally extreme
Protestant candidates will have a strong incentive to appeal more widely, and
to adopt a policy more open to compromise, one which is more acceptable to
moderate Protestants, and to Catholics. Similarly, the more extreme Catholic
candidates will have an incentive to be more responsive to moderate Catholics,
and to Protestants.
Binary STV has a similar incentive effect, in policy dimensions
other than the ethnic one. For example, two right-wing candidates will have to
compete with each other for the lower-preference votes of centre and left-wing
electors, and thus each of them will have the incentive to be more responsive
to their concerns.
Thus in general, in each party, candidates will have the
incentive to appeal more widely, outside the party's previous supporters, and
it will tend to be the most widely responsive candidates, the ones most
actively seeking ethnic compromise, who get elected as the party's
representatives. Parties can thus be expected to converge towards each other in
policy, moving towards a policy more equitably responsive to the different
sections of the electorate, and to the different ethnic groups. In other words,
this new system of Binary STV can be expected to promote ethnic compromise,
providing a strong pan-ethnic incentive, both for the individual candidates,
and for each party as a whole.
3: BINARY PROPORTIONAL REPRESENTATION
The Binary Competition principle can also be applied to parties,
by a modification of Party List Proportional Representation, so as to give to
each party as a whole, the incentive to be more pan-ethnic. An outline of this
new system, Binary PR, is as follows.
The basic structure of Binary PR, is that parties join together
into two-party "associations". (A party does not have to join an
association, but if it does not, it effectively loses three-quarters of its
seats, that is, it gets a percentage of seats equal to one-quarter of its
percentage of first-preference votes.) Any association thus formed gets seats
according to the total of first-preference votes of the two parties it
contains. Each party in the association then gets a share of the association's
seats according to the number of voters who prefer it to the other party in the
association (except that the party cannot get less than a certain guaranteed
minimum, a percentage of seats equal to one-quarter of the party's percentage
of first-preference votes).
For example, let us suppose that in Northern Ireland, the two main
Protestant-oriented parties, the DUP and the UUP, form an association. How big
a share of the association's seats the DUP will get, will depend on how many of
the Catholics, and of those Protestants who give their first preferences to
neither of these two parties, prefer the DUP to the UUP. Thus the DUP (and also
the UUP) will have an incentive which it did not have before, to appeal to
Catholics and to moderate Protestants.
Thus the operation of Binary PR has these two main features: (1)
it gives each party, whether in an association or not, a guaranteed minimum
percentage of seats equal to one-quarter of its percentage of first-preference
votes; (2) it gives any party which joins an association, the chance of greatly
increasing its seats above its minimum (normally to about four times its
minimum, but just possibly to about seven times its minimum), by competing with
the other party in the association, for the non-first-preference votes of the
rest of the voters, the ones who do not vote first preference for either party
in the association. Thus each party is given the incentive to appeal more
widely, and in particular to appeal to those ethnic groups to which it did not
appeal before, since under the new system it is now able to increase its seats
by doing so.
A more detailed account of Binary PR is as follows. The parties
join up into two-party associations as described above. The ballot paper
carries the names of the parties, and any two parties which are in the same
association are indicated, and placed next to each other on the ballot paper.
The associations are placed in order of the first-preference votes which they
obtained in the previous election, those with more votes being placed higher on
the ballot paper. However, within any one association, the party which obtained
fewer seats in the previous election is placed higher, in order to give it some
advantage, and so strengthen the competition between the two parties. The
electors vote by putting the parties in order of preference, as under STV,
writing "1" against their first preference, "2" against
their second preference, and so on. As in the case of Binary STV, an optional provision
of the system is for electors to be given an incentive to put a greater number
of the parties in order of preference. Thus a one-half value could be given to
any vote in which, for more than one of the parties, no preference has been
expressed. This provision is not essential, but if adopted, it would strengthen
the pan-ethnic incentive provided by the system.
The seats are allocated between the parties by a two-stage
process: (1) each association gets seats according to the first-preference
votes cast for the two parties in it; (2) in each association, each of the two
parties gets a share of the association's seats according to the number of
voters who prefer it to the other party in the association.
The first stage, the allocation of seats between the
associations, is as follows. If in any association the party with fewer
first-preference votes has at least two-thirds as many votes as the party with
more votes, then the association gets seats in proportion to the total
first-preference votes of the two parties. If the smaller party has less than
two-thirds as many votes as the larger party, then the association is penalised
by getting seats in proportion to 2.5 times the votes of the smaller party.
However, there are three exceptions to this. First, if, for any
association, there are any areas of the country in which one or both of the
parties have not run a candidate, then the seats of the association are reduced
proportionately (for example, if these areas without candidates contain
one-third of the electorate, the association loses one-third of its seats).
Second, there is a maximum limit on an association's number of seats: it must
not get more than one-half of the total seats. Third, there is a minimum limit
to an association's seats: it must not get less than the sum of the two
parties' guaranteed minimum percentages of seats, that is, one-quarter of the
sum of the two parties' percentages of first-preference votes. (See note 1.)
The second stage of the process is the allocation of the seats
of each association between its two parties, to be explained as follows. In an
association, let A be the party which is preferred to the other by more voters,
and B the party which is preferred by fewer voters. B gets a percentage of the
seats of the association equal to the following: twice the percentage
preferring B to A (calculated as a percentage of those voters who are not
indifferent between A and B) minus 50 per cent. However, there is one exception
to this: if this formula would give B less than its guaranteed minimum (that
is, less than a percentage of total seats equal to one-quarter of B's
percentage of first-preference votes), then B receives its guaranteed minimum,
and A receives the remaining seats of the association.
For example, suppose the association consisting of parties A and
B has 40 per cent of seats, B has 18 per cent of first preferences, and B is
preferred to A by 45 per cent of voters. Thus B receives a percentage of seats
equal to whichever is greater, either 0.4 x ((2 x 0.45) - 0.5) = 16 per cent,
or 0.18 x 0.25 = 4.5 per cent. Thus B gets 16 per cent of the seats and A gets
40 - 16 = 24 per cent.
Let us now consider a possible scenario of how this Binary PR
system might operate if used in Northern
Ireland. A typical party (let us call it P)
can be expected to react to the system as follows. To avoid losing
three-quarters of its seats, P must join an association. To maximise its seats,
P therefore seeks as an "associate", a party which (1) is unlikely to
get less than two-thirds as many first-preference votes as P does itself, and
(2) is unlikely to be preferred to P by significantly more than half the
voters. If P cannot find an acceptable associate, then P is likely to split
itself into two parties of similar size, which can then register as one
association.
If parties follow this pattern of behaviour, the result might be
as follows. The extreme Catholic Sinn Fein expects that, if it joins an
association with any other party, there will be very few electors other than
its own first-preference supporters who prefer it to its associate. Therefore,
in order not to lose seats, it splits into two parties, which form one
association. The moderate Catholic SDLP would probably lose seats if it joined
in an association with the much smaller non-sectarian Alliance Party, and so
this party also splits into two parties, which form one association. A new
Centre Party is formed by moderate politicians from both communities, and forms
an association with the Alliance Party. A somewhat reduced UUP (having lost
support to the Centre Party) forms an association with the DUP.
In each association thus formed, the two parties in it will be
competing strongly for the lower-preference votes of those electors who give
neither of them first preference, since a party which does not so compete, is
likely to lose about three-quarters of its seats to the other party. But this
assumes that an elector will be willing to vote, albeit with a non-first
preference, for a party of the other community--for example, that many
Catholics will be willing to express a preference for the UUP in what they will
regard as the pro-Protestant UUP-DUP association. Is this assumption justified?
It should be noted that these preferential votes will not increase the total
seats of the association, which will depend only on its first-preference votes.
All that these Catholics will be doing is to transfer seats from the Protestant
party they prefer less to the Protestant party they prefer more. Note also
that, unlike in the present situation, each of these parties will have the
incentive to actively campaign for Catholic lower-preference votes, and to
adjust its policy to attract them. Thus it seems justified to assume that many
voters will be prepared to express a preference for a party which originally
has not been of their community, if they prefer it to the other party in the
association.
Thus formerly mono-ethnic parties will need to appeal to, and
adopt a policy responsive to, the other ethnic group--the UUP and DUP to the
Catholic electors, the ex-Sinn Fein and ex-SDLP parties to the Protestants.
Even the already pan-ethnic Alliance
and Centre parties will need to broaden their appeal, to be more responsive to
the concerns not only of the moderate voters, but also of the voters more to
either extreme, both those who are Catholics and those who are Protestants, in
order to compete for their votes.
A majority coalition of the most moderate parties will then be
formed. This coalition is likely to include at least one formerly
pro-Protestant party, and at least one formerly pro-Catholic party, since in
order to attract Catholic votes, a formerly pro-Protestant party can be
expected to publicly support the inclusion of a pro-Catholic party rather than
another pro-Protestant party, and vice versa.
Thus in conclusion, Binary PR can be expected to have the
following effects. (1) It gives each of the two parties in an association the
incentive to broaden its appeal, and in particular to appeal to that ethnic
group to which it did not appeal before. (2) It gives the parties of different
associations an incentive to converge towards each other in policy, that is,
towards a similar moderate policy responsive to each section of electors and to
each ethnic group. (3) Binary PR tends to increase the number of parties, since
sometimes (as in the above Northern
Ireland example) a party will be motivated
to form its own association by splitting into two parties.
In most electoral systems, greater fragmentation should be
regarded as a disadvantage, since it tends to increase the sectionalism of
parties, and the divergence between parties in policy. However, under Binary
PR, this fragmentation has the opposite effect--the party competition within
each association promotes wider and more pan-ethnic responsiveness, and
convergence in policy between parties, not only between those parties in the
same association, but also between those in different associations. Hence under
Binary PR any tendency to fragmentation is to be regarded as an advantage,
since when it occurs, it provides an incentive towards greater moderation,
pan-ethnic responsiveness, and convergence.
4: THE TERRITORIAL LIST SCHEME
Binary PR as so far described, has done no more than determine
each party's number of seats. But to qualify as a complete electoral system, it
must also allocate the seats won by a party between its candidates, and
allocate these candidates to the areas of the country for which they are to
have responsibility. It would be possible to do this by means of party lists in
multi-member constituencies, just as in the usual forms of Party List PR.
However, this paper proposes an alternative scheme of party lists combined with
single-member territories, which will be referred to as the Territorial List
Scheme. The advantages claimed for this scheme are (1) that it would give much
better contact between the individual representatives and the electors each one
represents, and (2) that it would promote greater unity and cohesion in each
party. This party cohesion is especially important for Binary PR, which aims to
transform the currently more or less mono-ethnic parties, into pan-ethnic
parties, responsive to each ethnic group. This transformation is more likely to
be achieved, if each party can be made to function as a coherent and
disciplined team of representatives, united behind the objective of obtaining
votes from each ethnic group, and so maximising the party's seats. Thus the
Territorial List Scheme is put forward as an important part of Binary PR. Its
rules are as follows.
Shortly before the general election, each party registers its
provisional list of candidates, which normally will include the party's elected
representatives from the immediately previous term. A candidate selection
ballot is then carried out, separately for each party, under the auspices of
the electoral authority. In this ballot, all persons on the party's provisional
list are placed as candidates on the ballot paper, and only persons on this
provisional list may vote, voting for as many candidates as they like. The
ballot is secret. To be accepted on to the party's final list, a person must
obtain votes from at least one-half of the members of the provisional list.
Further persons may be voted on to the final list, by the assenting vote of a
majority of those already on the final list, in secret ballot. A party will
place on its list more candidates than the number it expects to elect, in order
to fill vacancies occurring during the legislative due to the death or
retirement of its representatives. Thus in effect the representatives of the
party act as an electoral college, to decide which of their number should be
selected to stand as candidates of the party for election for a further term.
The final list of candidates is then placed in the order in
which they are to receive seats, again by vote of the candidates on that list,
in another secret ballot. In this, the voting is preferential, that is, the
person votes by writing "1" for a first preference, "2" for
a second preference, and so on, for as many candidates as they wish. The candidates
on the list are then placed in order by successive applications of the
Alternative Vote method. (See note 2.) That is, one candidate is chosen by AV,
to occupy the first place on the list. From the remaining candidates, one is
chosen by AV to occupy the second place, then one for the third place, and so
on.
Under the Territorial List Scheme, the country is divided into
small electoral areas referred to as "tracts". The number of tracts
could suitably be about three times the number of seats in the parliament to be
elected, or more than this if preferred. From these tracts, each party creates
a number of "territories", a territory being a cluster of adjacent
tracts, as many as the party chooses, in which the party places one of its
candidates. The party thus creates as many territories as the approximate
number of seats it thinks it is going to get. The only restrictions are that
the party may not put more than one candidate in the same territory, and that
its territories may not overlap each other (but they will, of course, overlap
the territories of other parties).
The decisions about how many territories to create, about which
tracts are to go into them, and about which candidate is to stand in which
territory, should be taken by the candidates of the party themselves, acting
collectively. The procedure proposed is that they elect a committee to draft a
territory plan for the whole country, which is finally passed by majority vote
of the candidates of the party, in secret ballot.
There is a separate ballot paper for each tract, and it carries
the name of each candidate standing in the tract, together with the name of his
or her party. Any two parties/candidates which are in the same association are
indicated, and placed next to each other on the ballot paper. Electors vote by
putting the parties/candidates in order of preference, and the number of seats
for each party is calculated, by the rules of Binary PR.
If a party gets more seats than the number of territories it has
created, then all its candidates who have territories are elected, and the
extra seats are filled by the election of those of its candidates without
territories who are highest placed on the party's list. If on the other hand
the party gets fewer seats than it has territories, then the appropriate number
of candidates with territories, those placed lowest on the list, are denied
election, starting with the lowest-placed candidate, except that a candidate is
not denied election if his territory is adjacent to that of another candidate
who has already been denied election. Each tract of the territory of a
non-elected candidate is then included in one of the nearest adjacent
territories. Thus whether the party has created too few territories or too
many, each tract in the country will be the responsibility of one of the
elected representatives of the party, that one in whose territory it is
included.
If a representative dies or retires during the legislative term,
then a candidate is elected from the top of the party's list of non-elected
candidates.
Thus the territories of the representatives of any one party
will normally cover the whole country. If the party is large, it will have a
large number of territories, and each territory will be a small one, consisting
of a small number of tracts. If it is a small party, it will have a small
number of territories, and each territory will be large, containing many
tracts.
Each tract of the country will thus have an elected
representative of each party to take responsibility for it. At the same time,
the territory of each representative will not be any bigger than is necessary
for his party to cover the whole country. Thus this scheme would appear to
optimise the link between representative and elector, maximising the
opportunity of each representative to maintain contact with the electors who
support them.
In order to strengthen the parties, it is important that a
substantial amount of state funding should be paid to the parties as a regular
income throughout the legislative term, each party getting a share in
proportion to its seats. This party income should be paid, not to the individual
representatives of the party, but to the party's team of representatives as a
whole, or to an organisation they set up to receive it. The object of this
subsidy is to help establish each party as an independent team of
representatives, able to finance its own campaigning, and as far as possible
without financial or other ties to outside groups, and in particular without
ties to any one ethnic group. Such an independent team will be better able to
seek and attract votes from each ethnic group, and so retain or increase its
seats.
Another factor helping to establish each party as an independent
and coherent team, is the procedure for drawing up the party list. For the
individual representatives of a party will be fully aware that, before the next
election, they will face a candidate selection ballot, at which the other
representatives of the party can, in secret ballot, with no risk to themselves,
and without need to explain or justify, simply expel them without appeal from
the party's list, or place them so low down on the list that they are very
unlikely to get a seat. Thus any representative who wants to be re-elected will
have a strong incentive to cooperate with the rest of the party team, in their
efforts to develop and promote a policy which will get votes from each section
and from each ethnic group, and so retain or increase the party's seats. Any
representative who refuses to abandon the traditional ethnocentric policy and
rhetoric which presumably was used by the party under the previous systems, is likely
to be expelled from the party list. If a party has split into two parties to
form its own association (as Sinn Fein and the SDLP were assumed to do in the
above Northern Ireland scenario), then this factor, and also the party subsidy,
will enable each of the two new parties to establish itself as an independent
team, competing strongly with the other party in the association for votes and
seats.
Thus in conclusion, the provisions of the Territorial List
Scheme can be expected to strengthen and unify each party, building up its
independence, and encouraging its representatives to cooperate as a team, one
which is responsive to each section of electors and to each ethnic group. Thus
the Territorial List Scheme will reinforce the effectiveness of Binary PR, in
transforming the currently more or less mono-ethnic parties, into pan-ethnic
parties, responsive to each ethnic group.
5: THE DISTRIBUTED VOTE
Binary PR is one general method for protecting the ethnic
minorities. Another is that of the Distributed Vote. The principle of the
Distributed Vote is that it rewards a party for getting an even distribution of
votes as between the different ethnic groups, and penalises it for getting an
uneven distribution. For example, suppose that in Northern Ireland, party A gets the
votes of 20% of the Catholics, and 24% of the Protestants, while party B gets
the same number of votes as A, but gets them all from the Protestants. The
Distributed Vote principle requires that A should get more seats than B, thus
providing an incentive for any party to get votes from each ethnic group.
There are three systems, referred to as Tract DV, Set DV, and
Ethnic-Roll DV, which could be used to implement this Distributed Vote
principle. The way in which, under each system, the number of seats of each
party is determined, is set out below. The way in which a party's seats thus
determined are to be allocated to its candidates, could be either by Party-List
PR, or preferably by means of the Territorial List Scheme, as described in
section 4 above.
5.1: Tract Distributed Vote
Under Tract DV, the country is divided into small
"tracts", as with the Territorial List Scheme, except that the tracts
will tend to be smaller, and of irregular size and shape. The reason for this
is that the boundaries of each tract are drawn so that it contains as high as
possible a percentage of electors of one or other of the ethnic groups, that
is, so that the tract is as mono-ethnic as possible. For example, in Northern Ireland,
each tract should be drawn up so as to consist of as high a percentage as
possible, either of Protestants, or of Catholics.
Then, in order to determine each party's number of seats, the
following procedure is carried out. For each tract, each party's percentage of
the votes is calculated, according to the following formula: the party's votes
in the tract, divided by the sum of all parties' votes in the tract, multiplied
by 100. From these tract percentages, each party's "distributed
percentage" is calculated, this being its percentage of the votes in that
one-fifth of tracts in which the party obtains its lowest percentages of votes.
(For this purpose, one-fifth of tracts is defined as a set of tracts containing
one-fifth of all voters.) Each party then receives a share of the seats in proportion
to its distributed percentage. In the unlikely event that each party gets a
distributed percentage of zero, parties receive seats by normal PR, in
proportion to their totals of votes in all tracts.
For example, suppose that there are three parties A, B and C,
which respectively have overall percentages of votes of 30, 40 and 30, and
distributed percentages of 20, 15 and 5. They get seats in proportion to their
distributed percentages. Thus A gets 20/(20 + 15 + 5) = 50%, B gets 37.5% and C
gets 12.5%. Thus B, the party with most votes, gets fewer seats than A because
in its worst one-fifth of tracts it gets a lower percentage than A does, while
C loses more than half its seats by getting a very low percentage in its worst
tracts.
The effect of Tract DV can be seen by considering its use in Northern Ireland.
If for example a party gets very few votes from the Catholics, it will get a
very low percentage in those tracts which are mainly or entirely Catholic. Thus
its "distributed percentage"-its percentage in its worst one-fifth of
tracts-will be very low, and if the ratio of its distributed percentage to its
overall percentage of votes is lower than that for the other parties, it will
lose seats.
Tract DV thus gives each party the incentive to get votes as
evenly as possible from all tracts, and hence, in so far as each ethnic group
has tracts in which it is concentrated, it gives each party the incentive to
get votes from each ethnic group. Thus Tract DV is suitable for a country where
the ethnic minority or minorities are relatively large, and where it or they
tend to be concentrated in particular localities.
It thus appears that Tract DV would be effective in Northern
Ireland, since the ethnic minority, the Catholics, is large, being over 40%,
and since surveys have shown that "about one-half of the province's 1.5
million people live in areas more than 90 per cent Protestant and 95 per cent
Catholic. Fewer than 110,000 people live in areas with roughly equal numbers of
Catholics and Protestants." (A. Pollak, A Citizen's Enquiry: The Opsahl
Report on Northern Ireland, Lilliput Press, 1993, page 42). Thus it should not
be difficult to draw up tracts such that about half the electors are in tracts
which are either nearly all Catholic, or nearly all Protestant.
However, if one (or more) of the ethnic groups is small, and/or
is dispersed more or less evenly over most tracts, Tract DV will provide little
or no extra incentive above the normal, for a party to seek its votes. For this
reason, the other two forms of the DV system, Set DV and Ethnic-Roll DV, are
put forward, which are likely to be more effective in protecting such small or
dispersed ethnic groups.
5.2: Set Distributed Vote
In Set DV, the electors are registered on separate ethnic
electoral rolls, each roll containing one ethnic group. A "general"
roll may be created, for those persons who do not belong to any of the
specified ethnic groups. This has already been done in some countries, for
example in Fiji,
where there is a Fijian roll, an Indo-Fijian roll, and a "general"
roll for other races. In a country where such ethnic rolls need to be created,
the necessary procedure should be started well before the election, let us say
at least a year before. Electors then indicate which roll they wish to be
placed on, and a provisional list of names for each roll is made public. Any
elector's registration can then be contested by other electors, on the grounds
that they do not belong to the ethnic group with which they have registered. To
decide any contested cases, local tribunals will need to be set up, each
tribunal consisting of well-known persons of high standing in the area, drawn
from each ethnic group.
(For example, in Northern
Ireland, a pro-Protestant party which
expects to get very few Catholic votes and so to lose seats, might have an
incentive to persuade some of its Protestant supporters to register as
Catholics. But if these tribunals are set up, their registration can be
contested by other electors, and the prospect of this might deter the party
from pursuing this strategy in the first place.)
The country is divided into tracts, but in this case, unlike
under Tract DV, the tracts are given any convenient boundaries, and no attempt
need be made to make them mono-ethnic. A "set" is defined as a set of
electors in one or more tracts, who are all registered on the same ethnic roll.
(If an ethnic roll has only a small percentage of the electors in some area,
then in order to create a set of adequate size, for that roll electors from
several tracts will need to be included in one set.) A party's
"distributed percentage" is then defined as its percentage of the
votes in that one-fifth of sets in which the party obtains its lowest
percentages of votes. Each party then receives a share of the seats in proportion
to its distributed percentage (unless each party's distributed percentage is
zero, when each gets seats in proportion to its total of votes in all tracts).
The effect of Set DV is that if a party gets a much lower
percentage of votes from any one ethnic roll than it gets from the electors as
a whole, then it will get a low percentage in the sets of that roll. Thus it
will get a low distributed percentage, and so, if the ratio of its distributed
percentage to its overall percentage of votes is lower than that for the other
parties, it will lose seats.
This gives the party the incentive to appeal to each ethnic
group, irrespective of whether the group is concentrated in some localities, or
is spread out widely over the country. Thus in a country where one or more of
the ethnic minorities is widely dispersed, Set DV should be used rather than
Tract DV, the more so the smaller is the minority, since this means that it has
a low percentage of the votes even in those areas where it is most
concentrated. However, with the smallest minorities, let us say below 10 per
cent, the following system of Ethnic-Roll DV should preferably be used.
5.3: Ethnic-Roll Distributed Vote
In Ethnic-Roll DV, as in Set DV, the electors are registered on
separate ethnic electoral rolls, each roll containing one ethnic group. A
"general" roll may be created, for those persons who do not belong to
any of the specified ethnic groups. In order to allocate the seats, each
party's number of "points" is calculated. A party's number of points
is whichever is less, either one-half of its percentage of the votes in all
rolls, or its lowest percentage of votes in any one roll (other than the
general roll, if there is one). One-quarter of the total seats are then
allocated between the parties in proportion to their votes, and three-quarters of
the seats are allocated in proportion to their points (except when each party
gets zero points, in which unlikely case all seats are allocated in proportion
to votes).
The effect of Ethnic-Roll DV is that if in each roll a party
gets at least half its overall percentage (that is, if it gets what might be
regarded as adequate support in every roll), then it is given seats in
proportion to its votes, and it loses no seats at all. A party loses seats only
if in any roll it gets less than half its overall percentage, and it can lose
anything up to three-quarters of its seats if its lowest percentage in any one
roll is low enough. This gives the party the incentive to appeal to each ethnic
group, and especially to that group from which it formerly obtained the lowest
support.
But the most important point is this: a party loses the same
number of seats by getting a given lowest percentage, however few electors
there might be in the ethnic roll concerned. Thus Ethnic-Roll DV offers the
strongest protection to a numerically small ethnic minority. Provided the
minority can be put on a separate roll, each party will be motivated to respond
to its needs, however few electors the group contains.
At first sight, it might appear that this system is at risk of
giving a small ethnic group excessive influence over the policy of each party.
But in fact the system only gives the group the capability of deterring
discrimination against it; it does not empower the group to obtain privileged
treatment for itself. For a party loses seats only if its percentage of votes
from the group is less than half its overall percentage from all groups. If
this is so, it strongly indicates that the party has been discriminating
against that ethnic group. Thus the use of Ethnic-Roll DV should be seriously
considered, in any country where there are small ethnic groups which need
protection.
6: BINARY DISTRIBUTED VOTE
The Distributed Vote can be used on its own, as another distinct
electoral system, as was described in section 5 above. However, in some
situations, it might be preferable that the Distributed Vote principle should
be used in combination with Binary PR, to strengthen the incentive of each
party to obtain votes from each ethnic group, above what would be obtained by
the use of either system used alone. In this combined system, one or other of
the three forms of the Distributed Vote is used to determine the number of
seats of each association, and the rules of Binary PR are used to determine
what share of an association's seats should be given to either of the two
parties in the association. Which of the three forms of the Distributed Vote
should be used, whether Tract DV, Set DV or Ethnic Roll DV, depends (as was
explained in section 5) on the size of the minority ethnic group or groups, and
on the degree to which it or they are concentrated in particular locations, or
are spread out more or less evenly over the country.
To illustrate this, let us consider the example of Tract DV
added to Binary PR, as the combined system. The way the systems are combined is
as follows. The Binary PR system as described in section 3 above, is modified
as follows. Tracts are drawn up which are as mono-ethnic as possible, just as
for Tract DV. Parties form two-party associations, the parties are placed on the
ballot paper, and electors vote preferentially for them, just as for Binary PR.
Each party's "distributed percentage" is then
calculated, on the basis of its percentage of first-preference votes in each
tract. Thus the party's distributed percentage is its percentage of the
first-preference votes in that one-fifth of tracts in which the party obtains
its lowest percentages of first-preference votes.
The seats for each two-party association are then calculated by
the rules of Binary PR, except that an association gets seats according to the
two parties' distributed percentages as calculated above, instead of according
to their percentages of first preferences. Thus if in any association the
smaller of the two distributed percentages is at least two-thirds of the larger
one, then the association gets a share of total seats in proportion to the sum
of the distributed percentages of its two parties. If the smaller distributed
percentage is less than two-thirds of the larger one, then the association is
penalised by getting seats in proportion to 2.5 times the smaller distributed
percentage.
The seats of an association thus calculated, are then divided
between its two parties according to the number of voters preferring one to the
other, exactly as under the rules of Binary PR.
This combined system of Binary Distributed Vote thus provides a
double pan-ethnic incentive, combining the pan-ethnic incentive of the
Distributed Vote with that of Binary PR. As under Distributed Vote, a party
needs to get (first-preference) votes from each ethnic group, in order to avoid
getting a low distributed percentage, and thus losing seats for its
association. As under Binary PR, the party needs to get wide support from all
ethnic groups, in order to compete effectively with the other party in the
association, for a share of whatever seats the association obtains. This
combined system of Binary Distributed Vote thus provides a stronger and more
reliable pan-ethnic incentive, and stronger protection for the ethnic minority,
than either Distributed Vote or Binary PR used alone.
7: DISTRIBUTED STV
The three systems described in section 5, apply the
Distributed-Vote principle to the party-list form of proportional
representation. But it is also possible to apply it to the other form of PR,
that of STV (the Single Transferable Vote). In this way, a system can be
produced, which would give very much better protection for the ethnic
minorities than would normal STV. To construct this system, STV could be
combined with any one of the three forms of Distributed Vote, that is, Tract
DV, Set DV or Ethnic Roll DV. Which of these three forms of Distributed Vote
should be used, will depend (as was explained in section 5) on the size of the
minority ethnic group or groups, and on the degree to which it or they are
concentrated in particular locations, or are spread out more or less evenly
over the country.
To illustrate the construction of such a combined system, let us
consider the example of Tract Distributed STV, that is, Tract DV added to STV.
The rules of this system are as follows.
The country is divided into small tracts, the boundaries of
which are drawn so that they are as mono-ethnic as possible, just as under
Tract DV. The tracts are then combined into multi-member constituencies, let us
say ones of five members. Each constituency is ethnically balanced, in that the
percentage of its electors of any one ethnic group who are in mono-ethnic or
nearly mono-ethnic tracts, is about the same as that for the whole country. For
example, if 20 per cent of all electors are Catholics living in tracts which
are all or nearly all Catholic, then each constituency should be composed of
tracts such that about 20 per cent of its electors are Catholics in (nearly)
all-Catholic tracts. (To achieve this balance, it will probably be often
necessary to create divided constituencies, consisting of several
geographically separated sets of tracts.)
The electors vote preferentially between the candidates, as
under normal STV. To elect five candidates in any one constituency, the
procedure is as follows. Each vote is transferred to its first-preferred
candidate. Each candidate's "distributed percentage" is found (just
as for a party under Tract DV)--that is, his percentage of the votes in that
one-fifth of tracts in which he obtains his lowest percentages of votes. That
candidate is excluded who has the lowest distributed percentage. Each vote of
the excluded candidate is transferred to the candidate it indicates as next
preference, and the distributed percentage of each remaining candidate is
recalculated, to take account of the votes transferred. That candidate who then
has the lowest distributed percentage is excluded, and so on, if necessary,
until five candidates are left, who are declared elected.
What then is the likely effect of Tract Distributed STV? Each
candidate will have the incentive to get votes evenly from each tract, since if
he gets few or no votes from some of the tracts, he will get a low
"distributed percentage", and will risk being excluded. Thus if any
ethnic group has a significant number of tracts where it is concentrated, he
will have a strong incentive to appeal to it and get votes from it, since
otherwise he will get few votes in these tracts. Thus Tract Distributed STV,
like Tract Distributed Vote, is suitable for a country where the ethnic
minority or minorities are relatively large, and where it or they tend to be
concentrated in particular localities. In such a country, the system will
provide a strong pan-ethnic incentive, both for the parties and for each of
their individual candidates, and the ethnic minority will receive strong
protection.
8: ETHNIC-ROLL ALTERNATIVE VOTE
So far in this paper, the Distributed-Vote principle has been
applied only to systems of proportional representation. But it is also possible
to apply it to non-proportional systems using single-member constituencies,
such as AV (the Alternative Vote). Thus AV can be combined with Ethnic-Roll DV,
in order to produce a system which gives much better protection for the ethnic
minorities than would straight AV. The rules of this combined system, referred
to as Ethnic-Roll AV, are as follows.
Just as in Ethnic-Roll DV, the electors are registered on
separate ethnic electoral rolls, each roll containing one ethnic group. A
"general" roll may be created, for those persons who do not belong to
any of the specified ethnic groups. But unlike in Ethnic-Roll DV, each roll,
including any general roll, is divided into "electorates", an
electorate being a small set of electors of that ethnic group, normally a set
residing in one locality. There are as many electorates in a roll as there are
representatives to be elected, and the electorates in the same roll are roughly
equal in the number of electors each contains. Constituencies are then created,
each constituency consisting of one electorate from each roll. It will often be
necessary for the electorates of the same constituency to be at a considerable
distance from each other, but there seems no great problem about this.
The electors vote by putting the candidates in their
constituency in order of preference, as under normal AV. In any one
constituency, one candidate is elected as follows. Within each electorate of
the constituency, each vote is allocated to the candidate for whom it expresses
first preference. Each candidate's number of "points" is then
calculated. A candidate's number of points is whichever is less, either
one-half of his percentage of the votes in all rolls, or his lowest percentage
of votes in any one roll (other than the general roll, if there is one). The
candidate with fewest points is excluded, and each of his votes is transferred
to the continuing candidate for whom it expresses next preference, if there is
one. The points of each continuing candidate are recalculated, to take account
of the votes which have been transferred, and the candidate who then has the
lowest points is excluded. And so on, if necessary, until only one candidate is
left, who is elected.
What then is the likely effect of Ethnic-Roll AV? Any candidate
who at any stage in the vote counting, gets a very low percentage in one ethnic
roll, however small this roll may be, is likely to be excluded, even if his
overall percentage in all rolls is higher than that of any other candidate.
Thus each candidate will have a strong incentive to get adequate votes from
each ethnic roll, and more specifically, to get a percentage of votes in each
roll which is at least half of his overall percentage in all rolls. If all
candidates manage to do this, then candidates will be excluded or elected on
the basis of their overall percentages, just as in ordinary AV. But the use of
the ethnic rolls will still have had the desired incentive effect, in
motivating the candidates, and in particular the elected representative, to
respond to the concerns of each ethnic group. Thus the addition of Ethnic-Roll
DV to the Alternative Vote offers stronger protection to an ethnic minority
than would be offered by AV used alone. This is especially important for a
numerically small ethnic minority, whose needs might be neglected under
straight AV. But provided the minority can be put on a separate roll, and the
required "electorates" can be created from it, each candidate and
each elected representative will be motivated to respond fairly to its needs,
however few electors the group contains.
9: ETHNIC-AREA ALTERNATIVE VOTE
Ethnic-Area AV is a system very similar in structure to
Ethnic-Roll AV, using the Distributed-Vote principle, but in this case without
need to put the electors on separate ethnic electoral rolls. For Ethnic-Area AV
to be used in a country, it is necessary that the different ethnic groups
should to a considerable extent be geographically segregated. The way it
operates is as follows.
For each ethnic group which it is sought to protect, more or
less mono-ethnic areas are found, in which all or nearly all the electors
belong to this ethnic group. In these areas, the electors are divided into
local "electorates" as under Ethnic-Roll AV, there being as many
electorates for this ethnic group as there are representatives to be elected.
For the remaining areas where the ethnic groups are mixed, the electors are
divided into local electorates, again as many as there are representatives to
be elected. Constituencies are then created, each constituency consisting of
one mono-ethnic electorate from each ethnic group, and one mixed electorate. In
each constituency, the election is then held as under Ethnic-Roll AV, by the
exclusion of the candidates with fewest "points". Here, a candidate's
number of points is whichever is less, either three-quarters of his percentage
of the votes in all electorates in the constituency, or his lowest percentage
of the votes in any one mono-ethnic electorate. Used in this way, provided the
country is to a considerable extent ethnically segregated, Ethnic-Area AV could
offer the ethnic minorities much greater protection than they would be likely
to obtain under AV used alone.
In some countries, it might be more suitable to use a hybrid
system, combining Ethnic-Area and Ethnic-Roll AV. In this system, some ethnic
groups, those which have sufficient mono-ethnic areas, are treated as under
Ethnic-Area AV; and other ethnic groups, those which are small and widely
dispersed, and without significant mono-ethnic areas, are placed on separate
ethnic electoral rolls, and treated as under Ethnic-Roll AV. In this way, a
hybrid system could offer a small and dispersed ethnic minority much greater
protection than it would have obtained if Ethnic-Area AV had been used.
10: COALITION GOVERNMENT
So far in this paper, thirteen alternative electoral systems have
been put forward: Binary STV, Binary PR, three forms of the Distributed Vote,
three corresponding forms of Binary Distributed Vote and of STV, Ethnic-Roll AV
and Ethnic-Area AV. With Binary PR and Binary Distributed Vote, it is
impossible for any one party to get a majority of the seats, and with the other
systems, absence of a majority appears to be quite likely. Hence with Binary PR
and Binary Distributed Vote, there will always be coalition government, and
with the other systems there will be a considerable chance of it. How far then
will the parties produced by these systems be able to cooperate in a coalition
government, and how far will this government be stable and effective?
Now any of these new systems can be expected to exert over time
on each party and each candidate a steady incentive towards moderation, towards
a policy responsive to the needs of each ethnic group, and towards a centre
position, not only on the ethnic dimension, but also on any other
dimension-such as the left-wing/right-wing dimension-which is of importance to
the electors. Thus these systems encourage the parties to converge towards each
other, both in policy (towards this moderate centre position), and (in that
each party appeals to all groups) in the groups to which they seek to appeal.
Thus these systems are very different in their effects from the usual forms of
proportional representation, which tend to encourage in each different party a
narrow focus on the interests of one segment of the electorate, and provide the
incentive for parties to diverge in policy, rather than to converge.
Even in the first parliament elected under the new system, an
extremist party will have an incentive to become more moderate and to broaden
its appeal, in order to get more seats in the next election. It thus seems
likely that, even in this first parliament, there will be at least a majority
of the representatives who come from moderate pan-ethnic parties, so that a
majority coalition of these parties can be formed. Thus although under Binary PR
or Binary Distributed Vote, there might be more parties than before, it can be
expected that, since these parties will tend to be moderate and similar to each
other in policy, a stable and effective coalition government can be formed, one
which is widely acceptable to the electorate, and responsive to each ethnic
group.
11: DIRECT ELECTION OF GOVERNMENT
The new electoral systems so far described, have been ones which
in general will require government by coalition. In contrast, this section puts
forward a new system which also is designed to encourage pan-ethnic parties,
but which actually guarantees that the government will be a single party. This
new system, which is referred to as DEG, or Direct Election of Government, uses
the same basic mechanism as does Binary PR, for providing this pan-ethnic
incentive, that is, it uses the mechanism of Binary Competition between two
parties. But the difference is that, under DEG, the election serves two
functions: not only the usual function of determining each party's number of
seats, but also that of electing one of the parties to be the sole government
party.
The government party is thus elected by direct vote of the whole
electorate, which votes preferentially, each voter placing the competing
parties in order of his or her preference. The method proposed for electing the
government party is a variant of the Condorcet method. (See note 3.) This can
be expected to provide the competing parties with the incentive to follow a
moderate centre policy, equitably responsive to each section of the electors,
and to each ethnic group.
But how is this elected party, which will normally be far short
of a majority of seats, enabled actually to govern? This is achieved by the
following two provisions: (1) if the government party has fewer than 30% of the
seats, extra seats are created and given to it, equal to the number by which it
fell short of 30%; (2) each of its representatives is given a weighted vote,
sufficient to provide the party as a whole with a 55% majority of the votes in
the parliament. (See note 4.)
One possibility is to combine this scheme for the election of
the government party, with some form of proportional representation, which
would give each party seats in proportion to its first-preference votes. However,
PR has the drawback that it tends to give parties the incentive to become more
sectionalist and mono-ethnic, and to diverge from each other in policy, as each
party seeks the votes of and responds to a different section of the electors,
or a different ethnic group.
It is desirable that, in contrast, the electoral system used to
give each party seats, should reinforce the incentive which is provided by
Condorcet in electing the government party, for each party to be responsive to
each section of electors. This paper therefore puts forward a new electoral
system, All-Preference Representation (APR), which resembles Binary PR, except
that a party's seats depend not, as in the latter system, on a competition with
one other party, but on a series of two-party competitions with each of the
other parties. APR thus gives a party seats, not only according to its number
of first-preference supporters, but also according to the rankings given to it,
by all the other voters, the ones who do not give it first preference. Under
APR, as with Binary PR, a party can lose up to three-quarters of its seats, by
getting a low enough ranking from these latter non-first-preference voters.
This gives each party the incentive to appeal more widely, to sections of
electors and to ethnic groups to which it did not appeal before, in order to
obtain a higher non-first ranking from them. Thus the use of APR to determine
each party's seats, reinforces the incentive provided by the Condorcet election
of the government, for each party to be responsive to each section of electors
and to each ethnic group.
But with APR, how are the seats which a party wins, allocated
between its candidates, and how are these candidates allocated to the areas of
the country for which they are to have responsibility? Just as with Binary PR
or the Distributed Vote, this could be done either by means of Party-List PR,
or preferably by means of the Territorial List Scheme, as described in section
4 above.
11.1: Incumbency Condorcet
Under DEG, to elect the government party, a method of election
is required, which gives the government thus elected the incentive to be
responsive to each section of electors, and to each ethnic group. Perhaps the
best way of providing this incentive to all-round responsiveness, is to use one
of the variants of the Condorcet method. The incentives which Condorcet
provides can be seen by considering an example where there are three parties A,
B and C, in more or less equal competition, that is, each of them has a
significant chance of winning. Then any one of these parties, let us say A,
will need to appeal to the first-preference supporters of B in order to get a
majority of voters to prefer A over C, and to the first-preference supporters
of C in order to get a majority to prefer A over B. Thus in order to get itself
elected under Condorcet, party A will have the incentive to appeal to all
sections of the electorate, and to all ethnic groups.
But although Condorcet gives very desirable incentives, it has
the drawback that sometimes there is no Condorcet winner to elect. For example,
if 39% of voters vote ABC, 31% BCA, and 30% CAB, then there will be a cycle of
parties, where A beats B by 69 to 31, B beats C by 70 to 30, and C beats A by
61 to 39, so that there is no party which beats each other party. Thus
Condorcet needs to be supplemented by some other rule of election, such as the
Alternative Vote, to use when there is no Condorcet winner, so as to ensure
that in all situations, some party will be elected. (The Alternative Vote is
explained in note 2.)
However, this only leads to another problem. Consider the
following example, where the percentages of votes are: 40 ACB, 25 CAB, 35 BCA.
Then C is the Condorcet winner, and A is the party which would win under the
Alternative Vote. This gives the supporters of A the incentive to vote
strategically, by voting ABC, or just by voting for A alone without expressing
any more preferences. This will create a cycle, where A beats B, which beats C,
which beats A, so that the Alternative Vote comes into use, and A is elected
instead of C.
This possibility of strategy has important implications. As a
result, the fact that a party has become the Condorcet winner, genuinely
preferred by a majority to each other party, is no longer a guarantee that it
will be elected. There is thus no guarantee that parties will always strive to
become the Condorcet winner, or that the system will give them the desirable
incentives to all-round responsiveness which we originally associated with
Condorcet.
What then can be done to modify the system, so as to guarantee
these desirable incentives? A possible solution to the problem is to narrow
down our focus, from the attempt to provide all parties with the required
incentives, to the more manageable task of providing these incentives to only
one party.
Now the party in which it is important to ensure all-round
responsiveness, is clearly the government party, since it is on the actions of
this party, not those of the opposition parties, that the welfare of the
electors depends. And for such an incumbent government party, the election
which provides it with these incentives to respond, is not the past one at
which it was elected, but the future one at which it may or may not be
re-elected. Thus in any election, what is important is the incentives which
that election provides to that party which has been the government party in the
immediately previous term. Let us refer to this party as "the
incumbent".
A method of election is therefore put forward, which provides
the required Condorcet-type incentives to the incumbent party. That is, the
incumbent can be re-elected only if it is truly the Condorcet winner (in other
words, it is sincerely preferred by a majority to each other party), and it
cannot be re-elected if it is not the Condorcet winner. The rules of this
method of election, which will be referred to as Incumbency Condorcet, are as
follows. (Note 5)
(1) The incumbent is defined as that party which contains the
largest number of candidates who were representatives of the government party
or parties in the term immediately preceding the election, and which contains
enough such candidates to fill at least 20% of the seats.
(2) If no party beats the incumbent, then the incumbent is
elected. (A is said to "beat" B if more voters prefer A to B than
prefer B to A.)
(3) If only one party beats the incumbent, then this party is
elected.
(4) If more than one party beats the incumbent, then the party
which beats the incumbent by the largest margin is elected. (The margin by
which A beats B is defined as: the number of voters who prefer A to B, minus
the number of voters who prefer B to A.)
(5) If there is no party which conforms to the definition of
incumbent given in (1), then one party is elected as the government party by
means of the Alternative Vote method.
It will be clear from these rules that if the incumbent is the
CW (Condorcet winner), then the incumbent must be re-elected, and no strategic
voting, or indeed sincere voting, by the supporters of some other party P, can
prevent it from being re-elected, and elect party P instead. The reason is of
course that the supporters of P are only a minority, so that nothing they can
do will elect P.
Similarly, if some other party is the CW, then the incumbent
cannot be re-elected, and there is no strategy by which its supporters can
prevent its defeat. For let us suppose that it is expected that, on sincere
voting, a party P will beat the incumbent and will be elected. Only those who
prefer the incumbent to P will wish to elect the incumbent instead, but as a
minority they will be unable to do it.
But what if the incumbent is not the CW, not because some other
party is the CW, but because preferences are cyclical, and no CW exists? In
this situation, the party which beats the incumbent is elected. There is then
no strategy by which the incumbent can be elected instead, because again, only
a minority of voters prefer the incumbent. (Note 6)
Thus the only way the incumbent party can get re-elected is to
become the CW, or to continue to be the CW, that is, the party sincerely
preferred by a majority over each other party, and there is no strategy by which
its supporters can get it re-elected if it is not the CW. In order to beat any
one opposition party, and get itself preferred by a majority over it, it will
need to appeal to the first-preference supporters of each other opposition
party. Thus the system provides the government party with the incentive to be
responsive to each section of the electorate, and to each ethnic group.
Similarly, the system gives the government party the incentive,
in order to have a better chance of beating each opposition party, to move
towards the centre of the left-to-right policy dimension, or any other policy
dimension which is of interest to the electors.
What then can be said in conclusion, about the DEG system, this
device of electing the government party by means of Incumbency Condorcet?
First, however fragmented the legislature has been before the system is
adopted, the system will provide government by a single party with at least 30%
of seats, and with a majority of parliamentary votes. Second, this government party
will tend to be a moderate centre party, with strong incentives to be
responsive to each section of the electorate and to each ethnic group. Third,
even if the opposition has previously been extremely fragmented, it is likely
that, probably by merger of smaller parties, one or more substantial opposition
parties will be formed, in order to get a better chance of achieving a majority
over the government party.
11.2: All-Preference Representation
As was explained above, APR gives a party seats, not only
according to its number of first-preference supporters, but also according to
the non-first rankings given to it, by all the other voters. The rules of APR
are as follows.
If any party has more than 35% of first preferences, then the
value of each of the votes giving it first preference is reduced, so that it
has 35% of what is then the total value of first preferences. (See note 7)
Basically, APR is proportional representation, plus a correction
for each party's lower preferences. Thus each party is given an initial
allocation of a percentage of seats equal to its percentage of the
first-preference votes in the whole country, just as under proportional
representation. Each possible pair of parties is then considered, and for each
pair, if one party is preferred to the other by more voters than the other is
preferred to it, an appropriate percentage of seats is transferred from the
party preferred by fewer voters to the party preferred by more voters.
This can be expressed more precisely, as follows. In any pair of
parties, let A be the party which is preferred by more voters, and B be the
party which is preferred by fewer voters. The percentage of seats to be
transferred from B to A, is calculated as follows. Let us refer to the
percentage preferring A minus the percentage preferring B, as the
"preference difference" between A and B. (Note that these percentages
are not percentages of all voters, but only of those voters who are not
indifferent between A and B.) The percentage of seats to be transferred from B
to A is equal to whichever is less: either (1) B's % of first preferences,
multiplied by A's % of first preferences, multiplied by twice the preference
difference between A and B; or (2) three times B's % of first preferences,
multiplied by A's % of first preferences, divided by four times the total % of
first preferences of all parties other than B. In other words, a percentage of
seats equal to (1) is transferred from B to A, except when (1) exceeds the
upper limit on seats transferable from B to A, this upper limit being (2).
For example, if party B has 12% of first preferences, and party
A has 35%, and 25% of voters prefer B to A, then the proportion of total seats
transferred from B to A, is equal to whichever is less, either 0.12 x 0.35 x 2
x (0.75 - 0.25) = 0.042, or 3(0.12 x 0.35) /(4(1 - 0.12)) = 0.0358. That is,
the proportion of total seats transferred from B to A, is 0.0358, or 3.58%.
This APR formula guarantees each party a minimum percentage of
seats, equal to one-quarter of its percentage of first preferences, however few
lower preferences it obtains. (See note 8 for a demonstration of this.)
Note that unlike Incumbency Condorcet, APR is completely
strategy-proof, in that there is no way in which the supporters of a party can
increase its seats above what they can obtain by voting sincerely. For example,
if voters whose true preferences are ABC, vote ACB instead, this does not
affect the seats of A, their first-preferred party, but only transfers seats
from B to C, contrary to their true preferences.
What then will be the general effect of APR on a party's seats?
First, let us suppose that all parties are moderate centre parties, responsive
to all sections of the electorate, such that any one party is preferred to any
other party by about 50% of the voters. Thus few or no seats will be
transferred from one party to another, and each party will get a percentage of
seats about equal to its percentage of first preferences. But if one of these
parties moves away from a centre position, let us say to the right (while the
other parties stay at the centre), the percentage of voters preferring it to
any other party will be reduced, and it will lose seats to the other parties.
The following numerical examples illustrate the strength of the
effect on a party's seats, of a change in lower preferences under APR. If there
are two parties A and B, each with 30% of first preferences, and A moves down
past B on the ballot papers of 1% of voters (for example, if this 1% of voters
change from voting CABD to voting CBAD), then A loses 0.36% of seats, and B
gains 0.36%. If a party has 25% (or less) of first-preference supporters, and
all the 75% (or more) of voters who do not give it first preference give it
their last preference, then it will lose three-quarters of its initial
allocation of seats, that is, an allocation of 25% (or less). If it has 35% of
first preferences, and all other voters give it last preference, it will lose
almost two-fifths of its initial allocation of 35% of seats. This illustrates
the difference between ordinary PR and APR. With ordinary PR it makes no
difference to a party's seats, whether those voters who are not its
first-preference supporters, give it their second preference or their last
preference. Thus the party may well neglect the interests of many of these
voters, particularly if they already give the party a very low preference. But
in contrast, with APR, the party can lose up to three-quarters of its seats if
they give it their last preference. The party therefore has a strong incentive
to take their concerns and interests into account.
Thus in conclusion, APR gives each party the incentive to
respond and appeal to electors of all sections and all ethnic groups, even to
those who previously have given it their lowest preference. If it gets a low
ranking from enough of them, it could lose up to three-quarters of its seats.
Similarly, APR gives each party the incentive, in order not to lose seats, to
move towards the centre of the left-to-right policy dimension, or any other
policy dimension which is of interest to the electors. APR thus strongly
reinforces the incentives provided by Incumbency Condorcet, to become a
moderate centre party, and to be responsive to each section of the electorate,
and to each ethnic group.
11.3: Distributed-Vote DEG
In section 6 above, the Distributed Vote was added to Binary PR,
to strengthen the incentive of each party to obtain votes from each ethnic
group, above what would be obtained by the use of either system used alone.
Similarly, the Distributed Vote can be added to the DEG system, again to
strengthen the parties' incentive to obtain votes from each ethnic group, and
to increase the protection which the system provides for the ethnic minorities.
In this case, the Distributed Vote can be used in two ways: along with
Incumbency Condorcet to elect the government party; and along with the
All-Preference Representation system to determine each party's number of seats.
As in the case of Binary PR, which one of the three forms of the Distributed
Vote should be used, whether Tract DV, Set DV or Ethnic-Roll DV, will depend
(as was explained in section 5) on the size of the minority ethnic group or
groups, and on the degree to which it or they are concentrated in particular
locations, or are spread out more or less evenly over the country.
To illustrate this, let us consider the example of Ethnic-Roll
DV added to DEG, as the combined system. The way the systems are combined is as
follows. The electors are registered on separate ethnic electoral rolls, each
roll containing one ethnic group, as in normal Ethnic-Roll DV. Each party's
number of "points" is then calculated. A party's number of points is
the following: whichever is less, either one-half of its percentage of the
first-preference votes in all rolls, or its lowest percentage of the
first-preference votes in any one ethnic roll.
The DEG system consists of two parts, Incumbency Condorcet to
elect the government party, and APR (All-Preference Representation) to
determine each party's seats. The way in which Ethnic-Roll DV is used to modify
Incumbency Condorcet, is as follows. If there are more than three parties, then
the party with the lowest "points" is excluded. The votes of the
excluded party are transferred, each going to the party it indicates as next preference.
The points of each remaining party are recalculated, to take account of votes
transferred to it. The party which then has the lowest points is excluded. And
so on, if necessary, until three parties are left, which are referred to as the
"major" parties. The votes of the party last to be excluded are
transferred to these major parties, and each ones percentage of votes in each
ethnic roll is calculated. A major party is then said to have a
"deficit" if its lowest percentage in any roll is less than 17 per
cent. For example, if party A's lowest percentage is 12, then A's deficit is 17
- 12 = 5 per cent.
If the incumbent party (the one which has been the government
during the previous term) is one of the major parties, then the government
party is elected from these three parties as follows. For any two major parties
A and B, A is said to beat B if the percentage of voters preferring A to B
minus any deficit of A, is greater than the percentage of voters preferring B
to A minus any deficit of B. If no party beats the incumbent, the incumbent is
re-elected. If only one party beats the incumbent, then that party is elected.
If two parties beat the incumbent, then that one which beat it by the greater
margin is elected.
If the incumbent has been excluded, then the government party is
elected from the three major parties by a variation on the Alternative Vote
method, as follows. Each major party's overall percentage of first-preference
votes is found (that is, votes which give it the highest preference out of the
three major parties). That party is excluded for which its percentage of first
preferences minus its deficit, if any, is lowest. Of the remaining two major
parties, that one is elected which beat the other in the way explained above.
We now consider the second part of the DEG system,
All-Preference Representation, and the way in which Ethnic-Roll DV is used to
modify it. First, each party's "percentage score" is calculated, from
the party's number of "points" as defined above. A party's percentage
score is the following: its points multiplied by 100 divided by the total
points of all parties. (In other words, a party's percentage score is its
points, calculated as a percentage of all parties' points.)
One-quarter of the seats (the "PR seats") are shared
between the parties in proportion to each ones first-preference votes, just as
in ordinary PR. The other three-quarters of the seats (the "APR
seats") are shared between the parties by means of a modified APR formula,
as follows. Each party first receives an initial allocation of the APR seats,
in proportion to its percentage score. Each possible pair of parties is then
considered, and the appropriate percentage of seats is transferred, from the
party preferred by fewer voters, to the other party, as follows.
In any pair of parties A and B, let A be the party which is
preferred by more voters, and B be the party which is preferred by fewer
voters. The percentage of APR seats to be transferred from B to A, is
calculated as follows. Let us refer to the percentage of voters preferring A
minus the percentage preferring B, as the "preference difference"
between A and B. (Note that these percentages are not percentages of all
voters, but only of those voters who are not indifferent between A and B.) The
percentage of APR seats to be transferred from B to A is equal to whichever is
less: either (1) B's % score, multiplied by A's % score, multiplied by twice
the preference difference between A and B; or (2) B's % score, multiplied by
A's % score, divided by the total % score of all parties other than B. (Thus as
with normal APR, the expression (2) sets an upper limit on the percentage of
seats which can be transferred from B to A. But with this modified APR, the
limit is higher.)
To compare this modified APR formula with the previous normal
one (set out in section 10.2), let us consider an example with the same figures
as before, that is, where party B has a percentage score of 12%, party A has a
percentage score of 35%, and 25% of voters prefer B to A. Thus the proportion
of total seats transferred from B to A, is equal to whichever is less, either
0.12 x 0.35 x 2 x (0.75 - 0.25) = 0.042,
or (0.12 x 0.35)/(1 - 0.12) = 0.0477. That is, the proportion of
total seats transferred from B to A, is 0.042, or 4.2%. Note that with modified
APR, because the upper limit on seats transferred is higher, more seats are
transferred than with normal APR--4.2% instead of 3.58%.
What then is the likely effect of thus modifying the DEG system
by the addition of Ethnic-Roll DV? Let us first consider the effect on the
working of Incumbency Condorcet, the system for electing the government party.
Any party which gets a very low proportion of votes in any one of the ethnic
rolls, however small this roll may be, and however high the party's votes may
be in the other rolls, will get low "points", and so is likely to be
excluded at an early stage in the process, and not be selected as one of the
three "major" parties from which the government party is elected.
Even if it does manage to become one of the major parties, it is unlikely to be
elected as government, since this election takes into account the party's
"deficit", as well as the percentage preferring it to the other
party.
This modified system thus provides a strong second line of
defence against the election of a government party which discriminates against
one of the ethnic groups. For with Incumbency Condorcet used alone, a party
which discriminates against a small ethnic group will lose some preferential
votes, but could possibly make them up by a stronger appeal to the rest of the
voters. But with Incumbency Condorcet fortified by Ethnic-Roll DV, a party
which discriminates against an ethnic roll, however small, has no chance of
election against another party which is responsive to all ethnic groups. Thus
the inclusion in the system of Ethnic-Roll DV, greatly strengthens the
protection of each ethnic group, the additional protection afforded being
greater, the smaller is the ethnic group concerned. Any party which seeks to be
elected as government is thus given a much stronger and more reliable incentive
to respond to each ethnic roll, than it would have had if Incumbency Condorcet
had been used alone.
Let us now consider the effect of using Ethnic-Roll DV to modify
All-Preference Representation, the system used to determine each party's seats.
In the modified system, one quarter of the seats, the "PR seats", are
allocated to the parties by ordinary PR, in proportion to first-preference
votes; and three-quarters of the seats, the "APR seats", are
allocated by APR modified by Ethnic-Roll DV. With the latter, each party is
given an "initial allocation" of APR seats in proportion to its
"percentage score" (which in turn depends on the party's lowest
proportion of votes in any of the ethnic rolls). It is possible for the party
to lose all its initial allocation of APR seats if it gets very low preferences
from enough of the voters. It is also possible for the party's initial
allocation to be zero seats, if the party gets no first-preference votes in one
of the ethnic rolls. Thus a party might get no APR seats, and get only its PR
seats, a percentage of seats equal to one-quarter of its percentage of first
preferences. In other words, under this modified APR system, a party can lose
up to three-quarters of its seats. And it can lose this up-to-three-quarters of
seats in two basic ways: not only by getting sufficiently low preferences from
many of the voters (as it could do under normal APR), but also by getting few
enough first-preference votes in just one of the ethnic rolls.
With APR used alone, a party which discriminates against a small
ethnic group will thereby lose some seats, but not many of them. It can indeed
lose three-quarters of its seats, but only if it gets a last preference from at
least three-quarters of the voters. But with APR fortified by Ethnic-Roll DV,
the party loses many more seats by discriminating against a small ethnic roll.
Indeed, if it gets no first-preference votes from any one ethnic roll, large or
small, and however small it might be, it thereby loses three-quarters of its
seats. Thus the use of Ethnic-Roll DV along with the All-Preference
Representation system, greatly strengthens the protection given to each ethnic
group, the additional protection afforded being greater, the smaller is the
ethnic group concerned. Each party will thus be given a much stronger and more
reliable incentive to respond to each ethnic roll, than would have been
provided if APR had been used alone.
11.4: Overview of DEG
This new system, DEG or Direct Election of Government, has the
following properties. (1) It guarantees a single-party government with an
adequate majority of parliamentary votes, however fragmented the legislature
has been before the system is adopted. (2) At least if the Territorial List
Scheme is used (as set out in section 4), this government party, and indeed
each opposition party, is likely to be internally unified, functioning as a
strong and coherent team. Thus the system can be expected to provide stable and
effective government. (3) However uncompromising has been the political culture
of the country which has adopted it, each party, and especially the government
party, will be subject to powerful incentives to adopt a moderate centre
policy, and one which is equitably responsive to the concerns of each section
of the electorate, and of each ethnic group. (4) Any party which consistently
follows such a moderate policy, can expect to achieve considerable stability in
its number of seats, and so will be able to provide relatively safe seats for
all but a small proportion of the individual representatives on its list. Thus
the new system may well prove to be politically acceptable to the present
power-holders, the representatives in the current legislature.
However, the system may not be suitable for introduction in a
country where the parties have already divided on ethnic lines, and become
mono-ethnic, as for example is the case in Northern Ireland. For then at the
first election, it is very likely that one of these formerly mono-ethnic
parties will be elected as the single-party government. Admittedly, the
government party, and all the other parties, will be subject to powerful
incentives towards moderation, and towards pan-ethnic responsiveness.
Nevertheless, it may take some time for the parties to transform themselves
into ones which are seen by the voters as truly pan-ethnic. Hence it may not be
acceptable that the government should be entrusted to one party, a party which
might still be primarily associated with one ethnic group. For example, if DEG
were used in Northern
Ireland, in the first election, then
probably the UUP, which so far has been a Protestant-oriented party, would be
elected as the single-party government. No doubt, in order to increase its
chances of being re-elected, it would move towards a much more pro-Catholic
policy, and adopt Catholics on to its list of candidates. But it still might
not be acceptable to the Catholics to have as government a single party which
they continued to regard as belonging to the other community.
The DEG system might therefore be more suitable for an
ethnically divided country where not yet all the main parties have become
mono-ethnic. In particular, DEG might be useful where a more or less pan-ethnic
main party is threatened with the loss of votes and seats to one or more
mono-ethnic parties, as appears sometimes to have been the case in such
countries as India and Malaysia. If
DEG were adopted, the likely effect in such a situation would be that a
pan-ethnic party would be elected as the government, and that the pan-ethnic
party or parties would gain seats at the expense of the mono-ethnic parties, at
least until the latter were able to convince the electors that they had changed
to a more pan-ethnic position. Thus it might well happen that DEG will be
introduced by such a pan-ethnic main party, motivated not so much by the merits
of the system, but by its own electoral self-interest.
12: DISCUSSION
A range of seventeen new electoral systems has been put forward
in this paper, designed to protect the ethnic minorities, by giving each party
the incentive to be responsive to each ethnic group. The systems differ in the
strength of the pan-ethnic incentive they provide. The systems which provide
the strongest pan-ethnic incentives are the combined systems, Binary
Distributed Vote and Distributed-Vote DEG, and out of the three types of
Distributed Vote, the Ethnic-Roll Distributed Vote. Whether a system is the
most suitable, or is good enough, for a particular situation, will depend on
the intensity of the ethnic conflict, on how far the parties have become
mono-ethnic, on the percentage size of the ethnic groups which need to be
protected, and on the degree to which the ethnic groups are widely dispersed,
or are concentrated in particular localities.
A number of questions have been raised about these systems, in
previous presentations of them. Perhaps the most frequent question concerns the
possible complexity of the systems, either for the voter, or for the officials
counting the votes. Let us first consider the different systems from the point
of view of the voter. With the three forms of the Distributed Vote used alone,
an elector casts a single vote for one party or candidate. With all the other
systems, the electors vote by putting the parties or candidates in order of
preference. This is simple enough for the electors to do. But will the electors
understand what is being done with their votes? As with the Single Transferable
Vote, the system used in Ireland
and Malta, and in Australia for
the Senate, few people are likely understand all the details of the working of
the system. However, most people will understand the basic principle, that the
higher the preference they give to a party, the more likely is the party to get
an extra seat, and under DEG, the more likely it is to be elected as
government.
With the Distributed-Vote systems, the general principle of the
systems, that a party should be rewarded for an even distribution of votes
between the different tracts or ethnic groups, might or might not be
understood, or if understood, be approved of, by the voter. But this
understanding or approval is not necessary for the system to work effectively.
Provided the electors simply cast their votes, the system will give each party
the incentive to respond to the needs and concerns of each ethnic group, more
equitably than they would have done under systems that do not use the
Distributed Vote.
Concerning the counting of the votes, under any of the proposed
systems, this need not be excessively complex. A suggested method of counting,
applicable to the systems using preferential voting, is as follows.
The votes in each tract, in each "set" (under Set DV),
in each ethnic roll (under Ethnic-Roll DV), or in each "electorate"
(under Ethnic-Roll or Ethnic-Area AV), are counted separately. To count the
votes in a tract, etc., the ballot papers are first sorted into batches of
papers with the same votes on them. Thus papers are sorted into batches voting
first preference for the same party, then each batch is sorted into smaller
batches according to second preferences, and so on. Then the number of papers
in each batch is counted. Thus each paper will be sorted several times, once
for each vote on it, and it will be counted only once. On the basis of the
number of papers in each batch, all the calculations to determine each party's
number of seats, which of its candidates are to be elected to them, and (in the
case of DEG) to decide which party gets elected as government, can be carried
out, either on paper or by computer, without need for any further counting.
Especially if it is carried out in this way, the counting of the
votes for these systems appears to be at about the same level of complication
as that for the Single Transferable Vote system, which has been used for many
years by various countries, and is clearly quite manageable.
Another criticism of the proposed systems is that they
constitute a form of affirmative action, or positive discrimination, in favour
of the ethnic minorities. However, this seems to be based on a misconception of
the systems and their working. In fact, the systems simply provide an incentive
for a party to level up its appeal to the different ethnic groups, and to cease
discrimination against any group (using as an indicator of discrimination, that
the party is getting a low vote from it). The systems do not favour minority
groups as such, but favour any group, minority or majority, from which the
party has been getting a lower percentage of votes than from the others, and
give the party the incentive to raise this percentage up to the level of the
others.
Another point which should be made clear, is that the proposed
systems do not achieve their protection of the minorities by giving them some
kind of veto, as in a consociationalist power-sharing scheme. The parliament
elected by one of these systems will make its decisions by simple majority
vote, not by special majority, and there will be no minority veto. Thus there
seems no reason to expect any greater than normal tendency to deadlock. Indeed,
since the parties which the systems produce are likely to be more similar in
policy, the government can be expected to be less subject to deadlock than
coalition governments formed under one of the usual systems.
It has been objected that these systems might sometimes deviate
considerably from strict proportionality of representation. This is true, but
they deviate for a purpose-in order to penalise a party by loss of seats for
neglecting some group or other. But if a party responds to the incentives of
the system, and draws votes about equally from each group, it will get seats
about in proportion to its total of first-preference votes, as under normal
proportional representation.
A related objection is that these systems deny an ethnic group
the opportunity to be represented by a party solely devoted to defending its
interests, a truly mono-ethnic party which is not distracted by the need to
take into account the concerns of any other group. But it is the constant
argument of this paper that each group, and in particular a minority group, is
better served and protected by a system which produces pan-ethnic parties, such
that any one ethnic group is in effect represented by, and has its interests
defended by, each of the parties.
What then is the likelihood that one of these new systems will
be adopted? This obviously depends to a great extent on whether the present
parties and politicians will see it as in their interests. The more pan-ethnic
parties will gain seats under the new system, and the more mono-ethnic parties
will lose seats. If the parties can adjust to become about equally pan-ethnic
and moderate, then the parties will get seats about in proportion to their
first-preference votes, as under ordinary proportional representation. Each of
the new systems (except for Binary STV, Ethnic-Roll AV and Ethnic-Area AV) is
essentially a list system, and as with ordinary List PR, there will probably be
little change from election to election in each party's number of seats, at least
once the parties have adjusted to the new system. If so, the system will
provide safe seats for nearly all the representatives of the party, that is,
for all except a few at the bottom end of its list. Thus the system may well be
acceptable on these grounds to most of the parties and most of the
representatives. It is also possible that a considerable number of politicians
will welcome the opportunity to change to a more pan-ethnic and inclusive
policy, without having to take the risk of losing seats by doing so.
But will the new system be acceptable to the ethnic group (or
groups) which have been dominant under the present system? Such a group will
still have influence under the new system, indeed, it will probably have the
major influence. Also, its politicians will still have a high probability of
obtaining office, but they will share it with the politicians from the
previously excluded minority. Thus it may well be that they will accept a
modest reduction in their influence, in exchange for the prospect of greater
security in a more stable and peaceful political system.
A final question is whether these systems are too specialised,
whether they are dedicated systems designed to deal with the problem of ethnic
division, at the cost of their performance in dealing with more normal issues.
The answer is that the systems are designed to deal with any division of
interest in the electorate, whatever its basis, and to give each party the
incentive to adopt a compromise policy responsive to the needs of each section
of electors. These new systems can be expected to exert over time on each party
and each candidate a steady incentive towards moderation, towards a policy
responsive to the needs of each group, and towards a centre position, not only
on the ethnic dimension, but also on any other dimension-such as the
left-wing/right-wing dimension-which is of importance to the electors. It is
therefore argued that the proposed new electoral systems, though designed for
situations of ethnic division, can provide stable and effective government,
responsive to each section of the electorate, both when ethnic divisions exist,
and in more normal situations where ethnic divisions are absent.
ENDNOTES
Note 1. The purpose of these provisions is as follows: (1) the
penalty on an association with too small a party, gives each party the
incentive to join in association with another party which is expected to get at
least two-thirds as many first-preference votes as it gets itself, and which
therefore is likely to be able to compete effectively with the first party for
lower-preference votes as well; (2) the penalty on not running candidates in
all areas, gives each party the incentive to stand as a nation-wide party,
placing its candidates in all areas of the country, and in particular,
competing with the other party in the association in all areas of the country
and for the votes of all ethnic groups; (3) the maximum gives a large party the
incentive to split (perhaps into two roughly equal parties which come together
to form one association), so that it can avoid the maximum limit of one-half on
an association's seats.
Note 2. By the Alternative Vote, the candidate with fewest
first-preference votes is excluded, and his or her votes are transferred, each
vote going to its next preference, if it expresses one. The candidate then with
fewest votes is similarly excluded, and his or her votes are transferred, and
so on, until one candidate has more than half the total votes held by all the
non-excluded candidates, and is elected.
Note 3. A Condorcet method of election is one which elects the
Condorcet winner, if there is one. The Condorcet winner is that party which,
when all parties are compared in two-party comparisons, is preferred to any
other party by a majority of voters. For example, suppose the percentages of
voters with each preference ordering are: 40 ACB, 21 CAB, 39 BCA. Then although
C has fewest first preferences, C will be the Condorcet winner, because C beats
A by 60 to 40, and beats B by 61 to 39.
Note 4. For example, if there are 600 seats, and the government
party has 175, then five extra seats are created, and given to it. Each
representative of the government party is given a weighted vote in the parliament,
equal to (425 x 55)/(180 x 45) = 2.886. Thus the government has 180 x 2.886 =
519 votes out of a total of 519 + 425 = 944, which is a 55% majority.
Note 5. An alternative version of Incumbency Condorcet, which
would have much the same properties, is as follows. If there is a CW, elect it.
Otherwise, elect that one of the parties in the cycle which beats the
incumbent, or (if more than one in the cycle beat the incumbent) which does so
by the largest margin.
Note 6. However, if the incumbent party is not the CW, but some
opposition party is, the system cannot guarantee that the CW will be elected.
This is shown by the following example, where N is the incumbent, and A and B
are two opposition parties, and where the percentages of votes are: 17 ABN, 19 ANB,
13 NAB, 17 NBA, 17 BNA, 17 BAN. Thus B is the CW, since B beats N by a margin
of 51 - 49 = 2, and B beats A by 51 to 49. But A is elected, since A beats N by
a larger margin, one of 53 - 47 = 6.
Similarly, the strategy-proofness which applies in respect of
the incumbent, does not apply in the case of the competition between opposition
parties. For example, if 3% of B supporters were to change from voting BAN to
voting BNA, this would reduce A's margin over N to 50 - 50 = 0, thus electing
B.
How far then does it create a problem, that there is at least a
possibility that, if the CW is an opposition party, it will not be elected, or
that an opposition party can be elected by strategic voting? In this context,
the following considerations should be borne in mind. First, strategy will be
possible only between opposition parties, where supporters of one opposition
party seek to get it elected instead of the other. Second, this strategy, or
this non-election of the CW, is possible only if there are two or more opposition
parties all of which beat the government party, which situation is not itself a
very likely one. Third, if in fact an opposition party which is not the CW is
elected to government, or if an opposition party is elected by strategic rather
than sincere voting, this will do little or nothing to distort governmental
incentives, since the only way that the new government party can get itself
re-elected, is itself to become the Condorcet winner before the next election.
Thus however the new government party won the last election, the need to win
the next election will still provide it with the required incentives, that is,
ones to be responsive to all sections of the electorate.
Note 7. For example, if the party has 51% of first preferences,
each of the votes giving it first preference is given a value of (35 x (100 -
51))/(51 x (100 - 35)) = 0.5173. The party will then have (51 x 0.5173)/((51 x
0.5173) + 49) = 35% of the first preferences. The intention of this provision
is to give any large party with significantly more than 35% of first
preferences the incentive to split itself into two smaller parties, to avoid
losing seats. This is likely to achieve a party system where the parties are
more equally competitive, and where most probably there are three major parties
rather than two.
Note 8. This can be shown as follows. The maximum total loss of
seats a party B can incur, that is, the maximum amount of seats which can be
transferred from party B to all other parties, is shown by the above formula
(2), but changed so that "all parties other than B" is substituted
for "party A". Thus B's maximum seats loss will be: three times B's %
of first preferences, multiplied by the total % of first preferences of all
parties other than B, divided by four times the total % of first preferences of
all parties other than B. This cancels out to: Three times B's % of first
preferences, divided by four. In other words, the maximum amount of seats a
party can lose by getting a low preference from the first-preference supporters
of other parties, is three-quarters of its initial allocation of seats, the
allocation in proportion to its first-preference votes. Thus whatever its
performance in respect of lower preferences, the party is guaranteed to get a
final percentage of seats which is at least one quarter of its percentage of
first-preference votes.
Return to top of
page
Return to home page
Please e-mail comments on this paper to David Chapman at: chapman@democdesignforum.co.uk,
or (if you are a member) to the Reinventing Democracy discussion group at: redemoc-L@newciv.org
Democracy Design Forum, Coles House,
Buxhall, Stowmarket, Suffolk
IP14 3EB, UK
Tel [int. 44] (0)1449 736 223
Fax [int. 44] (0)1449 612 274
e-mail: chapman@democdesignforum.co.uk