You are on page 1of 10

Lijphart & Grofman - Electoral Laws and their political consequences(2003)

CHAPTER

10

Degrees of Proportionality of
Proportional Representation Formulas
Avend

t is a well-established proposition in the literature on electoral systems that proportional representation (PR) is generally quite
cessful in achieving its principal goal - a reasonably proportional
translation of votes into seats - especially in comparison with plurality
and majority formulas. One of Raes (1971, 96) differential propositions is that proportional representation formulae tend to allocate
seats more proportionally than do majority and plurality formulae. It
is also known that different PR formulas are not equally proportional,
but students of PR disagree about which of the formulas are more and
which are less proportional.
The purpose of this analysis is to establish a rank order of the
principal PR formulas according to their degree of proportionality. I
shall use Blondels (1969, pp. 186-191) ranking of the
able vote, Sainte-Lague,
and largest remainder systems - the
most important attempt to rank order PR formulas undertaken so
far-as my preliminary hypothesis. In addition to the four formulas
analyzed by
I shall also try to include the two Imperiali
formulas in my ranking.
The degree of proportionality may be defined in terms of two elements. One is the degree to which the seat percentages of the different parties correspond to their vote percentages. The second is the
170

Proportionality of PR Formulas

272

degree to which large and small parties are treated equally. It is the
second element that provides a clear criterion for judging the proportionality of PR formulas, because deviations from proportionality are
not random: They tend to systematically favor the larger and to discriminate against the smaller parties.

Blondel's Ranking
Blonde1 (1969, p. 191) ranks four PR formulas in the following decreasing order of proportionality:
1. Single transferable vote (STV)
2. Sainte-Lague
3.
4. Largest remainders

STV is therefore the most, and the largest remainder method the
least, proportional formula, according to Blondel. Other, more limited, attempts to determine the proportionality of PR systems tend to
(1)ignore the STV method, (2) agree with Blondel's judgment that the
formula is less proportional than the Sainte-Lague formula,
and ( 3 ) disagree with Blondel's placement of the largest remainder
formula at the bottom of the list.
Loosemore and Hanby (1971)consider three of Blondel's four formulas, and they arrive at the following rank order:

1. Largest remainders
2. Sainte-Lague
3.
Their relative placement of Sainte-Lagueand
is in agreement
with Blondel's, but they conclude that the largest remainder formula
is the most proportional of the three. Rae (1971, p. 105) also finds that
the largest remainder method is more proportional than the other two
lumped together. Balinski and Young (1980) confirm the LoosemoreHanby finding that the Sainte-Lague formula, equivalent to the
ster method of apportionment of the U.S. House of Representatives,
yields more proportional results than the
formula, which is
the equivalent of the Jefferson method of apportionment.
Most of the literature confines itself to a
comparison of the
and largest remainder formulas. The consensus is that
disproportionally favors the larger parties and that the

172

Arend Lijphart

est remainder formula is more proportional and more favorable to the


smaller parties (Van den Bergh, 1955, pp. 24-26; Mackenzie, 1958,
78-80;
1974,
93-97; Berrington, 1975,
366-368;
Nohlen, 1978, pp. 77-80; Bon, 1978,
This consensus
deviates from Blondels ranking. In the next section, I shall show that
the consensus is right and that Blonde1 is wrong.

Comparing dHondt and Largest Remainders


A
comparison of the dHondt and largest remainder formulas is a good starting point for our exercise because it also supplies
us with the key we need for the ranking of the other PR methods.
How can we explain the different results of these two basic PR formulas?
The initial difficulty is that the dHondt and largest remainder formulas appear to use completely different methods for allocating seats
on the basis of the parties votes. Table 10.1gives a concrete illustration. The largest remainder formula first calculates the electoral quota
or quotient (often called the Hare quota): the total number of valid
votes cast
divided by the total number of seats in the district (s).
The parties votes are divided by this quota, and each party receives a
seat for every whole number in the result. The remaining seats are
then awarded to the largest of the unused remainders or remaining
votes. The dHondt formula does not require the calculation of an
electoral quota. As Table 10.1shows, each partys votes are divided by
the series of divisors 1, 2, 3, and so forth, and the seats are given
successively to the highest of the resulting values, usually referred to
as averages.
It is possible, however, to interpret the
highest average
formula in such a way that it becomes comparable to the largest remainder formula. The purpose of the dHondt formula may be said to
be the improvement of the largest remainder formula by finding an
electoral quota, lower than the Hare quota, which allows us to allocate
all of the seats exactly according to the largest remainder rule but
without having to take any remaining votes into account (Van den
Bergh, 1955,
68-72). This lower dHondt quota is equal to the last
of the averages to which a seat is awarded. In the example of Table
10.1, the dHondt quota is 14 votes. When the parties votes are divided by this quota, party A is entitled to 2 seats, to 2, C to 1, and
and E to none; all of the seats have been allocated and the remaining
votes can be ignored.

Proportionality of PR Formulas

TABLE 10.1. Hypothetical Example of the Operation of the Largest Remainder,


and Pure Sainte-Lague Formulas
a District
with 100 Votes, 5 Seats, and 5 Parties

Largest
Initial
Parties

Votes

Allocation
of Seats

Remaining
Votes

A
B

36
30
14
12
8

1
1
0
0
0

16
10
14
12
8

u12

u13

Total

18 (3)
15 (4)
7

12
10

2
2
1
0

D
E

Parties
A
B

36 (1)
30 (2)
14 (5)
12
8

C
D
E

Allocation of

Remaining Seats
1

0
1
1
0

Final Seat
Allocation
2
1
1
1
0

Pure
Parties
A
B
C
D
E

36 (1)
30 (2)
14 (3)
12 (5)

u13

u15

Total

12 (4)
10
4.67
4

7.2

2
1
1
1

electoral quota is 10015 = 20.


numbers in parentheses indicate the sequential order of the allocation of seats.

The reason for the disproportionality of


now becomes
clear. The remaining votes that it disregards are a relatively small
portion of the votes of the larger parties but a very large portion of the
small parties votes - and, of course, the entire vote total of a party that
does not receive any seats. As a result, the seat shares of the larger
parties will tend to be systematically higher than their vote shares,
and the smaller parties will tend to receive seat shares that are systematically below their vote shares. In contrast, the largest remainder
method treats large and small parties equally: The initial allocation of
seats is exactly proportional, and small and large parties compete for
the remaining seats on an equal basis.

174

Arend Lijphart

Sainte-Lague
Like dHondt, the Sainte-Lague formula is a highest average
method but, in Sainte-Lagues original proposal, the divisors are
the odd integers 1, 3, 5, and so forth, instead of the dHondt divisors
1, 2, 3, and so forth (for an English translation of his article, written in
1910, see Lijphart and Gibberd, 1977, pp. 241-242). Table 10.1 illustrates this method of translating votes into seats. The Sainte-Lague
formula can also be interpreted as a variant of the largest remainder
method in that it tries to find a quota that it considers more suitable
than the Hare quota. Unlike dHondt, however, it aims to be completely proportional and even-handed as between large and small
parties. The Sainte-Lague quota equals twice the last of the averages
to which a seat is given. In Table 10.1, for example, it is 24. For each
quota of votes the parties receive 1seat, and all remaining votes of half
a quota or more are also honored. Party A with 36 votes receives 1seat
for its first 24 votes and a second seat for its remainder of 12 votes,
which is exactly half of the Sainte-Lague quota. Party Cs remaining
votes are 14, more than half of the quota and hence good for a remaining seat.
The crucial difference between dHondt and Sainte-Lague is that the
latter does honor some of the remainders. If all remainders were
honored with a seat, a strong bias in favor of the small parties would
result-just as the dHondt rule of ignoring all remainders entails a
bias against the small parties. By setting a boundary of half a quota
above which remainders do, and below which they do not, qualify for
a seat, Sainte-Lague treats all parties in an even-handed manner. In
this respect, it resembles the largest remainder formula, and both
have to be regarded as equally proportional. In most cases, they also
yield exactly the same results (as in Table
but this is not necessarily always the case.
In practice, the Sainte-Lague formula is not used in its original and
pure form. The Scandinavian countries use a modified formula in
which the first divisor is raised to 1.4 in order to make it more difficult
for small parties to win a first seat. The divisor series 1.4, 3, 5, and so
forth, can be made comparable to the dHondt and pure Sainte-Lague
divisors by dividing each of these divisors by 1.4. The modified
Sainte-Lague divisor series can therefore also be stated as 1, 2.14, 3.57,
5.00, 6.43, and so forth. It is clear that the distance between these
divisors is greater than in the dHondt series but smaller than in the
pure Sainte-Lague series. The modified Sainte-Lague formula is therefore less proportional than the pure form-and hence also less

Proportionality of PR Formulas

tional than largest remainders -but more proportional than


.
It should be noted that, as Rosensweig (1981; see also Grofman,
1975, p. 316) has pointed out, this conclusion also applies to the
chance that a small party has to win its first seat. The 1.4 divisor
makes modified Sainte-Lague less proportional than pure
Lague, but it would have to be raised to 1.5 to be equivalent to
as far as the winning of the first seat is concerned. Hence Rae
(1971, 34) is mistaken when he argues that modified Sainte-Lague
entails a higher cost of the initial seat than
and Taylor and
Johnston (1979, 67) erroneously state that modified Sainte-Lague is
relatively more severe on small parties than
Elder (1975,
187) commits the same error.
The relative proportionality of the three PR formulas considered so
far can be summarized as follows, with the proviso that here and
henceforth Sainte-Lague signifies the modified formula as used in
Scandinavia:
1. Largest remainders
2. Sainte-Lague
3.

The Single-Transferable Vote


The difficulty of comparing STV with these three formulas is that in
STV systems the voters cast their votes for individual candidates instead of for party lists. There are two ways to solve this problem. One
is to assume that all voters will vote for the candidates of only one
party, so that intraparty, but no interparty, transfers will take place.
Under such conditions, the STV rules become virtually equivalent to
those of the largest remainder formula. The second solution is to look
at the votes in the final round of counting, which is the basis on which
the seats are allocated. Applying the largest remainder rule to this
final count will yield the same results in all but very exceptional circumstances.
There is one important difference between the largest remainder
formula and STV The former uses the Hare quota while the latter
normally uses the Droop quota, defined as the total number of votes
divided by the total number of seats plus
+ 1)-and then
usually rounded up to the next higher integer. The consequence of
using the Droop instead of the Hare quota as well as its relationship
with the
formula is stated succinctly in one of the oldest

176

Arend Lijphart

treatises on PR by Humphreys. After discussing the largest remainder


method, Humphreys writes:
The rule subsequently devised aimed at reducing the importance of remainders in the allotment of seats. The total of each list was divided by the
number of seats plus one. This method yielded a smaller quota than the
original rule and enabled more seats to be allotted at the first distribution.
The final improvement, however, took the form of devising a rule which
should so allot the seats to different parties that after the first distribution
there should be no seats remaining unallotted. This is the great merit of the
Belgian or
rule (Humphreys, 1911, 188).

Humphreyss comment makes clear that the Droop quota is lower


quota. The Droop
than the Hare quota but higher than the
quota and STV based on it are therefore less proportional than the
largest remainder method with the Hare quota, since it honors fewer
remainders, but more proportional than
since it usually
does honor some remainders. In order to rank it among the three
previously ranked formulas, we also have to compare it with SainteLague. Unfortunately, no unambiguous relative placement is possible
here; it depends on the numbers of parties, the number of seats, and
the sizes of the small parties. If there are small parties that would be
barred from representation by the
the Droop quota is more
favorable; it should be remembered that the 1.4-divisor makes SainteLague almost like
as far as the first seats are concerned. Once
the small parties have gained their first seats, however, Sainte-Lague
becomes more favorable to them, because it is, from this point on,
completely proportional.
With this reservation, we can now rank the four methods as follows:
1. Largest remainders
2. Single-transferable vote
3. Sainte-Lague
4.

overlapping

The Imperiali Formulas

Two other formulas - variously referred to as the Imperiali and


Imperial
have been discussed in the literature on electoral systems. One is the Imperiali largest remainder formula (Rae,
1971,
34-36; Grofman, 1975, 309; Wertman, 1977, pp. 45-47;
Brew, 1981,
and the other is the Imperiali highest average

Proportionality of PR Formulas

77

formula (Van den Bergh, 1955, p. 25; Nohlen, 1978,


78; Laakso,
1979, 162).
The distinctive feature of the Imperiali largest remainder method is
that it uses a quota that is lower than the Hare and Droop quotas. It
equals the total number of votes divided by the total number of seats
plus 2:
t 2). As it is used in Italy, its other special characteristic is
that the remaining votes and seats are not handled at the district level
but gathered into a national pool. For our purposes, it is important to
recognize that the Imperiali quota is lower, and hence less proportional, than the Droop quota but usually higher and more proportional than the
quota. It is incorrect to state, as Rae (1971, p.
34) does, that Imperiali is a variant of the largest remainder formula
intended to lower the price of the initial seats, helping weak parties.
Similarly, Wertman (1977, p. 45) is mistaken when he argues that the
(s t 2) denominator lowers the quota and thus increases the small
parties chances to gain seats in the Chamber of Deputies. As Carstairs (1980, p. 159) points out, the quota used in the 1948 and 1953
Italian elections was even lower than the Imperiali quota:
+ 3)
instead of
t 2). The change to the higher quota was made in
response to the demands of the smaller parties: The denominator
used for calculating the quota of votes necessary for election was
reduced from seats plus
to seats plus two.This meant that there
would be more remaining seats to be allocated.
A more difficult question is what the relative ranking of Imperiali
and Sainte-Lague is. The answer is analogous to the relationship that
we found between the Droop quota and Sainte-Lague. With regard to
the first seats, Sainte-Lague behaves very much like dHondt and is
therefore less proportional than Imperiali, although the Imperiali
quota is usually close to the dHondt quota. After the first seats have
been obtained, Imperiali obviously becomes much less proportional
than Sainte-Lague. In the overall ranking, the former should therefore
be placed below the latter, but with the proviso that there is some
overlap.
The second Imperiali formula is completely different from the first,
in spite of the similarity of the name. It is a form of highest averages
that looks deceptively similar to dHondt since it uses the divisors 2,
3, 4, 5, and so forth; the crucial exception is that the first divisor of 1is
omitted. In order to make these divisors comparable to the dHondt
divisors, they have to be divided by 2. The Imperiali divisors then
become 1, 1.5, 2, 2.5, 3, 3.5, and so forth. It is immediately clear that,
because the distance between the divisors is much smaller than in
the Imperiali highest average formula is considerably less

178

Arend Lijphart

proportional than
In the example of Table 10.1, Imperiali
would give party A 3 seats, instead of the 2 awarded under the
rule, and it would take away party Cs 1
Our final ranking, including the two Imperiali formulas, is as follows:
1. Largest remainders
2. Single-transferable vote
3. Sainte-Lague
4. Imperiali largest remainders
5.
6 . Imperiali highest averages

overlapping
overlapping

Notes
1. A formula that is also sometimes compared with the other PR formulas is the so-

called Danish method (Laakso, 1979, p. 162). It is a highest average formula that
and so forth. The distance between the divisors is greater
uses the divisors
than in the pure Sainte-Lague method, and the Danish method would therefore be
even more favorable to small parties. I have not included it in my ranking for two
reasons. (1)It would have to be ranked above the largest remainder formula because
it favors the small parties more. However, this characteristic derives not from the
basic proportionality of the formula, as in the case of the largest remainder and pure
Sainte-Lague formulas, but from the fact that the Danish method is disproportionally
favorable for small parties. (2) In Denmark, this method is not used for the translation of votes into seats: It has nothing to do with the allocation of supplementary
seats among parties. It is solely related to the geographic distribution within parties
over regions and constituencies (Johansen, 1979, 47).
2. There are two crucial qualifications that must be added to the above ranking of PR
formulas according to proportionality.
First, the proportionality of PR systems is not only a function of the kind of
formula that is used. A more important factor is the magnitude of the electoral
district. As James Hogan has forcefully pointed out, the decisive point in PR is the
size of the constituencies: the larger the constituency, that is, the greater the number
of members which it elects, the more closely will the result approximate to proportionality. On the other hand, the smaller the constituency, that is, the fewer the
number of members which it returns, the more radical will be the departure from
proportionality (Hogan, 1945, p. 13).
Second, although I find that the largest remainder formula is the most proportional of the six, my conclusion should not be read as meaning that it is therefore
also the best method. One may be in favor of PR without wanting to maximize
proportionality. A special disadvantage of largest remainders is that it may give rise
to the Alabama paradox: the phenomenon that a party would lose a seat if the total
number of seats available in the district would be increased (Brams, 1976,
137-166).This problem does not occur in any of the highest average systems, since
these allocate seats sequentially. If one should want to maximize proportionality
while avoiding the Alabama paradox, the pure Sainte-Lague formula is preferable to

Proportionality of PR Formulas

179

largest remainders. However, both methods suffer from a more serious weakness:
They may encourage party splits. In the example of largest remainders in Table 10.1,
party B could win an additional seat, at the expense of party D, if it would present
two separate lists, each of which would get 15 votes. Pure Sainte-Lague would have
the same effect. This tendency declines as PR formulas become less proportional,
rule-certainly a very powerful arguand it disappears entirely under the
ment in favor of this not maximally proportional formula (cf. Balinski and Young,
1983).

You might also like