Content uploaded by Bernard Grofman
Author content
All content in this area was uploaded by Bernard Grofman on Dec 24, 2014
Content may be subject to copyright.
PROOF COVER SHEET
Author(s): Jon Fraenkel
Article Title: The Borda Count and its real-world alternatives: Comparing scoring rules in Nauru and
Slovenia
Article No: CAJP900530
Enclosures: 1) Query sheet
2) Article proofs
Dear Author,
1. Please check these proofs carefully. It is the responsibility of the corresponding author to check
these and approve or amend them. A second proof is not normally provided. Taylor & Francis
cannot be held responsible for uncorrected errors, even if introduced during the production process.
Once your corrections have been added to the article, it will be considered ready for publication.
Please limit changes at this stage to the correction of errors. You should not make trivial changes,
improve prose style, add new material, or delete existing material at this stage. You may be charged
if your corrections are excessive (we would not expect corrections to exceed 30 changes).
For detailed guidance on how to check your proofs, please paste this address into a new browser
window: http://journalauthors.tandf.co.uk/production/checkingproofs.asp
Your PDF proof file has been enabled so that you can comment on the proof directly using Adobe
Acrobat. If you wish to do this, please save the file to your hard disk first. For further information
on marking corrections using Acrobat, please paste this address into a new browser window: http://
journalauthors.tandf.co.uk/production/acrobat.asp
2. Please review the table of contributors below and confirm that the first and last names are
structured correctly and that the authors are listed in the correct order of contribution. This
check is to ensure that your name will appear correctly online and when the article is indexed.
Sequence Prefix Given name(s) Surname Suffix
1 Jon Fraenkel
2 Bernard Grofman
Queries are marked in the margins of the proofs, and you can also click the hyperlinks below.
AUTHOR QUERIES
General points:
1. Permissions: You have warranted that you have secured the necessary written permission from
the appropriate copyright owner for the reproduction of any text, illustration, or other material in
your article. Please see http://journalauthors.tandf.co.uk/permissions/usingThirdPartyMaterial.asp.
2. Third-party content: If there is third-party content in your article, please check that the
rightsholder details for re-use are shown correctly.
3. Affiliation: The corresponding author is responsible for ensuring that address and email details
are correct for all the co-authors. Affiliations given in the article should be the affiliation at the
time the research was conducted. Please see http://journalauthors.tandf.co.uk/preparation/
writing.asp.
4. Funding: Was your research for this article funded by a funding agency? If so, please insert
‘This work was supported by <insert the name of the funding agency in full>’,followedbythe
grant number in square brackets ‘[grant number xxxx]’.
5. Supplemental data and underlying research materials: Do you wish to include the location
of the underlying research materials (e.g. data, samples or models) for your article? If so,
please insert this sentence before the reference section: ‘The underlying research materials for
this article can be accessed at <full link>/ description of location [author to complete]’. If your
article includes supplemental data, the link will also be provided in this paragraph. See
<http://journalauthors.tandf.co.uk/preparation/multimedia.asp> for further explanation of
supplemental data and underlying research materials.
6. The CrossRef database (www.crossref.org/) has been used to validate the references.
Mismatches will have resulted in a query.
QUERY NO. QUERY DETAILS
AQ1 Please check whether this shortened title is correct.
AQ2 Please check whether the inserted abbreviation “PR”is correct.
AQ3 Please check whether the inserted abbreviation “SNTV”is correct.
AQ4 The reference “Black, 1968”is cited in the text but is not listed in the references list.
Please either delete in-text citation or provide full reference details.
AQ5 The sentence “….if we were electing a large enough set of candidates …” has been
changed. Please check whether the change conveys the intended meaning and amend if
necessary.
AQ6 Figure [3] has not been cited in the text. Please confirm whether the inserted citation is
correct. If not, please delete the citation and insert a citation at the appropriate place in
the text.
AQ7 The reference “Tabarrok and Spector, 1999”is listed in the references list but is not
cited in the text. Please either cite the reference or remove it from the references list.
How to make corrections to your proofs using Adobe Acrobat/Reader
Taylor & Francis offers you a choice of options to help you make corrections to your proofs. Your
PDF proof file has been enabled so that you can edit the proof directly using Adobe Acrobat/Reader.
This is the simplest and best way for you to ensure that your corrections will be incorporated. If you
wish to do this, please follow these instructions:
1. Save the file to your hard disk.
2. Check which version of Adobe Acrobat/Reader you have on your computer. You can do this by
clicking on the “Help” tab, and then “About”.
If Adobe Reader is not installed, you can get the latest version free from http://get.adobe.com/reader/.
3. If you have Adobe Acrobat/Reader 10 or a later version, click on the “Comment” link at the
right-hand side to view the Comments pane.
4. You can then select any text and mark it up for deletion or replacement, or insert new text as
needed. Please note that these will clearly be displayed in the Comments pane and secondary
annotation is not needed to draw attention to your corrections. If you need to include new sections
of text, it is also possible to add a comment to the proofs. To do this, use the Sticky Note tool in the
task bar. Please also see our FAQs here: http://journalauthors.tandf.co.uk/production/index.asp.
5. Make sure that you save the file when you close the document before uploading it to CATS using
the “Upload File” button on the online correction form. If you have more than one file, please zip
them together and then upload the zip file.
If you prefer, you can make your corrections using the CATS online correction form.
Troubleshooting
Acrobat help: http://helpx.adobe.com/acrobat.html
Reader help: http://helpx.adobe.com/reader.html
Please note that full user guides for earlier versions of these programs are available from the Adobe
Help pages by clicking on the link “Previous versions” under the “Help and tutorials” heading from
the relevant link above. Commenting functionality is available from Adobe Reader 8.0 onwards and
from Adobe Acrobat 7.0 onwards.
Firefox users: Firefox’s inbuilt PDF Viewer is set to the default; please see the following for
instructions on how to use this and download the PDF to your hard drive: http://support.mozilla.org/
en-US/kb/view-pdf-files-firefox-without-downloading-them#w_using-a-pdf-reader-plugin
The Borda Count and its real-world alternatives:
Comparing scoring rules in Nauru and Slovenia
JON FRAENKEL
Victoria University of Wellington
BERNARD GROFMAN
University of California
This article examines strategic elements of voter behaviour in parliamentary
elections where the voting method is a scoring rule other than plurality: the
Borda Count, which is used for the election of ethnic minorities in Slovenia,
and the Dowdall rule, which is used in the Pacific island state of Nauru in
multi-seat districts. After first examining the general properties of scoring rules,
and generating theoretical differences between the two rules, we look at
empirical evidence from Nauru and Slovenia. This casts a doubt on predictions
based simply on a voting rule’s mathematical properties and on the accuracy of
assumptions of sincere rank ordering.
Keywords: electoral systems; Borda Count; preferential voting systems; Nauru;
Slovenia
Introduction
Political scientists have long been concerned about principles of social choice, and
have investigated alternative rules governing the translation of individual preferences
into group decisions (Black 1958; McLean and Urken 1995). Many have preferred
methods that allow fuller expression of preferences than is the case with simple cat-
egorical systems such as single-member plurality (first-past-the-post) or closed-list
proportional representation (PR AQ2). The ‘method of marks’, invented by 18th century
CE: SPA QA: COLL:
Jon Fraenkel is Professor of Political Science and International Relations, Victoria University, Welling-
ton. Bernard Grofman is a professor of Political Science at the University of California, Irvine. We are
indebted to Nauru MP Kieren Keke, University of the South Pacific Centre Director Alamanda Lauti,
and former Nauru Parliamentary Counsel, Katy Le Roy, for assistance in obtaining the electoral data
for Nauru; to the National Electoral Commission of Slovenia and to Jure Toplak for assistance in obtain-
ing electoral data for, and other information regarding, the Hungarian and Italian constituencies in Slo-
venia; and to the Australian National University Cartographic Services for constructing our maps of
Slovenia and Nauru. Work done on this project by the second named author was supported by the
Jack Peltason Chair and the Center for the Study of Democracy at the University of California,
Irvine. During the preparation of an earlier draft he served as a Straus Research Fellow at the Straus Insti-
tute for Advanced Study in the Law, New York University Law School.
Australian Journal of Political Science, 2014
http://dx.doi.org/10.1080/10361146.2014.900530
5
10
15
20
25
30
35
40
45
CAJP900530 Techset Composition India (P) Ltd., Bangalore and Chennai, India 3/17/2014
© 2014 Australian Political Studies Association
French mathematician Jean–Charles de Borda has aroused particular interest, both
because it permits a full ranking of voter preferences and because of its method of
determining victors (Borda 1781; Saari 1994;1995; Young 1974). Unlike some
other preferential voting systems, Borda’s method takes into account all voter prefer-
ences, allotting a value to each, and establishes victors by a simple tallying of the total
each candidate obtains. Despite considerable theoretical interest in such systems,
real-world use of the Borda Count has been restricted –in part because of concerns
about potential strategic voting. Electoral reformers more commonly advocate other
electoral systems making use of ranked preferences, such as the alternative vote (AV)
(in the USA, promoted under the name ‘the instant runoff’) or the single-transferable
vote (STV).
There are, however, two countries that use sophisticated non-elimination-based
preferential voting systems. The classical Borda method is used for national elections
to special reserved legislative seats for Hungarian and Italian ethnic minorities in Slo-
venia. The small Pacific Island state of Nauru, since 1971, has used a unique points
system named after its inventor –the Dowdall method,
1
that can be considered a rela-
tive of the Borda rule, though it has quite distinctive features and effects.
2
This
method has been widely described as a ‘modified’form of the Borda Count
(Golder 2005: 109; IFES 2013; Reilly 2002; Reynolds et al. 2005), but it has distinc-
tive properties and, in practice on Nauru, generates different outcomes to the Borda
Count. One key difference between the Borda Count and the Dowdall method is that,
for Borda, the scores allotted to each candidate vary with the number of candidates,
whereas with the Dowdall method a first preference is always worth one, a second
preference half, a third preference one-third and so on.
Others have pointed to the existence of these forms of preferential voting in Nauru
and Slovenia, but until now no data has been available to permit investigation of how
such systems work in practice. The purpose of this article is to examine how these two
different scoring rules operate in legislative elections in Nauru and Slovenia, and to
compare results under the different systems. We are particularly concerned to ascer-
tain how the potential manipulability of such systems (an issue regularly raised in the
theoretical discussions) plays out in practice, and whether real-world resort to stra-
tegic voting amounts to a formidable argument against such rules.
We examine results from nine legislative elections on Nauru, occurring between
1997 and 2013, and outcomes in the two minority districts in four Slovenian elections
from 2000 until 2011. We compare actual outcomes under Dowdall (or Borda) to
simulated outcomes with the same preference rankings under the other rule, and
we compare these rules to two other voting methods whose results we can recreate
from aggregate-level preference data on elections in these two countries: plurality
voting in either single-seat constituencies (for Slovenia) or multiple-seat constituen-
cies for Nauru, and the single non-transferable vote (SNTV AQ3) for the multi-seat
constituencies in Nauru.
1
The country’s Secretary for Justice, an Irishman, Desmond Dowdall, devised the system in 1971. The
Nauru government in 1971 indicated a preference that its system be described as the ‘Dowdall system’
instead of the previously used term ‘exhaustive ballot system’(cited in Supreme Court of Nauru 1977).
2
A similar method was also used (until 2002) for the parliamentary selection of nominees for presidential
elections in another small Pacific Island nation, Kiribati (Reilly 2001a,2001b,2002; Reilly and
Gratschew 2001; Van Trease 1993).
2J. FRAENKEL AND B. GROFMAN
50
55
60
65
70
75
80
85
90
Theoretical work on the Borda Count and the potential for manipulation
Social scientists have differed about the merits of practical use of the Borda Count.
Saari (1994;1995) has been the leading contemporary academic exponent of the
Borda rule as the most desirable way to make collective choices. Saari’s arguments
are several, but arguably the most important is that it guarantees a transitive ordering
of preferences, and thus eliminates voting paradoxes. Others have suggested that the
Borda rule is likely to choose an alternative that is highly regarded by a substantial
number of voters, and that this system ‘reduce[s] the chances of divisive candidates’,
and favours candidates ‘occupying moderate positions’or at least ones who do not
evoke either strong hostility or strong support (Dummett 1997: 161; 1998: 290,
292; Emerson 2013).
3
Moreover, simulations show it to have a high probability of
choosing the Condorcet winner AQ4when one exists,
4
but it is also without irresoluteness
in the absence of a Condorcet winner (Merrill 1984). And there are formal results spe-
cifying the (relatively plausible) conditions under which the Borda winner and the
Condorcet winner will coincide (Tangian 2013).
On the other hand, several scholars have worried about the potential for strategic
manipulation under the Borda Count by deliberate misrepresentation of preferences.
For this reason, Jean–Charles de Borda himself famously commented that his system
was only suitable for ‘honest men’(Black 1958: 182). In the 18th century, the Académie
Française, of which Borda was a prominent member, experimented with this system in
its deliberations on internal matters. Its archives show that ‘the voters found how to
manipulate the Borda rule: not only by putting their most dangerous rival at the
bottom of their lists, but also by truncating their lists’(McLean and Urken 1995:40).
Duncan Black, whose 1958 Theory of Committees and Elections did much to revive
modern interest in Borda’s electoral thought, warned that the practical application of
such systems might prove troublesome: ‘even to the unsophisticated voter, the Borda
count is an invitation to strategic voting …It would be exceedingly dangerous to
employ any scheme with this property, in elections to parliament or congress’(Black
1976: 15). Even enthusiasts for the Borda Count have acknowledged that ‘there is a
danger in adopting [Borda’s] system for parliamentary and local elections when we
cannot be confident that sincere voting will be the general rule’(Dummett 1997:86).
Consider the following example. Table 1 assumes four voters and four candidates:
a, b, c and d. Voter 1 backs the four candidates in the order abcd. Voter One prefers
bacd, and under sincere ranking his first choice bwins. If Voter 1 instead strategically
ranks adcb, by shifting bto last place, he ensures victory for his first choice candidate.
3
The reasons for adopting the Borda Count in Slovenia were connected to its potential for generating
election of a more moderate candidate:
There are two or three interest groups in each of the two national minorities (Italian, Hungarian). If
plurality or majority system were used, these groups would confront each other and one would win
over the other. With Borda, usually the winner is the person who is most acceptable to all and who
is not an extremist. (Jure Toplak, personal communication, 28 February 2013)
4
The Condorect winner, also known as the majority winner, is that alternative, if any, which can defeat
all others in a paired contest (or, for an even number of voters, it is that alternative which cannot be
defeated by any other alternative in a paired contest). When it exists, many authors (Black 1968)
regard the Condorcet winner as the normatively desired choice. Black, however, proposed that the
Borda winner be chosen when no Condorcet winner existed. Nanson (1882) proposed a complex
scheme based on the Borda rule that has the property that it always yields the Condorcet winner
when one exists (see McLean and Urken 1995:57–60; Tangian 2013: 203).
THE BORDA COUNT AND ITS REAL-WORLD ALTERNATIVES AQ1
3
95
100
105
110
115
120
125
130
135
In total, aand bhave the same number of both first and second preferences: with
sincere voting, b’s third preference from Voter 4 trumps a’s fourth preference from
Voter 3, but that advantage vanishes if Voter 1 votes strategically. Under the non-
strategic ranking, the Borda Count is a= 7 and b= 8, whereas if Voter 1 votes
strategically, the outcome is a= 7 and b= 6. To achieve this end result deliberately,
Voter 1 must be aware of the preferences of all other voters, and of the workings of
the electoral system.
Other types of strategic manipulation are possible under the Borda Count in
addition to strategic manipulation of revealed preferences. Since the scores attributed
to each preference depend on the number of contestants, political strategists may field
‘dummy’or ‘red herring’candidates in an effort to manipulate outcomes. In our
hypothetical example above, third and fourth preferences decide between the top
two candidates. Hence, it might be possible to strategically sponsor other candidates
that acquire those preferences and so devalue some points earned by one or other of
these top two contenders by introducing what we might call ‘weak clones’. For this
reason, the Marquis de Condorcet, a contemporary of Borda, argued that the Borda
Count ‘relies on irrelevant factors to form its judgments’and was consequently
‘bound to lead to error’(Condorcet cited in Young 1995: 57).
Before we examine the operation of such electoral systems in real-world settings,
and how strategic voting operates in these contexts,
5
we need to explore the broader
properties of scoring rules.
Defining weights for scoring rules
The Borda Count and the Dowdall method, along with the far better known plurality
rule (first-past-the post), are examples of what the social-choice literature calls
scoring rules (Fishburn 1973; Saari 1994;1995). A scoring rule operates over a
set of ballots in which each voter provides a (partial or complete) ranking of a set
of nalternatives. Each scoring rule can be identified with a vector of ranking
weights (w
1
,w
2
,…,w
n
) where w
i
is the weight attached to an alternative located
at the ith rank by any given voter. These values must be monotonically
Table 1. Hypothetical scoring under sincere and strategically chosen preference ranking
Voters
1234
Preferences Points
Voter 1
(strategic)
Adjusted
points
a1242 7 1 7
b2123 8 4 6
c3314 5 3 6
d4431 4 2 5
Note: Example entails a first preference worth 3 and last preference worth 0.
5
There has been a large body of empirical work measuring strategic voting in real-world single-seat plur-
ality contests in countries such as Canada and the UK, and for some elections, including party primaries,
in the USA, but little empirical work has been done measuring strategic voting in multi-seat systems
(see, however, Irwin and Van Holsteyn 2012).
4J. FRAENKEL AND B. GROFMAN
140
145
150
155
160
165
170
175
180
non-decreasing if ties are allowed and monotonically decreasing if ties are not per-
mitted, which is the case for the rules used in Slovenia and Nauru. For any scoring
rule, the cumulative score for the ith alternative is given by ∑w
ij
, which is the
weighted sum over all voters, j, corresponding to the rank each voter has assigned
to the ith alternative.
6
Plurality is the simplest scoring rule, since voters only indicate
their first choice, that is, the first item in their rank ordering of alternatives. All scoring
rules, including Dowdall as well, are vulnerable to Condorcet’s criticism (discussed
above) that they fail to satisfy the principle of ‘independence of irrelevant
alternatives’.
If we are choosing among nalternatives, a standard way of representing the Borda
rule is as the vector of weights (n–1, n–2, …, 2, 1, 0). With the Dowdall rule, in con-
trast to the Borda rule, a first preference is worth 1, a second preference ½, a third
preference ⅓, a fourth preference ¼ and so on. In other words, for the Dowdall
rule, we have a weighting vector of (1, ½, ⅓,…,1/n).
7
In plurality, we have the
weights vector (w
1
,0,0,0,…, 0), with w
1
> 0, that is, only first preferences are
given any weight.
Implementation options
Scoring rules such as Borda and the Dowdall method can be implemented either by
allowing all voters to rank as many candidates as they choose, or by requiring voters
to rank order only a fixed number of candidates –to cast a fixed truncated ballot –or
by requiring voters to rank order all candidates. On the island of Nauru, if there are n
candidates, it is compulsory for voters to assign a unique rank to each candidate, from
1ton: that is, to record as many preferences as there are candidates. Otherwise the
ballot is discarded as invalid.
Borda, the Dowdall rule and plurality can each be used either for selecting a single
alternative from among several, or for selecting Malternatives, M>1, with the M
alternatives receiving the highest weighted vote being chosen. It is most common
to think of plurality in the context of single-member plurality elections, but we can
also have plurality-based outcomes in a multi-seat district, with the most common
form of such an electoral rule called plurality bloc voting, in which each voter
may cast a single vote for M(or up to M) candidates. The potential for a Borda
Count to operate effectively either in single-member or multi-member districts is
one reason why it has been considered for local-government elections in Australia’s
Northern Territory (Reilly 2011: 23, 28). The Slovenian uses of the Borda rules are
for single-seat elections, but the Dowdall rule in Nauru is used in districts either with
two, three or four seats.
6
For example, if there were three alternatives and three voters, then if the weight vector for the scoring
rule were (1, ¾, 0) and Voter 1 ranked the alternatives abc, Voter 2 ranked the alternatives abc, and
Voter 3 ranked the alternatives cba, the scores corresponding to each alternative would be a= 2 (2 ×
1+1×0),b= 9/4 (3 × ¾) and c= 1 (1 × 1 + 2 × 0), which gives us the ranking of bac. For the
same example, if the weights were (1, ½, 0), the scores corresponding to each alternative would be a
= 2 (2 × 1 + 1 × 0), b= 1.5 (3 × ½) and c= 1 (1 × 1 + 2 × 0), which gives us the ranking of abc. For
the same example, if the weights were (1, ¼, 0), the scores corresponding to each alternative would
be a=2(2×1+1×0),b= ¾ (3 × ¼) and c= 1 (1 × 1 + 2 × 0), which gives us the ranking of acb.
7
In its original 1971 formulation, the Nauru system used the fractions 1, ½, ⅓, etc., but this was sub-
sequently converted to a decimal basis. For mathematical tractability in deriving exact analytic
results, we begin with the fractional representation, and then report results as decimals.
THE BORDA COUNT AND ITS REAL-WORLD ALTERNATIVES AQ1
5
185
190
195
200
205
210
215
220
225
Theoretical comparisons among scoring rules
For a fixed n, there are an infinite number of different ways of representing the vector
of weights for any given scoring rule, since any two sets of weights that may be rep-
resented as linear transformations of one another can be shown to be equivalent (Saari
1994;1995).
8
If we wish to compare weights across scoring rules, we need a common
way to represent these rules. There are two obvious approaches. First, for any given
scoring rule, we might look at the representation of the weights vector for that scoring
rule that has the property that all weights sum to one. Second, following Saari, we
could look at the representation of weights that has the property that the first-
ranked alternative is scored one and the last-ranked alternative is scored zero.
We can always find such representations. For example, the plurality rule can be
represented as the vector (1, 0, 0, 0, …, 0). This vector has the property that its
entries sum to one and the property that the first entry is one and the last entry is
zero. Now consider the Borda Count. The sum of the weights in its vector of (n–1,
n–2, …,2,1,0)is(n)(n–1)/2. Consider n= 3, for which the sum of these weights
is 3. We may thus represent the Borda vector for n= 3 as either (2, 1, 0) or as (⅔,
⅓, 0), so that its values sum to one. Moreover, we may take the latter representation
and divide by ⅔to get the representation (1, ½, 0), which has the property that the first
value is one and the last is zero. To find alternative weights representations for the
Dowdall rule for n=3, we first find that ∑1/n=1+½+⅓= 11/6. Now, instead
of (1, ½, ⅓), we can specify the weights vector for the Dowdall rule as (6/11, 3/
11, 2/11), which sums to one. Then, we can subtract 2/11 from all entries to get
the new vector (4/11, 1/11, 0), and then normalise by dividing though by 4/11 to
get the vector (1, ¼, 0) in which the first entry is one and the last entry zero.
In this latter representation, comparing the vectors (1, ½, 0) and (1, ¼, 0), it is
apparent that, at least for n= 3, the Borda rule gives more relative weight to
second-place preferences than does the Dowdall rule. This turns out to be true for
all values of n. The Borda Count weights relative preferences differently depending
on the number of candidates, however, according to a principle that ensures that the
gap between each weight is equal. If there are three alternatives, a first-place vote
under Borda is worth 2 and a second-place vote is worth 1, while if there are four
alternatives a first-place vote is worth 3 and a second-place alternative is worth 2,
and so on, but the last-place vote is always given as zero weight.
9
By contrast, as
noted earlier, the Dowdall rule is not dependent on how many candidates there are,
since the weight it gives to any rank is fixed.
One consequence of this difference between the two formulas is that nincreases the
relative importance of the weight attached to the first-ranked alternative relative to
the weight attached lower-ranked alternatives under the Borda rule, but not under
the Dowdall rule. For example, the first-ranked alternative relative to the weight
attached to the second alternative decreases under the Borda rule from a ratio of 2
to 1 for n= 3 to, for example, a ratio of 5 to 4 for n= 6; with a limiting value of
n/(n–1), which goes to 1 as ngoes to infinity. In contrast, the ratio between the
8
A similar non-uniqueness result is found when we look at weight vectors for weighted voting rules (Fel-
senthal and Machover 1998).
9
As noted earlier, there are many alternative ways to assign Borda weights, and Borda himself used a
different scheme from the ‘modern’instantiation of the Borda Count (Emerson 2013).
6J. FRAENKEL AND B. GROFMAN
230
235
240
245
250
255
260
265
270
weight given by each voter to first-ranked and second-ranked candidates stays the
same under the Dowdall rule: 2 to 1.
There are several useful ways to look at the hypothetical comparisons between the
two rules. One thing we can do is to calculate the sum of the (absolute) weighting
differences between the two rules as we vary n. That value, along with the average
value of the discrepancy is shown in Table 2.
For the sum of the total (absolute) differences between the two rules, except for the
comparisons between value for n= 3 and n= 4, differences monotonically increase
with n. In contrast, when we look instead at average (absolute) differences
between the two rules, since we are steadily increasing the divisor, we find that the
average discrepancy between the two rules will go to zero as nincreases. In other
words, if AQ5we were electing a large enough set of candidates the differences in rankings
between Borda and Dowdall would cease to be important. For the small district mag-
nitudes, we observe in Nauru, however, differences are still quite substantial.
Another important comparison is the following: as nincreases, the relative weight
Borda attaches to first-place preferences shrinks as compared to the weight the
Dowdall rule attaches. As Figure 1 indicates, where there are three candidates, the
weights attached to the first- and second-placed positions remain constant (at
1.22). Where there are four candidates or more, this relative weight under Borda
rises but then falls.
Having established these theoretical differences, we are now interested in how
these rules might differ when applied to actual elections. There is a great deal of
work done showing the extent to which, under particular distribution assumptions,
voting rules are likely to yield similar or identical outcomes (Merrill 1984;
Tangian 2013: Chap 4). We are concerned, however, that assumptions often used
in such analytic or simulation results, the uniform distribution, for example, or sym-
metric distributions such as the univariate or multivariate normal, bear little resem-
blance to the distribution of ballots we observe in real-world elections, which tend
to be skewed and often rather sharply peaked (Feld and Grofman 1992; Tsetlin,
Regenwetter, and Grofman 2003). Seminal as such analytic results are, showing
how to find preference distributions that have the property that they will give any
desired pattern of disagreement across common voting rules, including scoring
rules (Saari 1990a;1994;1995), does not tell us how likely such disagreements
are to be in practice.
Elections in Nauru and Slovenia using the Dowdall rule and the Borda Count
Nauru, 1997–2013
The Republic of Nauru consists of only a single island, covering eight square miles. It
is located close to the equator in the midst of the Pacific Ocean, roughly equidistant
Table 2. Comparisons of the total and average (absolute) differences between the Borda rule and the
Dowdall rule for various values of n
N3 4 5 6 7 8 9 10 11 12 13 14
Absolute discrepancy .36 .24 .27 .32 .34 .36 .37 .38 .39 .43 .43 .45
Average discrepancy .12 .06 .054 .053 .049 .045 .042 .038 .036 .036 .033 .032
THE BORDA COUNT AND ITS REAL-WORLD ALTERNATIVES AQ1
7
275
280
285
290
295
300
305
310
315
from Hawaii and Australia. The island was brought under German control in 1888,
but with the onset of World War I, became a League of Nations-mandated territory
under Britain, New Zealand and Australia, but in practice administered by Australia.
It was occupied by the Japanese during the Pacific War and, after the conclusion of
hostilities, became a United Nations Trust territory, again under Australia, until it
obtained independence in 1968. Phosphate exports underpinned the island’s
economy for most of the 20th century, but the industry is now close to exhaustion.
Over recent years, Nauru has become the location of a detention centre for asylum
seekers who try to enter Australia by boat.
Nauru has a parliamentary system, with a president selected through the legisla-
ture, rather than by direct popular election. Since independence, voters have cast a
rank-ordered ballot in one of eight multi-member constituencies, based on Nauru’s
14 traditional districts or (in the case of Ubenide, Anabar and Anetan) amalgamations
of these (Figure 2). Until 2010, seven constituencies returned two members to parlia-
ment, and one constituency, Ubenide, returns four members. As of the 2013 election,
an extra member was added to the two-member Meneng constituency. In total, the
Nauru parliament now has 19 members, up from 18 during the period 1968–2010.
The rule used to tally these rank-ordered ballots has changed over time. STV was
used for the country’sfirst election, in 1968, also in two- and four-member districts.
10
Figure 1. Ratio of the kth place weight under the Borda scoring rule to the kth place weight under the
Nauru scoring rule for various values of n
10
Reilly (2002: 154, 366) and Reilly and Gratschew (2001: 699) report use of the AV in the 26 January
1968 election. They draw on the Nauru Supreme Court’s1977 decision which mentions the 1965 Elec-
toral Act, which provided for use of the AV both in single-member by-elections and, at general elections,
in the two- and four-member constituencies (Supreme Court of Nauru 1977). Available electoral data for
1968 records quotas, however, which are clearly STV quotas calculated for each constituency. The
records also note that the election was conducted in accordance with Australia’sCommonwealth Elec-
toral Act 1918–66. Since independence came on 31 January 1968, that is, four days after the election, we
believe that the election was conducted under the Australian law using STV, perhaps because of an
awareness that this is a better method of handling elections in constituencies where M>1 than via
usage of multi-member AV. Australia had a negative experience with use of multi-member AV
8J. FRAENKEL AND B. GROFMAN
320
325
330
335
340
345
350
355
360
Since 1971, as noted earlier, the voting rule used in Nauru has been the inverse to the
rank-scoring rule, called the Dowdall rule, and there has been a requirement for full
ranking in order to cast a valid ballot. We will report data from elections held in 1997,
2000, 2003, 2004, 2007, 2008, April 2010 and a second election in June of that year
Figure 2. Electoral and traditional districts on Nauru
Notes: Figures in brackets show the number of seats per district.
during the interwar years (Lijphart 1997). The Dowell system was introduced by way of the Electoral
(Electoral System) Regulations 1971, gazetted 22 January. The new system was used for the general
elections of 23 January 1971 (Supreme Court of Nauru 1977).
THE BORDA COUNT AND ITS REAL-WORLD ALTERNATIVES AQ1
9
365
370
375
380
385
390
395
400
405
and 2013.
11
In these years, as Table 3 indicates, the number of candidates contesting
elections in Nauru ranged from 5 to 22 with a mean per seat of 4.0 (and a standard
deviation of 9.5). There exists no party system, but sharp competition prevails
between the parliamentary factions over the capture of government.
Whereas Nauru uses the Dowdall rule in multi-member constituencies, the Slove-
nian system uses a classical Borda system in single-member districts.
Slovenia, 1992–2011
The Republic of Slovenia is located between Italy, Austria, Hungary and Croatia. It
has a population just over 2 million spread across just over 20,000 km
2
of land area.
Slovenia broke away from Yugoslavia in 1991, and joined the European Union in
2004. The country has a parliamentary system, but with a popularly elected president.
Since 1992, the National Assembly has had 90 elected members, 88 of whom are
selected by a list PR system from eight 11-member constituencies using a 4 per cent
threshold and a four-year electoral cycle.
12
The remaining two members, representing
Table 3. Correlations between outcomes and rankings under the Dowdall system and simulated
alternative systems in Nauru, 1997–2013
1997 2000 2003 2004 2007 2008
2010
April
2010
June 2013 Average
Candidates (No.) 61 72 84 73 79 65 86 60 68 4.0
Invalid votes (%) 3.4 4.0 4.7 4.1 4.4 3.6 4.8 3.3 3.6 4.0
Disagreements between outcomes (percentages of total)
Dowdall/Borda 38.9 22.2 55.6 33.3 22.2 16.7 38.9 16.7 26.3 30.1
Borda/PBV 33.3 22.2 55.6 27.8 22.2 16.7 38.9 22.2 21.1 28.8
Borda/SNTV 44.4 33.3 55.6 38.9 27.8 22.2 50.0 27.8 26.3 36.2
Dowdall/PBV 16.7 5.6 5.6 0.0 0.0 0.0 0.0 5.6 10.5 4.9
Dowdall/SNTV 5.6 16.7 0.0 5.6 16.7 5.6 11.1 11.1 0.0 8.0
Correlations between outcomes
Dowdall/Borda 0.48 0.70 0.29 0.56 0.71 0.77 0.51 0.76 0.75 0.62
Borda/PBV 0.55 0.70 0.22 0.56 0.71 0.77 0.51 0.76 0.75 0.62
Borda/SNTV 0.40 0.56 0.29 0.48 0.71 0.77 0.51 0.68 0.67 0.56
Dowdall/PBV 0.78 0.93 0.93 0.93 1.00 1.00 1.00 0.84 0.92 0.92
Dowdall/SNTV 0.93 0.78 1.00 0.93 1.00 1.00 1.00 0.92 1.00 0.95
PBV/SNTV 0.70 0.85 0.93 0.85 1.00 1.00 1.00 0.84 0.92 0.90
Correlations between rankings
Dowdall/Borda 0.60 0.73 0.76 0.53 0.77 0.76 0.68 0.81 0.87 0.70
Borda/SNTV 0.53 0.62 0.57 0.43 0.63 0.64 0.44 0.70 0.76 0.57
Dowdall/SNTV 0.97 0.96 0.90 0.96 0.94 0.91 0.88 0.92 0.94 0.93
Notes: Ties for second place under Borda in Buada in April 2010 and under PBV. PBV, plurality bloc
vote; SNTV, single non-transferable vote.
11
In 2010, a second election was held in June in an effort to resolve an impasse generated by a 9 versus 9
MP split between government and opposition (Le Roy 2010).
12
Prior to 2000, the threshold was three parliamentary seats. The system works like a ‘closed list’system
at the level of the district. The parties field single candidates in each of the 88 districts (11 for each of the
8 districts), but voters effectively select a party by choosing a candidate.
Mandates are first allocated within each constituency to the eligible candidate lists using the Droop
quota. After the first allocation, the overall proportional calculation of the number of mandates
each party is entitled to on a nationwide basis is done, using the d’Hondt method. The mandates
10 J. FRAENKEL AND B. GROFMAN
410
415
420
425
430
435
440
445
450
the Hungarian and Italian minorities, are elected using the Borda Count, although
minority voters are also entitled to cast a second ballot in one of the 88 list PR
seats (Toplak 2006: 826–7).
13
Overall, the population is roughly 88 per cent Slove-
nian, with Hungarians and Italians, respectively, at 0.32 per cent and 0.11 per cent
and enjoying a special protected constitutional position (Republic of Slovenia
2002). Representatives of the two minority groups have regularly held the balance
of power, and decided election outcomes in favour of left-leaning governments
(Toplak 2006: 827). Figure 4 shows the location of the Borda-using Italian and Hun-
garian constituencies, as well as the eight broader list PR constituencies. The Italian
district covers the three Adriatic port towns of Koper, Isola and Piran in the consti-
tuency of Postojna in the southeastern part of the country, whereas the Hungarian
constituency is located in the Ptuj constituency in eastern Slovenia (Figure 3 AQ6).
In the version of the Borda system used in Slovenia, with ncandidates, a first pre-
ference is worth n, a second preference is worth n–1 and so on with the last preference
worth n–(n–1): that is, it is worth 1. Of course, we may readily re-normalise to score
the last place candidate zero instead of one, to get the more common Borda scoring
rule.
14
As Table 4 indicates, the Borda system has been used for the minority districts
over six elections (1992, 1996, 2000, 2004, 2008 and 2011), but the scoring details
were not included in the official results for minority constituencies in 1992 and 1996,
and for other years, the Italian constituency was usually uncontested (2000, 2004 and
2011). Hence, we report outcomes only for the Hungarian constituency from 2000
until 2011 and for the Italian constituency in 2008.
In these elections, it was not legally necessary for voters to provide a full ranking.
15
Many voters mark only a single preference. In the Hungarian constituency, ballots
included a second preference on average in less than half of cases. Because many
ballots were incomplete, in 2008 the Slovenian Electoral Commission decided to
give one point for each unranked position for all candidates (personal communi-
cation, Jure Toplak, 28 February 2013). The outcome of this change is discernible
in the 2008 election results, with numbers with only a single vote correspondingly
inflated by around 20 per cent. Several times over the past decade, representatives
of the Hungarian minority have exerted pressure for the Borda method to be
dropped in favour of a plurality system.
not already distributed are then allocated to candidate lists among all the constituencies, in the
order of the highest remainder of votes in proportion to the quota used in each constituency. Man-
dates within a list are assigned to specific candidates on the basis of the percentage of the vote each
received in his/her district. That is, candidates within a list are ranked on the basis of the percen-
tage of votes they received in comparison to the overall total of valid votes in their respective dis-
tricts. The overall equality of the vote is ensured in the Slovenian system because mandates are
awarded to parties proportionately to their nationwide vote total. However, because the allocation
of mandates to candidates within a party list is done on the basis of the percentage of the vote
received by each candidate, the voting weight of voters in a small district can be greater than
that of voters in a large district in allocating mandates within a party list. The Constitutional
Court decided in 2000 that this did not violate the principle of equality of the vote. (OSCE
2012:4)
13
The Office for Democratic Institutions and Human Rights concludes that such dual voting rights
diverge from principles ‘regarding equality of the vote and are at odds with international good practice’
(OSCE 2012: 1).
14
In 2008, unranked candidates were scored with ones rather than zeros.
15
§73 of the National Assembly Elections Act 1992 (amended 2000) provides that ‘A voter may vote for
only one candidate’(Republic of Slovenia 2000).
THE BORDA COUNT AND ITS REAL-WORLD ALTERNATIVES AQ1
11
455
460
465
470
475
480
485
490
495
Figure 3. The Hungarian and Italian minority districts and the eight broader list PR constituencies in
Slovenia
Figure 4. Top-placed victors’share of preference votes by number of candidates contesting, two-seat
Nauru constituencies 1997–2013
Notes: Y-axis is percentage share of preferences. X-axis is the ranking order. The chart shows all two-
seat Nauru contests between 1997 and 2013. There were no contests with fourteen canmdidates.
12 J. FRAENKEL AND B. GROFMAN
500
505
510
515
520
525
530
535
540
Invalid voting in Nauru and Slovenia
A common criticism of preferential voting systems is that they encourage a high
number of invalid ballots. In Slovenia, numbers of invalid votes were substantial
in some elections, but this was usually in the Italian constituency in contests where
there was only one candidate, suggesting some deliberate spoiling of ballots.
Invalid ballots stood at 13.2 per cent, 12.7 per cent and 11 per cent in 2000, 2004
and 2011, all elections when there was only a single candidate standing, whereas
they stood at 3 per cent and 3.5 per cent in 1992 and 2008 when more than one can-
didate contested. In the Hungarian constituency, which had multiple candidates at all
elections between 1992 and 2011, invalid ballots ranged from 1.7 per cent to 3.9 per
cent: that is, close to the lower end seen in the competitive Italian elections (Table 4).
On Nauru, despite the requirement to rank order all candidates, the share of invalid
votes was small. Over the 21 elections since 1968, the share of ballots ruled invalid
has averaged 2.9 per cent, with very low variation (from slightly below 2 per cent to
slightly above 4 per cent). And there was no statistically significant change in the pro-
portion of invalid ballots witnessed after the 1971 electoral shifted from the STV
system to the present-day Nauru rule. Some economists have argued that Nauru’s
voting system is too complex for such a small country (Hughes 2006). Nauru’s count-
ing system is computerised, however, and the administration of the process has raised
few difficulties. Reformers have indicated concerns about the voter-registration
process, about overseas proxy voting and about the usage of a distinct system (the
AV) for single-member by-elections, but not about the usage of the Dowdall
points system at general elections (Cain 2005; Pacific Islands Forum 2010).
Actual and simulated elections in Nauru and Slovenia under alternative
voting rules
Our available data are much more extensive for Nauru than for Slovenia. We have
details of the scoring system that determined the election of 163 candidates from
Table 4. Outcomes in the Hungarian and Italian minority districts, Slovenia, 1992–2011
1992 1996 2000 2004 2008 2011
Hungarian
Registered voters 6838 6229 N/av 6262 7063 6661
No. candidates 7 4 5552
Invalid votes % 3.9 N/av 4.9 3.6 1.8 1.7
Second preferences lodged (%) N/av N/av 40.9 41.7 44.5 36.5
Victors points 12.8 10.4 11 5.1
Dowdall points (simulated) 23.6 19.4 20 25.7
Italian
Registered voters 1913 2604 2750 2737 1385 2712
No. candidates 2 4 1151
Invalid votes % 2.97 N/av 13.2 12.7 3.5 11.0
Second preferences lodged (%) N/ap N/ap N/ap N/ap 45.5 N/ap
Correlations between rankings 100% 100% 100% 100%
Correlations between outcomes 100% 100% 100% 100%
Note: N/av, not available and N/ap, not applicable.
THE BORDA COUNT AND ITS REAL-WORLD ALTERNATIVES AQ1
13
545
550
555
560
565
570
575
580
585
72 constituency contests (over 9 general elections) in Nauru, but for Slovenia we only
have data indicating the points obtained by 5 candidates in 5 single-member consti-
tuency contests (over 4 general elections). No data are available from either country
to illuminate how either Borda or the Dowdall rule might operate in a political party-
based system (for example, as regards votes/seats disproportionality), though empiri-
cal scrutiny of the practical operation of both systems offers some useful pointers
about how these might work in mass elections.
16
Nauru’s use of multi-member con-
stituencies, with a single vote as compared to the single-seat contests in Slovenia,
limits the comparisons we can make between countries. So we focus on within-
country comparisons across rules, not just between Borda and the Dowdall rule,
but also other electoral rules, such as SNTV and plurality bloc voting (PBV) for
the Nauru data.
Given differences in the scores attached to preferences under the two electoral
systems, we might expect substantial variations between outcomes under the
Dowdall and Borda rules. This was true for Nauru, but not for Slovenia.
Outcomes under the Dowdall system in Nauru differed from those simulated under
Borda in 30.1 per cent of cases. By contrast, in Slovenia, there were no differences in
outcome between the actually used Borda system and outcomes simulated under
Nauru’s Dowdall system in any of the five contests. This difference, as regards Slo-
venia, was not a result of any intrinsic difference between the two scoring rules. It
stemmed primarily from the use of optional ranking in the Slovenian minority dis-
tricts, coupled with the reluctance of voters to indicate lower-order preferences. Slo-
venian minority voters tend to cast truncated ballots indicating only a first preference,
but no subsequent choices, whereas Nauru voters are compelled to lodge full prefer-
ences. Dowdall (simulated) and Borda may have yielded identical outcomes in Slo-
venia, but these were also the exact same outcomes that would have arisen under
single-member plurality. Indeed, in all the Slovenian results, there was such high
agreement on a single candidate that virtually every set of electoral rules would
have delivered the same result.
More broadly, and based on the larger Nauru data set alone, simulated outcomes
under Borda varied from those under PBV or the SNTV to a much greater degree
than do the Dowdall results. Table 3 shows the percentage share of outcomes that dif-
fered between Dowdall and other systems at the nine elections. On average, the
Dowdall rule gets a conflicting outcome to PBV only 4.9 per cent of the time, as com-
pared to 28.8 per cent for the Borda Count, and a different outcome to SNTV in only
8 per cent of cases, as compared to 36.2 per cent of cases with the Borda Count. Next,
Table 3 shows correlations between actual and simulated outcomes under the
Dowdall system, the Borda Count, SNTV and PBV. Pair-wise correlations
between outcomes under systems other than Borda (Dowdall, PBV and SNTV)
ranged on average between 0.90 and 0.95, whereas those of other systems with
Borda averaged between 0.56 and 0.62.
A similar finding emerges if attention is focused on correlations between rankings
(shown in the bottom part of Table 3), rather than between outcomes (though here the
contrast is necessarily a three-way one). Again, the comparison between Dowdall and
16
Borda elections in Slovenia are disconnected from regular party competition while Nauru has no pol-
itical parties, only loose and regularly changing parliamentary factions.
14 J. FRAENKEL AND B. GROFMAN
590
595
600
605
610
615
620
625
630
SNTV, 0.93, is much closer than that between Dowdall and Borda (0.7) or Borda and
SNTV (0.57).
These findings cast doubt on the common classification of the Dowdall system as a
‘modified form’of the Borda Count (Reilly 2001b: 154; Reynolds et al. 2005). That
label is earned on account of similarity in ballot structure. If instead electoral systems
were identified by the kind of outcomes they generate, the Dowdall system might be
seen as a more distant relative of Borda. With district magnitudes of 2–4, the Dowdall
system should be seen as standing somewhere between plurality and the Borda
Count, but as veering more towards plurality.
17
Strategic voting
Nauru’s compulsory ranking system may preclude Slovenian-style truncation of pre-
ference ranking, but it does not prevent other forms of strategic voting. As Duncan
Black shows, under Borda, electors may also vote for a favourite candidate first,
and rank all others in the inverse order to their popularity (Black 1976:13–15;
Cox 1997; Ludwin 1978: 85; Saari 1990b). If strategic voting of this type took
place, we would expect top-placed candidates not only to obtain large numbers of
first-preference ballots, but also a significant share of last preferences. On average,
top-placed victors in Nauru had 30.1 per cent of first preferences, 20 per cent of
second preferences and 15.1 per cent of last preferences, but only 7.5 per cent of
the other intermediate preferences. Second-placed victors had 22.2 per cent of first
preferences, 16.8 per cent of second preferences, 15.3 per cent of last preferences,
but only 8.3 per cent of intermediate preferences.
Figure 4 examines the distribution of preferences for top-placed victors in each of
the Nauru two-seat constituencies from 1997 until 2013. Between 5 and 15 candi-
dates contested these seats, and there is one panel for each of these configurations.
There were numerous contests with between 5 and 11 candidates, but there was
only a single case where 12, 13 and 15 candidates contested.
No clear pattern is discernible in the five-seat contexts. As the number of candi-
dates rises, however, a U-shaped trajectory becomes increasingly apparent. Why?
Less polarised constituencies tended to have fewer candidates, and the same parlia-
mentary faction tended to win both seats. In many cases, electoral support on
Nauru remains highly personalised, and closely connected to kinship or clan linkages.
Different factions able to strategically direct voters to place major rivals last usually
contested those constituencies with larger numbers of candidates. Candidates distri-
bute ‘how to vote’cards at Nauru elections, and citizens are familiar with the tactic of
strategically placing key adversaries in the last position (personal communication,
Kieren Keke, Nauru MP, 23 August 2013).
Figure 5 shows that this characteristic U-shaped trajectory was also evident in the
three- and four-seat constituencies, which a larger number of candidates tended to
contest. Unlike Figure 4, which only shows the top-placed victor and groups
results from different elections, Figure 5 shows all of the victors for each of 10 sep-
arate constituency contests. Separate investigation of those exceptional series in
Figures 4 and 5which are not U-shaped indicates that there was usually a well-
17
Ben Reilly has remarked that ‘the Nauruan system [is] much more majoritarian than a standard Borda
count’(2002: 365). Here we make that statement much more precise.
THE BORDA COUNT AND ITS REAL-WORLD ALTERNATIVES AQ1
15
635
640
645
650
655
660
665
670
675
placed incumbent, expected to win, who obtained a U-shaped preference schedule,
and/or that actual victors were not polarising figures. The data in Figures 4 and 5
are consistent with strategic voting, but similar patterns could potentially arise with
sincere ranking and simply reflect polarisation in electorates.
Neither the Slovenian nor the Nauru results provide evidence of intermediate
candidates, promoted up the scale by the practice of ranking major challengers
last, becoming serious contenders, as Dummett (1997: 85) fears may arise under
Borda. In Slovenia, there is no need to allocate preferences in such a way to
avoid strengthening major rivals. Voters can instead more simply avoid submitting
full rankings. In Nauru, factions competing in the two-, three- or four-seat consti-
tuencies put their major rivals as last preference, and thereby potentially promote
middle-ranked contestants. If Nauru used an elimination-based system, such as the
AV or STV, these intermediate candidates might stand a better chance. With the
Dowdall rule, they rarely perform strongly, owing to the low weighting of inter-
mediate preferences.
For the same reason, the incentives for fielding ‘red herring’or ‘dummy’candi-
dates are greater under Nauru’s compulsory ranking system than under optional
Borda ranking in Slovenia (although not as great as they would be under a Borda
system with compulsory ranking). In Nauru, the factions tend to run two candidates
in the two-member districts but also to encourage ‘buffer candidates’–who are not
expected to win –to soak up intermediate preferences, and thereby lower the vote
tallies of their major rivals (personal communication, Roland Kun, Nauru MP, 20
August 2013). These were not difficult-to-engineer ‘clones’intended to split rivals’
votes, but ‘irrelevant alternatives’in Condorcet’s sense, designed to diminish the
value of preference votes allocated to arch-rivals. With the Dowdall rule, at least
the scoring of those lower-order alternatives remained fixed, irrespective of the
greater number of candidates. Had the Borda Count been in place, the inclusion of
these ‘irrelevant alternatives’would have adjusted the overall vector of ranking
weights.
Figure 5. Victors’share of preference votes by number of candidates contesting, three and four seat
Nauru constituencies 1997–2013
Notes: as for figure 4. The chart shows the distribution of preferences for all victors in contests for the
four seat Ubenide constituency and for the constituency of Meneng in 2013, which was increased from
two to three members ahead of that election
16 J. FRAENKEL AND B. GROFMAN
680
685
690
695
700
705
710
715
720
Discussion
Our investigation shows major differences between the Slovenian, Nauru and Borda
systems in theory and in practice, but there is one significant commonality. Both the
Slovenian Borda variant and the Dowdall systems tend to entail ultimate election of
first-count leaders, but for different reasons. On Nauru, this arose due to relatively
low weighting of lower preferences. In Slovenia, it stemmed from substantial agree-
ment on the strongest candidate and voters choosing to cast truncated ballots, as
allowed by an optional ranking scheme. Since so few indicated second or lower pre-
ferences, outcomes were often identical to those that would have arisen under the
single-member district plurality. This evidence should not be interpreted as support
for the case for categorical choice-based electoral systems: it only shows that one
type of real-world choice, under rank-order scoring rules, is to express a common
indifference to all but one’s most-favoured candidate.
The system used on Nauru is not merely a ‘modified form of Borda’, but an impor-
tant rule in its own right. The usual classifications of electoral systems tend to derive
from their ballot structure, their methods of determining victors or their broad reper-
cussions. Terms like ‘first-past-the-post’refer to the method of determining victors;
while the term ‘proportional representation’or when we say a system is ‘majoritar-
ian’, these refer to the kind of outcomes that are delivered. Dowdall has a similar
ballot structure to Borda, and methods of determining victors closely resemble
each other but, given real-world preference distributions, outcomes can and will
differ between the two methods. As it operates in Nauru, the Dowdall rule is in
fact closer to the SNTV than to the Borda system, at least for the larger (three- and
four-member) districts.
The Dowdall rule can be viewed as a corrective to preferential voting systems such
as the AV or the STV which, it is sometimes argued accord second preferences exces-
sive weight by counting these as of equal value to first preferences. By rapidly dis-
counting second and lower preferences, the Dowdall rule meets this objection. In
addition, AV and STV are sometimes criticised because only a restricted share of
voters’second or lower preferences are ever counted, whereas the Dowdall rule
counts all preferences. It also removes the need for time-consuming or costly elimin-
ations in successive rounds or counts, but without losing sensitivity to lower
preferences.
By introducing multiple members per constituency, the Dowdall system increases
the likelihood that second or lower preferences will determine at least some out-
comes, since voters cast only a single (ranked) ballot. This adds an element of poten-
tial proportionality that would be absent in single-member ‘winner takes all’variants.
In the Nauru context, where there are no political parties, that proportionality is not as
apparent as would be the case if that system were used in larger constituencies and in
a party-centered system. On the negative side, the Dowdall system potentially has
some of the features of SNTV, but these are only fully apparent in the three- and
four-member constituencies, not the more widely used two-member districts. If pol-
itical parties were to operate under such a system, as with SNTV (see Grofman et al.
1999), they would potentially have to gauge their voter support accurately, and field
exactly the right number of candidates –with possible heavy penalties for over-
nomination.
Supporters of the Borda Count, or other ranking-based methods, have often been
attracted to such systems because of their potential to express preferences fully. That
THE BORDA COUNT AND ITS REAL-WORLD ALTERNATIVES AQ1
17
725
730
735
740
745
750
755
760
765
such systems may be used strategically to ensure victory of a favoured candidate by
placing his or her most popular challengers last is often counted as a major defect
(Black 1976; Ludwin 1978). In Nauru and Slovenia, we found that voters tended
to manipulate such systems with the objective of maximising the likelihood of
defeat of a rival to their first choice. In Slovenia, where lodging full preference rank-
ings was optional, we found that on average, only 41.8 per cent of voters ranked a
second candidate in our five contests. The majority of voters engaged in a form of
‘bullet voting’, generating in response efforts by the Slovenian Electoral Commission
to enforce a fuller ranking. On Nauru, that difference was maximised by placing
major rivals as last preference and/or by introducing weak candidates to mop up inter-
mediate preferences.
The Borda Count and similar scoring rules cannot be simply dismissed on account
of susceptibility to strategic manipulation. All electoral systems are potentially vul-
nerable to voter-expressed insincerity about true preferences (Gibbard 1973; Sat-
terthwaite 1975). Where there are only three candidates, the Borda Count has been
shown to be less susceptible to micro-manipulation than other electoral systems
(Saari 1990b). Manipulation, at least of this type, is only a major problem for
social-choice theorists insofar as the primary justification adopting such systems is
that they better express the ‘felt preferences’of voters (Black 1976: 15; Dummett
1997: 63, 182). If voters submit ballots that record their preferences sincerely
when doing harms their likely ability to achieve results more to their liking, strategic
voting seems desirable rather than problematic (Mckelvey and Niemi 1978). Also, as
Dowding and Van Hess argue, democratic processes are more than mere counting
devices: ‘the fact that aggregative systems are manipulable encourages the very delib-
eration that makes politics what it is’(2007: 11).
Strategic use of these scoring rules in Nauru and Slovenia has not generated
chaotic or unrepresentative outcomes that might warn against their adoption else-
where. Nauru has used the Dowdall system for over 40 years, during which time
compulsory ranking has not fostered a high rate of invalid voting. Hungarian and
Italian reluctance to specify second or lower preferences may have frustrated the Slo-
venian Electoral Commission, but this was primarily a consequence of voter choices
together with the law on optional ranking, not of the Borda Count itself. Those reac-
tions to the real-world presence of such rank-order preferential systems, both by the
voters and the responsible institutions, must be an important focus of inquiry by pol-
itical scientists. The evidence that we report here warns against judging electoral rules
simply by their mathematical properties, and suggests that similarities and differences
across different rules can, and will, vary with political context.
References
Black, D. 1958. The theory of committees and elections. London: Cambridge University Press.
Black, D. 1976. Partial justification of the Borda count. Public Choice 28: 1–15.
Borda, J.–C. 1781. Memoire sur les elections au scrutin (Memoir on the ballot elections). Historie de
l’Academie Royale des Sciences. URL: < http://gerardgreco.free.fr/IMG/pdf/MA_c_moire-Borda-
1781.pdf>. Consulted 16 May 2013.
Cain, F. 2005. Electoral systems –pros and cons. Unpublished paper No. 19. Fifth Forum Presiding
Officers and Clerks Conference.27–29 April, Samoa.
Cox, G. 1997. Making votes count: Strategic coordination in the world’s electoral systems. New York:
Cambridge University Press.
Dowding, K. and Van Hees, M. 2007. In praise of manipulation. British Journal of Political Science 38:
1–15.
18 J. FRAENKEL AND B. GROFMAN
770
775
780
785
790
795
800
805
810
Dummett, M. 1997. Principles of electoral reform. Oxford: Oxford University Press.
Dummett, M. 1998. The Borda count and agenda manipulation. Social Choice and Welfare 15: 289–96.
Emerson, P. 2013. The original Borda count and partial voting. Social Choice and Welfare 40: 353–8.
Feld, S. L. and Grofman, B. 1992. Who’s afraid of the big bad cycle? Journal of Theoretical Politics 4:
231–7.
Felsenthal, D. and Machover, M. 1998. The measurement of voting power: Theory and practice, pro-
blems and paradoxes. Northampton: Edward Elgar.
Fishburn, P. 1973. The theory of social choice. Princeton, NJ: Princeton University Press.
Gibbard, A. 1973. Manipulation of voting schemes: A general result. Econometrica 41: 587–601.
Golder, M. 2005. Democratic systems round the world, 1946–2000. Electoral Studies 24: 103–21.
Grofman, B., Lee, S., Winckler, E. and Woodall, B. eds. 1999. Elections in Japan, Korea and Taiwan
under the single non-transferable vote: The comparative study of an embedded institution. Ann
Arbor, MI: University of Michigan Press.
Hughes, H. 2006. Submission to the Constitutional Review Committee, Republic of Nauru. Unpublished
paper.
International Foundation for Electoral Systems [IFES]. 2013. Election guide. URL: < http://www.
electionguide.org/countries/id/150/>. Consulted 13 November 2013.
Irwin, G.A. and van Holsteyn, J.M. 2012. Strategic electoral considerations under proportional represen-
tation. Electoral Studies 31: 184–91.
Le Roy, K. 2010. Nauru’s parliament in crisis. The Parliamentarian 91: 240–3.
Lijphart, A. 1997. Disproportionality under alternative voting: The crucial –and puzzling –case of the
Australian senate elections, 1919–1946. Acta Politica 32: 9–24.
Ludwin, W.G. 1978. Strategic voting and the Borda method. Public Choice 33: 85–90.
Mckelvey, R.D. and Niemi, R.G. 1978. Multistage game representation of sophisticated voting for
binary procedures. Journal of Economic Theory 18: 1–22.
McLean, I. and Urken, A.B. 1995. Introduction. In I. McLean and A. Urken, eds. Classics of social
choice. Ann Arbor, MI: University of Michigan Press.
Merrill, S. 1984. A comparison of the efficiency of multicandidate electoral systems. American Journal
of Political Science 28: 23–48.
Nanson, E.J. 1882. Methods of election. Transactions and Proceedings of the Royal Society of Victoria
18 L: 197–240.
Office for Democratic Institutions and Human Rights [OSCE]. 2012. Republic of Slovenia: Early elec-
tions of the national assembly, 4 December 2011. OSCE/ODIHR Election Assessment Mission Final
Report. URL: < http://www.osce.org/odihr/elections/84559>. Consulted 28 February 2014.
Pacific Islands Forum. 2010. Report of the Pacific Islands forum election observer mission to the April
2010 Nauru general elections, 14 May. URL: < http://www.forumsec.org/resources/uploads/
attachments/documents/Forum_Election_Observer_Report_Nauru_2010.pdf>. Consulted 16 May
2013.
Reilly, B. 2001a. Democracy in divided societies: Electoral engineering for conflict management.
Cambridge: Cambridge University Press.
Reilly, B. 2001b. The Borda Count in the real world: The electoral system in the Republic of Nauru.
Working Paper. Canterbury: Macmillan Brown Centre for Pacific Studies.
Reilly, B. 2002. Social choice in the South Seas: Electoral innovation and the Borda count in the Pacific
Island countries. International Political Science Review 23: 355–72.
Reilly, B. 2011. Electoral system reform options for Northern Territory local government elections.
Report prepared for the Northern Territory Department of Housing, Local Government and
Regional Services. URL: < http://www.localgovernment.nt.gov.au/__data/assets/pdf_file/0017/
121319/Electoral_System_Reform_Options_for_Northern_Territory_Local_Government_Elections.
pdf>. Consulted 28 February 2014.
Reilly, B. and Gratschew, M. 2001. Nauru. In D. Nohlen, F. Grotz and C. Hartmann, eds. Elections in
Asia: A data handbook. Oxford: Oxford University Press.
Republic of Slovenia. 2000. National assembly elections act (1992, amended). URL: < http://www.
legislationline.org/documents/action/popup/id/3825>. Consulted 28 February 2014.
Republic of Slovenia, Statistical Office. 2002. Population by ethnic affiliation, Slovenia, census 1953,
1961, 1971, 1981, 1991 and 2002. URL: < http://www.stat.si/popis2002/en/rezultati/rezultati_red.
asp?ter=SLO&st=7>. Consulted 20 May 2013.
Reynolds, A., Reilly, B. and Ellis, A. 2005. Electoral system design: The new international IDEA hand-
book. Stockholm: Institute for Democracy & Electoral Assistance.
Saari, D.G. 1990a. The Borda dictionary. Social Choice and Welfare 7: 279–317.
Saari, D.G. 1990b. Susceptibility to manipulation. Public Choice 64: 21–41.
Saari, D.G. 1994. Geometry of voting. New York: Springer Verlag.
Saari, D.G. 1995. Basic geometry of voting. Berlin: Springer Verlag.
Satterthwaite, M.A. 1975. Strategy-proofness and arrow’s conditions: Existence and correspondence
theorems for voting procedures and social welfare functions. Journal of Economic Theory 10:
187–217.
THE BORDA COUNT AND ITS REAL-WORLD ALTERNATIVES AQ1
19
815
820
825
830
835
840
845
850
855
Supreme Court of Nauru. 1977. In the matter of the Electoral Act 1965–1973 and in the matter of an
election in the constituency of Ubenide. Miscellaneous Cause No. 15. URL: < http://www.paclii.
org/nr/cases/NRSC/1977/4.html>. Consulted 20 May 2013.
Tabarrok, A. and Spector, L. 1999. Would the Borda Count have avoided the civil war? Journal of
Theoretical Politics 11: 261–88. AQ7
Tangian, A. 2013. Mathematical theory of democracy. New York: Springer.
Toplak, J. 2006. The parliamentary elections in Slovenia, October 2004. Electoral Studies 25: 825–31.
Tsetlin, I., Regenwetter, M. and Grofman, B. 2003. The impartial culture maximizes the probability of
majority cycles. Social Choice and Welfare 21: 387–98.
Van Trease, H. 1993. The presidential election. In H. Van Trease, ed. Atoll politics: The Republic of
Kiribati. Christchurch: Macmillan Brown Centre for Pacific Studies, University of Canterbury.
Young, H.P. 1974. An axiomatization of Borda’s rule. Journal of Economic Theory 9: 43–52.
Young, H.P. 1995. Optimal voting rules. Journal of Economic Perspectives 9: 51–64.
20 J. FRAENKEL AND B. GROFMAN
860
865
870
875
880
885
890
895
900
