How the World Votes
The Political Consequences of Ballot Design, Innovation and Manipulation
Andrew Reynolds and Marco Steenbergen
University of North Carolina, Chapel Hill
Abstract
There has been limited research into the impact of some aspects of voting procedure but
very little research into ballot paper design and the act of voting in itself. This article
gives a historical overview of the evolution of voting procedures from ancient times to
the modern day, describes the results of a survey of 134 paper ballots used over the last
decade for national legislative, executive and referenda elections in 107 countries, and
gives the findings of a laboratory experiment testing the impact of ballot paper design
conducted with student subjects.
The findings demonstrate evidence about the relationship between ballot paper design,
spoilt votes and voting behavior. The evidence suggests three things: first, ballot
friendliness (elaborated ballots incorporating colors, symbols and photographs) is more
likely where literacy is lower and competitive multi party elections are a new
phenomenon. But friendliness is not significantly related to the level of democracy, the
effective number of parties, or the type of electoral system. Second, there is little
evidence to suggest that elaborate and costly ballots reduce spoilt ballot rates or are
essential tools for illiterate voters. Rather, the negative correlation between ballot design
and spoilt ballots, combined with the weight of historical evidence which shows that
ballots are often a highly manipulative tools of political symbolism, implies that ballot
papers symbols, photographs, layout, and color are of most interest as political cues (for
both literate and illiterate voters). This intuition is confirmed by the results of a vote
simulation experiment conducted on 401 students in April and October 2002 in which
ballot design had a pronounced effect on voting behavior.
How the World Votes 1
The Political Consequences of Ballot Design, Innovation and Manipulation
It is perhaps remarkable that one of the most important processes operated in some way
by almost all of human kind Ð that of how to elect leaders Ð has hardly altered in two and
a half thousand years of history. Technological innovation in the way voters demonstrate
their preferences in the ballot booth, i.e., the design of ballot papers (or other voting
methods), was almost non-existent until the turn of the twentieth century and the methods
brought into practice then largely remained unreconstructed during that century.
However, the start of the 21st century is witnessing a new era of ballot design and
innovation. To facilitate the act of voting for illiterates, to reduce voting errors, and to
increase the amount of information on the ballot, many new democracies are
experimenting with multi-colored ballots, symbols, and photographs of candidates.
These innovations are not merely aesthetic, they can influence voting behavior and
election outcomes.
There has been limited research into aspects of voting procedure this century, but the
bulk of that research has been concerned with issues such as secrecy,2 types of ballot
marking,3 name ordering (see below), the impact of machine voting,4 and the effect of
registration and voting procedures on turnout.5 There have also been studies of the
effects of ballots that allow for Ôsplit-ticketÕ voting, but there has been very little research
into ballot paper design and the act of voting in itself.6 This is despite the fact that the
use of symbols and color was prevalent in 19th century American elections and that many
2
developing world countries have used symbols on their voting papers for much of the
post-war period.
This article begins with an historical overview of the evolution of voting procedures from
ancient times to the modern day. After highlighting the most consequential areas of
ballot paper design we then describe the results of a survey of 134 paper ballots used over
the last twenty years for national legislative and executive elections in 107 countries.
The findings demonstrate the relationship between ballot paper design, spoilt voting
paper rates, literacy, country wealth, level and age of democracy, the competitiveness of
the party system and regional design trends. The evidence suggests three things: first,
ballot friendliness (elaborated ballots incorporating colors, symbols and photographs) is
more likely where literacy is lower and competitive multi party elections are a new
phenomenon. But friendliness is not significantly related to the level of democracy, the
effective number of parties, or the type of electoral system. Second, there is little
evidence to suggest that elaborate and costly ballots reduce spoilt ballot rates or are
essential tools for illiterate voters. Rather, the negative correlation between ballot design
and spoilt ballots, combined with the weight of historical evidence which shows that
ballots are often a highly manipulative tools of political symbolism, implies that ballot
papers symbols, photographs, layout, and color are of most interest as political cues (for
both literate and illiterate voters). The way a party/candidate/issue is presented Ð at the
very Ôpoint of saleÕ Ð can have a substantial impact on voting behavior and subsequent
candidate/party success.
3
A short history of voting
For two and a half thousand years voting was conducted largely by a show of hands or by
voice approbation (viva voce). This was the dominant method used for elections all the
way through from those held in Athens in 500 B.C. to Prussian, British and American
elections in the mid to late 19th century. The Church was largely responsible for the
introduction of the written, and in some cases secret, ballot in the middle ages.
Beginning in 1562 popes were elected by a two-round written ballot of the cardinals
present. Romans would gather in the VaticanÕs square to look for wisps of smoke from
the burning of ballots from the first round, which would tell them how the vote was
progressing.7 Nearly 500 years on the practice has been ceremonially formalized into the
releasing of white smoke to signify the election of a new Pope.
Before the 20th century great variety and innovation in voting methods was to be found in
the United States. In the royal colonies of pre-independence America the vote (based as
elsewhere on a highly restricted franchise) was open Ð viva voce or by a show of hands,
but if a candidate or voter demanded it, a poll was taken. Virginia utilized perhaps the
most Ôvoter friendlyÕ method (if not particularly secret) where the local sheriff would ride
out to the planterÕs gate and write down his vote. Nine out of ten State constitutions
framed between 1776-1780 required a secret ballot for some offices but the majority of
elections were still viva voce.8 In Kentucky, which only abandoned the voice vote in
1890, elections for Sheriff were conducted by lining up the supporters of one candidate
on one side of a road, the supporters of another on the opposite side Ð the longest line
won.9
4
All of these methods were open to manipulation, intimidation, bribery and corruption,
and for these reasons the British colony of South Australia introduced a secret, preprinted single ballot paper in 1856.10 Secret voting of this sort became known as the
ÔAustralian BallotÕ and was adopted in nearly all constituencies in Britain under the 1872
ballot act. The abandonment of the old and open method of viva voce was regarded with
disgust by leading politicians such as Gladstone, Russell, and Shaftesbury, who
considered secret voting as cowardly and an encouragement to the less educated to
express their unsophisticated base views. The simple candidate ballot adopted in Britain
in 1872 is effectively the ballot used today, although after 1970 candidates were allowed
to identify themselves with a political party.11 The blandness of the British ballot may
not have necessarily been the case had history followed a different path. Secret voting
was tested in the constituencies of Manchester and Stafford in 1869 with candidatesÕ
names printed in different colors, the voter being asked to score out names he did not
support, and the use of envelopes for casting the vote.12 Sixteen years later in 1888
Massachusetts became the first American state to adopt the ÔAustralian ballotÕ for all
statewide elections.13
Over two millenniums three main methods of secret voting developed. First, the mark on
a single ballot paper which is then deposited by a voter into a single box. Second,
preprinted party or candidate ballots (or some other method of identification such as a
colored disc) which are placed into a single ballot box. Third, uniform ballots, marbles,
or rods, which are placed into individual candidate boxes.
5
Single ballot, single box
Before printed ballot papers, the most common form of the secret ballot was to simply
write down the name of a chosen candidate, fold it, and then place it in the box. In
Syracuse of ancient Greece voting was by petalism where candidates names would be
written on olive leaves.14 In the time of Pericles, votes to ostracize would be taken by
scratching the name of the proposed victim on a broken piece of pottery.15 The first
reported use of a secret ballot in the Americas was to elect the pastor of the Salem
Church, Massachusetts on July 20, 1629. After 1859 voting for the Italian Chamber of
Deputies was conducted by writing the candidateÕs name on official blue paper, and
Japan and Jordan to this day request the voter to write the candidateÕs name on what is
little more than a blank piece of paper.16 For Senate elections in the Philippines voters
are given a strip of paper with 24 blank lines on it and are required to write down the
names of up to 24 candidates taken from the lists of candidates posted on the walls.
Again this gives huge discretion to election officials who can void ballots for minor
infractions.17 At the other end of simplicity is the single mark that the Belgian elector has
been asked to make for well over a century. Here the voter merely blackens the white
ÔbullÕs eyeÕ next to the party, or candidate, of their choice.18 The vast majority of ballots
surveyed for this paper (85 per cent) ask the voter to make an X, tick, line, or thumbprint
by the candidate(s)/party of choice.
While making a mark on the ballot paper by hand is today by far the most common form
of voting, voting machines have long been a feature of US elections (as they are in the
Netherlands).19 Modern day US voting machines stem from Britain where in the 1830Õs
6
a Mr. Grote designed and advocated a voting machine in which the voter would puncture
the card of their favored candidate through a hole without ever touching the paper.20
Multiple party/candidate ballots, single box
The chief voting characteristic of Franco-phone countries is that they use a system where
voters deposit a pre-printed party or candidate ballot in the ballot box Ð usually no mark
is made on the pre-printed ballot. Oftentimes this system is accompanied by the use of
envelopes for the ballots placed in the box. While multiple candidate/party ballot
systems are associated with Franco-phone countries they are also used in Scandinavia,
Spain, Greece, Latin America and Malawi, and were in widespread use in Europe (e.g.,
Bulgaria, Romania and Germany) and the United States before the First World War.
Sometimes the ballots were printed and distributed by the parties, sometimes a voter was
free to draw up their own, and sometimes the ballots were available in the polling station.
The extent of regulating the production of ballots varied widely. At the turn of the
century Bulgaria, Romania and France regulated the color and size of the party ballots
which could be cast, but in the US before 1900 parties were free to play with colors,
ballot size and even the nature of the paper Ð furthermore, voters could make their own
ballots.21 The lack of consistency led to a multitude of abuses and ultimately the system
was abandoned in favor of the single printed ballot. In France and Turkey at the turn of
the century one could strike out names from the pre-printed party list and in Sweden there
has always been a degree of freedom to prepare ones own list of names or adapt the party
list.22
7
The multiple ballot system originated in ancient Rome where under the Lex Gabiana of
139 B.C. voters would be given a series of carved wooden ballots for each candidate and
then drop their chosen one into an urn.23 France first used the system for elections under
Napoleon III Ð voters were required to bring a pre-prepared ballot to polls and cast it
under the watchful eye of the local Mayor.24 This practice was re-instituted during the
Third Republic but after 1919 the State printed the ballots and would mail them out to
eligible voters.25 Another variation on the theme operated in Botswana which until 1999
used a colored disc system. Each of the individual candidates in a constituency would be
assigned a specific color and voters would choose that disc and drop it in the box. All of
these methods reduce the occurrence of Ôspoilt ballotsÕ but as the American and French
experience demonstrated, the fewer regulations on the production and distribution of
ballots the much higher chances of corruption and intimidation.26
Uniform ballot, separate candidate boxes
The final method of voting is most faithful to the literal meaning of the term ballot which
comes from the Italian diminutive of ballotta (ball). Here uniform papers (or balls or
marbles or rods) are placed by the voter in separate and individual candidate ballot boxes.
The box with the most ballots wins. This method Ð which is now effectively extinct Ð
was used widely in colonial India, Pakistan, Anglophone west Africa, western Nigeria,
Uganda, Kenya, British Guiana, Zanzibar and the Sudan in the mid part of the twentieth
century.27 Serbia used a party list system with separate ballot boxes in the decades
spanning 1900 but the most remarkable use of the single ballot-multiple box system
8
occurred in Hungary from 1848 til 1874. Here each candidate had a large box in his
colors which displayed his name. Upon entering the polling station the voter was given a
rod of 4-6 feet in length (to avoid the smuggling in of non-official rods) which they then
placed in the candidate box of their choice.28
The use of balls and multiple boxes was also a characteristic of voting schemes which
allowed for both a positive and negative vote. In ancient Athens the dicasts used balls of
stone or metal. Those pierced or black in color signified condemnation, those unpierced
or white signaled acquittal Ð a brass box received the votes and a wooden box the
discarded balls.29 Greece brought back a version of this method in 1864 in which
candidates would be allocated tin boxes split into two, one half with a white sack (for
yes) one half with a black sack (for no). Voters would drop a ball or leaden bullet in their
preferred compartment in each box.30 In Massachusetts after 1643 assistant legislators
were elected by the casting of Indian beans or corn Ð white for the candidate, black
against.31 In 18th century France a negative vote was possible although not required, but
any candidate receiving an absolute majority of negative votes cast was eliminated from
the election.32 Chinese village elections retain the option of a negative vote today Ð a
circle in the candidateÕs box indicates a positive vote, an X is a negative. Problems
arranging a secret vote with separate candidate ballot boxes and the difficulty of using
such a system when more than one vote needed to be cast meant that the uniform ballot,
separate candidate boxes system was phased out from its last vestiges in anglophone
Africa before most nations became independent in the 1960s.
9
The first two types remain in use today (89 countries use single ballot systems, 15 use
multiple party/candidate ballot methods, three countries operate both methods for
different elections). Single ballots with single boxes predominate, partly because all
other schemes had a high propensity for corruption and intimidation.
Nevertheless, even when the printed ÔAustralian ballotÕ is used corruption can find an
opening in special arrangements for illiterates. In Tanganyika, the Gold Coast and other
parts of colonial Africa the Ôwhispering voteÕ where illiterates whispered their vote to the
presiding officer who then marked the ballot paper was fraught with abuse in the 1940s
and 1950s. In Jordan today voters who claim they are illiterate whisper their vote to the
polling officer who then announces to the assembled party observers for whom the ballot
is being cast. In practice this allows candidates to buy a vote and then be reassured that
the bargain was kept. Apparently a number of highly educated voters mysteriously claim
illiteracy in the polling station.33
Proposals for rectifying these flawed methods of illiterate voting increasingly focused on
how to make the ballot papers themselves more ÔfriendlyÕ to illiterate voters and in the
second half of the twentieth century this entailed increased information on the ballot.
Such as the incorporation of party colors, symbols, and photographs along with the
broader issue of the lay-out of the ballot. It is the consequences of these innovations to
which we now turn.
10
The political consequences of ballot paper design
Layout and name ordering
The range of literature which directly deals with ballot paper design effects is limited but
there are a number of potentially useful texts which address related subjects and provide
insights into the probable motivations for symbol/color choice, their possible effects on
voter behavior and the psychology of politics and symbolism. The closest related ballot
design issue, which gives us some idea about the potential influence of symbols on ballot
papers, concerns name ordering or Ôalphabetic voting.Õ It is reasonable to hypothesize
that if the order of candidate name appearance on a ballot paper influences voting
behavior, then the much more politically charged practice of using logos, colors and
photographs is likely to have an even greater influence.
A number of studies have noted the impact of alphabetic voting, i.e., there is an
advantage in being at the top of a ballot paper as opposed to the bottom,34 and that the
effects are strongest when voters are less knowledgeable about politics and there is less
descriptive information on the ballot.35 This is due to the occurrence of ‘donkey voting’
where an elector will either, list the candidates (1 through n) from top to bottom on a
ballot paper in a preferential system of voting, or simply place an X by the top candidate
in a non-preferential system. Of the ballots surveyed for this paper 38 listed candidates
or parties alphabetically, 65 listed them randomly or by a non-alphabetical method, 24
ballots print the party’s ordered list of candidates. Austria, Brazil, the Netherlands and
Venezuela list parties by their performance in the previous election, South Africa holds a
11
lot to see which party will be at the top of the ballot and then lists parties alphabetically.36
Tasmania and many American states rotate the order and print different versions of the
ballot to ensure no single candidate benefits from alphabetical voting.37 There is also a
marked difference in voting behavior when voters can pull a single lever or mark a single
box to vote a straight party ticket for all the races being contested. In the U.S. some
states allow for party ticket voting and some do not.38 The three cases of the single
transferable vote – Ireland and Malta and the Australian Senate – offer an interesting
comparison. In Ireland candidates are listed alphabetically and not by party. In Malta
and Australia candidates are grouped by party (and in the case of the Australian Senate
given a straight party ticket voting option) and consequently there are far fewer transfers
between parties than there are in Ireland.39
Beyond name ordering there are a number of interesting cases which illustrate the
importance of the way the ballot is constructed. In NicaraguaÕs 1990 election the
eventual winners, the Union Nacional Opositora (UNO), were positioned first on a ballot
paper that listed parties from left to right (see figure 1). On the extreme right was a small
urban Trotskyite party (the Movimento de Unidad Revolucionaria [MUR]) which had
previously demonstrated only a few supporters in the capital Managua. However, in the
election itself the MUR polled its highest vote (and three times its national average) in
the poorest districts on the Caribbean side of the country where population concentrations
were low and illiteracy high. Because of the high illiteracy levels the UNO had asked
voters to make a cross on the first party on the ballot Ð however, apparently a number of
voters picked up the ballot the wrong way around (which was easy to do if the words
12
meant little) and thus the first party on an upside down ballot was the MUR. In
Mozambique in 1994 the parties were randomly listed from top to bottom but there were
separate draws for the Presidential and Parliamentary ballot. On the presidential ballot the
eventual winner, Joaquim Alberto Chissano of the Frente de Liberta•‹o de Mo•ambique
(FRELIMO), was listed third and FRELIMO used much of their publicity to urge voters
to vote for the third party from the top. But in the parliamentary ballot a previously
unknown party, the Uni‹o Democr‡tico (UD) was listed in third place. To the great
surprise of all observers the UD were the only party outside of the two main movements
FRELIMO and the Resist•ncia Nacional Mo•ambique (RENAMO) to surmount the 5%
threshold (they polled 5.1%) and win seats in the national assembly.40 In the second
elections of 1999 the ballot paper confusion was removed and the UD lost all their seats
by polling only 1.5% of the vote.41
Symbols
It is interesting to note the evolution of party symbols on ballot papers in the United
States. It has historically not always been the case that the Democrats are uniformly
represented by a donkey and the Republicans by an elephant. By the 1890s Colorado,
Delaware, Indiana, Maryland, Michigan and Ohio all had emblems on their ballot papers
and by 1916, 12 out of the 33 states used symbols. However, party symbols varied
widely from state to state. In 1906 in most places the Republican party used an eagle and
the Democrats a gamecock; but the Republicans used statue of Vulcan in Alabama, a log
cabin in Kentucky, an elephant in Louisiana, and a portrait of Lincoln with the flag as
background in Michigan, while the Democrats utilized a plough, flag, and a star in
13
various states.42 Some counties in some states continue to use symbols to this day. In
Alabama, Kentucky and Oklahoma the Democrats are represented by a rooster but in
Missouri they use the Statue of Liberty. In Alabama, Kentucky and Missouri the
Republicans use an elephant but in Oklahoma they are represented by an eagle. The
Libertarian party uses the statue of liberty in both Alabama and Oklahoma but the liberty
bell in Missouri.
For a number of decades, party and candidate symbols have been used to aid voting in
Africa, Asia and the Caribbean. Today over 50 countries print party logos on the ballot
and Ethiopia, Grenada, India, Bangladesh, Haiti, Pakistan, Nepal, Sri Lanka, St. Lucia,
and St. Vincent and the Grenadines use everyday objects as symbols to represent
individual candidates. In Zimbabwe the choice and allocation of ballot paper symbols
has been highly politically charged and the process exposed to a significant level of
abuse. All parties contesting elections in Zimbabwe since 1980 have been required to
register a symbol which appears on the ballot paper and on campaign materials. The
registration process is strictly controlled by the government and, displaying a good
understanding the impact of symbols, the President may prohibit any symbol without
being obliged to give any justification. In 1990 President Robert Mugabe unilaterally
declared 18 symbols prohibited which included representations of a lion, leopard, owl,
sword, cobra, flaming torch, elephant, cockerel and bull. However, the cockerel is the
symbol of Robert MugabeÕs ruling Zimbabwe African National Union (Patriotic Front)
party and has been since 1980. Indeed, many observers attributed ZANU(PF)'s success
14
in those first non-racial elections in part to the cultural significance and attractiveness of
their cockerel symbol on the ballot paper.43
Tanzania offers the best evidence on the impact of seemingly neutral candidate symbols
in elections. In 1965 two candidates from the ruling Tanzanian African National Union
(TANU) party ran in each of the 111 constituencies. One candidate was allotted the
symbol of a hoe (kjembe in Swahili) and the other a picture of a house (nyumba). Molnos
found that there was clear evidence that the symbol allocation affected voting behavior.44
The National Executive Committee (NEC) of TANU ranked the candidates from each
constituency and then alternated allocation of the hoe and house symbols between top
ranked candidates as they went down an alphabetical list of constituencies. The only
exception to this rule was that the symbols were to be evenly distributed between
government ministers. While these guidelines were kept to in the 1965 election in
subsequent elections, once the impact of symbols became known, the NEC's allocation
became more politically charged and biased.
The hoe and house symbols were supposed to be neutral and simply represent each
candidate “not by exact resemblance (but by an) accidental and conventional relation.”
i.e., they were to have no meaning in themselves rather they were to merely denote, for a
temporary period, one candidate or another. However, the hoe and the house already had
their own established cultural meanings and symbolic functions and could not perform a
simply denominative function. In a society orientated around subsistence farming, the
hoe was a powerful symbol of practical everyday life. The vast majority of Tanzanians
15
were farmers who relied on their hoe for a living and would only lend it to a close friend.
Throughout Tanzania the hoe was seen as an important symbol of work, manhood and
good citizenship and as essential key to obtaining wealth. In the Bukoba region the
people of the Haya had a proverb ÒEnfuka efukaÓ which stated that Òthe hoe is father of
everything.Ó The hoe became a more overt symbol of wealth in some regions as they
were used as a form of currency for bartering and as an essential part of a dowry - the
going rate was approximately six hoes for a wife. Furthermore, in the early post-colonial
period the hoe became closely associated with the President Julius Nyerere and the
founding of the new state.45
In contrast the other symbol, the house, had familiar associations with security and
shelter but it also conjured up ideas of cattle, fertility and the spirits of the dead. But the
house represented on the ballot paper was a Western style rectangular house with four
walls, a tiled sloping roof and windows. This was widely seen as a foreign modern
house, unsuitable for local conditions and viewed as a superfluous luxury beyond the
means of most Tanzanians. For many the drawing represented a shop or an official
building rather than a place to live. Molnos argued that “for many voters this house
became an anti-symbol of the traditional family system. The house standing alone, out of
sight of the other houses, without the cattle-byre…is either something incomplete or it
represents the modern nuclear family and with it the total breakdown of a still strong
tradition.”46
16
Molnos’s survey found that in Dar es Salaam over 50% of respondents could name their
representative in the National Assembly (which compared with an 15% equivalent level
of recognition in the United States at that time). But remarkably 93% of voters could
name the symbol used by the winning candidate and when illiterates were taken as a
separate polling group the symbol recognition factor still reached 75%.47 Candidates in
the 1965 election sought to play up their symbol association in agricultural areas if they
had the hoe and try and play down its significance if they were allotted the house. In
1965 hoe candidates won 53% of the time.
Three more elections followed between 1970 and 1980, all using the same hoe and house
symbols and displaying increasing success from those candidates allotted the hoe symbol.
In 1970 the hoe candidates won 62% of the time, in 1975 74% of the time and this was
not related to incumbency. By the 1980 election the public began to believe that the hoe
symbol was being deliberately allotted to candidates who were favored by the TANU
NEC. In 1980 all cabinet ministers (with the exception of one) were given the hoe
symbol and none were defeated.48 The evidence of symbol bias became so great that the
government was forced to drop the hoe and house symbols and move to a ballot paper
which showed a photograph alongside the names of the candidates.
Referendums
Single issue referendums often have the highest propensity for symbolic manipulation Ð
no more so than if the referendum deals with questions of secession, independence or
sovereignty. Such elections also highlight the question Ð who is to determine the ballot
17
format and choice of symbols? The March 1993 Eritrean referendum on secession or
retention within a greater Ethiopia, gave a massive 99.8% to 0.2% victory to those who
wished for independence. This victory was not unexpected but the huge margin may
have been partly influenced by ballot paper design. The ballot consisted of an instruction
section (in red), a No vote section (in red) and a Yes to independence section (in blue).
To vote the elector had to tear off one of these sections and place it in the ballot box.
United Nations election monitors argued that the dramatic red No section on the left hand
side of the ballot subtly disposed people to tear off the blue Yes section and cast a vote for
independence. Probably more decisive was the fact that for a right handed person it was
a natural action to tear off the right hand blue side of the ballot paper, as opposed to the
added manual dexterity it took to remove the right hand red No vote section. In East
Timor in 1999 the ballot design was less of an issue than the way the question was
drafted by the Indonesian authorities (although symbols were used). With the belief that
Asian voters were more inclined to say Yes rather than No, the Indonesian government
posed the question so that voting for independence required a No vote. Voting Yes to
autonomy meant remaining part of Indonesia, only a No vote would eventually lead to
independence.49
A number of African one-party states have held referendums over whether to move to a
multi-party democracy or not. In Sierra LeoneÕs 1991 referendum a Yes to multi-party
democracy was symbolized by three umbrellas (engendering a feeling of protection,
comfort) a No was represented by a vicious looking cutlass (perhaps less comforting) Ð
multi-partyism won with 60%. In Malawi in 1993 multi-partyism was represented by the
18
cockerel (which is closely associated with the new dawn, which has such significance in
Malawi that the currency is even called kwacha Ð chiChewa for Ônew dawnÕ). While,
retention of the one-party state was represented by a lantern (which rarely are as
illuminating as the sun) Ð again the multi-party option won with 65 per cent.
Photographs, Color and Supplementary Information
In developing world elections countries are increasingly using photographs on their ballot
papers, often as a supplement to party symbols, in order to make it absolutely clear to
whom the vote is going. Over thirty countries in the survey allow for the use of candidate
photographs and it is particularly prevalent for the presidential ballot papers analyzed –
22 out of 32. Photos on ballots are predominantly found in Africa and Latin American
countries but also in Cyprus, Papua New Guinea, Haiti and since 2001 the Republic of
Ireland. Papua New Guinea offers the most elaborate photographic clues on their ballot
paper. Since 1987 candidate photographs have appeared on the ballot and if the
candidate is endorsed by a party the party leader’s face appears alongside the candidate.
The double photo system was introduced to strengthen an exceptionally weak party
system but to date it has not succeeded in dramatically reducing party fragmentation.
Beyond symbols and photographs there are a number of other items which may be
included in ballot papers to provide further cues to influence voting behavior. Chief
among these is color which has long been used to denote party tickets. In 19th century
America parties used colors to differentiate the pre-printed party ballots Ð the
Republicans used a flaming pink border with rays projecting toward the center (a design
19
unlikely to be used today).50 Hungry, Bulgaria, France, Sweden, and Norway similarly
used different colors for each party ballot. In Pakistan prior to 1956 candidates were
assigned colors and the voter had to put their ballot in the colored box of their choice but
this system was abandoned because of color-blindness. 41 of the 133 papers categorized
in the appendix are printed in full (or multi) color and many of the other ballots in black
with one other color.
Psychological experiments have found that different colors can influence reactions;51
responses are conditioned by gender,52 and by culture.53 Rubinoff and Marsh conducted
an experiment to test for the impact of color on ballots and found that the color
presentation of candidates effected perceptions of the candidateÕs attributes.54 White on
blue and white on green tended to engender positive ratings, for males the highest ratings
for extroversion and experience were given to those candidates printed in white on blue,
while orange on blue received the lowest rating. For women, white on blue was
associated with honesty, the lowest ratings on this dimension were again given to orange
on blue candidates. Lastly, males found candidates most competent when they were
presented as black on yellow, least competent for when white on red. Garrett and Brooks
studied the influence of ballot color on sex of candidate and sex of voter and found that
male voters had a bias for candidates on green paper while female voters preferred
candidates on pink paper.55
Language can also be an issue in ballot presentation. Belgium, Bosnia, Canada, East
Timor, Estonia, Eritrea, Finland, Hong Kong, India, Ireland, Israel, Malta, Namibia, New
20
Zealand, Norway, Papua New Guinea, Russia, South Africa, Spain, Sri Lanka, Sweden,
the United Kingdom, the United States,56 and Zimbabwe all run instructions in multiple
languages on the ballot paper or print separate ballots in different languages as needed.
Some countries (such as Austria and Russia) print the age of the candidate, while many
anglophone countries list the candidateÕs profession and address. In the US, ballots in
Iowa, Maine and Vermont print the presidential candidateÕs home state. Keeley and
McAllister found that the printing of honorary titles in the UK (as is also the case in
Austria, Germany, and Papua New Guinea) garners the candidate an appreciable
differential vote - although, unfortunately, an academic title is of no benefit.57
Psychological elements of ballot design
Over the last few decades a body of literature has accumulated which highlights the
psychological impact of political symbolism,58 and the influence of Ôvoting cuesÕ in
multi-candidate and multi-proposition elections.59
Zaller and Feldman found voters
answer survey questions by sampling, from the top of their minds, the most salient
preferences they have on a particular issue. Their Ôambivalence axiomÕ notes that
individuals have competing preferences which are often contradictory. The Ôresponse
axiomÕ describes how the design of survey questions can influence which preferences
become salient in the respondents mind and therefore help determine the respondents
answer to any given question. The model helps explain inconsistency in survey response
over time from the same individuals.
21
Ballot paper design - symbols, photographs, colors - are a form of Ôresponse moldingÕ
applied directly at the point of sale when it comes to voting preferences. Symbols can
influence which preferences become salient to a person at their moment of voting. For
example, a voter, who happens to be Christian, is faced with a mix of Christian, Muslim
and secular parties on the ballot. Zaller and Feldman would argue that, if symbols were
introduced on the ballot and the Christian party used a particularly obvious symbol such
as a crucifix, then the Christian voter would be more likely to choose the Christian party.
The voterÕs Christianity being made salient to them just before they voted, and the more
ambivalent they were the more likely this Ôpreference saliencyÕ would have an influence
on their choice.
Popkin identifies symbols (defined in the wider sense) as an important source of
information to potential voters and describes how in America socio-political symbols
“serve as information short cuts, as cognitive placeholders and focal points,” and are used
as a direct channel to sway the voting public.60 This builds on Downs’s earlier
observation that voters rely upon information shortcuts because they have neither the
time nor inclination to research each candidate and party to acquire full information.61
Popkin argues that these shortcuts are “by no means devoid of substantive content”
because they are rooted in a whole concoction of “past experiences, daily life, the media,
and political campaigns.”62 Lupia found that information shortcuts (such as symbols)
allowed badly informed voters to emulate the behavior of relatively well informed voters,
while McDermott notes that race and gender can also act as powerful cues (which might
accentuate the impact of photographs on the ballot). Likewise symbols act as ‘fire
22
alarms’ in triggering negative experiences which guard a voter from choosing a party
which had wronged them in the past.
Popkin recognizes that symbols do not always influence voters to act in ways which we
might predict and that the importance of particular symbols are varied across time and
between different segments. But he notes that Òjust as our party identification can remind
us what to do when we have no other information, our symbols and principles orient us
when we have no other information. The inconsistencies that exist between the general
and the particular are what make campaigns and the choice of issues and symbols so
important.Ó63 At their most powerful symbols can Òre-create our memories, literally
ÔinventingÕ our history.Ó64 A good example was the strategy used by the RussiaÕs Choice
party which contested the December 1993 Russian parliamentary elections under the
leadership of First Deputy Prime Minister Yegor Gaidar with the public backing of
President Yeltsin. Many of the RussiaÕs Choice leadership were former officials of the
Communist Party but they sought to counter both this connection and their newness by
designing a party logo which was a dramatic representation of Peter the Great on
horseback - presumably hoping that this would define them as anti-communist and
inheritors of a proud and powerful Russian historical tradition. Very few reformed
communist parties in Europe continue to use the hammer and sickle.
Explaining the occurrence of friendly ballots
Appendix A illustrates that the inclusion of symbols, color, and photographs are
predominantly a characteristic of Latin American countries while party or candidates
23
symbols are also found in the former British colonies of Asia and the Caribbean and
European nations which have a Mediterranean coastline. Multi-lingual provisions are
more a characteristic of established western democracies and those new multi-party
systems where language has been a salient and divisive issue (e.g., Estonia, Bosnia, and
South Africa). OLS regression estimates (n=133) suggest that ballot friendliness65 is
higher where literacy is lower and competitive multi party elections are a new
phenomenon. Friendliness is not significantly related to the level of democracy, the
effective number of parties, or the type of electoral system.
Ballot papers for presidential elections are on the whole more elaborate than those for
legislative elections (presidential systems are disproportionately used in developing
world countries). A full 22 of the 32 (69%) presidential ballots categorized in appendix
B show candidate photographs (only 9 of 90 [10%] legislative ballots include photos).
Presidential ballots are slightly more likely to include symbols (22/32 [69%] as opposed
to 56 of 90 [62%] for legislative ballots) and be in full color (16/32 [50%] as opposed to
22 of 90 [24%] for legislative). In legislative elections only 10 per cent of FPTP ballots
are in color, while 34 per cent of List PR ballots are in full color. 60 per cent of FPTP
ballots utilize party or candidate symbols, while 65 per cent of List PR ballots show party
symbols.
Do friendly ballots reduce spoilt rates?
Using data on ballots and recent elections in 117 nations, we analyze the impact of ballot
friendliness on rates of spoilt ballots. An OLS model which regressed ballot friendliness,
literacy rates, the electoral system (a continuum of proportionality), dummy variables for
24
regions, and compulsory voting found that the effect of ballot friendliness on spoilt
ballots66 was substantively small and not statistically significant (see Table 1).67 While
literacy was significant, its effect on spoilt ballots was dwarfed by the impact of
compulsory voting and the dummy variable for South America. The electoral system
also played a part, with more majoritarian systems being associated with less spoilage -but this was not related to complexity of the vote.
The lack of a more substantive correlation with ballot friendliness suggests that the
incorporation of symbols, photos and color does little to affect the number of ballots
which are marked improperly. The fact that the South American dummy shows up so
strongly Ð and there is a well developed literature that points to the exceptionally high
spoilt rates on that continent as indicators of political alienation Ð suggests that spoilt
ballots have much more to do with protest than error.68
Table 1: Explaining Spoilt Ballots
Coefficient
Friendly
Electoral System
Compulsory
South America
Literacy
Constant
-0.1543
-0.3347
5.897
3.287
-0.0902
11.0925
Standard
Error
0.2948
0.1468
1.329
1.365
0.0249
2.455
P>|t|
0.602
0.025
0.000
0.000
0.018
0.000
Implied
Effect, 1
Standard
deviation
change
-0.3581
-1.1302
-2.057
N=117
F statistic=10.84 (Prob > F = 0.0000)
Adjusted R-squared: 0.30
All tests are two-tailed.
25
The Impact of Ballot Design on Voting Behavior: A controlled experiment
Through April-October 2002 we conducted a ballot design experiment with
undergraduate students at the Universities of North Carolina, Chapel Hill and Notre
Dame. In nine separate 35 minutes sessions a total of 401 students went through the
process of privately reading a script (appendix C), voting in three separate elections
(Presidential, Parliamentary and a referendum) and filling out a short questionnaire about
themselves. To control for existing political preferences we invented a nation, its history
and geography, two ethnicities and two political parties.
The island nation of Kamuzu is a plantation society divided between Blues and Pinks
who are also to some degree separated by geography and income (see the country profile
in appendix C). Blues are the dominant group on the mainland but Pinks are in the
majority on the offshore island of Babu (see the map in appendix D). There are two
political parties. The Kamuzu PeopleÕs Party (KPP) predominantly draws its support
from wealthy Kamuzuians and those on the mainland (both codes for Blues). It has won
every election since independence. It believes in, (a) a capitalistic free market economy,
and (b) opposes the secession of the island of Babu from the Kamuzu mainland. The
Babu Front (BF) predominantly draws its support from inhabitants of Babu island (a code
for Pink). It has lost every election since independence. It believes in, (a) a socialistic
State run economy with subsidies for the poor working class, and (b) independence for
the island of Babu. The participants were randomly assigned an ethnicity (Blue or Pink),
an economic status (wealthy or poor) and a location (mainland or the island of Babu).
26
We presented the students with three ballots (they either received all black and white
ballots, all coloured, or all colour plus symbols ballots, See appendices E, F and G).
There were three elections -- one for the presidency and one for the parliament -- and a
referendum on independence for the island of Babu. In the presidential election voters
were asked to place an ÔXÕ by the candidate of choice. In the Parliamentary election they
had 3 votes for three candidates. The referendum was posed: ÒDo you believe the island
of Babu should become independent? Ð Yes or No?Ó
Our aim was to engender ethnic block voting but assign some of participants with cross
cutting characteristics which would give them a reason to defect from their ethnic party.
These involved economic and geographical traits which we hypothesized would effect
the presidential/parliamentary and referendum votes respectively. We hoped to create
colour and colour/symbol ballots which clearly reinforced and made salient the
participants ethnicity on the ballot paper. On the referendum ballot we went further and
attempted to design the symbols representing independence for the island of Babu and
retention of the national state in a way which would make a YES vote less appealing than
a NO vote. As can be seen in the ballot (appendix G) voting Yes to independence was
represented by a shattered map of the islands with a large red X through the center.
Voting NO, against independence for Babu, was represented by a less jarring and warmer
blue encircling both land masses in a show of togetherness. Participants were randomly
assigned to these three ballot types Ð black and white, colour, and colour/symbol.
27
Our overall hypotheses were as follows:
(1) Blues are likely to vote KPP and Pinks are likely to vote BF. This is even more likely
if their ethnicity is made salient to them on the ballot paper through colour and
symbols.
(a) The most likely defectors should be Poor Blues and Wealthy Pinks.
(2) Wealthy Blues and Poor Pinks are most likely to vote a straight party ticket in the
parliamentary ballot. They are even more likely to do this is their ethnicity is made
salient through colours and symbols.
(3) Blues are likely to vote against BabuÕs secession and Pinks in favour. This is even
more likely if their ethnicity is made salient to them on the ballot paper.
(a) Pinks might defect from the BF party position at a higher rate because of the
negatively of the symbol representing the YES (to independence) vote.
We should emphasize that we do not expect to find massive effects of ballot design. Our
experiment follows in the tradition of the Òminimal groups paradigmÓ of Tajfel.69 That is,
the group identities that we created were based on fairly arbitrary distinctions Ð
participants were randomly assigned to an identity, acquiring this identity for the
relatively short duration of the experiment. While Tajfel has shown that minimal groups
can create distinct social identities, which influence behavior toward ingroups and
outgroups, it is less clear that we should expect large effects on voting behavior.
Specifically, allusions to these identities in the ballot designs (the use of blue/pink colors)
should produce relatively subtle effects because of the short time that participants
28
acquired their identities. Thus, we expect certain patterns in voting behavior across
ballots, but we do not expect massive differences in voting behavior.
Results
The results confirm (to varying degrees) all of our hypotheses. In both the presidential
and parliamentary elections we noted a significant degree of ethnic block voting Ð 73%
(82% Blue and 64% Pink) in the presidential race, while in the parliamentary race 40% of
voters gave all three of their votes to their ethnic partyÕs slate (36% of Blues, 43% of
Pinks). We can assess the impact of ballot design and the salience of ethnicity by
measuring defectors in the presidential and referendum ballots and straight party ticket
voting in the parliamentary block vote election. Hypothesis (1)(a) Ð that Poor Blues and
Wealthy Pinks would be more likely to defect Ð was dramatically confirmed throughout
all three elections and the rates of defection were differential with ballot design. As we
expected black and white ballots (without the visceral colour message of ethnic identity)
consistently allowed the space for the most defections while ballots in colour and with
colour symbols increased ethnic party loyalty (see Tables 2-4).
In the presidential vote Poor Blues and Wealthy Pinks were much more likely to defect
regardless of ballot paper but in both cases they were most likely to defect when voting
on the black and white ballot paper (although the disparity is more significant among
Pinks rather than Blues) [see Table 2]. Indeed a full 76% of Wealthy Pinks defected to
the KPP on the black and white ballot.
29
Table 2: Defectors in the Presidential Vote
Trait
Blue-Wealthy
Blue-Poor
Pink-Wealthy
Pink-Poor
Black & White
12.5 % (N = 48)
45.8% (N = 24)
76.0% (N = 25)
28.0% (N = 50)
Color
5.5% (N = 91)
35.0% (N = 40)
47.7% (N = 44)
22.8% (N = 79)
Note: Table entries indicate the percentage of defectors within a trait category.
The parliamentary election followed a similar pattern. Here we tracked loyal Ôyellow
dogÕ party voting where a Blue would vote for all three candidates of the KPP or a Pink
would vote for all three candidates of the BF. While the multiple vote system did give
voters more room to cut across ethnic lines by giving one (or more) of their three votes to
a candidate of a different ethnicity, straight party ticket voting was still significant. We
found highest levels of Ôyellow dogÕ voting among Pinks and Blues when using the
colour ballots and the lowest levels of block loyalty when using the black and white
ballots (see Table 3). 59.5% of Poor Pinks (the BFÕs expected most loyal voters) voted a
straight party ticket when using the colour ballot.
Table 3: Straight Ticket Voting in Parliamentary (Block Vote) Election
Trait
Blue-Wealthy
Blue-Poor
Pink-Wealthy
Pink-Poor
Black & White
45.8% (N = 48)
8.3% (N = 24)
16.0% (N = 25)
47.9% (N = 50)
Color
47.3% (N = 91)
15.0% (N = 40)
25.0% (N = 44)
59.5% (N = 79)
Note: Table entries indicate the percentage of straight-ticket voters within a trait category.
The results of the referendum ballot were less dramatic but demonstrated evidence for
hypotheses (3) and (3)(a) (see Table 4). Blues were more likely to vote against BabuÕs
secession (79%) and Pinks in favour (48%) and defections from this position were found
more commonly on black and white ballots. When ethnicity was given salience on
30
colour ballots the levels of defection were the lowest. Hypothesis (3) (a) was also given
credence. Pinks defected from the BF party position at a higher rate than Blue defectors
and defected at almost as high a rate using the colour and symbols ballot as they did with
just the colour ballot.
Table 4: Defectors in the Referendum Vote
Trait
Blue-Wealthy-Mainland
Blue-Poor-Mainland
Blue-Wealthy-Island
Pink-Poor-Island
Pink-Wealthy-Island
Pink-Poor-Mainland
Black & White
26.1% (N = 23)
25.0% (N = 24)
76.0% (N = 25)
48.0% (N = 25)
72.0% (N = 25)
52.0% (N = 25)
Color
18.2% (N = 22)
22.2% (N = 18)
85.7% (N = 21)
27.8% (N = 18)
55.0% (N = 20)
52.6% (N = 19)
Color & Symbols
16.7% (N = 24)
33.3% (N = 21)
79.2% (N = 24)
47.8% (N = 23)
41.7% (N = 24)
31.6% (N = 19)
Conclusion
There is scant evidence to suggest that elaborate ballots have a great impact on the ÔeaseÕ
of voting but plenty of evidence that they are politically charged cues to voting behavior.
If that is indeed the case psephologists might usefully pay attention to ballot design as a
response molding factor which influences vote shares and pay attention to who actually
designs the ballot in the first place. The choice of party political symbols for ballot
papers will be increasingly important as developing world countries move to multi-party
democratic electoral systems. A symbol may be most electorally attractive if it contains
some, or all, of these properties: (i) it carries a high recognition factor among the society
as a whole. (ii) It is easily communicable and is rooted in some political or cultural
tradition which is at the very least neutral in the eyes of the electorate. (iii) It is attractive
enough to win votes if other political considerations were removed. (iv) It is not a
Ôculturally insensitiveÕ symbol which would actually lose the party votes of people whom
it would have won if no symbol had been used at all.70
31
The evidence suggests that future innovation in ballot design and modes of voting will
have electoral consequences. Voting by mail is already a fixture in Oregon, Toronto, and
for a multitude of non-governmental elections but postal votes may not use different
ballots than those given out at a polling station. Telephone voting, as used in Nova
Scotia’s provincial leadership election in 1992, the Reform party’s presidential
nomination process, and parts of Toronto, may by-pass entirely the need to present
candidates/parties on a sheet of paper.
However, the development of
computerized/internet voting may dramatically open up the way candidate options can be
presented to voters – computers will be able to present candidates in a more extravagant
and potentially interactive way. The US Reform Party already allows for email voting
and Arizona Democrats tested a system in their presidential primary of 2000. Both
California and Minnesota have set up commissions to investigate the viability of internet
voting. But perhaps of more realistic import to the vast majority of new democracies
who are light years away from computer voting is the expense of sustaining expensive
ballots once the wealthy donor community has left the euphoric scene of the ‘first’
democratic election.
32