Mlaga Economic Theory

Research Center
Working Papers

Indicators of Electoral Victory

Pablo Amoros and M. Socorro Puy

WP 2008-8
April 2008

Departamento de Teora e Historia Econmica

Facultad de Ciencias Econmicas y Empresariales
Universidad de Mlaga

Indicators of Electoral Victory

Pablo Amors and M. Socorro Puyy
Universidad de Mlaga
April 14, 2008

Abstract
We study a two-party contest where candidates strategically allocate their campaign resources between two salient issues. We analyze
to what extent the following indicators of a partysuccess predict the
electoral victory: (1) the pre-campaign advantage, (2) the advantage
on every salient issue, and (3) the advantage on campaign resources.
We show that the electoral victory is guaranteed only when a party
has a su ciently large advantage on every salient issue. Otherwise
no combination of these indicators ensures the electoral victory.
Key-words: Election campaign, salient issues, majority voting.
JEL classication numbers: D72, C70

A previous version of this paper was entitled A model of political campaign manipulation. The authors thank Enriqueta Aragons, Carmen Bevi, and Luis Corchn for
their helpful comments. Financial assistance from Ministerio de Educacin y Ciencia under project SEJ2005-04805, and Junta de Andaluca under project SEJ01645 is gratefully
acknowledged.
y
Corresponding author: M. Socorro Puy, Dpto. Teora e Historia Econmica, Facultad
de Econmicas. Campus El Ejido, Universidad de Mlaga, 29013, Mlaga, Spain. Fax:
+34 95 213 1299. E-mail: [email protected]

Introduction

Political parties success indicators measure the chances that a party has
of achieving a majority on election. Having more chances of winning the
elections is usually attributed to (1) the pre-campaign advantage, (2) the advantage on every salient issue, and (3) the advantage on campaign resources.
In this paper we analyze from a theoretical viewpoint whether any of these
indicators predicts the electoral victory of a political party. We consider
a two-party, two-dimensional spatial model of political competition. For
the duration of the political campaign, partiesplatforms have already been
decided. We follow the emphasis theory to describe partiescompetition in
the political campaign. The emphasis theory, as described by Page (1976),
argues that "political information is imperfect and there are limits on the
number of messages that candidates can transmit or that the average voter
can or will receive. Candidates must allocate their emphasis (in time, energy,
and money) among policy stands and other sorts of campaign appeals". As
proposed by Page and formalized by Simon (2002) and Amors and Puy
(2007), parties strategies for the duration of the political campaign aim
at emphasizing the feature (in terms of issues) of the political party that
attracts more votes.1 In our model, parties campaign strategies consist of
emphasizing some political aspects more than other.
According to our model, the electorate can be partitioned into three
groups: partisan voters (those who will vote for one of the parties irrespectively of the parties campaign strategies), issue voters (those whose vote
depend on the campaign strategies, and whom the parties aim at inuencing
via campaign expenditures), and abstention voters (those who are indierent
between both political parties).
We show that the only indicator that can be used to predict the electoral
victory is the advantage on the salient issues. A su ciently largeadvantage
on every salient issue guarantees the electoral victory as far as it guarantees
that a party has a majority of partisan voters. When no party has a majority
of partisan voters, the issue voters come into the scene and every kind of
mischief regarding the electoral results can occur. In particular, we nd
that a party may be majority-defeated in the elections even if it is the precampaign winner, has more campaign resources than its opponent, and has
1
Among others, Laver and Hunt (1992), Budge (1993), Riker (1993), Petrocik (1996),
Simon (2002), and Sigelman and Buell (2004) show that there is empirical evidence of
candidates emphasizing some issues more than others.

an advantage on every salient issue (as long as the sum of these advantages
is not too large).
The later result is related to the literature on the Ostrogorskis paradox
(Daudt and Rae 1976), originating from Ostrogorski (1902). This literature postulates that the candidate with a majority on every single issue can
be majority-defeated in a representative democracy.2 We depart from the
assumptions of this paradox in two main respects: voters preferences and
partiesbehavior. While the Ostrogorskis paradox assumes that voters vote
for the party with which they agree on more rather than fewer issues, we
assume that voterspreferences have intensities over issues and they vote for
the party that match their own ideal policy more accurately. Concerning the
political parties, while the Ostrogorskis paradox considers static parties, we
consider that parties behave strategically allocating emphasis across issues.
The rest of the paper is organized as follows. Section 2 introduces the
model. Section 3 analyzes the success indicators. Section 4 provides the
conclusions.

The model

There is a society with a nite set of voters, N = f1; :::; ng, which will select
by popular election a representative to serve in the legislature. There are two
political parties, A and B, that compete for winning a majority of the votes
by spending campaign resources. There are two political issues, 1 and 2.3
Each party j 2 fA; Bg has a political position xj = (xj1 ; xj2 ) 2 [0; 1]2 ;
where xjr 2 [0; 1] is the political position of party j on issue r 2 f1; 2g. We
assume that parties political positions are dierent on both issues: xA1 6=
xB1 and xA2 6= xB2 .4 Each party j is endowed with some xed campaign
funds cj
0. Campaign funds are devoted to the advertising campaign.
A campaign strategy of party j is a vector cj 2 Cj = f(cj1 ; cj2 ) 2 R2+ :
cj1 + cj2
cj g, which indicates how the party allocates its funds between
2

There are several posterior analyses that compare the results obtained in the elections
under representative democracy and those obtained by majority voting in issue-by-issue
elections (e.g., Bezembinder and Van Acker, 1985, Kelly, 1989, and Laond and Lain,
2006).
3
The results of this paper can be generalized to the case of m 2 issues. For simplicity
of exposition we focus on the case of two issues.
4
If the political positions of both parties on issue r are identical, then issue r becomes
innocuous according to the emphasis theory.

the two dierent issues. Let c = (cA ; cB ) 2 CA CB = C denote a prole of

campaign strategies. For each c 2 C and each r 2 f1; 2g, let cr = cAr + cBr
be the total funds spent on issue r.
Each voter i 2 N has an ideal political position i = ( i1 ; i2 ) 2 [0; 1]2 .
Preferences of voter i over political parties are represented by the following
utility function:
ui (j; c) =

1 (c1 )[xj1

i1 ]

2
i2 ] :

2 (c2 )[xj2

(1)

For each issue r, r (:) is a strictly increasing function of the campaign

expenditures on that issue, with r (0) > 0. We refer to r (:) as the inuence
function on issue r. It indicates the weight that voters assign to that issue.
The inuence functions on each of the issues may be dierent.
Voter i casts his ballot for party j 6= k when ui (j; c) > ui (k; c): We
suppose that those voters who are indierent between the two parties abstain from voting. Given any prole of campaign strategies c 2 C, let
Vj (c) be the number of votes that party j obtains in the elections, i.e.,
Vj (c) = # fi 2 N : ui (j; c) > ui (k; c)g.
The utility function of voter i can be rewritten as:
ui (j; c) =

T (c) [xj1

i1 ]

[xj2

i2 ]

(2)

1)
is the relative intensity of the preferences of voters over
where T (c) = 12 (c
(c2 )
issue 1 when the prole of campaign strategies is c. The greater T (c), the
more relevant is issue 1 compared to issue 2 in voterspreferences.
Voter i is indierent between the two parties (and then he abstains from
voting) when his ideal political position satises the following condition:

i2

T (c) [x2A1 x2B1 ] + [x2A2

2 [xA2 xB2 ]

x2B2 ]

T (c) [xA1 xB1 ]

[xA2 xB2 ]

i1 :

(3)

The line dened by Equation (3) divides the policy space [0; 1]2 into two
regions based on the voters most preferred party. Since each party j can
+cB )
spend at most cj , we have 2 (c1A(0)
6 T (c) 6 1 (c2A(0)
for all c 2 C. We
+cB )

+cB )
denote Tmin = 2 (c1A(0)
and Tmax = 1 (c2A(0)
the minimum and maximum
+cB )
values of T (c).
Every voter i such that ui (j; c) > ui (k; c) (with j 6= k) for all c 2 C always
votes for party j, no matter what the prole of campaign strategies is. We
call these voters partisan voters of party j.

There can be voters such that ui (A; c) > ui (B; c) for some c 2 C and
ui (B; c0 ) > ui (A; c0 ) for some c0 2 C. These voters cast their ballots for one
or the other political party, depending on the campaign strategies. We call
these voters issue voters.5
Given any prole of campaign strategies c 2 C, party j wins the elections
if Vj (c) > Vk (c) (and therefore, party k is majority-defeated). For simplicity,
we assume that when parties are involved in a tie party A governs.

Tmax

Issue 2
1

T1

Tmin
(xA2-xB2)/2

A
1

1
(xA1-xB1)/2
B

Issue 1

Tmin

T1

Tmax

)[

Figure 1. An example of winning partition

Interval [Tmin ; Tmax ] can be partitioned into dierent subintervals according to the party that wins the elections. We call it the winning partition of
[Tmin ; Tmax ]. Figure 1 provides an example to illustrate this concept. There
5

Some authors assume that the set of voters inuenced by campaign expenditures is a
xed fraction of uninformed voters (see, e.g., Baron, 1994, and Grossman and Helpman,
1996).

are three voters with ideal political positions 1 , 2 and 3 . The political
positions of the parties are represented by points A and B. Abusing notation we use Tmin (resp. Tmax ) to denote the line dened by Expression (3)
when T (c) = Tmin (resp. T (c) = Tmax ). The winning partition of [Tmin ; Tmax ]
consists of two subintervals: [Tmin ; T1 ) and [T1 ; Tmax ]. To see this note that
for all T (c) 2 [Tmin ; T1 ) the winning party is B (since voter 1 casts his ballot
for party A and voters 2 and 3 cast their ballots for party B). Similarly, for
all T (c) 2 [T1 ; Tmax ] the winning party is A (if T (c) = T1 there is a tie).
Political parties aim at winning the elections. Preferences of each party
j 6= k are represented by the following utility function:
1; if Vj (c) > Vk (c)
0; if Vj (c) < Vk (c)
and wA (c) = 1; wB (c) = 0 if VA (c) = VB (c):

(4)

wj (c) =

We are interested in analyzing under which circumstances a party can

ensure its victory in the elections. This is the idea behind the concept of
dominant strategies. A campaign strategy of party j, cj 2 Cj , is a (weakly)
dominant strategy for that party if wj (cj ; ck ) > wj (c0j ; ck ) for all c0j 2 Cj
and all ck 2 Ck (j 6= k). Note that, in particular, cj 2 Cj is a dominant
i
h
1 (cj1 +ck )
1 (cj1 )
strategy for party j that ensures its victory if and only if 2 (c +ck ) ; 2 (c )
j2

j2

is included in a subinterval of the winning partition where party j wins.6

If a party achieves a majority of the votes for every T (c) 2 [Tmin ; Tmax ] the
problem is trivial. From now on, we assume that the distribution of voters
ideal political positions is such that the winning partition has at least two
subintervals.7
6

The concepts of dominant strategies and Nash equilibria are closely linked: (cj ; ck ) is a
Nash equilibrium of the campaign game where party j wins if and only if cj is a dominant
strategy that ensures js victory.
7
In particular, if the winning partition has two subintervals, then there is T1 2
[Tmin ; Tmax ] such that party j wins for all T (c) 2 [Tmin ; T1 ] and party k wins for all
k)
T1 , cj = (0; cj ) is a dominant strategy for party
T (c) 2 (T1 ; Tmax ]. In this case, if 12(c
(cj )

j that ensures its victory, and if

k that ensures its victory.

1 (ck )
2 (cj )

> T1 , ck = (ck ; 0) is a dominant strategy for party

Results

Next, we analyze to what extent three success indicators ensure that a political party cannot be majority-defeated in the elections. These indicators
are: the advantage on campaign resources, the pre-campaign advantage, and
the advantage on every salient issue.
3.1 The advantage on campaign resources
Having more campaign funds than the opponent does not guarantee the
electoral victory, even if the dierence is extremely large. To see this note
that when a party has a majority of partisan voters, such party wins the
elections for all c 2 C, even if it has no campaign resources.
3.2 The pre-campaign advantage
The pre-campaign advantage measures the percentage of votes that a
party would obtain if there were no political campaign. We write c = 0 to
denote the situation where there is no campaign expenditures. Let T0 = 12 (0)
(0)
be the relative intensity of voterspreferences over issue 1 in this situation
(note that Tmin < T0 < Tmax ). The party that achieves a majority in this case
is the pre-campaign winner. The pre-campaign advantage does not guarantee
the electoral victory.
Proposition 1 There is no pre-campaign advantage that guarantees that a
political party wins the elections.
Proof: Suppose that all votersbliss points coincide, and that there is T1 2
[Tmin ; Tmax ] such that all voters vote for party j when T (c) 2 [Tmin ; T1 ), and
vote for party k when T (c) 2 (T1 ; Tmax ]. Suppose that T1 < T0 . Then party
k would obtain 100% of the votes if there where no political campaign (i.e.,
party k has the maximum pre-campaign advantage). However, there always
exist some inuence functions and amounts of campaign funds such that
1 (ck )
< T1 . In this case cj = (0; cj ) is a dominant strategy for party j that
2 (cj )
ensures its victory.
3.3 The advantage on every single issue
We say that party j has a majority on issue r when it wins the hypothetical election where individuals only care about that issue. Let njr be the
number of voters that, on issue r; strictly prefer party j to party k. Formally,
7

2
2
njr = # i 2 N : [xjr
ir ] > [xkr
ir ] . Party j has a majority on issue
r when njr > nkr (j 6= k).8
Having a majority on both issues does not guarantee the electoral victory.
The electoral victory is guaranteed only if a party holds a su ciently large
majority on both issues.
n +n

Proposition 2 Suppose that party j has a majority on both issues. If j1 n j2 >

3
, then party j wins the elections for all c 2 C (whatever the campaign funds
2
cA ; cB
0 and the inuence functions 1 and 2 are). Otherwise, the electoral victory of party j is not guaranteed.9
Proof. (See Figure 2) Let nAr;Bs
0 be the number of voters that prefer
party A on issue r, but prefer party B on issue s, i.e., nAr;Bs = #fi 2 N :
2
2
2
2
[xAr
ir ] < [xBr
ir ] and [xAs
is ] > [xBs
is ] g. Let nj = #fi 2
2
2
N : [xjr
[xkr
ir ]
ir ] for all r 2 f1; 2g, with strict inequality for
some r 2 f1; 2gg. Note that, since individuals take into account both issues
simultaneously, the nj voters will always vote for party j (i.e., the nj voters
are partisan voters of party j). Suppose that, for some party j, nj1 +nj2 > 32 n:
Since nj1
nj + nj1;k2 , nj2
nj + nj2;k1 , and nj + nj1;k2 + nj2;k1
n, we
have n2 < nj and therefore party j will win the elections for every c 2 C.10
n +n
3
In Figure 2 we provide an example showing that, if j1 n j2
, then the
2
electoral victory of party j is not guaranteed even if it has a majority on
both issues. The political positions of the parties are given by points A and
B. There are four groups of voters: 50% of the voters have ideal political
position 1 , 25% of the voters have ideal political position 2 , and 25% of the
voters have ideal political position 3 . Note that nnB1 = 0:75 and nnB2 = 0:75.
Suppose that the campaign funds, cA ; cB
0, and the inuence functions,
1 and 2 , are such that Tmin and Tmax are as depicted in Figure 2. In this
case, for every c 2 C, party A obtains 50% of the votes.
Proposition 2 shows that a su ciently large advantage on each of the
issues guarantees the electoral victory. If the sum of votes that a party
8

We suppose that, if [xAr

ir ] = [xBr
ir ] , then voter i abstains from voting in
the hypothetical one-issue elections.
9
The fact that nk1 + nk2 < n2 does not necessarily imply that nj1 + nj2 > 23 n, since
some voters could abstain in the hypothetical one-issue elections.
m
P
10
This result can be easily generalized to the case of m 2 issues: if
njr > 2m2 1 n,
r=1

then party j wins the elections, no matter how much campaign funds the parties have.

obtains in the hypothetical one-issue elections is greater than 32 n, there is no

way of defeating this party in the elections since it has a majority of partisan
voters. Note that a particular instance for that condition is given when the
advantage of a party on every single-issue is above the qualied majority of
three-quarters.

Issue 2
Tmax
n*B

nA1,B2
Tmin

(xA2-xB2)/2

B
1(50%)

2(25%)
A

nA2,B1

n*A
3(25%)

(xA1-xB1)/2

Issue 1

Figure 2. Illustration of Proposition 2

When the advantage of a party on single issues does not satisfy the condition stated in Proposition 2, the paradox presented by Ostrogorski reappears.
The Ostrogorskis paradox postulates that the party that has a majority on
every single issue may be majority-defeated in the elections by its rival.
If we put together the three success indicators that we are analyzing,
we nd a stronger version of the Ostrogorskis paradox: a party may be
majority-defeated in the elections even if it has a majority on every single
issue, is the pre-campaign winner, and has more campaign funds than its
rival.
9

Example 1 (Stronger version of Ostrogorskis paradox). The parties have

the following political positions and campaign funds:
Party A : xA = (0:3; 0:3) cA = 16
Party B : xB = (0:7; 0:7) cB = 9
There are three voters with ideal political positions 1 = (0:1; 0:1), 2 =
(0:2; 0:77), and 3 = (0:8; 0:44). The inuence functions are 1 (c1 ) = 1 +
p
p
c1 ; 2 (c2 ) = 1 + 2 c2 . The relative intensity of voters preferences over
+cB )
1
issue 1 may vary between Tmin = 2 (c1A(0)
= 11
and Tmax = 1 (c2A(0)
= 6,
+cB )
depending on the campaign strategies c 2 C. Given the preferences of voters
over political parties, we have that:
Voter 1 is the only partisan voter of party A.
Voter 2 votes for party B when T (c) 2 [Tmin ; 0:9), and votes for party A
when T (c) 2 (0:9; Tmax ].
Voter 3 votes for party A when T (c) 2 [Tmin ; 0:2), and votes for party B
when T (c) 2 (0:2; Tmax ].
The winning partition of [Tmin ; Tmax ] represented in Figure 3 has three
= 1. Therefore party A is the pre-campaign
subintervals. Note that T0 = 12 (0)
(0)
winner. Moreover, party A has majority on both issues: on issue 1 it is
supported by voters 1 and 2, while on issue 2 it is supported by voters 1 and
for party
3.11 Nevertheless, cB = (cB1 ; cB2h) = (2; 7) is a dominant
i hstrategy
p
p i
1 (cB1 )
1 (cA +cB1 )
1+p2 1+ p
B that ensures its victory, since 2 (c +cA ) ; 2 (c )
= 1+2 23 ; 1+2 187
B2

B2

(0:2; 0:9).12

In the previous example, two issue voters are crucial to win the elections.
As we have shown, there is a subinterval of the winning partition where
party B can win the elections. This subinterval can be achieved by slightly
inuencing preferences on both political issues. The key idea is that by
means of slightly inuencing preferences in this way, party B captures voter
2 without losing voter 3: If instead party B spends all its campaign funds
on issue 1 or on issue 2, then party A wins. It is important to note that a
necessary condition for the above result is that on one of the issues (issue 2),
Since 23 + 23 < 32 n; the su cient condition to win the elections does not hold.
While three issues and ve voters are the minimal requirements to prove the Ostrogorskis Paradox, when we account for voters intensities on political issues and for parties
strategically emphasizing issues, two issues and three voters are the minimal requirement
to prove this paradox.
11

12

10

the inuence function is more vulnerable than on the other issue (issue 1).13
Thus, although party A may try to compensate the strategy of party B, it
is unable to reach the pre-campaign result.

T0

Issue 2
1(c*A1+ cB)

2(c*A2)

0.7

1(c*A1)
2(c*A2+ cB)

0.5

3
A

0.3
1
0.3

0.5

0.7
Issue 1

A
Tmin

B
0.2

A
0.9 T0

](

)[

Tmax

Figure 3. Illustration of Example 1

Conclusions

In this paper we introduce a model that, in line with the emphasis theory,
considers that candidates in the political campaign can emphasize some issues more than others. We analyze to what extent the electoral victory can
13
We say that issue r is more vulnerable to campaign expenditures than issue s 6= r
when the growth rate of r when cr changes from 0 to cB is greater than the growth
B)
A)
rate of s when cs changes from 0 to cA (where cB < cA ), i.e., when rr(c(0)
> ss(c(0)
: In

Example 1,

2 (cB )
2 (0)

=7>5=

1 (cA )
1 (0)

11

be predicted. We show that neither the pre-campaign advantage nor the

amount of campaign resources can guarantee the electoral victory. Electoral
victory is guaranteed only when a party holds a su ciently large advantage
on every salient issue, as far as its advantage guarantees that the party has a
majority of partisan voters. The strategic emphasis of issues, however, does
not prevent the misleading conclusion of the Ostrogorskis paradox: we nd
that a party that has an advantage on every salient issue may be majoritydefeated in representative democracy (even if it is the pre-campaign winner
and has more campaign resources than its opponent), as long as the sum of
the one-issue advantages is not too large.

12

REFERENCES
Amors, P. and Puy, M.S. (2007) Dialogue or Issue Divergence in the Political Campaign?, Core Discussion Paper 2007/84.
Bezembinder, T.H. and Van Acker, P. (1985) The Ostrogorski Paradox and
its relation to nontransitive choice, Journal of Mathematical Sociology 11,
131-158.
Baron, D.P. (1994) Electoral Competition with Informed and Uninformed
Voters, American Political Science Review 88, 33-47.
Budge, I. (1993) Issues, Dimensions, and Agenda Change in Postwar Democracies, in William H. Riker ed., Agenda Formation, The University of Michigan Press, 41-80.
Daudt, H. and Rae, D. (1976) The Ostrogorski paradox a peculiarly of
compound majority decision, European Journal of Political Research 4, 391398.
Grossman, G.M. and Helpman, E. (1996) Electoral Competition and Special
Interest Politics, Review of Economic Studies 63, 265-286.
Kelly, J.S.(1989) The Ostrogorskis Paradox, Social Choice and Welfare 6,
71-76.
Laond, G. and Lain, J. (2006) Single-switch preferences and the Ostrogorski paradox, Mathematical Social Sciences 52, 49-66.
Laver, M. and Hunt, W.B. (1992) Policy and Party Competition, Routledge.
Ostrogorski, M. (1902) Democracy and the Organization of Political Parties.
Paris.
Page, B.I. (1976) The Theory of Political Ambiguity, American Political
Science Review 70, 742-752.
Petrocik, J.R. (1996) Issue Ownership in Presidential Elections, with a 1980
Case Study, American Journal of Political Science 40, 825-850.
Riker, W.H. (1993) Rhetorical Interaction in the Ratication Campaign, in
William H. Riker ed., Agenda Formation, The University of Michigan Press
81-123.

13

Sigelman, L. and Buell, E. (2004) Avoidance or Engagement? Issue Convergence in Presidential Campaigns, American Journal of Political Science
48 (4), 650-661.
Simon, A. (2002) The winning Message: Candidate Behavior, Campaign Discourse, and Democracy. Cambridge. Cambridge University Press.

14