Unit 4: Models of Voters’ Behavior

4.1 A classification of voters

4.2 Proximity voting
4.3 Directional voting
4.4 Conclusions

M. Socorro Puy
(Universidad de Málaga)
• We showed the impossibility of aggregating individual preferences to obtain a
social preference (Arrow’s Impossibility Theorem) (Unit 1).

• We defined the concept of voting rules and studied different voting procedures
as an alternative to aggregate social preferences (Unit 2).

• The constitutions of modern States in the world opt for representative democracy
by which sovereign power is delegated to national and regional parliaments.
Presidents, prime ministers, and members of parliament are elected either by
majority voting or proportional representation (Unit 3).

• This unit explores how citizens vote in a single-winner majority voting election.
(Unit 4)
4.1 A classification of voters (Lau and Redlawsk 2006)
• Rational voters
• They are well informed about the electoral programs, they compare candidates and opt for the
party (or candidate) that better fits their policy preferences

• Retrospective voters
• They look back and evaluate their well-being with one or another incumbent party in power. Their
past experience is used to evaluate candidates and to make a decision.

• Phsico-social voters (partisans)

• Their family tradition or their social networks makes them opt for the same candidate (or party),
no matter how it performed in the past or what political campaign it proposes.

• Voting by political issue

• These voters pay attention to particular issue dimensions (education, religion, environmentalist,… )
and opt for the party that in this issue offers the most appealing policy platform.

• Non-informed voters
• When the voter has an stereotype and seeks confirmation about it, strongly affected by the
electoral campaign.
M. Socorro Puy 3
A classification as a function of voters’ information:

Information

Non-
informed Rational voters
voter Phsico-social Retrospective Issue-voters
voters voters

Sections 2 and 3 pay specific attention to rational voters.

M. Socorro Puy 4
 Voter Turnout in Spain (over total voting-age population)

• General Elections:
2023 66.59 %
2019 (A) / 2019(N) 71.7% / 66.2%
2015 (D) /2016 (J) 69.7% / 66.5%
2011 71,7%
2008 75.3%
2004 75.6%
2000 68.7%
• European Parliament Elections:
2019 50,6%
2014 45,8%
2009 44,9%
2004 45.47% Table: Voter Turnout, Spain and the EU Parliament (199-2023)
1999 49.51%
M. Socorro Puy 5
Electoral Results: General Elections Spain
 4.2 Proximity Voting (Downsian model) Downs 1957
Suppose that voters only evaluate parties in one dimensión (left-right):
Let xA and xB be the policy positions of Party A and Party B.
Let (xi , yi) be the ideal policy of voter i.

The preferences of each voter i over Party j where j =A or j=B are measured by
the single-peaked representation:

Ui(j) = – [xj – xi]2

max utility is
when xj - xi = 0 ,
because then
their is a party xi = the most prefferenced , the closer the
who completes options to xi the higher the utility/welfare
100% of your
expectations
Suppose that xA < xB: someone who is in the middle between the two parties = indifferent voter

• There is a voter that is indifferent between the two parties.

and for whom: mathematically : -(Xa-Xi)² = - (Xb - Xi )²
Ui(A) ― Ui(B) = 0. --> dubbel product uitwerken
--> - aan beide kant schrappen
--> uitwerken en dan bekom je Xi = (Xa+Xb) /2

• The above equality implies that the voter with ideal policy
equal to (xA + xB)/2 is indifferent between Party A and Party B.
Graphically:

Ui(A) = Ui(B)

xA xB
 By proximity voting:

- Voters with ideal policy to the right of the indifferent voter prefer Party B
- Voters with ideal policy to the left of the indifferent voter prefer Party A.

Vote for A Vote for B

𝑥 𝑥 x
what with 3 parties : 𝑥 𝑥 vote C
vote A 2
vote B

(Xa + Xb)/2 (Xb + Xc ) / 2

Suppose that a party has a valence advantage vA >0: valence advantage = other
carasteristics of the canditate on
which all the voters agree.

• Preferences for Party A are represented by for example : Kamala is more

appealing because she is a
woman
Ui(A) = vA – [xA – xi]2

• Calculate the new indifferent voter: Ui(A) ― Ui(B) = 0.

• Calculate the new graph:

Xi = -Va / (Xa-Xb) +(Xa+Xb)/2

valence for Va means that A will have a larger

fraction of votes compared to B so Xi will shift
Ui(A) = Ui(B)
to the right

xA xB
Consider two issue-dimensions X=[0,1] and Y=[0,1].
For example, X is the position in the left-right dimension, and Y is foreign policy or degree of
decentralization (division of power between national and regional governments, i.e., level of
centralization).

Let (xA , yA) ϵ [0,1]2 and (xB , yB) ϵ [0,1]2 be the policy positions of Party A and Party B.
Let (xi , yi) ϵ [0,1]2 be the ideal policy of voter i.

The preferences of voters over Party j (where either j =A or j=B) are represented by:
left- right = 0 --> 0 corruption is the most preferred point
in issue Y
Ui(j) = – α [xj – xi]2 – β [yj - yi] 2
where α, β are the salience of issues X and Y, respectively. See the graphical representation:

= 0.5 =0.5
 Y
Y

Voter’s yi yi
Indifference
Curves

xi X
xi X
(a) When issue X is more salient α > β (b) When issue Y is more salient α < β
care more about left right then about corruption

o α and β are the average concern for voters. This captures how society cares, on average, about one issue over
another.
o In (a), each unit (quadratic) distance over issue X is more relevant for each voter than each unit (quadratic)
distance over issue Y.
When evaluating competing parties, voters assess the differential utility:

△Ui = Ui(A) ― Ui(B) = – α [xA – xi]2 – β [yA - yi] 2 + α [xB– xi]2 + β [yB - yi] 2

Simplifying:
△Ui = k₀ + k₁ xi + k₂ yi where

k₀ = – α(xA² – xB²) – β(yA² – yB²); k₁ = 2α(xA – xB); k₂ = 2β(yA – yB)

Optimal decisions:
If △Ui > 0 vote for A
If △Ui < 0 vote for B
If △Ui = 0 indifferent
 Division of votes for Party A and Party B
All the voters with an ideal policy on the blue line are indifferent between Party A and Party B. All the voters
with an ideal policy below the blue line prefer Party B, and all those with an ideal policy above the blue line
prefer Party A. Y

= vote B

= Vote A
EXERCISE ON PROXIMITY VOTING

Consider the proximity model of voting described in class. Suppose a one-

dimensional policy space equal to the unit interval [0,1] and where the policy
positions announced by three political parties, A, B and C, are xA = 0,3, xB = 0,5 and
xC = 0,8.
a- Calculate the location of the indifferent voter between Party A and Party B, and
between Party B and Party C.
b- Suppose that voters’ ideal policies are distributed according to a uniform
distribution over the interval [0,1]. Calculate the percentage of votes to each
political party when considering that no voter abstains from voting.
simpel : gewoon gemiddelden berekenen

1. teken de lijn met de verschillende opties

2. voeg de gemiddelden toe
3.
4.3 Directional Voting Rabinowitz and Macdonald (1989)
The utility of the voter is defined by:
Ui(j) = (xi – n) (xj – n)
with n representing the ideological middle or the point of neutrality between left and right.

Directional theory uses a two-step rationale:

• First, the voter explores whether there are parties on the side that she took on the left-
right scale (side rule).
• Second, if there are, the voter prefers the party that holds the side with the most intensity
(party intensity rule).

Overall, the highest utility is reached when the voter and the party are on the same side of
the issue and they are both most extreme.
By directional voting, all the voters to the right of n (where n may not
𝑥 𝑥
coincide with 2 ) prefer Party B and all those below n prefer Party A.

Vote for A Vote for B

𝑥 x
𝑛 𝑥

Ideological middle
One implication of directional voting is that the non-polarized parties obtain
no vote if there are more than two parties.

Exercise: Apply in this example the side and the intensity rules:

Vote for C
Vote for A

𝑥 𝑥 𝑥 x
𝑛

Ideological middle
EXERCISE ON DIRECTIONAL VOTING

Consider the directional voting model. Suppose that the policy position of the
parties competing to win the elections are xA = 0,3, xB = 0,5 and xC = 0,8 and that
the ideological middle is n= 0,55
a- Calculate the voting decisions of the voters as a function of their ideal policy.
b- Suppose that voters’ ideal policies are distributed according to a uniform
distribution over the interval [0,1]. Calculate the percentage of votes to each
political party when considering that no voter abstains from voting.
4.4 Conclusion
• We can describe voters’ decisions from a behavioral perspective: what they care
about, how partisan they are, or how well-informed or manipulable they are.

• We can provide several models that try to map voting decisions as a function of i)
parties’ announced positions and ii) voters’ ideal policies.

• There are additional factors that affect vote choice that we have not explored: iii)
voters’ perception about the quality and talent of the candidates; iv) identity of the
candidate (religion, race,…etc.)

• What model better fits voting decisions is a question that needs empirical scrutiny.
EXERCISE OF PROXIMITY AND DIRECTIONAL VOTING

Suppose that the policy position of the parties competing to win the elections are
xA = 0,5, xB = 0,6 and xC = 0,8 and that the ideological middle is n= 0,55.
Suppose that voters’ ideal policies are distributed according to a uniform
distribution over the interval [0,1].

a- Calculate the percentage of votes to each political party according to Proximity

Voting.
b- Calculate the percentage of votes to each political party according to
Directional Voting.