Transcript EITM 2005 Institutions Week

EITM 2007
Institutions Week
John Aldrich
Duke University
Arthur Lupia
University of Michigan
Two-Candidate Competition in a
Spatial Model






Downs
Davis-Hinich-Ordeshook
McKelvey
Aldrich and McKelvey
Aldrich
Poole and Rosenthal
Testable Spatial Models

Downs develops an intuitive mostly one-dimensional
spatial model

Davis, Hinich, Ordeshook formalize and extend
Downs.

McKelvey develops aggregate level conditions that are
testable.

Aldrich specifies McKelvey conditions, given newly
available suitable data, Aldrich-McKelvey developed.
Poole Rosenthal use WARP to infer policy space from
votes.

7/20/2015
3
Davis, Hinich, and Ordeshook

M: Why formalize?

NH: Same as Duncan Black in Downsian setting?

P: Key decisions in formalization:




Voter Utility and Choice
Abstention
Candidate Goals
Equilibrium Concept

C: Results: Convergent equilibrium at multivariate median

Value of the Exercise?
7/20/2015
4
McKelvey’s Generalization

The What – Probabilistic Formulations

The Why – What was Gained Theoretically?

The Success – How General? How Empirically
Useful?
7/20/2015
5
Aldrich-McKelvey Scaling

M: Estimate “true” candidate positions from perceptions with
error and with what else embedded in responses?

NH: Public perceptions unrelated to positions candidates adopt,
or at least so biased that true positions unrecoverable.

P: True position reported, unique up to linear transformation,
with i.i.d. error.

C: True positions inferable from noisy perceptions; ideal points
fit into candidate space.
Aldrich-McKelvey Scaling
Poole and Rosenthal

M: Estimate ideal points of voters and cut points of bills in a
unidimensional policy space, using actual votes cast.

NH: Members of Congress vote on grounds other than the utility of policy.

P:
Unidimensionality
No Strategic Voting
Induced policy preferences are just like “naturally occurring” policy
preferences.

C:
Most multi-dimensional results (e.g., Clausen) are artifactual.
Major dimension is liberal-conservative
Lib-con does poorly on pork barrel, regional, and special interest bills
Spatial positions are largely stable over time, with changes being
idiosyncratic.
7/20/2015
8
Strategic Voting:
Wasted voting has been understood by politicians for at least 150
years.
Politicians understand it in rational choice terms, just as they so
understand the calculus of candidacy.
Social scientists struggle to understand it. (Duverger believes that
his rational choice explanation in the early 1950s is wrong and
that voters are not rational; theorists do not agree on what model
to use; those that they use are wrong; social scientists’
explanation predicts no strategic voting, along with no turnout).
Strategic Voting:
Core Evidence, Example 1
Percentage of Respondents Who Intend to Cast Votes for Their Favorite Party or Candidate:
Preferences and Choice among U.S. Presidential Candidates,
1968, 1980, 1992, 2000
Voted for
Highest Preference
Republican
Democrat
3rd Party
Republican
95.6%
2.9
1.9
Democrat
3.6
95.2
1.1
Third Party
19.8
22.9
57.1
Strategic Voting:
Core Evidence, Example 2
Probability of Respondent Intending to Vote for Most Preferred Candidate,
Israeli Prime Ministerial Election, 1999
Probit Estimates, Given Normalized,
Multiplicative Utility and Viability
Measures
PB12
PB13
PB23
N=
Coef. (Std.
Err)
2.23
(0.72)
1.67
(0.41)
-1.53
(0.61)
689
P>|z|
0.002
0.000
0.012
Strategic Voting

Two Models:

Strategic Voting as Expected Utility Maximization
Riker and Ordeshook, “A theory of the calculus of voting
 McKelvey and Ordeshook, “A general theory ….”


Strategic Voting as Nash Equilibrium Strategy
Farquaharson, Theory of Voting
 McKelvey and Niemi, “A Multistage Game Rep….”

McKelvey and Ordeshook




M: To generalize Riker and Ordeshook’s exepcted utility
calculations of the turnout decision.
NH: Everyone rationally votes for their most preferred choice
P: Turnout is a decision theoretic choice
Voters are expected utility maximizers
Voters make plausible calculations of probabilities
“P” terms are not so small (relative to “B” terms) as to
render turnout implausibly rare.
C: Duvergerian “wasted vote” thesis follows rationally
McKelvey and Ordeshook
Equations of Interest
Riker and Ordeshook “calculus”
Generalization
Where, E(U) for an candidate is:
McKelvey and Ordeshook,
“Strategic” Voting
McKelvey and Niemi

M: Provide a “modern” game theoretic underpinning to
Faquaharson and his account of strategic voting

NH: Strategic voting is non-Nash

P:

C: Strategic voting is everyone playing subgame perfect
equilibrium strategies
Voting is binary
Voters are all fully informed and rational
Voting structure is fixed and, like preferences, is common
knowledge
McKelvey and Niemi:
Farquaharson Game
McKelvey and Niemi:
Multistage Game
DO VOTERS CARE ABOUT GOVERNMENT
COALITIONS? TESTING DOWNS’
PESSIMISTIC CONCLUSION

M: To show that voters in proportional representation systems
strategize over outcomes, rather than vote “sincerely” for parties

NH: Downs points out that in such cases the rational voter needs
to determine what coalitions are possible, to ascertain their
probability, and to anticipate the policy compromises that they
entail. Downs adds that this may too complex a task and
concludes that “most voters do not vote as though elections
were government-election mechanisms”.
Preferences and Choices in
First-Past-the-Post and Proportional Elections
by Abramson, Aldrich, Blais, Diamond, Diskin, Indridason Lee, Levine



M:
Voters care about outcomes and vote accordingly,
regardless of electoral system
NH: In PR voters care (only) to have their views represented on
the floor of the legislature
P:



Voters are expected utility maximizers
Voters weigh probabilities relatively, but not absolutely, correctly
C:



Voters care about likelihood of outcomes, prime ministerial candidates,
and government coalitions
Voters weigh expectations, do not pursue spe strategies
PR may well have worse connections between public preferences and
democratic outcomes than FPTP voting.
Table 2A. U.S. Presidential Candidates, 1968, 1980, 1992, 1996, and 2000
Voted for
Highest Preference
Republican
Democrat
Third-Party or
Independent
Total
(Number)a
Republican
96.0%
2.4
1.8
100.2%
(2,061)
Democrat
2.4%
96.2
1.4
100.0%
(2,195)
Third Party
18.2%
20.2
61.6
100.0%
(656)
3A. Israeli Knesset Parties, 1999
Vote intention for
Highest Preference
One Israel
Likud
Center
Total
(Number)
One Israel
98.4%
1.6
0.0
100.0%
(186)
Likud
1.9%
97.1
1.0
100.0%
(210)
Center
27.3%
6.5
66.2
100.0%
(77)
Probit Estimates of the probability of voting for most preferred party or candidate
USA
Israel
1988a
1999
2003
2006
Primaries
PM^i
Knessetii
Partyii
Partyiii
Partyii
0.694 (0.54)
0.79** (0.06)
0.80** (0.07)
0.77** (0.06)
1.15** (0.06)
Cons.
1.09** (0.06)
PB12
5.35** (1.38)
1.26**
(0.46)
2.42** (1.01)
0.82 (1.91)
2.17 (1.30)
0.86 (0.63)
PB13
2.93** (0.65)
1.01**
(0.27)
3.4** (0.74)
5.34** (1.47)
2.38* (1.06)
1.17** (0.43)
PB23
-1.99** (0.83)
-1.52** (0.41)
-1.32 (1.23)
-10.5** (3.02)
-5.28** (1.98)
-0.78 (0.74)
970
851
701
498
611
916
100.47
56.82
56.48
29.02
34.13
21.42
-
0.078
0.092
0.074
0.062
0.036
N
Chi2
Pseudo R2
Netherlands
Expected Utility Model of
Vote Choice Including
Preference over Prime Minister
United Kingdom
1998
2002
England 2005
Polls
Polls
R’s District
Expectation^
0.429**
(0.06)
0.384**
(0.07)
0.799** (0.06)
2.475
(2.07)
3.397
(2.5)
0.74* (0.41)
1.062** (0.35)
10.107**
(1.83)
12.831**
(1.7)
2.028** (0.3)
1.915** (0.26)
-8.894**
(2.77)
14.056**
(2.91)
-3.116** (0.56)
-2.828** (0.49)
0.394** a
(0.13)
0.788** d
(0.11)
0.466** a (0.07)
0.487** a (0.07)
853
883
2244
2334
99.24
114.78
180.01
222.81
0.098
0.1428
0.121
0.138
R’s Gen. Election
Expectations^
0.767
(0.77)
Probability of Voting
on Basis of
Coalition Considerations
Full sample
Only voters asked
about expectations
Constant
-2.492**
(0.59)
-1.173 (1.06)
-0.353
(1.15)
Intensity of
Preference for a
Party
-0.152**
(0.06)
-0.291**
(0.14)
-0.280*
(0.15)
Intensity of
Preference for a
Coalition
0.397**
(0.14)
0.314 (0.27)
0.252
(0.25)
-0.031 (0.47)
-0.292 (0.77)
-0.712
(0.72)
Information
Coalition
Expectations for
Favorite Party
N=
-7.63**
(1.91)
847
322
322
DO VOTERS CARE ABOUT GOVERNMENT
COALITIONS? TESTING DOWNS’
PESSIMISTIC CONCLUSION

P:
Voters have preferences over government coalitions
All who vote for their most preferred party vote “sincerely”

C: A small (1 in 10) but significant percentage vote against
their party preference so as to realize more preferred outcomes,.
The least informed were as prone to vote on the basis of
coalition preferences as the most informed.
DO VOTERS CARE ABOUT GOVERNMENT
COALITIONS? TESTING DOWNS’
PESSIMISTIC CONCLUSION
Table 4: Determinants of Coalition Voting: A Logit Estimation
Independent Variables
Coefficients
Intensity of Preference for a Party
Intensity
of
Preference
for
-0.15** (0.06)
a
0.40** (0.14)
Coalition
Information
Constant
N = 847
1
**Significant at p=0.01; *Significant at p=0.05
-0.03 (0.47)
-2.49** (0.59)