Noncompetition and Collusion in Competitive Party Systems

1
Non-competition and Collusion in Competitive
Party Systems

• Robert Weiner, Berkeley (Cornell)
• Presented by Sona Nadenichek Golder, NYU

2
Motivation

• Why do parties form surplus coalitions?

3
Empirical Puzzle

• In many Japanese gubernatorial elections, all
  non-Communist parties join in a bandwagon
  coalition to back a single candidate. After the
  candidate wins, the parties legislators maintain
  the coalition.

4
Null Hypothesis

• Forming coalitions is costly, so we should only
  see them when necessary to attain a majority (or
  other relevant threshold).

5
Robs Hypothesis

• Coalitions have positive advantages as well as
  costs, so under certain circumstances we should
  expect to see coalitions form even if they are
  not necessary to reach a certain threshold

6
Premises (game structure)

• Two parties, L (large) and S (small). L has a
  majority of expected votes/seats.
• Third party, C, not a strategic actor.
• L chooses to make a coalition offer, run alone,
  or stay out of the race.
• S chooses to accept the offer (joint candidate
  runs), run alone, or stay out.

7
Premises (payoffs)

• Payoffs p(b-cl) ce
• p probability of winning
• b office benefits
• cl legislative costs
• ce electoral costs

8
Premises (payoffs)

• Payoffs p(b-cl) ce
• Additional vote and/or seat support from
  coalition decreases b (a bad thing), BUT
• A larger coalition increases p and decreases cl
  and ce (all good things)

9
When do we expect coalitions?

• Depends on how large costs are relative to
  benefits, and on the different rates at which
  additional vote/seat support decreases b and
  increases p, cl and ce

10
Example

• Think about a within-legislature coalition.
• Party L has 90 of the seats, and Party S has 5.
  If each L legislator is now getting a 1/90 share
  of office benefits, forming a coalition yields
  1/95
• Does reducing the size of the opposition from 10
  to 5 outweigh the loss in benefits (1/90 1/95)?

11
Payoff structure

• Payoffs p(b-cl) ce
• Each of these four terms is a function of
  parameters vote shares (vi), seat shares (si),
  electoral uncertainty parameter (?), etc.

12
Possible outcomes

• Coalition
• Partisan
• Stay out

13
Predictions

• For a given set of parameter values, the payoff
  structure tells us the players preference
  ordering the 3 possible outcomes
• the game converts the preference orderings into a
  predicted equilibrium outcome

14
Testing

• For each case, measure the measurable parameters
  (vote share, seat share, uncertainty parameter)
• For all cases, posit a single value for each
  unmeasurable parameter (legislative cost
  elasticity, electoral entry cost, benefits
  accruing to executive rather than legislators)

15
Testing

• For each case, use these values to determine
  which of the three outcomes is predicted.
  Express it via two dummy variables.
• For each case, use a categorical variable (three
  values) to express the actually-observed outcome.

16
Testing

• Test correlation between predicted and actual
  outcomes via multinomial logit (including
  additional control variables). Do
  statistical/substantive significance tests.
• As robustness check, posit new combinations of
  unmeasurable parameters values and repeat.

17
Results

• Prediction dummy variables have a positive,
  statistically significant effect on the relevant
  dependant variable
• Robustness checks show that different values for
  the parameter values set by the researcher
  (costs, etc.) yield similar results

18
Next step?

• QRE
• Simplify model
• Measure the unmeasurable parameters