How the Electoral College Influences
Campaigns and Policy : The
Probability of Being Florida
by David Strömberg (2008)
Political Economics: explaining public policy
Teacher: Aldashev G.
Sophie Maldague
Arthur Piret
Antoine Sluysmans
I.a - Election process
Each state is allocated a number of electors
Step 1 - Voters vote for a candidate in their respective state
Step 2 - The candidate with the majority of votes in each state “wins the state”
Step 3 - All the electors from the state vote for their state’s candidate (winner
takes it all)
Step 4 - The candidate having the absolute majority in the electoral college (all the
electors from all states) becomes president elect (>270 electors).
If there is a tie the house of representative selects the president.
2
I.b - State power
-
-
-
President is elected by the
electoral college
The electoral college consists of
538 presidential electors from
the 50 states and Washington
DC
Each state number of electors is
equal to a number of
representatives + 2 senators.
Elector are technically not
bounded to their state choice
3
4
I.c - Some specifications
-Maine and Nebraska have a per district method
-Winner of the presidential election does not have the majority in the popular vote
(1824(?),1876, 1888, 2000, 2016)
-Very controversial system: 700 modification proposals over the last 200 years.
5
Existing literature on the topic
-Part of the burgeoning literature on comparative political economy (linking political
institutions to outcomes in politics and economic policy):
-Torsten Persson and Guido Tabellini (2000)/ Alessandro Lizzeri and Nicola
Persico (2001)/ Persson and Tabellini (1999)/ and Gian Maria Milesi-Ferretti,
Roberto Perotti, and Massimo Rostagno (2002),
-Compare taxes, spending, and transfers in polities with majoritarian and
proportionally representative electoral rules
6
-Timothy Besley and Anne Case (2003) - Impact of US political institutions on
policy
-Literature on campaigning:
-Brams and Davis (1974): resources should be allocated disproportionately in
favor of large states
-Claude S. Colantoni, Terrence J. Levesque, and Peter C. Ordeshook (1975):
a proportional rule, modified to take into account the ex post closeness of the state
election, predicts actual campaign allocations better.
-James M. Snyder’s (1989): Model of two-party competition for legislative seats
7
New in this paper:
-Complete integration of theory and empirics
-Allows states to have different partisan leanings
-Allows uncertainty regarding the election outcome
-Has candidates maximizing the probability of winning the election instead of the
expected number of electoral votes that they win.
-Allows for different state sizes
-Link the model to the literature on voting power (probability that a vote is decisive)
8
II. Model
●
Theoretical model of presidential campaigning
○
Which states are important for election outcome (why Florida is so important?)
Constructed probabilistic-voting model
●
Assumptions :
○
○
○
2 candidates (D and R) want to maximize their probability of winning
Ignore the impact of minority-party candidates
Campaign matters and affects the voters
9
Setup
I days to use to visit each of s states.
Each candidate J choose the number of campaign day dJs under the constraint :
The number of electoral votes esi is obtained by the candidate in state Si if he gets
the majority in this state. The candidate who gets more than half those votes in the
entire US wins the election.
10
Setup (cont’d)
The increasing popularity of candidate J is captured by the increasing and
concave function u(dJs ).
Voters are also affected by other factors : ideological preferences
○
○
○
Ri
ηs
η
predictable
unpredictable at the state level
unpredictable at the national level
Voter i in state s will vote for D if
The share of votes ys that D receives in state s on election day is :
11
Approximative probability of winning
The candidate wins if
The probability of this event, conditional on the aggregate popularity η and the
campaign visits is :
The winning of the candidate D, depending es is :
Ds = 1, with probability Gs ( . )
Ds = 0, with probability 1 - Gs ( . )
12
Approximative probability of winning (cont’d)
Probability that D wins :
=> Difficult to maximize, number of such combinations is of the order of 251
Solution by the Central Limit Theorem of Liapounov :
Approximate probability of D winning (given national popularity shock η)
13
Equilibrium
The Nash equilibrium with (dD*, dR*)
For each d, X the set of allowable campaign visits :
This game has a unique interior pure-strategy equilibrium
NE => dD = dR = d*, for all s and for some λ>0.
14
Equilibrium (cont’d)
Where
Candidates should spend more time in states with high Qs
Qs depends only on the parameters es, μs, and σs
15
III. Estimation : Relation between Qs and actual
campaigns
Data : visit data collected by D. Shaw in 2000 and 2004
How the actual allocation of presidential candidate visits to states in those two
elections corresponds to the theoretical equilibrium campaigns?
State S share of equilibrium visits =
Forecasted vote outcomes, ŷst in Table 3
16
17
18
Relation between Qs and actual campaigns (cont’d)
The actual and equilibrium shares are shown in Figure 1
●
The actual campaigns is close to the model’s equilibrium campaigns
○
●
●
The correlation between equilibrium and actual visit shares is 0.9 in both years.
Candidates should concentrate on close races (like Florida)
The distribution of visits seems more concentrated in 2004 than in 2000 (due
to electoral incentives).
19
20
21
Relation between Qs and actual campaigns (cont’d)
An OLS regression of actual visit shares in 2000 and 2004 on equilibrium visit
shares
, shown in Table 4 :
22
●
●
The equilibrium visit shares are strongly correlated with actual visits shares,
with an estimated coefficient of around 1.
Once Qs is added to the regression, the other variables don’t contribute to
explaining visits at all (the R-squared is still 0.77) => Qs seems to be a
sufficient statistic in explaining candidate visits.
23
IV. Interpretation
A)
-
B)
-
What is Qs?
Qs ≈ Pr(decisive swing state)
Link to the “voting power”: Qs ≈ (# votes cast)s X (“voting power”)s X (marginal
voter density, conditional on tied state election)s
Effect of Electoral Votes on Influence
Qs roughly proportional to es
Qs/ es = Qsμ/es + Qsσ / es
24
C) Influence per Electoral Vote and Forecasted Vote Shares ( Q sμ/es )
µ
25
D) Incentives to Affect the Variance in the Electoral Vote Outcome (Q sσ / es)
(even if decreasing of es)
26
E) Additional Issues
-
influence of campaigns more on voter turnout rather than vote choice
existence of minority candidates
electoral votes in Maine and Nebraska allocated in a special fashion
27
V. Electoral reform
Electoral College (EC) : sharp incentives to target only a few states
1)
Direct Vote (DV):
Would loose from
the reform
Losers in both
cases
Winner under both
systems
Would gain from
the refom
28
Distribution of visits
Gain or loss of attention because of (A) electoral size per
capita and (B) influence relative to electoral size
29
Lodge-Gossett Amendment : keeping electoral size per capita constant, only
changing the influence per electoral size
30
Other reform concerns
-
concentration of minorities in large and competitive states, and higher
influence than average on the election → inconsistent with the data
presidents without a majority of the popular vote
razor-thin victories
31
VI. Conclusion of the paper
1)
2)
3)
4)
5)
6)
More resources should be devoted to states that are likely to be decisive swing states
probability of being a decisive swing state equals the number of voters in the state
multiplied by the “voting power” in the state, multiplied by the marginal voter density
conditional on the state election being tied
Probability of being a decisive swing state is roughly proportional to the number of
electoral votes
Probability of being a decisive swing state per electoral vote is higher for states that
have a forecasted state election outcome that lies between a draw and the forecasted
national election outcome
More precise state-election forecasts make the optimal allocation of resources more
concentrated.
The presidential candidate who is lagging should try to increase the variance in
electoral votes.
32
Main difference with the model studied in class
33
2016 US elections - facts checking
Clinton won the popular
vote by more than 2.5
million votes (<1% of the
population, 2% of the
number of voters)
Source: http://www.270towin.com/
34
In the 5 states with the highest population per electorate
California: Clinton won by 28.3%
New York: Clinton won by 21.3%
Texas: Trump won by 9.2%
Florida: Trump won by 1.3%
Illinois: Clinton won by 16%
Has Clinton allocated its rally as recommended by our model?
35
“I would have won the popular vote if I was campaigning for the popular vote”
-Donald J. Trump, january 2017
2 conditions to be a swing state:
1)
Closeness between both candidates in the poll
Both candidates ended up winning all the states where they had more than a 10%
lead.
Trump won all the states where he had a lead between 3% and 10%, among
those states Clinton only lost Pennsylvania
2)
Having a determinant impact on the final output
Which states would have changed the election output?
36
States with a lead within 5%
Trump won 11 out of the 13 states where the elections were within a 5% margin.
Only Texas was decisive taken alone, but most combination between those states
could have given Clinton a win: Texas (38), Florida (29), Pennsylvania (20), Ohio
(18), Michigan (16), Georgia (16), North Carolina (15). Trump ended up winning all
those states
37
With data from Fair Vote between end of july and election day:
5 out of the 7 states studied are amongst the 6 most visited states (by both
candidates).
State
Visits by Trump
Visits by Clinton
Texas
1
0
Florida
35
36
Pennsylvania
28
26
Ohio
30
18
Georgia
3
0
Michigan
14
8
North Carolina
31
24
38