Political Participation 1 / 42

Lecture 1: Political Participation

Mateusz Stalinski

EC340: Topics in Applied Economics (3a)

University of Warwick

Autumn 2023
Political Participation 2 / 42
Introduction

Political Participation

▶ Electoral choices are highly consequential as they have impact

on government policy and regulation.
▶ It is critical to understand individual motivation to participate
in political processes.

In particular, we would like to know:

1. Why do people vote in elections?
2. What factors shape the dynamics of election turnout?
3. What strategies can political parties and other stakeholders
use to encourage turnout?
Political Participation 3 / 42
Introduction

Let’s Start with Some Data

▶ We will look at election turnout in the U.S. from 1789 to 2022.

▶ Turnout is defined as the proportion of voting-eligible
population (VEP) that voted.
▶ Excludes people below the voting age, felons, and others who
did not meet requirements for suffrage.
▶ The latter category covers criteria such as property ownership,
and matters mostly for elections in 1800s and 1900s.
▶ Voter turnout data is available, for example, through
University of Florida’s Election Lab.
▶ What patterns can you notice by analyzing the plots below?
Political Participation 4 / 42
Introduction

Election Turnout Patterns

Figure 1: Election Turnout in the U.S. from 1789 until 2022

Political Participation 5 / 42
Introduction

Midterm vs. Presidential Elections

Figure 2: Election Turnout in the U.S. from 1789 until 2022

Political Participation 6 / 42
Introduction

Stakeholders

▶ Many NGOs encourage political participation and promote

civic society.
▶ But they are not the only intuitions interested in turnout.
▶ Political parties care deeply about what types of individuals are
more likely to turn out to vote.
▶ In some environments, encouraging turnout among party’s
base is more important than persuading the undecided.
▶ This can lead to more extreme policy proposals.
▶ Ability to encourage turnout and motivate voters may be key
for electoral victory.
Political Participation 7 / 42
Introduction

Elasticity

▶ We can create a measure to think about differential turnout

vs. persuading the undecided.
▶ Elastic state: relatively responsive to changes in nationwide
political conditions (538 article, Update, GitHub Estimates).
▶ Elasticity x: 1 pp change in the national party support would
change state support by x pp.
▶ If Wisconsin’s elasticity is 1.1, then a 1 pp. swing nationally
would change Wisconsin’s numbers by 1.1 pp.
▶ In inelastic (and swing) states, motivating the party base to
vote is essential.
Political Participation 8 / 42
Introduction

Elastic vs. Inelastic States

Figure 3: State Elasticity with Respect to Nationwide Political Conditions

Political Participation 9 / 42
Introduction

Examples

New Hampshire (elasticity: 1.15, Clinton won by 0.37 pp in 2016):

▶ An example of a persuasion swing state.
▶ A large number of independents.
▶ A lot of people with demographics that are weaker predictors
of voting behavior (white, not evangelical).

Georgia (elasticity: 0.9, Biden won by 0.23 pp in 2020):

▶ An example of a turnout-battle swing state.
▶ A rigid Republican coalition with a lot of evangelicals.
▶ An equally strong Democratic coalition with many
African-American voters and Atlanta residents.
Political Participation 10 / 42
Introduction

Research Questions Related to Election Turnout

1. Why do people vote even though the chance that their vote
decides the result is infinitesimally small?
2. Do election competitiveness and electoral systems
(first-past-the-post, proportional, etc.) matter for turnout?
3. What strategies (e.g., TV and online ads, canvassing) are
effective in encouraging turnout?
▶ What are the key moderators?
▶ A moderator is a factor that affects the success rate
(treatment effect)?

We will shed light on these questions by studying several key papers

on the topic.
Political Participation 11 / 42
Spenkuch (2018)

Spenkuch (2018)

▶ Spenkuch (2018) empirically tests two key theories of voting.

▶ Pivotal voting: despite the low chance of a draw (the voter is
pivotal), the stakes are high enough to make voting optimal.
▶ Such voters are likely to vote tactically, i.e., avoid voting for
their most preferred candidate if they are not in contention.
▶ Expressive voting: voters derive utility from voting for their
most preferred candidate.
▶ Low prevalence of tactical voting.
Political Participation 12 / 42
Spenkuch (2018)

Setting (1)
▶ The study relies on the unique structure of the German
electoral system.

Figure 4: German Election Results in 2021, BBC article

Political Participation 13 / 42
Spenkuch (2018)

Setting (2)

▶ Germans cast two votes: (1) candidate vote and (2) list vote.
▶ (1): 299 single-member districts, first-past-the-post system
(plurality sufficient for winning).
▶ (2): proportional representation to ensure that the overall seat
allocation reflects % of votes for each party.
▶ The total number of Bundestag seats allocated in (2) is 598.
▶ Caveat: 5% threshold is required to win any seats.
Political Participation 14 / 42
Spenkuch (2018)

Setting (3)

▶ dp,s is the number of districts party p won in state s.

▶ lp,s is the number of mandates party p is allocated according
to the proportional representation in state s.
▶ If dp,s < lp,s , the first lp,s − dp,s candidates from the party list
are elected too.
▶ Overall, the total number of seats won by p in s is
np,s = max{dp,s , lp,s }.
▶ The total number of Bundestag seats won by p is s np,s .
P

▶ Occasionally, lp,s > dp,s , which leads to overhang mandates.

▶ This means that the total number of seats can exceed 598.
▶ The number of overhang mandates is low in recent elections.
Political Participation 15 / 42
Spenkuch (2018)

Key Idea

▶ Controlling for candidate characteristics, preferences for the

party and candidate votes should be very strongly aligned.
▶ Expressive voters always vote for their preferred candidate.
▶ They are unlikely to “split” their vote between two parties.
▶ Those driven by the chance of being pivotal vote tactically.
▶ At least some of the time, they will “split” their vote.
▶ They would not vote for candidates out of contention.
▶ Looking at the proportion of “split” votes for contenders
vs. non-contenders disentangles the two.
Figure 5: Graphical Illustration of the Empirical Strategy
Political Participation 17 / 42
Spenkuch (2018)

Some Caveats

▶ There are three additional assumptions required for this

empirical strategy to work.
▶ (1) Voters understand this complex electoral system well.
▶ (2) Voters know which candidates are in contention.
▶ (3) Voters are not indifferent between candidates other than
their preferred candidate.
Political Participation 18 / 42
Spenkuch (2018)

Data and Estimation

▶ The sample restricted to 2005 and 2009 federal elections (all

main parties exceeded the 5% threshold).
▶ The party has to be expected to cross the threshold to ensure
optimality of picking the top choice in the party vote.
▶ 299 electoral districts and almost 89,000 precincts.
▶ Data on precinct level but races on district level.
▶ Explores within-candidate variation (candidate fixed effects).
▶ Conditioning on observable and unobservable characteristics.
▶ Regress candidate vote share on party vote share and FEs.
▶ Standard errors clustered at state level.
▶ Allows for dependence between districts and within districts.
Political Participation 19 / 42
Spenkuch (2018)

Results (1)

Figure 6: Results Restricting Attention to the Eventual Winner and

Runner-up of Each Race.
Political Participation 20 / 42
Spenkuch (2018)

Results (2)

Figure 7: Results Restricting Attention to Those Who Finished in Third

Place or Worse.
Political Participation 21 / 42
Spenkuch (2018)

Results (3)

Figure 8: Results Restricting Attention to Candidates at Least 10

Percentage Points Behind the One in Second Place,
Political Participation 22 / 42
Spenkuch (2018)

Discussion

▶ At least 61% of voters do not behave in accordance with the

canonical pivotal voter model.
▶ Expressive voting alone is not a good theory of voting behavior
either, as 1/3rd votes tactically.
▶ Possibly, we face a mixture of types of voters.
▶ The author further explores whether the types are
endogenously shaped.
▶ Evidence that tactical voting requires experience.
Political Participation 23 / 42
Spenkuch (2018)

Tactical voting and experience

▶ Consider voters from the former German Democratic Republic

(GDR).
▶ After the reunification, voters from GDR had less experience
with the electoral system.
▶ The reunification offers a natural experiment inducing
variation in voter experience.
▶ However, are we sure that other relevant factors did not differ
between East and West?
▶ Test whether the proportion of nonstrategic voting between
East and West Germany varies over time after the event.
Political Participation 24 / 42
Spenkuch (2018)

Results: Experience

Figure 9: Difference in the Incidence of “Wasted Votes” between East and

West Germany, 1990–2009.
Political Participation 25 / 42
Spenkuch (2018)

Summary

▶ Two-thirds of individuals violate the predictions of the pivotal

voter model.
▶ At the same time, one-third does not behave expressively.
▶ Neither theory alone explains voting behavior well.
▶ Voters cannot be easily classified into two types: strategic vs.
expressive.
▶ The types may be endogenously shaped, including by voter
experience.
Political Participation 26 / 42
Kamenica and Egan Brad (2014)

Kamenica and Egan Brad (2014)

▶ Kamenica and Egan Brad (2014) used a lab experiment to

exogenously assign the likelihood of pivotality.
▶ They tested whether it matters for voting decisions.
▶ They compared pivotality to other key factors, including
ideology.
▶ Context: voting on redistribution policies.
Political Participation 27 / 42
Kamenica and Egan Brad (2014)

Experiment Protocol (1)

▶ The study was conducted in groups of nine participants.

▶ Survey questions eliciting beliefs and policy preferences.
▶ Government redistribution, gun rights, death penalty, etc.
▶ Beliefs on whether the poor are unmotivated or unlucky.
▶ Five rounds of rock-paper-scissors.
▶ Points for beating the experimenter who uniformly randomized
choices.
▶ This ensured random assignment of performance.
▶ Earnings according to three tiers of performance ($5, $10,
$15).
Political Participation 28 / 42
Kamenica and Egan Brad (2014)

Experiment Protocol (2)

▶ Question: should the money be redistributed so that everyone

earns $10?
▶ Randomization at a group level:
▶ (a) majority voting,
▶ (b) one dictator decides.
▶ All members of (a) groups are voters.
▶ For (b) groups, one player is randomly selected to be a
dictator.
▶ Other (b) players are peons, who are nonetheless asked about
their choice even though it won’t be implemented.
▶ Choices: REDISTRIBUTE, KEEP SAME.
Political Participation 29 / 42
Kamenica and Egan Brad (2014)

Results (1)

Figure 10: Balance and Main Result.

Political Participation 30 / 42
Kamenica and Egan Brad (2014)

Results (2)

Figure 11: Taxation Preferences and Voting Decisions.

Political Participation 31 / 42
Kamenica and Egan Brad (2014)

Results (3)

Figure 12: Skill vs. Luck and Voting Decisions.

Political Participation 32 / 42
Kamenica and Egan Brad (2014)

Discussion

▶ The chance of being pivotal did not significantly affect voting.

▶ However, other factors played a role.
▶ Preference for high/low taxes.
▶ Beliefs about performance in the game being skill vs. luck.
▶ What are the pros and cons of this study?
▶ Consider the role of randomization for causal inference.
▶ But also think about the artificial setting and sample size.
▶ Did the study protocol mimic some real-life considerations
pertaining to inequality and redistribution?
▶ Considering the results together with those by Spenkuch
(2018), we are inclined to reject the pivotal voter model.
▶ Results are partially consistent with the expressive voter theory.
Political Participation 33 / 42
DellaVigna et al. (2016)

Voting to Tell Others

▶ It is estimated that 25–50% of non-voters lie when asked

about their past turnout.
▶ DellaVigna et al. (2016) postulate and test a novel explanation
for why people vote in elections.
▶ Voting to tell others: people vote to avoid lying when later
asked by others.
▶ The authors conducted an experiment with three levels of
randomization to provide evidence supporting the theory.
Political Participation 34 / 42
DellaVigna et al. (2016)

Setting

▶ Time and location: suburbs of Chicago in 2011.

▶ Visited households and asked whether they were willing to
answer a short survey.
▶ The survey included a question on whether they voted in the
2010 election.
▶ Used voting records to target households where either:
▶ (a) all registered members voted, or
▶ (b) none of the registered members voted.
▶ This was done unbeknownst to the participants.
▶ What is more, the researchers knew who lied about voting, but
the participants were not aware of that.
Political Participation 35 / 42
DellaVigna et al. (2016)

Variation 1

▶ Left flyers at the door of some households prior to the visit.

▶ The information on the flyer was randomly assigned:
▶ (a) they will come the next day to conduct a survey, or
▶ (b) they will come the next day to conduct a survey
specifically about the participation in the 2010 elections.
▶ Look at the design of the flyers on the next slide.
▶ When can we interpret differential refusal rates as a sign of:
▶ Utility of proudly sharing information on one’s participation,
▶ Disutility of lying about voting?
▶ How to ensure that the results are not driven by people’s
reluctance to talk about politics?
▶ There is data for both voting and non-voting households.
Political Participation 36 / 42
DellaVigna et al. (2016)

Flyers

Figure 13: Examples of Flyer Designs Used in the Experiment

Political Participation 37 / 42
DellaVigna et al. (2016)

Results (1)

Figure 14: Survey Completion in Voting Households

Political Participation 38 / 42
DellaVigna et al. (2016)

Results (2)

Figure 15: Survey Completion in Non-Voting Households

Political Participation 39 / 42
DellaVigna et al. (2016)

Discussion
▶ Little overall evidence of utility of sharing the fact of voting.
▶ However, the paper offers clear support for high social image
costs associated with lying about voting.
▶ To quantify these costs, the researchers cross-randomized
incentives ($0 or $10) and completion time (5 or 10 minutes).
▶ A $10 incentive induces a 6 percentage point increase in the
share of opening the door.
▶ Similarly, receiving an “election flyer” reduced the likelihood by
6 pp among non-voters in comparison to a normal flyer.
▶ This indicates large social image costs of lying about voting.
▶ Through introducing further variation, the authors estimate
that the disutility of lying to a surveyor corresponds to $7.
Political Participation 40 / 42
Bursztyn et al. (2023)

Election Closeness

▶ Recent literature finds positive effects of election closeness

(polls) on election turnout (Bursztyn et al., 2023).
▶ It is tempting to assume that this, contrary to our previous
discussion, favors the pivotal voter theory.
▶ At least for a fraction of voters.
▶ But election closeness could also increase the expected number
of interactions in which people may get asked about voting.
▶ Then, the result could be explained by “voting to tell others”.
▶ This remains one of the topics to be explored in the near
future.
Political Participation 41 / 42
Encouraging Turnout

Encouraging Turnout

▶ Today, we looked at different theories on why people vote.

▶ We also explored various empirical strategies to test them, and
estimate the prevalence of tactical voting.
▶ In the student presentations (Week 2), we will discuss methods
of encouraging turnout and persuading voters.
▶ Presentation 1: Broockman and Green (2014)
▶ Impact of online ads on candidate’s name recognition and
favorability.
▶ Presentation 2: Spenkuch and Toniatti (2018)
▶ Effects of TV political ads on election turnout and vote shares.
Political Participation 42 / 42
References

References
Broockman, D. E. and D. P. Green (2014): “Do online advertisements increase
political candidates’ name recognition or favorability? Evidence from
randomized field experiments,” Political Behavior, 36, 263–289.
Bursztyn, L., D. Cantoni, P. Funk, F. Schönenberger, and N. Yuchtman (2023):
“Identifying the Effect of Election Closeness on Voter Turnout: Evidence
from Swiss Referenda,” Journal of the European Economic Association,
jvad038.
DellaVigna, S., J. A. List, U. Malmendier, and G. Rao (2016): “Voting to tell
others,” The Review of Economic Studies, 84, 143–181.
Kamenica, E. and L. Egan Brad (2014): “Voters, dictators, and peons:
expressive voting and pivotality,” Public Choice, 159, 159–176.
Spenkuch, J. L. (2018): “Expressive vs. Strategic Voters: An Empirical
Assessment,” Journal of Public Economics, 165, 73–81.
Spenkuch, J. L. and D. Toniatti (2018): “Political advertising and election
results,” The Quarterly Journal of Economics, 133, 1981–2036.