DEBATESThe Impacts of Voter Knowledge Initiatives in Sierra Leone
Katherine Casey (Stanford GSB)Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL)
*** Preliminary Results ***
October 2013
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 1 / 21
Motivation
A poorly informed electorate limits democratic returns in terms ofselecting better candidates and holding elected politicians to account
The limited depth and reach of mass media exacerbates this problemin many developing countries
If true, then delivering information to voters could naturally be part ofthe solution
How to do so effectively - in a way that is credible and accessible topoor, illiterate and remote citizens - remains an open question
When taken to scale, the endogenous response of candidates andparties is likely to be important
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 2 / 21
Voters are poorly informed about politics in Sierra Leone
Election Day (Nov 2012) exit polls in control villages reveal that:
Only 28% of voters could name the two Parliamentary candidates
64% could not name a single MP job responsibility
3% knew the amount of the MPs’constituency facilitation fund
35% knew the proposed quota for women’s representation ingovernment and 17% knew the candidates’positions on the bill
70% of voters have zero years of schooling, 31% have no radio
Research question: can giving voters information about candidates andpolicy facilitate more informed voting and greater accountability?
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 3 / 21
Could debates be part of the solution?
NGO partner Search for Common Ground hosted and filmed debatesbetween Parliamentary candidates, with a standardized structure:
Moderator introduction
Get to know you questions
First priority for govt spending
Constituency facilitation fund (CFF)
Strategy to uplift the youth
Position on gender equity bill (GEB)
Free healthcare implementation (FHC)
Two local policy issues
Constituency 63, Yilleh, Tonkolili
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 4 / 21
Selections from related literature
Electoral information interventions in developing countries
3rd party information: Ferraz and Finan (2008), Banerjee et al(2011), Banerjee et al (2010), Humphreys and Weinstein (2012)
Campaign strategy: Wantchekon (2003); Fujiwara and Wantchekon(forthcoming)
Some parallels in American elections
Debates: Abramowitz (1978), Geer (1988), Jackman (2012)
Campaign advertising: Gerber et al (2011)
“Thin slices:”Todorov et al. (2005), Benjamin and Shapiro (2009)
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 5 / 21
Large scale field experiment leading up to Election Day
Three levels of randomization:
Debates hosted in 14 of the 28 most closely contested constituencies
“Road show”of debate screenings in 112 of 224 polling centers plus85 in satellite villages, approximately 19,000 individuals exposed
2,600 individuals into 6 treatment arms in another 40 polling centers
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 6 / 21
Implementation framework
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 7 / 21
Hypotheses
Pre-analysis plan publicly (PAP) registered before exit polls completed
Domain A, Voters: Exposure to debates is expected to increase:
Political knowledge: in general, and of candidates and policy stances
Policy alignment
Voter openness to other parties
Votes for the debate winner
Voting across ethnic-party lines
Secondary: turnout, perceived legitimacy, interest in politics
Domain B, Candidates: Does debate publicity complement or substitutefor campaign spending?
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 8 / 21
Econometric framework
PAP specifies the hypotheses, outcomes and econometricspecifications (including test direction, sub-groups and heterogeneity)
Estimating intention to treat effects: 82% of treatment groupreported attending a screening and 4% of controls
Main specification:Yipc = β0 + δTpc + X ′ipcΠ+ Z ′pcΓ+W ′
ipcΨ+ cp + εipc
- X is voter characteristics (omitted here)- Z is polling center stratification bins (# registered, dist. to nearest)- W is voter stratification bins (age, gender)
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 9 / 21
Debates had strong positive effects on political knowledge
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 10 / 21
Debates had strong positive effects on vote choices
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 11 / 21
What drives policy alignment?
Persuasion versus choosing aligned candidates
For FHC and priority issue, similar TE for those who selected thesame MP party in 2007 and 2013 (~persuasion) as those whoswitched parties (~choosing aligned)
Learning versus party loyalty
For FHC and priority issue, members of affi aliated tribes aresignificantly more likely to align than unaffi liated
FHC is a heavily partisan issue, while priority issue and GEB are not:similar TE for FHC and priority issue, no effect for GEB
TE on alignment equally strong in areas where all candidatesarticulated the same position as where they diverged
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 12 / 21
Endogenous campaign response
Domain B: Candidates increased their campaign efforts in communitieswhere screenings were held
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 13 / 21
Which aspects of the debates do voters respond to?
Individual-level treatments unpack the different types of informationconveyed by the debates (at i-level, estimating TOT):
Debate, viewed on tablet (N = 399)
Get to know you video: Candidate personality (402)
Radio report: Journalist summary of policy and qualifications (392)
Lab experiment: "Thin slice" impressions (to be analyzed) (407)
Surveyed controls (601)
Pure controls: surveyed only in exit polls (399)
Research question: how much do voters care about personality, charismaand physical appearance, as compared to “hard facts” about policystances and professional qualifications?
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 14 / 21
Debate impacts combine both persona and policy effects
All 3 inform voters about candidates and general politics; debates andradio inform about policy; only debates impact vote choice and alignment
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 15 / 21
Surveying matters
How much of these effects can be attributed to the content of treatmentversus the experience of being surveyed (see Zwane et al 2011)?
Possible mechanisms of survey priming effects include:- Increased salience of policy issues- Social desirability or Hawthorne effects- Enumerator may explain policies during interview
Data collection framework allows us to parse survey priming effectsfrom the priming function of debates themselves in two ways
- Priming, individual level: surveyed vs. pure controls- Reinforcing, polling-center level: surveyed vs. pure treatment
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 16 / 21
Survey priming effects - Individual level treatments
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 17 / 21
Survey reinforcing effects - Polling center level treatment
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 18 / 21
Limited heterogeneity in effects
Women benefited less than mean in terms of knowledge acquisition- Negative heterogeneous effects for 17 of 20 outcomes- 2 of the 3 mean effects indices are negative and significant
Little evidence for heterogeneous effect by age, interest in politics orfluency in Krio
Results for “voting for best” are robust to excluding eachconstituency one by one
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 19 / 21
Conclusions thus far
Voters acquired significant political knowledge from watching thedebates, knowledge that persisted through Election Day andinfluenced their voting choices
Debates have the potential to make political contests morecompetitive, which is consistent with the observed increase incampaign effort
Information conveyed by debates is comprehensive - includingcharisma, professional qualifications and policy stances - and thecombination of factors is more powerful than each in isolation
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 20 / 21
Ongoing work
Descriptive analysis: correlates of debate performance, spillovers,dissipation over time, treatment saturation, electoral returns
Accountability: tracking the performance of elected MPs whoparticipated in the debates and those elected in control constituencies
Ease of inference: analyzing impacts of field lab experiments - visual,audiovisual, text - on voter inferences of candidate competence,election prediction and performance
Katherine Casey (Stanford GSB) Joint work with Kelly Bidwell (IPA) and Rachel Glennerster (JPAL) (*** Preliminary Results *** )DEBATES October 2013 21 / 21
Top Related