close× Call Us
Title Political Information and Electoral Choices: A Pre-meta-analysis Plan
Post date 03/08/2015
C1 Background and Explanation of Rationale

Civil society groups and social scientists commonly emphasize the need for high quality public information on the performance of politicians as an informed electorate is at the heart of liberal theories of democratic practice (Fearon, 1999). The extent to which performance information in effect make a difference in institutionally weak environments is, however, an open question. Specifically when does such information lead to the rewarding of good performance candidates at the polls and when are voting decisions dominated by nonperformance criteria such as ethnic ties and clientelistic relations? This project covers seven separate studies (in Benin, Mexico, Brazil, India, Burkina Faso and two in Uganda) that address the above questions using interventions that provide subjects with information about key actions of incumbent political representatives. A subset of studies employ a factorial design in which information is provided through a public mechanism. Details of individual studies are registered separately.

C2 What are the hypotheses to be tested?


  • H1a Positive information increases voter support for politicians (subgroup effect).
  • H1b Negative information decreases voter support for politicians (subgroup effect).

Secondary outcomes:

  • H2a Bad news decreases voter turnout.
  • H2b Good news increases voter turnout.


  • H3 Positive (negative) information increases (decreases) voter beliefs in candidate integrity.
  • H4 Positive (negative) information increases (decreases) voter beliefs that candidate is hardwork- ing.
  • H5 Politicians mount campaigns to respond to negative information

Substitution effects:

  • H6 Information effects are more positive for voters that do not share ethnic identities.1
  • H7 Information effects are more positive for voters with weaker partisan identities.
  • H8 Information effects are more positive for voters who have not received clientelistic benefits from any candidate.

Context heterogeneous effects:

  • H9 Informational effects are stronger in informationally weak environments.
  • H10 Informational effects are stronger in more competitive elections.
  • H11 Informational effects are stronger in settings in which elections are believed to be free and fair.

Design Heterogeneous Effects:

  • H12 Information effects|both positive and negative|are stronger when the gap between voters' prior beliefs about candidates and the information provided is larger.
  • H13 Informational effects are stronger the more the information relates directly to individual welfare.
  • H14 Informational effects are stronger the more reliable and credible is the information source.
  • H15 Informational effects are stronger when information is provided in public settings.
  • H16 Informational effects are not driven by Hawthorne effects.
C3 How will these hypotheses be tested? *

1. Regression controlling for a predefined set of controls (see detailed pre analysis plan) with clustering at the constituency level or equivalent and weighting by inverse propensity weights when relevant.

Let \(Q_j\) denote the quality of candidate j and \(P_{ij}\) denote the prior of voter i regarding j. Define \(L^+\) as the set of treatment subjects for whom \(Q_j>P_{ij}\) or \(Q_j=P_{ij}\) and \(Q_j \geq \hat{Q}_{j}\). These are subjects that receive good news | either the information provided exceeds priors or the information confirms positive priors. Let \(L^-\) denote the remaining subjects. Let \(N^+_{ij}\) denote the difference \(Q_j - P_{ij}\), defined for all subjects in \(L^+\) and standardized by the mean {and standard deviation} of \(Q_j - P_{ij}\) in the \(L^+\) group in each country (or relevant locality). \(N^+_{ij}\) is therefore a standardized measure of "good news'' with mean 0 {and standard deviation of 1}. Let \(N^-_{ij}\) denote the same quantity but for all subjects receiving bad news.

Then the two core estimating equations are:

\(E(Y_{ij} | i \in L^+) = \beta_0+ \beta_1 N^+_{ij} + \beta_2 T_i + \beta_3 T_i N^+_{ij} + \sum_{j=1}^k(\nu_k Z_i^k + \psi_k Z_i^kT_i) \label{eq.main1a}\)

\(E(Y_{ij} | i \in L^- ) = \gamma_0+ \gamma_1 N^-_{ij} + \gamma_2 T_i + \gamma_3 T_i N^-_{ij} + \sum_{j=1}^k(\nu_k Z_i^k + \psi_k Z_i^kT_i) \label{eq.main1b}\)

where \(Z_1, Z_2,...,Z_k\) are prespecified covariates, also standardized to have a 0 mean.

2. Primary analysis pools data; secondary analysis uses Bayesian hierarchical regression to estimate average treatment effects for each site as well as population parameters.

C4 Country
C5 Scale (# of Units) not provided by authors
C6 Was a power analysis conducted prior to data collection? For individual studies
C7 Has this research received Insitutional Review Board (IRB) or ethics committee approval? Individual studies will receive approval
C8 IRB Number not provided by authors
C9 Date of IRB Approval not provided by authors
C10 Will the intervention be implemented by the researcher or a third party? Mixed; generally by researchers in partnership with third parties
C11 Did any of the research team receive remuneration from the implementing agency for taking part in this research? No
C12 If relevant, is there an advance agreement with the implementation group that all results can be published? Yes
C13 JEL Classification(s) not provided by authors