NEW UPDATE: Completed trials may now upload and register supplementary documents (e.g. null results reports, populated pre-analysis plans, or post-trial results reports) in the Post Trial section under Reports, Papers, & Other Materials.
Central Government Performance and Local Election Outcomes: A randomized experiment
Last registered on October 25, 2017


Trial Information
General Information
Central Government Performance and Local Election Outcomes: A randomized experiment
Initial registration date
October 25, 2017
Last updated
October 25, 2017 11:31 PM EDT
Primary Investigator
Nova School of Business and Economics
Other Primary Investigator(s)
PI Affiliation
Imperial College Business School
PI Affiliation
Nova School of Business and Economics
PI Affiliation
Nova School of Business and Economics
Additional Trial Information
Start date
End date
Secondary IDs
There is a belief that local electoral outcomes are connected with voters’ perception about central governments’ performance and quality. We implement an experimental study, in the 2017 Portuguese municipal elections. We randomly exposed participants to positive, neutral and negative news about the central government in office. We use a sample of around 3 000 undergraduate and masters students of two business schools in Lisbon. Our goal is to assess how information affects turnout and voting decision – namely, local votes for the central government party – in local elections. Our sample includes subjects who never voted before the 2017 local election, together with more experienced voters. A by-product of our analysis is to test if the impact of information is differentiated across these two subsamples.
External Link(s)
Registration Citation
Carvalho, Bruno et al. 2017. "Central Government Performance and Local Election Outcomes: A randomized experiment." AEA RCT Registry. October 25. https://doi.org/10.1257/rct.2539-1.0.
Former Citation
Carvalho, Bruno et al. 2017. "Central Government Performance and Local Election Outcomes: A randomized experiment." AEA RCT Registry. October 25. http://www.socialscienceregistry.org/trials/2539/history/22694.
Experimental Details
The treatment questionnaire Q2 consisted of positive, negative and neutral information about the performance of central government office. Positive and negative information consisted of adapted news from reference newspapers. We included pieces of information on the following topics: public finance, social security, health, education, labor and road accidents. We created 6 different versions of each of the positive and negative treatments. That is, we had a total of 12 treatments, plus a control version with neutral information. Each version of the treatment questionnaires included two pieces of information on two of the above topics. The objective of creating different versions was to avoid that any effect found in the data is driven by a specific piece of information which, for reasons not foreseen by the research team, might drive the respondents’ behaviour.
The control group was presented with information about "Cão Serra da Estrela", an indigenous, not in danger of extinction, dog breed in Portugal.
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
Election turnout and voting decision.
Primary Outcomes (explanation)
Secondary Outcomes
Secondary Outcomes (end points)
Willingness/propensity to acquire new or additional information about municipal elections, candidates, electoral programs and/or the voting system as a result of participation in the study.
Secondary Outcomes (explanation)
Measured through direct question (e.g. do you think you had searched for new or additional information due to the participation in these questionnaires?).
Experimental Design
Experimental Design
We implemented a series of in-classroom questionnaires in two business schools in Lisbon, in the two weeks before the 2017 local elections and the week immediately after the election.
i) In one of them we conducted 3 rounds of questionnaires. The baseline (Q1) and treatment (Q2) questionnaires were implemented in each of the two weeks prior to the municipal elections, respectively, and the follow-up (Q3) in the week after the municipal election.
ii) In the other business school, we implemented 2 questionnaires. Therefore, we collapsed baseline (Q1) and treatment (Q2) surveys in one. For parsimony, the collapsed version has fewer questions than the joint Q1 and Q2. This was implemented in the week before the election. The follow-up Q3 was implemented in the week after the election.

The treatment questionnaire Q2 included either positive or negative information about the incumbent government at the central level. There was also a control Q2 version with neutral information, unrelated to the government. The treatment questionnaire was randomised at the individual student level.

The questionnaires were implemented in paper format. The baseline questionnaire Q1 was on sociodemographic data, as well as political orientation, alignment, past experience in elections, and perception about the central government's performance in different areas. Most importantly, we also get information about voting intention (whether subject intends to vote and whom is she going to vote for) in the upcoming municipal election. The final Q3 (follow-up) questionnaire was primarily to obtain information on whether the person voted and whom she had voted for.
Experimental Design Details
Randomization Method
The different versions of positive (P) and negative (N) papers were shuffled and put in different piles. The third pile was the control questionnaires (C). We then created bundles of 6 questionnaires, in the following order: CPPCNN.
Finally, we created a single pile with successive bundles of CPPCNN. During the implementation, questionnaires would be distributed by rows. Therefore, the format 'CPPCNN' prevented to the greatest extent possible positive and negative questionnaires being answered side by side, while it also maximized the probability that the same number of questionnaires was answered in each part in the classroom. Notice that there are no predetermined seats in the classrooms, so the way students seated on that day is another element of randomization for this experiment.
Randomization Unit
Individual (student) level.
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
2 Business schools, 80 classrooms.
Sample size: planned number of observations
3000 students.
Sample size (or number of clusters) by treatment arms
1000 Negative treatment (166 of each version), 1000 Positive treatment (166 of each version), 1000 Control.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB Name
IRB Approval Date
IRB Approval Number
Post Trial Information
Study Withdrawal
Is the intervention completed?
Is data collection complete?
Data Collection Completion Date
Final Sample Size: Number of Clusters (Unit of Randomization)
Was attrition correlated with treatment status?
Final Sample Size: Total Number of Observations
Final Sample Size (or Number of Clusters) by Treatment Arms
Data Publication
Data Publication
Is public data available?
Program Files
Program Files
Reports, Papers & Other Materials
Relevant Paper(s)