x

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.
Election Observers, Photo Evidence and Result-sheet Discrepancies: A Field Experiment During the 2019 General Elections in Malawi
Last registered on February 07, 2020

Pre-Trial

Trial Information
General Information
Title
Election Observers, Photo Evidence and Result-sheet Discrepancies: A Field Experiment During the 2019 General Elections in Malawi
RCT ID
AEARCTR-0005403
Initial registration date
February 06, 2020
Last updated
February 07, 2020 1:48 PM EST
Location(s)

This section is unavailable to the public. Use the button below to request access to this information.

Request Information
Primary Investigator
Affiliation
London School of Economics and Political Science
Other Primary Investigator(s)
PI Affiliation
London School of Economics and Political Science
Additional Trial Information
Status
On going
Start date
2019-04-01
End date
2020-09-30
Secondary IDs
Abstract
This study examines the effect of election observers taking pictures of result-sheets in constituency tally centres. We focus on the Malawi 2019 general elections, during which we cooperated with domestic election observers to examine the impact of different observer methods. We argue that by taking pictures at crucial points in the results-transmission process, election observers can deter possible attempts to alter vote figures and errors while results are being transmitted and tabulated. To investigate this further we randomly assigned observers to take pictures of result-sheets at constituency tally-centres during the 2019 elections. We further collected data on the election results in each polling station both prior to results being transferred (at polling station level) as well as upon the completion of the result-transmission process (as electoral commission-level). To estimate the impact of our treatment, we will compare the total deviation in results between those polling stations that were assigned to treatment and those that were assigned to the control group.
External Link(s)
Registration Citation
Citation
Ahlback, Johan and Ryan Jablonski. 2020. "Election Observers, Photo Evidence and Result-sheet Discrepancies: A Field Experiment During the 2019 General Elections in Malawi." AEA RCT Registry. February 07. https://doi.org/10.1257/rct.5403-1.0.
Experimental Details
Interventions
Intervention(s)
The intervention carried out in this research project was done as a collaborative effort between London School of Economics and Political Science (LSE) and the National Initiative for Civic Education (NICE). NICE has a long history of election observations and voter education in Malawi. For the 2019 Malawi General elections, NICE had prepared a large-scale polling station observation, complemented by ward-level observers whose task it was to report in real-time on election related incidents. In total, NICE assigned stationary polling station observers in all 5,002 polling stations across the country. These monitored voting throughout Election Day. Once voting was completed, polling station observers also copied the vote figures for the presidential race from the result-sheet in all polling station before they were sent onwards. In addition, NICE also assigned 462 mobile election observers to each ward. Upon completion of voting, one ward-level observer in each constituency was tasked to monitor the constituency tally centre. Ward-level observers were equipped with smartphones to enable them to report in real-time on possible election incidences throughout Election Day.

Given this context, an intervention under the supervision of both LSE and NICE was carried out to evaluate the effectiveness of new mobile phone-based election observation methods. In particular, we wanted to know whether having observers take pictures of individual result-sheets at the middle of the results-transmission process could deter possible attempts to alter vote figures. Previous studies have shown that the presence of election observers tend to raise the costs of electoral misconduct in those locations where they are present, making it more likely that election officials and local actors abide by rules and refrain from committing fraud (Hyde, 2007; Enikopolov, et al., 2012; Ofosu & Posner, 2015; Callen & Long, 2015; Asunka, et al., 2017). Callen and Long (2015) have also shown that taking pictures of result-sheets at polling station level can expose attempts of tabulation-fraud and that providing officials with information about the intention to take pictures can deter officials from even attempting such fraud.

To examine the effect of having observers take pictures of result-sheets during the results-transmission process we randomly assigned NICE ward-level observers in 97 out of the 193 constituencies to gather photo evidence at the constituency tally-centre using their smartphones. The randomisation was done using direct random draw without blocking. Prior to the election NICE had verified with the MEC that they had the legal right to employ such observer methods at the tally-centres. On Election Day, upon arrival to the tally-centre, the assigned election observers were instructed to inform the official in charge of his/her intention to take the pictures. Once all counting and tabulation was completed at the tally-centre, the assigned election observer would then proceed to take pictures of all polling station result-sheets that had arrived to the tally-centre, including for presidential, parliamentary and local councillor races. Pictures were sent electronically via the observers’ mobile phones and then collected by the research team.
Intervention Start Date
2019-05-21
Intervention End Date
2019-05-27
Primary Outcomes
Primary Outcomes (end points)
The primary outcome variable of interest is the potential deviation in vote-figures per polling station between those recorded at polling station level and those published by the electoral commission.

These will be measured by 1) the absolute deviation in votes per polling station i; 2) the increase in votes for candidate j in polling station I; and 3) the absolute deviation in the number of valid votes in polling station i.

Please see attached PAP for further details.
Primary Outcomes (explanation)
Please see attached PAP for further details.
Secondary Outcomes
Secondary Outcomes (end points)
Please see attached PAP for further details.
Secondary Outcomes (explanation)
Please see attached PAP for further details.
Experimental Design
Experimental Design
The experimental design takes into account the result-transmission process used in Malawi. Across the country, all 5002 polling station results are transmitted first to one out of 193 constituency counting centres, and then onwards to the central tabulation centre. To measure the impact of our treatment, we randomly assigned observers to take pictures in 97 constituency counting centres. The observer was instructed to inform the official in charge at the assigned counting centre and then proceed to take pictures using a smart-phone of all result-sheets arrived at that counting centre.

We expect the presence of observers taking pictures will deter potential attempts to fraudulently alter figures in the counting centre and reduce discrepancies due to human error. Therefore, we expect that H1 the mean absolute deviation in votes will be smaller in treatment versus control polling stations; H2 that the mean absolute deviation in the number of valid votes will be smaller in treatment than control constituencies; H3 that increase in votes for the incumbent will be smaller in treatment versus control polling stations; and H4 that decrease in votes for the leading opposition candidate will be smaller in treatment versus control polling stations.

Please see attached PAP for further details.
Experimental Design Details
Not available
Randomization Method
The randomisation was done using direct random draw without blocking.
Randomization Unit
Constituency counting-centre
Was the treatment clustered?
Yes
Experiment Characteristics
Sample size: planned number of clusters
193 constituency counting centres
Sample size: planned number of observations
5002 polling stations
Sample size (or number of clusters) by treatment arms
Please see attached PAP for further details.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
IRB Approval Date
IRB Approval Number
Analysis Plan

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information