2,340 rated or chosen profiles from 390 respondents, in up to 3 tasks of 2 profiles varying on 6 attributes.

About this table

Description	Online panel respondents in Kenya and Malawi (Qualtrics, March-April 2022; 390 respondents) compared images of polling-station vote tally sheets in 3 pairs, varying signatures of the presiding officer, ruling- and opposition-party agents and an observer, and crossed-out corrections on either party's tally. Outcomes: which sheet is more reliable, plus 1-5 ratings of misconduct likelihood and four reasons.
Reference	Mbozi, F. (2024). When do voters see fraud? Evaluating the effects of poll supervision on perceptions of integrity. Journal of Experimental Political Science.
DOI	10.1017/xps.2024.5
Licence	CC0 1.0
Source data	https://doi.org/10.7910/DVN/EZS44G

Notes

Conjoint table (irw_conjoint). Image conjoint of vote tally sheets; attributes are the author's coding of six manipulated features. choice: which of two sheets has "more reliable information about the vote outcomes" (forced, no opt-out); rating_misconduct 1 = very unlikely .. 5 = very likely misconduct occurred, and four 1-5 agree items (tally changed, party interests unprotected, party influence, officer pressure); all ratings: higher = more suspicion, kept as asked. Clean and corrected images came from separate pools (non-uniform combinations); 21 tasks with identical feature sets are kept (the author drops them, 2,298 rows). Vote numbers on the images not stored.

These notes describe the source data, which the IRW reproduces as released. All notes are listed on the Data Notes page.

Design

Topic	democracy institutions
A profile represents	message
Country	KE, MW
Language shown	en
Language of stored levels	en
Randomization restrictions	observed in the data (not documented by the source): images drawn from a pool in which 'clean' forms (no corrections, 16 signature combinations) and 'error' forms (at least one correction) are separate sets of roughly equal use, so both tallies uncorrected occurs in 40% of profiles (930/2340) rather than the 36% the margins imply, and each correction pattern with at least one correction ~20%; within pools combinations roughly uniform. Level weights observed: each tally attribute is 'No correction' in ~60% of profiles (1407 and 1383 of 2340) vs ~40% 'Crossed out and corrected'
Task order	recorded in the deposit
Profile position	recorded in the deposit

Because the attribute levels were not randomized independently, an estimator has to account for the restriction; a plain difference in means across levels can mislead.

Outcomes:

choice	Between these two vote tally sheets, click on the one that you believe has more reliable information about the vote outcomes at its polling station? [choice, forced]
rating_interests_unprotected	To what extent do you agree with the following statements: The interests of political parties were not protected at this station [rating, 1 (Strongly disagree) to 5 (Strongly agree)]
rating_misconduct	In your opinion, what is the likelihood that electoral misconduct occurred at this polling station? [rating, 1 (Very unlikely) to 5 (Very likely)]
rating_officer_pressure	To what extent do you agree with the following statements: The presiding officer may have been put under pressure from other groups in the polling station. [rating, 1 (Strongly disagree) to 5 (Strongly agree)]
rating_party_influence	To what extent do you agree with the following statements: It is very possible that a political party had too much influence at this polling station. [rating, 1 (Strongly disagree) to 5 (Strongly agree)]
rating_tally_changed	To what extent do you agree with the following statements: The vote tallies may have been changed in favor of a specific candidate. [rating, 1 (Strongly disagree) to 5 (Strongly agree)]

Columns

attr_observerattr_opposition_agentattr_opposition_tallyattr_presiding_officerattr_ruling_party_agentattr_ruling_party_tallychoicecov_agecov_countrycov_educationcov_gendercov_trust_embidprofilerating_interests_unprotectedrating_misconductrating_officer_pressurerating_party_influencerating_tally_changedtask

The layout (one row per respondent, task and profile; choice/rating outcomes; attr_ columns holding each attribute level as displayed) is defined in the IRW conjoint standard. What the outcomes asked is under Design above.

Get the data

Download CSVno account needed Browse on Redivisexplore and query Croissant metadataHugging Face, Kaggle, OpenML

Or load it directly in R or Python (both need a free Redivis account; the first call opens a browser to sign in):

# R
install.packages("remotes")
remotes::install_github("itemresponsewarehouse/Rpkg")
library(irw)
df <- irw_fetch("mbozi_2024_tally_sheets", source = "conj")

# Python
pip install irw

import irw
df = irw.fetch("mbozi_2024_tally_sheets", source="conj")

How to cite

@article{Mbozi_2024, title={When Do Voters See Fraud? Evaluating the Effects of Poll Supervision on Perceptions of Integrity}, volume={12}, ISSN={2052-2649}, url={http://dx.doi.org/10.1017/xps.2024.5}, DOI={10.1017/xps.2024.5}, number={1}, journal={Journal of Experimental Political Science}, publisher={Cambridge University Press (CUP)}, author={Mbozi, Fanisi}, year={2024}, month=July, pages={27–41} }

Respondents	390
Rows (one per respondent, task and profile)	2,340
Tasks per respondent (most)	3
Profiles per task	2
Attributes	6
Tasks with no profile chosen	0

IRW version	v584
Redivis dataset	irw_conjoint v4.2
Redivis dataset DOI	10.57761/qkye-q530
Manifest pin for this IRW version	v4.2
Metadata source	irw_meta v44.0

mbozi_2024_tally_sheets