Back to History

Fields Changed

Registration

Field Before After
Last Published April 13, 2026 09:09 AM June 06, 2026 11:25 AM
Study Withdrawn No
Study Withdrawal Date March 31, 2026
Intervention Completion Date March 31, 2026
Data Collection Complete Yes
Was attrition correlated with treatment status? No
Final Sample Size: Total Number of Observations 319 students (econometric panel: 319 × 5 scenarios = 1,595 observations). Note: 330 students completed the survey; 11 were excluded due to incomplete scenario data.
Final Sample Size (or Number of Clusters) by Treatment Arms Control group (video = 0): 151 students Treatment group (video = 1): 168 students Total: 319 students
Is there a restricted access data set available on request? No
Program Files No
Data Collection Completion Date March 31, 2026
Is data available for public use? No
Intervention (Public) The intervention consists of randomly presenting, through a short synthesized video, the results of the previous study. It also provides information on opportunities, promotion prospects, dismissal risks, and the advantages and disadvantages specific to each sector. Treatment group (video = 1): participants watched a short informational video (~3 min) presenting factual statistics on salary levels, working hours, layoff risk, and promotion prospects in the public and private sectors in Côte d'Ivoire, drawn from Méango & Girsberger (2025). Control group (video = 0): participants did not watch the video. After completing modules 1–3 (demographics, academic performance, stated employment preferences), they proceeded directly to the labor market beliefs and choice scenario modules. Both groups completed identical questionnaire modules; the only difference is exposure to the video.
Experimental Design (Public) The intervention consists of randomly presenting, through a short synthesized video, the results of the previous study. It also provides information on opportunities, promotion prospects, dismissal risks, and the advantages and disadvantages specific to each sector. This study uses two independent levels of randomization. (1) Video treatment (between-subjects): at the moment of connection to the online questionnaire (KoBoToolbox, March 2026), each participant was automatically assigned to the treatment group (video = 1, n = 168) or control group (video = 0, n = 151) via an independent Bernoulli draw (p = 0.5). The video was placed between module 3 (stated employment preferences) and module 4 (labor market beliefs), ensuring that all pre-treatment variables are exogenous. (2) Scenario attributes (within-subjects): each participant was presented with 5 pairs of hypothetical public vs. private job offers. Scenario 1 is a fixed symmetric reference (identical attributes in both sectors). For scenarios 2–5, the six attributes of each offer (monthly salary, weekly hours, annual layoff risk, promotion probability, and employer type) were drawn uniformly and independently at random from 3–4 pre-specified discrete levels, independently for the public and private offers. This within-individual variation is the source of identification for the fixed-effects preference models (WTA, TMS).
Randomization Method We used KoboToolbox as our data collection tool. This platform not only enables digital survey administration but also allows the integration of randomization mechanisms directly within the questionnaire. The randomization procedure is automatically implemented upon the first opening of the questionnaire by the enumerator. Specifically, a random draw is generated by the computer following a binomial distribution with probability 𝑝=0.5 X∼B(1,0.5) Thus, each respondent has an equal probability (50%) of being assigned to either group (treatment or control). This mechanism ensures random and independent assignment of participants, thereby supporting the internal validity of the study. Randomization was performed automatically by the KoBoToolbox platform upon each participant's connection to the survey. Each individual was assigned independently via a Bernoulli draw (p = 0.5) to the treatment or control group, with no possibility of self-selection or re-assignment. Scenario attributes (levels 2–5) were drawn uniformly at random by the platform for each participant × scenario pair, independently of the video assignment.
Randomization Unit Students Individual student. Randomization was conducted at the individual level — each student was independently randomized regardless of institution (ENSEA or INPHB). The treatment was NOT clustered at the school level. Both treatment and control students were recruited simultaneously from both institutions
Back to top