The Impact of Information on Japan’s Residence Status on Nepali Youths’ Migration Decision-Making: Evidence from a Randomized Controlled Trial

The intervention consists of a structured information session designed to provide factual, Japan-specific guidance on migration procedures, visa categories, required qualifications, labor rights, expected costs, risks, and living and working conditions in Japan. The session lasts approximately one hour and is delivered in classroom format using slides, printed materials, and short explanatory videos. Immediately after the session, students complete a post-intervention survey. The total time commitment for treated participants, including the survey, is approximately two hours.
Students in the control group attend a separate session of similar duration, but the content is limited to general migration trends and the current situation of foreign employment in Nepal, without country-specific information, no details about Japan, and no practical guidance on visa procedures or requirements. This ensures equal time and attention across groups while withholding the treatment-relevant information.
Both sessions are delivered once, in person, by trained facilitators. All participants complete a baseline survey prior to any session, and a follow-up survey is conducted approximately three months after the intervention.

2025-11-07

2026-05-31

The study aims to identify the key factors that shape migration decisions, assess students’ understanding of Japan as a potential destination country, and explore their expectations regarding work and living conditions abroad.

1. Change in perception and attitude toward Japan as a destination country, including perceived suitability for work or study, expected working environment, and overall favorability.
2. Change in intention to migrate to Japan, measured through self-reported likelihood of choosing Japan over other countries, intention to begin preparation within one year, and subjective probability of successfully reaching Japan.
3. Change in expected migration outcomes, including students’ estimates of the cost, time, income, and savings associated with migrating to Japan.
4. Change in preparation-related intentions, such as willingness to study Japanese language, register for required tests, search for verified agencies, or plan finances.
5. Behavioral follow-up outcomes, measured approximately three months after the intervention, including whether participants have taken concrete steps such as enrolling in language classes, contacting a recruitment agency, applying for tests, preparing documents, saving money, or making a migration timeline.
6. Change in awareness of fraud prevention and safe migration practices, including whether students can identify risks, verify agencies, and recognize misinformation.

We implement a cluster randomized controlled trial with pairing at the district level. In each of Nepal’s seven provinces, we select the district with the highest and the district with the lowest out-migration per 1,000 population (based on census “absent population”), yielding 14 districts. From CollegeNP we compile district-specific rosters of secondary schools and colleges and randomly choose one school and one college per district, alongside ranked backup lists (top five). We then randomize once at the district level which institution type (school or college) is assigned to Treatment, with the other type in the same district assigned to Control (targeting a 7–7 split across districts). If a first-choice institution declines, we replace it with the next highest-ranked institution of the same district and type, which inherits the pre-assigned treatment status; no re-randomization occurs. We survey a fixed number of eligible students per institution at baseline, administer the intervention/control session, conduct an immediate post-survey, and follow up at approximately three months. Primary analysis uses district fixed effects with standard errors clustered at the institution level, supplemented by wild-cluster bootstrap inference.

Not available

All randomization procedures in this study were conducted in Stata using a fixed random seed to ensure full reproducibility. The randomization took place in two steps:
District-level randomization:
A list of 14 districts (one high-migration and one low-migration district per province) was created, and Stata’s set seed and runiform() functions were used to randomly order the districts. The first seven districts in the randomized list were assigned to school-treatment and the remaining seven to college-treatment.
Institution-level randomization:
Separate lists of eligible secondary schools and colleges were compiled for each district. Using Stata, we randomly assigned a number to each institution and selected the first-ranked school and first-ranked college as the primary institutions for the study. Additional institutions were randomly ranked (top five) to serve as replacements in case of refusal. If a selected institution declines participation, the next institution in the pre-randomized list is selected without re-randomization, and it inherits the same treatment status.
The randomization seed, code, and output files (district assignment file and institution selection file) were archived prior to implementation to ensure transparency and replicability. No manual selection or post-hoc alteration was made to the randomized lists.

The unit of randomization is the institution (cluster). In each of the 14 selected districts, one secondary school and one college were identified, and treatment was assigned at the district level by randomly determining whether the school or the college in that district would receive the intervention. Thus, each district contains one treated institution and one control institution, resulting in 14 treatment clusters and 14 control clusters (28 institutions in total). From each institution, approximately 20–30 eligible students (Grade 12 or Bachelor-level) will be surveyed, yielding an estimated total sample size of 560–840 students. All randomization procedures were conducted in Stata using a fixed random seed.

Yes

The study includes a total of 28 clusters, consisting of one secondary school and one college in each of the 14 selected districts. In every district, one institution is randomly assigned to the treatment group and the other to the control group, resulting in 14 treatment clusters and 14 control clusters.

Estimated total sample size of 560–840 students.

20-30 students from each institutions.

Assuming 28 clusters (14 per arm), an average of 25 students per cluster (20–30), ICC = 0.03, two-sided α = 0.05, 80% power, and analysis with district fixed effects plus baseline ANCOVA (conservative R² ≈ 0.30), the minimum detectable effect size (MDE) on the standardized knowledge index is about 0.30 SD; for key binary outcomes with a baseline proportion near 50%, the corresponding MDE is approximately 12 percentage points.