The effect of peer earnings information on wage expectations

The experiment will focus on how individuals update their wage expectations in response to different types of information about the actual income distribution. Groups of students, who are part of an on-going phone survey which is tracking their progression from school to work, will be sent information on average earnings by SMS. There will be three different messages which are sent out:
1) A general message: summarises wage information from the entire sample of students
2) A school type specific message: summarises wage information from the sub-sample of participants that attended the same school type as the recipient (public/private)
3) A field-specific message: summarises wage information from the sub-sample of participants in the same study field as the participant.

The wage information shared will be the most recent results from the on-going tracer study of technical and vocational students across Mozambique. Note, the specific information contained in each SMS varied by survey round; and, in the second and third types of message, the information varied by individual according to the school type they had attended or their field of training. Mirroring variation in actual wages, this design provides for substantial variation in the underlying wage information received.

2020-04-06

2020-09-22

We have three primary outcome variables: beliefs about current peers' earnings, current reservation wage, and future own-earning beliefs (expected wage at the end of 2020).

Directly after the baseline survey, which was conducted in October 2019, individuals were randomly-allocated to one of five experimental arms (including the control). The four treatment arms are distinguished by the type of messages they would receive. Specifically:
• Group one (general): received the general (all-student) message in all relevant rounds;
• Group two (school-type): received the school-type-specific message in all relevant rounds;
• Group three (field): received the field-specific message in all relevant rounds; and
• Group four (mixed): received the general message in round two, the school-type specific message in rounds three and four, and the field-specific message in rounds five and six.

The full sample is being tracked as part of a broader tracer study, where, on a quarterly basis, the participants respond to a phone survey on their current employment situation. Through these phone surveys the outcome variables, as well as other valuable employment information, is tracked over time.

Randomisation conducted in STATA. Randomisation across the five experimental arms was stratified by study field and gender, with a
target of around 260 participants in the control group and 340 in each treatment arm.

Individual

No

1 control group (259 individuals), 4 treatment arms (each of approximately 340 individuals)

1,622 individuals

Control group: 259 individuals
Treatment group 1: 337
Treatment group 2: 343
Treatment group 3: 342
Treatment group 4: 341
Total: 1,622

Given the size of the experimental sample and the allocation into treatment arms, using an assumption that natural log of the expected wage in the control group equals 9.4 with a standard deviation of 0.7 (calculated from the baseline survey), the minimum detectable effect (at the 10% significance level) is 0.12 considering all treatment groups (combined). Or, for the smallest treatment group alone, the minimum detectable effect is a 0.14 log points. The former is about a three quarters the effect size found in Jones and Santos (2020).

The effect of peer earnings information on wage expectations: Design and pre-analysis plan