Job referrals and strategic network formation -- Experimental evidence from urban neighbourhoods in Ethiopia

The study is a field experiment on job referrals in day labor markets.

The study comprises three different referral interventions:
1. Referral treatment (open, control) (C)
2. Referral treatment (performance incentivised) (T1)
3. Referral treatment (partly anonymous) (T2)
4. Referral treatment (fully anonymous) (T3)


In detail, the treatments/interventions are defined as follows: The control referral treatment (C) is an un-incentivized, open job referral, similar to `standard' job referrals mostly observed in reality. Workers in this group can invite someone from their neighborhood network to the next work session without incentives and `in the field'. Treatment variation 1 (T1) introduces an incentive treatment, meaning that the worker making a job referral gets a financial reward which is linear in the performance of the invited worker.The second margin of variation shocks the visibility of the referral. Workers in the control referral condition can additionally make partly anonymous or fully anonymous referrals to other workers. Under partly anonymous referrals (T2), the invited worker does not know the identity of the inviting worker, but can reciprocate the referral, while under fully anonymous referrals (T3), reciprocation is now allowed.

In general, the day work I am providing is similar to office-style, workplace-related jobs. The task performance is incentivised for all participants, with a maximum payment of 200 ETB (~ 7 GBP). Work sessions take place on consecutive days or are spread over a whole day. The multiple rounds of the experiment are unannounced ex ante (to ensure comparability between other/same referrals).

2016-12-15

2017-02-15

I measure the effects of the referral type on three main outcomes of interest: i) worker i’s productivity, expressed as correctly entered work teams, P_i, ii) worker i’s network centrality, measured as the number of (in or out) links at baseline, and iii) reciprocity: an indicator that equals one if worker j refers i to the job in period t-1, and i refers j in return in period t .

Short summary:
- Baseline survey of over 700 individuals;
- Several rounds of a work experiments with three different referral treatments;
- (Short and long run) Follow-up phone survey of participants

Details:
- I randomly select 16 densely populated urban neighbourhoods in Addis Ababa, where I collect baseline data on young unemployed individuals and their social networks.
- Within the selected neighbourhoods, I take a census (door-to-door solicitation) of all eligible resident individuals, where eligibility is defined as follows: Permanently living in the selected neighbourhood; between 18 and years of age; not in permanent employment or education.
- These eligible individuals are surveyed with a baseline questionnaire, including questions on personal characteristics, labour market outcomes, behavioural questions, and a detailed social network section, containing a range of questions about every other eligible individual in the neighbourhood.
- I then conduct a work experiment in various locations in Addis Ababa: After the baseline survey, a random subsample from each neighbourhood is invited to a paid day job (the experiment). Sessions take place separate by neighbourhood
- The sessions consist of a work-related and quantifiable effort task. After the completion of the task, participants are randomly allocated to different work referral treatments (described above). The work experiment is repeated over three rounds.
- A few days after participating in work experiment, participants are called by phone and asked follow-up questions on the referral decisions they made.
- Eventually, this experiment tests what influences the linking decisions between participants through the different referral treatments, and whether these social job networks are formed strategically. A particular focus will be on the role of networks centrality in referring other workers. A further focus will be on reciprocity in referring (measured over the multiple rounds of the experiment) as well as permanent exclusion from the temporary day labour market opportunities I am generating.

Randomisation before sessions done in office by a computer (allocation to work sessions and treatment groups)

Individuals (young job-seekers)

No

16 neighborhoods, without clustering. All treatments are implemented in all neighborhoods

Approximately 700 individuals

150 referral treatment 1, 150 referral treatment 2, 150 referral treatment 3, 150 control group