An Internship Programme for Young Ethiopian Entrepreneurs
Last registered on March 12, 2018

Pre-Trial

Trial Information
General Information
Title
An Internship Programme for Young Ethiopian Entrepreneurs
RCT ID
AEARCTR-0002776
Initial registration date
March 12, 2018
Last updated
March 12, 2018 5:46 PM EDT
Location(s)
Region
Primary Investigator
Affiliation
Department of Economics, University of Oxford
Other Primary Investigator(s)
PI Affiliation
University of Oxford
PI Affiliation
Stanford University
PI Affiliation
Ethiopian Development Research Institute
Additional Trial Information
Status
Completed
Start date
2015-11-01
End date
2018-01-31
Secondary IDs
Abstract
We are currently running a field experiment in Addis Ababa, in which we place treated respondents in a four-week ‘management internship’, to work alongside middle and senior managers of established Ethiopian firms. This Pre-Analysis Plan outlines our primary hypotheses and accompanying identification strategy. At the time of writing, we are finalising 12-month follow-up interviews, and cleaning the data.
External Link(s)
Registration Citation
Citation
Abebe, Girum et al. 2018. "An Internship Programme for Young Ethiopian Entrepreneurs." AEA RCT Registry. March 12. https://www.socialscienceregistry.org/trials/2776/history/26552
Experimental Details
Interventions
Intervention(s)
We placed aspiring entrepreneurs into established Ethiopian firms, for them to be able to observe directly the process of management. Of our pool of applicants, we randomly gave half the opportunity to be placed into an established Ethiopian firm. Each participant was intended to shadow a middle manager or senior manager, observing how she or he works. Each participant was intended to spend one month in her or his host firm.
Intervention Start Date
2015-11-01
Intervention End Date
2017-01-31
Primary Outcomes
Primary Outcomes (end points)
Please see our pre-analysis plan document, which summarises this in detail.
Primary Outcomes (explanation)
Please see our pre-analysis plan document, which summarises this in detail.
Secondary Outcomes
Secondary Outcomes (end points)
Please see our pre-analysis plan document, which summarises this in detail.
Secondary Outcomes (explanation)
Please see our pre-analysis plan document, which summarises this in detail.
Experimental Design
Experimental Design
We placed aspiring entrepreneurs into established Ethiopian firms, to observe the process of management. Of our pool of applicants, we randomly gave half the opportunity to be placed into an established Ethiopian firm. Each participant was intended to shadow a middle manager or senior manager, observing how she or he works. Each participant was intended to spend one month in her or his host firm.
Experimental Design Details
Please see our pre-analysis plan document, which summarises this in detail.
Randomization Method
We randomised by drawing balls from a bag, having first constructed pairwise matches.
Randomization Unit
We randomised at the level of the pair; i.e. we paired interns and then assigned one to treatment and one to control.
Was the treatment clustered?
Yes
Experiment Characteristics
Sample size: planned number of clusters
800 pairs
Sample size: planned number of observations
1600 individuals
Sample size (or number of clusters) by treatment arms
800 interns assigned to treatment, 800 assigned to control
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
We used a design with 800 treated participants and 800 control, observed at six-month and 12-month follow-up. This design provides good power to detect even small treatment effects. To test and illustrate this, we consider a benchmark specification in which our control group has a 10% probability of transitioning into self-employment over the course of a year. We suppose that our treatment shifts this probability to 16%. This implies a value for “Cohen’s Phi” of 0.18, which is smaller than what is typically regarded as a ‘small’ effect for a difference in proportions (see Cohen (1988, p.184)). Using a significance level of 0.05, we obtain a power of 0.95. This value is obtained simply by a difference-in-means test at the 12-month mark (allowing for pair fixed effects). This implies that we can test separately for effects at the six- and 12-month marks, without needing to pool those periods to increase power. For example, suppose that, at the six-month mark, our treatment has increased the probability of starting a business from 5% to 10%; then we can still detect an effect with power of 0.96. We extend this analysis to a simple heterogeneous effects specification. We expect about 70% of our sample to be ‘economically active’ at baseline (see Table 4.1.10 of the Ethiopia Employment Survey 2012 (Central Statistical Agency of Ethiopia, 2012), considering respondents aged 20-24 in Addis Ababa Region). Suppose now that ‘economically active’ respondents have a probability of running a business of 30% in the control group at the 12-month mark, and that respondents ‘not economically active’ have a probability of just 5%. (These transition probabilities are drawn from the CSAE’s Ghana Urban Household Panel Survey (‘GUHPS’); we do not have transition data for urban Ethiopia, but the GUHPS provides a useful analogy for approximate figures.) Suppose that the treatment increases the probability of running a business from 30% to 35% (a very small effect), but increases the probability from 5% to 20% for those not earlier economically active. (This would be consistent, for example, with economically active respondents already being locked in to an activity – whether it be self-employment or wage employment – while economically inactive respondents are more open to new business opportunities.) In this hypothetical, the ATE is 0.08 (= 0.7 x 0.05 + 0.3 x 0.15), and estimation of equation (1) recovers this estimate with power 0.96 at a significance level of 0.05.
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
Central University Ethics Committee
IRB Approval Date
2013-03-04
IRB Approval Number
DREC 213/0013
Post-Trial
Post Trial Information
Study Withdrawal
Intervention
Is the intervention completed?
No
Is data collection complete?
Data Publication
Data Publication
Is public data available?
No
Program Files
Program Files
Reports and Papers
Preliminary Reports
Relevant Papers