Back to History Current Version

Competition and Cooperation in Small Firm Networks: Evidence from garment makers in Ghana

Last registered on July 27, 2020


Trial Information

General Information

Competition and Cooperation in Small Firm Networks: Evidence from garment makers in Ghana
Initial registration date
March 05, 2015

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
March 05, 2015, 1:31 AM EST

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated
July 27, 2020, 4:06 PM EDT

Last updated is the most recent time when changes to the trial's registration were published.



Primary Investigator

New York University Abu Dhabi

Other Primary Investigator(s)

PI Affiliation
UC Berkeley

Additional Trial Information

Start date
End date
Secondary IDs
Empirical research on firm-level network connections in developing countries has focused to date primarily on agricultural technology adoption. Several influential papers document that in the context of agricultural commodities, social networks can be leveraged to increase technology adoption, agricultural yields, and ultimately incomes (e.g. Bandiera and Rasul, 2006, Conley and Udry, 2010, Foster and Rosenzweig, 1995).

Transferring these findings to other income generating contexts is complicated by the fact that many manufacturing and services firms compete more directly over local demand (which is essentially fixed in the short term) than communities of agricultural producers compete given a wider market or world price. This direct competition puts potentially profit increasing functions of firm-level networks in conflict with co-insurance functions of firm-level networks. As Barr (1998) observes, networks maintained primarily to address uncertainty and income variability generate greater positive spillover effects, but limit the potential for network connections to improve firm performance. The co-insurance motive may also actually limit competitiveness within a network, as firm owners prefer to maintain co-insurance relationships rather than aggressively compete and alienate potential risk sharing partners.

Small and micro enterprise owners in Ghana, and throughout the developing world, maintain complex network relationships within their industries. Our sample, the universe of garment making firm owners in a single mid-size town in the Volta Region of Ghana, is no exception. Baseline data reveals patterns of skill sharing, information sharing, productive input sharing, and outsourcing that suggest these firm-level networks serve an important co-insurance function.

In this experiment, a random subset of our sample will receive training in a new fashion style. Training will include a technical component, discussion of marketing the new style, and marketing materials. Leveraging baseline heterogeneity in the number and nature of network connections, in combination with the randomized experiment, we seek to characterize the nature of network relationships in our context and measure the effects of different types of relationships on firm performance.

Ex-ante, the overall predictions are ambiguous, but broadly we hypothesize that we will observe both positive direct effects of training on income and positive indirect/spillover effects of training on the income of trained firms' network connections. In addition, we expect these positive spillover effects to be strongest (1) for network connections with a history of skill sharing, (2) in the later weeks of the follow-up panel, and (3) for competition-distant network connections. In addition, we hope to document negative spillover/competition effects of training on firms that are close competitors of trained firms but not co-insurance network connections. Documenting the dynamic ability of trained firms to capitalize on any competitive advantage associated with the new style in the early period, and whether/how that competitive advantage dissipates due to co-insurance skill-sharing norms is a key ambition of this study.

Focus groups and qualitative interviews before the design of the experiment revealed that skill-seeking behavior may be an important skill diffusion channel, as the norm among skill-sharing partners requires positive response to requests for skill-sharing but not proactive/preemptive skill-sharing. Consequently, the experiment will also randomize a subset of the sample to receive information and marketing material about the fashion style, but not direct technical training. We expect these firm owners to seek technical instruction from skill-sharing network connections. More skill-sharing to informed (but not trained) firms would be evidence that the co-insurance mechanism functions imperfectly, where the ability to hide a new competitive advantage allows firm owners to retain that advantage longer.

External Link(s)

Registration Citation

Hardy, Morgan and Jamie McCasland. 2020. "Competition and Cooperation in Small Firm Networks: Evidence from garment makers in Ghana ." AEA RCT Registry. July 27.
Former Citation
Hardy, Morgan and Jamie McCasland. 2020. "Competition and Cooperation in Small Firm Networks: Evidence from garment makers in Ghana ." AEA RCT Registry. July 27.
Sponsors & Partners

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
Experimental Details


Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
Firm performance (profits, sales, outsourcing, number of garments made/altered); Firm inputs (labor hours, shared workers, physical capital); Network activity (skill-sharing, outsourcing, referral, productive input sharing, discussion of prices/marketing)
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
The randomization is stratified by gender and assigns 15 men and 52 women (15\% of each gender strata) to the training group, and 15 men and 52 women (15\% of each gender strata) to the information/marketing group.
Experimental Design Details
Randomization Method
Randomization done in office by a computer
Randomization Unit
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
445 firms
Sample size: planned number of observations
445 firms, 3 pre-treatment observations, 7 post-treatment observations
Sample size (or number of clusters) by treatment arms
67 training treatment, 67 info treatment, 311 control; spillover effects to control group firms will be measured using number of baseline network connections in the training and info treatment groups
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Institutional Review Boards (IRBs)

IRB Name
Human Subjects Committee for Innovations for Poverty Action IRB-USA
IRB Approval Date
IRB Approval Number
Analysis Plan

Analysis Plan Documents

Ghana Firm Networks PAP

MD5: 36ff8ae2e1b6ea28e9540cc8b01c2c07

SHA1: 6a2b207de9cbc26c67bea6b7b6a3508b33c73bbe

Uploaded At: March 05, 2015


Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information


Is the intervention completed?
Intervention Completion Date
October 01, 2016, 12:00 +00:00
Data Collection Complete
Data Collection Completion Date
October 01, 2016, 12:00 +00:00
Final Sample Size: Number of Clusters (Unit of Randomization)
417 firms
Was attrition correlated with treatment status?
Final Sample Size: Total Number of Observations
417 firms
Final Sample Size (or Number of Clusters) by Treatment Arms
15% of firms invited to training, 50% of firms received demand shock (cross-cut)
Data Publication

Data Publication

Is public data available?

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

This paper reports on an experiment that brings insights from the literature on demand-side determinants of technology adoption to the study of peer-to-peer diffusion. We develop a custom weaving technique and randomly
seed training into a real network of garment making firm owners in Ghana. Training leads to limited adoption
among trainees, but little to no diffusion to non-trainees. In a second phase, we cross-randomize demand for
the technique. Demand shocks increase adoption of the technology in both groups and diffusion to untrained
firms, generated by a pattern in which trained firm owners teach approximately 400% more of their peers if
they are randomly assigned to the demand intervention. We find no evidence that our main effects are driven
by differences in ability (learning-by-doing) or other adoption-based mechanisms. Rather, our findings are most
consistent with the demand intervention generating differential willingness to diffuse among potential teachers.
Morgan Hardy, Jamie McCasland, It takes two: Experimental evidence on the determinants of technology diffusion, Journal of Development Economics, Volume 149, 2021, 102600, ISSN 0304-3878.

Reports & Other Materials