Data Quality: 7 Back-Office Data Checks to Improve Its Quality
October 28, 2020Programme Evaluation, Impact Assessment, and Impact Evaluation: Commonality and Differences
November 11, 2020RCTs: Best Practices in Managing a Randomized Controlled Trial
We’re hearing about Randomised Controlled Trials (RCTs) a lot more since Abhijit Banerjee, Esther Duflo, and Michael Kremer won the Nobel Prize in Economics in 2019. RCTs are no doubt the gold standard in proving attributable impacts caused by a development intervention.
They are a gold standard because randomization is the most unbiased method to find an ‘exchangeable’ control or comparison group where we measure what would be the outcome in absence of the programme (the counterfactual); in common parlance, this is also called obtaining ‘balance at baseline’ between treatment and control group. When the mean outcome in the control group is subtracted out from the outcome level in the treatment or intervention group, we get the attributable magnitude of impacts caused by the intervention. In RCTs, the design of the trial is the king (sample size, how to randomize, how we will measure the outcome), and not wizardry in statistical or econometric estimation!
If RCT design is the king, then the proper management and implementation of this design is the “king’s hand”. If the implementation and management are messed up, the nice theoretical properties of RCTs are compromised and the results can be invalid. Analytical solutions exist to subdue the adverse effect of non-adherence to the trial protocol, but not without compromises which might have made a non-randomized or observation evaluation a better choice in hindsight.
Read on to know the four most critical components of managing RCTs.
How to better manage RCTs

1. Stakeholder Engagement
Early discussions with stakeholders, primarily program implementers, funders, or government partners goes a long way in the successful implementation of a program. Discuss and decide on the following:
- start with understanding their needs, constraints, and intent for the program;
- assess if RCT is indeed the right design for the impact evaluation. Especially, don’t push it for proving the obvious/common sense results just because RCTs are publishable;
- anticipate and understand implementation challenges w.r.t. RCT and give options for sampling and randomization which will be least disruptive to the implementation plan, and
- transparently discuss what RCTs can and cannot do and justify costs (sample size and measurement tools).
The above discussion is necessary to build a strong relationship and mutual trust between evaluators and key stakeholders, which eventually helps in the smooth implementation of the programme. But, this is just a start, this relationship needs to be continuously nurtured while maintaining the independence and credibility of the evaluation team; a fine balance indeed.
2. Commission Process Evaluations and Monitoring Systems
If there is no intervention, then there is no RCT! Therefore, mechanisms to check and validate the progress of the programme are important. See if everything is working as intended and if the ‘theory of change’ is indeed working as expected, and in a rare case, whether the intervention design or implementation plans need to change. Process evaluations combined with monitoring systems are the best-suited methods for the above assessment.
3. Check Contamination
Just as the intervention is important to the RCT so is not having the same intervention in the control group. Such contamination of control groups happens when a part of the control group opts in for the intervention or gets the intervention from some other sources (and we cannot say no). Contamination will usually underestimate the impacts of the intervention. Therefore, it is in the programmes best interest to keep a tab on the control area through periodic random sample surveys, site visits, etc. These activities should be light touch. Sometimes, control areas are those which will be taken up by the programme at a later phase, and hence this data can help in programming activities in future as well.
4. Always learn from Grassroot levels
Learning with someone who has worked hands-on in programme implementation is never a bad idea; so what if a person doesn’t have a Ph.D.! Visit grass-root level implementers (with permissions of their management) to gain special insights on pathways to impact, key barriers, and enablers, beneficiary perspectives, comments on the theory of change. Trust us, you will get lots of insights that will help in better analysis and interpretation of findings. You will also be able to better assess any risk of serious deviation from RCT protocol which can put a question mark on the theoretical validity of your impact estimates.
We put these tips into practice in all evaluations we do at NEERMAN (of course subject to the funding constraints!)
SUGGESTED: Watch this webinar recording on the Design and Management of RCTs.
We’re also happy to answer any RCT related questions if Google has failed you. Set a micro-meeting with us.