8 steps for inferring Causality

by Sep 21, 2016

Here is another visual rescued from the past that had never seen the light until now.

As an evaluator, one of my first obsessions was how to claim causality: how can you be sure that one intervention (a program, a policy) has been the one and only cause of some (positive) changes in reality, so you can claim that the program had that effect or impact. This is one very common debate in the evaluation community, that involves many different concepts such as the theory of change of the program, social underlying mechanisms, credible evidence, confirmation bias, among many others.

Well, in 2014 I was trying to understand the theory behind this causation issue and which tools I should be aware of to be able to make such analysis in an evaluation, when I ran into this article “Causal inference for program theory evaluation” by Patricia Rogers and Jane Davidson (both within my top 5 of favorite evaluation authors), which made me see the light and suddenly realize that if you follow these 8 progressive methodological steps one can reasonably conclude that the intervention seems to be the most probable cause of the effect:

I’m not convinced I fully understood the whole concept back then, and I don’t think I totally grasp it yet, as I have never been involved in an impact evaluation that required me to go through these steps. However, I wish to thank Patricia and Jane for their guidance, as usual, and I’m sure I will revisit this when I’m confronted to a suitable opportunity in my practice.

Anyway, I wanted to share it in case any of you can find it useful… or even help me walk the extra mile with it! Cheers 🙂


  1. Omega_Masha!

    Common frameworks for causal inference are structural equation modeling and the Rubin causal model .

  2. Patricia Rogers

    Thanks for sharing this visualisation of causal inference. I have two suggestions – one is unpacking what we mean by attribution and an initiative being THE cause of specific outcomes. It’s much more likely that the program forms part of a causal packages – often neither sufficient to produce outcomes (eg favourable context needed) nor logically necessary (eg there are other ways of achieving the outcomes of improved education, housing, health etc). So contribution seems more often to be the relevant concept. I also think it’s important to recognise non-counterfactual strategies as being sometimes more appropriate and useful than counterfactual strategies – eg when it is not possible to construct or identify a credible counterfactual.

Submit a Comment

Your email address will not be published.

You want to see more Visuals?

Thanks for sharing it if you found it interesting!