Theory of Change for Development Evaluation

To conduct experimentation in development economics, practitioners use the method of monitoring and evaluation. These processes involve administering interventions to a target sample to measure impact after a pre-determined period. The theory of change is a tool used to define the research hypothesis for experimental or non-experimental studies. It is a mechanism that would cause the initial conditions to change. 

While correlation is a relationship between the initial conditions and outcomes, causation – the theory of change – specifies ‘how’ and ‘why’ these interventions lead to impact. It is the initial conceptualization, a roadmap, or a blueprint for development research. 

However, the term theory of change finds its use to denote different things by different organizations and practitioners. Words such as theory, assumption, hypothesis, and causal linkages, used interchangeably, bear no explicit explanation. There is no consensus of what the term concretely stands for and is just jargon. While it is used for strategic planning to map out the change and expected outcomes, it also is a way to articulate the process while monitoring development. 

One way to look at this is by focusing on the actors involved in the research. How will individuals act, given their perceptions, socio-cultural norms, and economic status? Does a shift in policy motivate people to behave in contradictory ways? Such questions inform us about the attributes and possible behaviors of our sample participants. 

Subsequent are the macro attributes of the society that can ex-ante evaluate the impact on structural, institutional, and systemic mechanisms. Network effects and cooperation between participants can act as powerful tools in fostering social changes. It is critical to identify and measure the outputs, assumptions, and outcomes of an intervention. Further, mapping out the underlying mechanism can build a generalized theory applicable to other research questions and settings. 

A word of caution. Though a generalized theory might be replicable to other research settings, the prospects are rare. For one, each context is unique in its way – economic, social, cultural, and political aspects differ from individual to individual, society to society, and region to region. 

Defining the theory of change also motivates the elaborateness of data required for analysis. If our assumptions expect different impacts on male and female participants, such as studies on women empowerment and its aspects, it is wise to collect gender-disaggregated data. However, in situations where religion drives cause and effect, caste-disaggregated data benefits analysis and conclusions. 

It is imperative to create a framework of relevant indicators that exhibit valid results and comply with our assumptions. The number and kind of indicators inform our data collection methodology – surveys, interviews, focus groups, or administrative data use. While administrative datasets are cheap, they are not reliable sources to analyze and derive conclusions. On the other hand, while instruments such as surveys and interviews are reliable, they are not easy to collect and analyze. 

Lastly, there are risks to potential impacts of the intervention. These might be internal or external, depending on the context of the study. Non-participation, self-reporting bias, and other factors are examples of risks and threats to the experimental investigation. 

We can understand this using a problem statement – Do mid-day meals increase student attendance and education outcome? Starting with our key audience, the children, we can specify an age group for which we want to measure impact. We can designate the setting for which we want to administer this intervention – a rural setting, in this case. Other indicators may include the household income, family members, number of siblings, parents’ occupation, etc. 

The input in this case-study is the mid-day meals at schools. The outcome we are interested in is student attendance and change in school performance, based on marks and grades. Increased attendance will perhaps be a behavioral change – meals will incentivize students to attend school more often as parents cannot afford to cook a meal for their children. The long-run impacts will be better nutritional outcomes as children eat meals rich in nutrients and minerals regularly. It will lead to better grades as children can concentrate better, maintain good health and attend school regularly. 

As promising as this framework sounds, there might be some threats to our expectations and assumptions. For one, it might be possible that students do not attend school for reasons such as traveling long-distance to go to school or the opportunity cost of working and earning to support the household is too high to give up. Even if students attend school on the pretense of receiving a mid-day meal, they might not necessarily concentrate during classes, resulting in no or little academic improvements. 

The theory of change motivates the chain of causality for our research in a structured way. The results change when a different chain of causal relations is applied, leading to diverse policy implications.

Thus, designing interventions and their implementation are crucial steps for the theory of change of monitoring and evaluation. It solves for theory failure if we align the outcomes and long-run goals to the intervention and outputs. It solves for implementation failure if we align inputs and interventions to behavioral outcomes and long-run effects.