What, exactly, are theories of change? It’s not intended as a trick question, but the wide range of views about theories of change causes a lot of problems. Despite conferences about theories of change, and a lot of academic literature, there is still no agreement on a single definition. Funders and implementers often require a project or programme to include a “theory of change” in planning, reporting or evaluation – but then expect their own specific version to be used. What should be a valuable tool that helps everyone to understand a programme or project too often becomes merely a chore when finishing the paperwork.
Summary of this article:
In a project, or programme, or any other intervention:
– a work-plan describes how results are to be achieved – it describes the actions and outputs that will lead to the results;
– a theory of change describes why it is expected that the actions and outputs will lead to the results.
In an ordered (or linear) situation a cause-and-effect type of theory of change is suitable, based on “if this – then that” relationships from the inputs through to the impacts.
In a complex adaptive situation, cause and effect are not repeatable and a theory of change based on people, process, participation and learning must be used. Adaptability is crucial, and adaptive programming provides one of several ways to address the complex domain.
The current discussion between proponents of different approaches to defining theories of change is, essentially, about knowing if the intervention is based in a chaotic, complex, or ordered context. Sense-making frameworks like the Cynefin framework provide a way to understand this and choose the right response for each situation.
Theories of change describe why a project or programme is expected to lead to a change from the current situation to something that is in some way “better”. There is nothing new about theories of change which have been used for about 50 years and are currently facing a resurgence of interest. Sometimes we know exactly where we want to end up, even before starting an intervention. But at other times we can do no more than describe the direction that we need to go, and then move gradually in the right direction to get closer a distant goal that may take many years to achieve. One example of this is when working in peacebuilding in complex situations. Trying to repeat what worked previously, or what worked in a different place may simply fail. In other situations repeating what is known to work may provide useful “good practice” to guide the work (e.g. installing a drinking water system in a town). The context is crucial to success.
To repeat the summary (above):
A work-plan describes how results are to be achieved by undertaking activities.
A theory of change describes why the required results are expected be achieved by the proposed actions and the outputs from the actions.
If we compare an intervention (a project or programme) to a journey, then a theory of change can be like a map, or like a compass bearing, or a set of turn-by-turn instructions, or even guidance as to who to ask in order to discover the route. There is no single answer. Nor is there a single way to share and explain the answer: a written document, a diagram or a map with – or without – notes, or a framework, are the common forms, but there are no restrictions. The flexibility makes theories of change a very powerful tool, but can cause a lot of confusion. To be useful, a theory of change must communicate understanding about why the intervention is expected to succeed.
Current definitions of theories of change often fall into one of the two main groups, summarised below. In practice there is now a chasm between two incompatible approaches. The different views are deeply held, and field staff may be told that they “must provide a theory of change for their work” without explanation of which version to use.
To understand “linear” theories of change it is useful to know about the widely used “results chain diagram”. This is often mistakenly called a theory of change, adding to the confusion. A results chain diagram is not a theory of change – it describes the sequence by which results will be achieved in an ordered or linear intervention, and is not relevant for a complex intervention.
For an in-depth look at theories of change the report by Isabel Vogel for DFID in 2012 gives a lot of information.
The two views of theories of change
The first view is that a theory of change comprises the cause and effect links that join the different results of an intervention, as well as the key assumptions.
This is a logical “if this – then that” description that moves along the results chain forwards, or describes what is needed from the previous stage of the results chain to enable progress.
A proponent of this approach is IRC who have made available on-line tools for the humanitarian sector. There are numerous other documents and “how to” guides published for this approach.
An example of this type of theory of change is when an agency runs a workshop to teach planning skills. The theory of change might include either:
IF we train people to plan THEN their planning skills will improve.
IF people have better skills THEN they will write better plans.
IF we want better plans THEN we need people with good planning skills.
IF people are to have better skills THEN training is required.
The reverse approach is preferred as it starts with the outcomes that are wanted, not the activities. In practice, a lot of interventions are planned by starting with the inputs.
To turn the results chain into a theory of change for a linear intervention the key assumptions that underlie the “if-then” logic must also be included. For example:
Assumption example 1: People will stay in their job after they are trained so that they use the new skills. If people leave the organisation immediately after finishing the training then there will be Outputs but no link to Outcomes.
Assumption example 2: There is sufficient political will to allow changes to the way planning is done. Better plans being produced is an Outcome, but if the plans are rejected by the government there is no link to an Impact.
The second view of theories of change is that change is about people, process, participation, and learning. Learning what works in a particular situation is at the core, and those who are involved in the change process are both the key actors in bringing about change as well as the people who are affected by the changes. Flexibility in implementation, based on continual learning, is essential as the situation part-way through the intended change will already be different to the situation at the start, and participants may have changed their views.
A leading proponent of this second approach is the UK Overseas Development Institute (ODI), with a guide by Craig Valters.
Four principles are set out in the ODI guide.
- Focus on process
- Prioritise learning
- Be locally led
- Think compass, not map.
This second view of theories of change is not logic-based. Adapting a theory of change based on process, people, participation and flexible learning, in order to make it match existing administrative and financial structures, that are designed for linear cause-and-effect, can be a challenge. When logframes are used as a planning tool frustration is a common experience. Defining an end-point that has to be reached, or clearly identifying “value for money” may not be feasible, especially in advance of starting work. More relevant measures may involve how much, and in what way the beneficiary community values the Impact of the intervention in their daily lives.
One way that this has been partially addressed is to use adaptive programming to provide a responsive linear model that more closely matches the complex issues.
However, matching the legacy of linear administration and finance to complex interventions in the short term is an imperfect fix and not a long-term solution.
How to reconcile the two opposing views ?
Once the difference between Complicated and Complex is clear, the two views both make sense and can be used as appropriate.
In everyday language, complicated and complex have almost the same meaning. However, just like Inputs, Outputs, Outcomes, have specific meanings for the results chain, Complex and Complicated also have specific definitions for Complexity Thinking. From a complexity viewpoint, there are three types of “system” (or “situation” or “intervention” or “context” or “domain”): (1) ordered or linear, (2) complex, and (3) chaotic. The rough equivalent in physics is solids, liquids and gases.
(1) In an ordered system cause and effect are linked, repeating the same actions leads to the same results each time, and the end goal and the route to get there can be set out before starting. Ordered systems can be divided into obvious systems where the cause and effect are obvious to most people, and complicated systems where cause and effect are still linked and repeatable, but it takes expertise and analysis to make the link.
Fetching a bucket of water is an obvious system. Find the bucket and the tap (faucet), put the bucket under the tap, turn on tap, wait until bucket is full, turn off tap. No need to analyse further. The series of actions is sense-categorise-respond: identify what is going on, select the best known solution, and apply it. For “obvious systems” there is often a single best solution that has been developed over years of experience, which can be described as “best practice”.
Installing a water supply system is (usually) in the complicated ordered domain. Cause and effect still dominate, but specific expertise is now also required: analysis of the hydrology, calculation of demand, slope, access, etc. Previous experience is still a good guide and – within constraints that can be accurately measured – the same basic solution can be repeated for different locations. The series of actions is sense-analyse-respond: identify what is going on, use expertise to make an analysis, and then apply the solution. There are usually multiple possible solutions based on knowledge and analysis so “good practice” is the guide (not “best practice”), and people with expertise will be part of the solution. But, this is still an ordered situation so linear cause and effect still applies and we can plan where we want to end up, and how to get there, before starting. A water supply system can usually be designed, budgeted and planned before starting work, and if the necessary skills and materials are available the plans can be realised on time and on budget.
An example: while filling a bucket with water, the tap breaks and can’t be closed; our “obvious” system just became “complicated”. Expertise will be needed (a plumber who knows how to fix or replace the tap) and there is more than one possible good response: find the main valve and turn it off, or jam something in the end of the pipe to stop the water flow, or if it is an outside tap, use the flow to water the garden until the plumber arrives.
The linear model of a theory of change that uses cause and effect is suitable for either an obvious or a complicated ordered project.
(2) In a complex situation repeating the same action does not get the same results each time. Cause and effect may be visible with hindsight but cannot be seen in advance. The full name for such a system is a “complex adaptive system” – any intervention changes the situation and affects the response. One well known example is the stock market: buying shares today and buying the same shares next year may produce widely different results. The cause and effect of the stock market crash of 2008 is now understood with hindsight, but very few people were able to predict it. Many humanitarian and development interventions, as well as a lot of peacebuilding actions, operate in complex environments. Linear cause and effect don’t apply – though all too often the planning and administration tools of donors (and implementers) rely on linear cause and effect. Clearly, in a complex situation the second type of theory of change that is based on people, process, participation, and learning is more appropriate. A description of the long term aim may not be helpful when starting out, it may be too distant and abstract. For example, peacebuilding that relies on defining an ideal situation where violent conflict no longer takes place and then trying to somehow reach that state, is generally less useful than identifying small moves towards a slightly better situation that can be taken in the short term. Small incremental steps can make an improvement to daily life and also build confidence so that each time the next step seems more feasible. Focussing on the “adjacent possible” rather than “long term impact” often works better in complex interventions. Because repeating an action may produce a different result each time, interventions can’t be replicated directly. Instead, small-scale probes are needed to find out what the reaction will be, and then a larger response can be used to build up actions that produced positive results. If the results of the quick probe are negative, then change course and damp-down the unwanted results. The theory of change can be described in terms of a compass direction for the next step, but not in terms of a distant (and theoretical) end-point that may never be reached.
The required sequence of actions for a complex situation is probe-sense-respond: find out what brings the adaptive situation to a better nearby state, build on that positive response to take a step in the right direction, then repeat the probe. If a probe produces unintended negative consequences it is not a failure, but an important lesson in what to avoid. Learning about what doesn’t work in a given situation is part of building success.
(3) In a truly chaotic situation there is no pattern of events, nothing can be predicted and the over-riding need is to bring about sufficient order to start to move away from chaos. Rapid response is needed, and the sequence becomes: act-sense-respond. Immediately steering the chaotic towards an increase in order is of more value than focussing on a longer term ideal. Failure to respond in time to chaos can lead to human suffering and death, or serve as a driver for extremism.
A useful sense-making framework that clearly sets out the obvious, complicated, complex, chaotic and disordered is the Cynefin framework – there is more information on this page.
Adaptive programming for complex interventions.
Adaptive programming and adaptive implementation are ways of addressing the issue of how to work, in practical terms, in complex interventions. Duncan Green’s blog post here gives an introduction and some useful links to DFID work. The adaptive “shopping list” for how to implement includes:
“adaptive, iterative, flexible, problem driven, politically smart and locally led”.
This is a useful approach with a lot of practical applications, but in many ways represents a users’ guide for “how to successfully hammer a square (linear) peg into a round (complex) hole”. In practice, this may be the best that can be done if funding and administration is locked to a “cause and effect” model that pretends all situations are predictable. Adaptive programming may be pragmatic, and it includes political and power issues which are crucial, but it risks spending a lot of effort to rediscover, through extensive field work, what is already well known and understood through formal complexity thinking. Complexity thinking is already used in other areas of work like complex computer software development; the humanitarian/development/peacebuilding sector does not need to reinvent the wheel when it can learn from the success of others. Setting aside what is already known and well documented and starting to rediscover complexity thinking by experimenting with new ways of “adaptive implementation” is likely to be a slow, difficult and costly route.
The value of a theory of change comes from increasing understanding (by participants, beneficiaries, donors, and the general public) about why it is expected that the chosen intervention will produce the desired results – the workplan describes how the activities and outputs will achieve the results.
There is no single form of a theory of change that has universal application. The two main versions: linear cause and effect, and the people, process, participation and learning model, each has application in a separate context. Complexity thinking describes these contexts as “ordered”, and “complex adaptive” domains, respectively. There is not yet a widely used theory of change model for the chaotic domain, experienced field practitioners are needed to lead the response to chaos as acting to reduce the chaos is the first step.
The two different theory of change approaches have been presented as a disagreement, and no single unifying model has been agreed. However, there is no point in arguing about the symptoms (incompatible views about theories of change) until we understand the root causes (the need for different responses to: complex environments, complicated but ordered environments, and chaotic environments). Once the context is understood the whole disagreement about theories of change reduces to deciding which is the context-appropriate method. A single theory of change for all circumstances is not possible, but also not desirable.
The humanitarian, development and peacebuilding sector does not, in general, know about complexity thinking and sense-making frameworks, and this gap in knowledge will continue to limit the value of using theories of change. The frustration (and even anger) that many practitioners feel when forced to use linear tools for a complex domain could be reduced by an understanding of complexity. Fortunately, available on-line resources about sense-making frameworks like the Cynefin framework are readily available, they are well established in other complex domains, and their use in humanitarian interventions, development and peacebuilding is increasing.
A more difficult challenge is the implementation of a probe-sense-respond complexity sequence when funding comes from traditional donors who insist on working only with linear cause and effect tools. The compromise of “adaptive programming” is pragmatic, but relies heavily on the insight and determination of individual project leadership.
Russell Gasser resultsbased.org September 2016.