Equivalent for What? (Post 2)
Comparing causal functions across difference — without pretending different interventions are the same
In the first post in this series, I used Samin Nosrat’s Salt Fat Acid Heat to explore a philosophy of learning: understand underlying principles, use recipes as scaffolding, and develop the judgment to respond to the ingredients and conditions in front of you.
That brought me back to an idea I had promised to explore years ago: functional equivalence.
The phrase can sound more technical than it needs to. But it addresses a very practical problem. How can we compare interventions, organizations, or other units that look different without pretending they are the same? And how can we avoid collapsing activities or organizations that look similar but are actually doing very different work?
The problem is not apples and oranges
People often say that a portfolio is “comparing apples and oranges” when its interventions or organizations are too diverse to aggregate.
But apples and oranges can be compared.
This is also a problem I first encountered through comparative politics. Locke and Thelen’s classic argument for contextualized comparison challenged the idea that comparison requires cases to look institutionally alike. What matters is identifying what is substantively comparable in context — not assuming that similar forms mean the same thing, or that different forms cannot illuminate the same problem.
We can compare apples and oranges by price, nutritional content, sweetness, acidity, shelf life, or suitability for a particular recipe. The problem begins when we fail to specify the dimension on which the comparison is meaningful.
The same applies to programs or other units of analysis.
A municipal committee and a trusted community intermediary are not the same institution. A digital platform and an embedded government adviser are not the same intervention.
Yet different arrangements may sometimes address a common causal blockage. For example, a system may contain valuable local knowledge that does not influence decisions because it is fragmented, expressed in politically unusable terms, held by people without access to decision makers, or unable to travel across organizational boundaries.
A formal committee, trusted intermediary, or embedded adviser might each help translate and route that knowledge so that someone with authority can act on it. That does not make the three arrangements equivalent in every respect. It gives us a specified function in relation to which comparison may be useful.
Nosrat’s framework helps because it trains us to ask what an ingredient is doing in the dish, not only what it is called. Functional equivalence asks a similar question of interventions — but with an important extra discipline: we have to specify exactly which causal function we are comparing, and how far that comparison can reasonably travel.
Functional equivalence is bounded
For this series, I use the following working definition:
Different arrangements are functionally equivalent when they perform a sufficiently comparable causal job within a specified and bounded pathway of change.
Different arrangements reminds us that similarity of form is not required.
Sufficiently comparable warns against claiming complete equivalence.
Causal job moves us from labels toward what the arrangement enables, changes, or constrains.
Specified and bounded requires us to say exactly which function, outcome, and part of the pathway we are comparing.
Comparative research has long wrestled with how concepts travel across settings without stretching beyond usefulness. Functional analysis also has a much older concern with the possibility that different arrangements can fulfil similar functions, while also emphasizing that the range of possible substitutes is constrained rather than unlimited. A recent study of functional equivalents in comparative education shows why resemblance is a poor shortcut: different operations and actors can perform comparable functions, while apparently similar operations may turn out not to be functionally equivalent at all.
Functional equivalence is not something we identify from a distance by noticing resemblance. It has to be investigated.
And it does not eliminate the need for contextual knowledge. If anything, it makes that knowledge more important, because we need to understand what a particular arrangement is actually doing before deciding whether the comparison holds. That is precisely the tension I often hold in research and evaluation — whether looking across countries and sectors or looking at sub-national variation within a sector in a single country.
Form, function, mechanism and conditions
To make the comparison defensible, I find it useful to distinguish four things.
Form
The visible arrangement or activity — what most colleagues are used to talk and report about: a committee; workshop; network; database; adviser; public hearing; grant; or digital platform.
Function
The causal job the arrangement needs to accomplish: connect; translate; authorize; legitimate; protect; remember; coordinate; create feedback and so much more!
The same form can perform different functions, and the same function can be performed through different forms.
Mechanism
The process through which the arrangement changes how people reason, relate, or act. A broker may reduce the perceived risk of cooperation. A public commitment may make inaction more costly. Recognition from a respected institution may give an actor confidence or legitimacy. Repeated interaction may build enough trust for people to share information they previously withheld.
Conditions
The circumstances that enable, modify, or block that mechanism: political authority; trust; timing; incentives; institutional rules; resources; social position; or the credibility of the people involved.
Different forms, related functions
Consider three arrangements:
- a statutory municipal committee;
- a trusted intermediary from a community organization;
- an adviser embedded in a ministry.
They differ in visibility, authority, accountability, and durability. But all three might help local knowledge become institutionally actionable. The mechanisms would not be identical. The committee may work through formal authorization and documented procedure. The intermediary may work through trust, relational legitimacy, and translation between social worlds. The adviser may work through proximity, repetition, and detailed understanding of administrative processes.
The comparison is therefore not:
Are committees, brokers, and advisers basically the same?
It is:
Under what conditions did each arrangement enable otherwise fragmented knowledge to influence a decision, through what mechanism, for whom, and with what consequences?
That can be a much more useful causal question.
I am encountering this problem in current research on young women’s education-to-work trajectories in Southeast Asia. Activities that initially sit in very different program tools — a database, an English-language activity, individual accompaniment, connections to employers — start to look more analytically comparable when we ask what work they perform at a transition point between high school and university or between university and employment. Do they make an otherwise distant next step more visible, reachable or navigable? The answer is not automatically yes. But asking the functional question changes what we know to look for in the evidence.
The same form can do different work
The inverse problem is equally important.
Get Florencia Guerzovich’s stories in your inbox
Join Medium for free to get updates from this writer.
Consider “training.” In one setting, training may transfer technical knowledge — English language skills for women. Or perhaps, it mainly creates relationships that allow dispersed students to contact one another later. The activity may provide a shared language or an opportunity to develop confidence or leadership skills. If its consequential function was network formation or development of leadership skills, counting workshops, participants, and test scores will tell only part of the story.
The same applies to other familiar activities. A database may store information. It may also create institutional memory, make previously invisible people legible to an organization, or allow opportunities to be matched with people who can use them. A convening may communicate knowledge. It may also negotiate disagreement, preserve relationships, or hold a coalition together until a political window opens. An intermediary may transmit information or funding. It may translate among political, technical, and administrative languages, protect relationships, or help route a problem toward someone able to act. It can also be a gatekeeper. A monitoring system may demonstrate compliance. It may also detect failing assumptions early enough for a team to adapt. The activity label does not tell us which function mattered.
Where causal research and evaluation meet
This is where I deliberately combine research and evaluation.
From qualitative causal research, I draw the discipline of reconstructing explanations:
- what happened first;
- what changed next;
- who acted differently;
- what evidence supports the proposed mechanism;
- and what else could explain the outcome.
Process tracing is one way of imposing that discipline. It directs attention toward within-case evidence that can strengthen or weaken a proposed causal explanation rather than simply observing that an intervention and an outcome occurred in the same place. Realist evaluation brings a closely related concern: interventions do not produce outcomes uniformly. We need to understand how causal processes interact with context, and for whom particular outcomes are generated.
The theoretical repertoires we carry can — and often should — inform our critical judgment, not unlike recipes in cooking. As in Post 1, I like to think of them as “training wheels”: useful for the inspiration, context and general guidance they offer, rather than something to follow to the letter. So, is our ability, muscle memory and experience for mixing and matching theoretical priors — including those that have become second nature.
The point is not to leave our prior theories and assumptions at the door. It is to put them into dialogue with what emerges — or, to keep the cooking metaphor going, with what is actually in the fridge. Used well, theories like recipes need not be infantilizing. They give us an initial menu of propositions to probe, challenge and revise rather than answers to reproduce. They preempt us from reinventing the wheel.
Iterative inductive–deductive approaches are one technical way of bringing this reasoning systematically out from under the rug, increasing transparency, and inviting reflexivity. So is the purposive examination of alternative explanations: asking not only whether the evidence fits the account we expected, but what else could plausibly explain what we are seeing.
The Nepali retrospective review that triggered this post series is pushing both me and stakeholders to think hard about the theoretical repertoires we bring with us, the assumptions embedded in them, and what those assumptions make easier — or harder — to see and how they affect how we value evidence. Work in progress, for another post, or perhaps another series.
Evaluation, especially utilization-focused one, adds another layer: The excercise must be fitted to a decision. The aim is not to produce the most elaborate possible causal account. It is to produce one that is credible enough, transparent about its limitations, and useful for the people deciding what to continue, change, fund, stop, or investigate further.
Functional equivalence and mid-level theory
I have previously argued for mid-level theory: explanations that are neither theories of everything nor descriptions so specific that they cannot inform action elsewhere.
Or think in Salt, Fat, Acid, Heat’s terms and hold several insights simultaneously: fat matters the world over, different types of fat make more sense in one cuisine or another — and there are clusters of cuisines!
Functional equivalence helps operationalize that middle ground.
It allows us to say:
- these cases are not identical;
- the intervention labels may differ;
- the mechanisms and conditions may vary;
- but they are confronting a comparable causal problem and attempting to perform a comparable function. Just like fat.
That creates a bounded space for learning — in research, evaluation, and cooking.
It also clarifies the relationship with the typologies, theories of change, and rubrics I have written about elsewhere. Typologies help group cases. Theories of change explain why the grouping may matter. Rubrics help assess quality or progress within a pathway. Functional equivalence addresses a prior question:
What common causal job makes it analytically meaningful to place these otherwise different cases in the same conversation?
How far does the equivalence travel?
This is where I think the discipline becomes especially important. Functional equivalence is not a yes-or-no property of two arrangements. It is a claim about a specified part of a causal pathway. The comparison may hold at one point and stop at the next. A committee and an intermediary might both succeed in routing information. That does not mean they both secure authorization. Securing authorization does not mean creating the capacity to act. Changing one decision does not mean altering the institutional rules that produced the blockage. And institutional uptake does not necessarily mean institutionalization.
Each additional claim crosses another evidentiary threshold. So the discipline is not merely to ask whether two things are functionally equivalent. It is to ask:
How far does the equivalence travel before the evidence — or the causal pathway — diverges?
This also changes how I think about “limits.” Functional comparison can become dangerous when researchers are too eager to find commonality. A formal committee and an informal intermediary may both route information. But they may differ radically in: whose knowledge is included; whether decisions are documented; whether participation creates risks; whether authority is accountable; whether the arrangement survives staff turnover; and/or whether the function depends on one unusually skilled person. These are not peripheral caveats. They may determine whether the arrangements produce comparable value at all.
So a functional-equivalence claim should always include a search for where it stops:
- Where did the expected function not occur?
- Who remained excluded?
- Which mechanism failed to activate?
- What unintended effects followed?
- What depended on unusually favorable conditions?
- Which apparent similarities disappeared under closer examination?
- And at what point did one pathway continue while another did not?
The boundary of the comparison is part of the finding.
Seeing the orchard without losing the fruit
In Invisible Weaving, I argued that mechanisms give us an altitude at which we can see patterns across diverse programs without forcing coherence to mean sameness.
Functional equivalence sharpens that argument. It tells us what exactly we are comparing at that altitude. Not the whole intervention. Not the organizational label. Not a generic claim that “relationships matter.” We compare a bounded causal function, investigate the mechanisms through which different arrangements perform it, and remain explicit about the conditions and consequences that differ.
That is how we avoid both bad aggregation and analytical paralysis.
In cooking, recognizing that miso and sea salt can both contribute salinity does not make them interchangeable. In evaluation, recognizing a common causal function should create the same kind of disciplined curiosity: similar for what, under what conditions, and up to what point?
Specifying a common function gives us only part of the answer. It does not mean that the same mechanism will activate — or activate in the same way — wherever that function is needed. Context still matters. So does timing. And sometimes, so does something less frequently captured in our frameworks: the collective mood through which people interpret what is possible, risky, or worth attempting. That is the next challenge.
