This piece concludes our analysis into Ethiopia’s failure to addressing clean cooking. We began by quantifying the economic and ecological cost of traditional fuels—forests lost, hours consumed, lungs blackened. But data alone couldn’t explain why interventions often failed. So, we dissected the bureaucracy: “Ethiopian Airlines Proves Bureaucracy Works” revealed how institutions can function; “Planners vs. Searchers” exposed why they usually don’t; and “How Bureaucracy Escapes the Planner’s Trap” isolated the rare anatomy of success—prioritization, autonomy, focus, and feedback—in outliers like Telebirr, GERD, eGP, INSA, and IFTMS.

We ended that last piece with a provocation: Can the same escape anatomy outlined in the last piece be engineered for clean cooking—a domain so thoroughly institutionally orphaned that politicians ignore it, ministries disown it, and planners ghost it?

This question is not trivial. Annually, 64,000 Ethiopians suffocate in their kitchens—not for lack of technology, but because the state lacks the institutional reflexes to solve cross-cutting crises. The losses are both intimate and tectonic: USD 72 billion each year in lost productivity, health burdens, forest depletion, and time poverty. At the same time, Ethiopia ranks third globally—after India and China—in potential climate mitigation through a clean cooking transition. Yet clean cooking remains politically weightless: visible in plans, invisible in budgets, absent in implementation.

This piece is the culmination of that inquiry. It confronts the decisive test: Can Ethiopia build adaptive institutions for problems that offer no ribbon-cuttings, no prestige, and no symbolic capital?

Clean cooking poses that test—not because it is about stoves, but because it forces governance to operate without visibility or applause. In this domain, declarations and donor funding are not enough. Reform must be structurally engineered to perform under stress. If Ethiopia is to solve problems that cross sectors, resist ownership, and fail to mobilize political theater, it must learn to build institutions that adapt by design, not by accident. The anatomy of escape we observed in Telebirr and IFTMS—prioritization, autonomy, narrow focus, hybrid finance, and embedded feedback—must now be built deliberately in a domain where no natural support structure exists. This is not a test of clean cooking. It is a test of whether the state can govern what it cannot see.

To claim that clean cooking needs adaptive institutions is to assert that adaptive institutions are the right tool. But that raises a question: adaptive to what? Not all policy failures stem from rigidity. Not every public problem demands institutional innovation. If adaptive institutions are the answer, we must first be precise about the kind of failure they are meant to resolve.

Adaptive institutions are not designed for tasks that benefit from central command, high-frequency execution, or elite attention. Prestige projects like GERD succeed through symbolic politics and centralized power. Routine services like vaccinations or salary disbursement require scale and procedural discipline. Crisis response demands speed and hierarchy. In each case, planning logic—clear targets, time-bound deliverables, and fixed chains of authority—is not a hindrance. It is a prerequisite.

But some domains defeat this logic. These are not problems that resist planning—they are problems that planning actively misrepresents. Solutions exist but do not scale. Outcomes depend not just on delivery, but on human behavior. Success cannot be captured by a single metric. No institution claims full ownership. And the issue remains structurally invisible—neither budgeted, nor coordinated, nor defended.

Clean cooking sits precisely in this space. It is not just difficult—it is unclaimed. It is what we have elsewhere called an institutionally orphaned domain: no lead actor, no clear accountability, and no structure to convert feedback into adaptation. The problem is not technical complexity—it is bureaucratic fragmentation and political neglect.

Some problems—like tax collection or pandemic response—are supported by metrics, models, and institutional memory. Others sit in a space where both data and understanding are weak. Clean cooking falls squarely in this space: we lack reliable information on usage, adoption patterns, and behavior under sustained intervention. We cannot predict failure, yet urgent decisions must be made. In this context, traditional tools fail. Planning assumes control where none exists; monitoring systems confuse distribution with impact. When data and understanding are both weak, the only viable strategy is institutional: to build structures that can learn, adapt, and act under conditions of uncertainty.

This is the question to which adaptive institutions are the answer. Not how to plan better, but how to act when the problem defies planning itself. Not how to command change, but how to build a structure capable of recognizing what is working before it collapses. These questions are central to clean cooking.

Clean cooking does not simply fit the profile of an orphaned domain—it defines it. It meets every criterion outlined in the previous section. The solution exists but does not scale. The outcome depends on behavior. The metrics are ambiguous. No single actor owns the result. And most critically, it remains institutionally invisible: present in strategy documents, but absent from operational structures, sustained budgets, or accountable delivery channels. Worse still, the domain is inherently dynamic. Technologies change, subsidies distort behavior, and adoption patterns shift with food prices, cultural norms, and seasonal labor. In such an environment, surprise is not a risk—it is the baseline condition. Static plans age faster than implementation cycles. Without structures that can respond to movement, every intervention decays into irrelevance on arrival.

This orphanhood is not incidental—it is engineered. Ethiopia’s Ministry of Health publishes health burden estimates but does not budget for interventions. The Ministry of Environment sets adoption targets for improved stoves but leaves implementation unfunded. The Ministry of Women and Social Affairs acknowledges fuel burdens in its gender strategy, yet allocates resources only to training workshops. No ministry denies the problem, but none is structured to act. Responsibility is distributed so thinly that it disappears.

This structural neglect has created a domain where failure is not just common—it is routine. Stove programs are launched with donor fanfare, then quietly abandoned. Monitoring focuses on distribution counts, not usage or durability. Behavioral adoption is reduced to information campaigns, while real-world evidence shows abandonment rates as high as 83  percent in regions like Oromia. The system rewards visibility—manuals printed, workshops held, stoves handed out—over results. Clean cooking becomes not a site of intervention, but of performance.

And yet, that is precisely why it is the necessary test. GERD succeeded through elite consensus and symbolic urgency. Telebirr benefited from political insulation and Prime Ministerial backing. Even IFTMS had the structure of World Bank funding to enforce focus and iteration. These projects were not easy—but they had protection. Clean cooking has none.

It is not shielded. It is not prioritized. It is not even clearly defined. It exists at the edge of the State’s attention, too technical for politics and too political for technocrats. That is what makes it the hardest case—and also the most honest one. If adaptive institutions can be deliberately designed to work here, they can be designed to work anywhere.

This is not a domain waiting for better stoves. It is a domain waiting for institutional architecture—testing whether governance can survive in conditions without visibility, ownership, or default legitimacy. If escape from planner logic can be built here, then the Ethiopian state will have done more than reform clean cooking. It will have demonstrated that adaptation is not the exception—it can be designed.

The failure of clean cooking in Ethiopia is not a puzzle of bad luck or flawed technology. It is a pattern—regular, visible, and structurally reproduced. Programs are launched, funds disbursed, trainings conducted, stoves distributed. But sustained usage rarely follows. Designs are not adapted. Feedback is not collected. Failures are recorded but never trigger institutional correction. Instead, each wave of failure is absorbed into the planner’s logic: blame is deflected to users, new models are proposed, and the cycle begins again.

Consider the Tikikil stove program, distributed across Southeastern Ethiopia under a Results-Based Financing (RBF) model. Under this structure, the stove’s retail price was subsidized—users paid 75 percent, and the remaining 25 percent was reimbursed to suppliers by the donor upon verified sale. But the model equated purchase with success. It offered no incentive to ensure durability, usability, or sustained use. Surveys revealed high dissatisfaction: the wood inlet was too small for typical fuel sizes, pots didn’t fit well, and many units cracked under regular use. Despite “adoption” on paper, 65 percent of households continued using traditional stoves.

The National Improved Cookstove Program scaled faster still—distributing over 9 million stoves by 2020. Yet one year after rollout in Merawi (Kolela district), 40 percent had been abandoned. Fragile construction, especially failing concrete joints, and a lack of maintenance support made repair nearly impossible. Functional adoption—actual usage beyond initial uptake—was below 50 percent. The poorest households, unable to afford replacements or understand stove care protocols, dropped off fastest. No system existed to adjust rollout based on observed failure.

Even Ethiopia’s most iconic stove, the Mirt, promoted for its fuelwood savings and distributed at scale (360,000+ units), suffered from contextual mismatches. Usage in SNNP lagged 33 percent behind Amhara and Oromia due to incompatible cooking styles and communal kitchen layouts. In some cases, recipients used the stove less if they had paid subsidized prices than if it was given free—an indicator that subsidy design distorted perceived value. After one year, no measurable emissions reductions were observed due to fuel stacking and inconsistent use.

The pattern repeats across other initiatives. The Lakech and Gonzie stoves, celebrated in lab trials, were later found in field evaluations to have thermal efficiencies below 20 percent and cooking times exceeding five hours. At Wereta University, electric mitads were installed without accounting for grid instability; when outages hit, cooks reverted to biomass. In Gambela’s refugee camps, imported stoves were incompatible with communal cooking norms and were abandoned by 80 percent of households before user-centered redesigns were even attempted.

These systemic failures—technical, contextual, and operational—were not incidental. They were coordinated by design. Distribution was treated as delivery; usage was inferred from presence. Ministries reported stoves distributed, not stoves used. Donors tracked workshops held, not time saved or air quality improved. Monitoring visits yielded compliance photos, not insight. The system rewards the visible: manuals printed, forms filled, targets met. But when data contradicts the plan—when stoves are abandoned or misused—it is either ignored or blamed on user behavior. This is not implementation failure. It is planning as performance: a ritual of activity without consequence, incapable of adaptation because it is not designed to learn.

A key reason for this institutional denial is the absence of feedback capacity. Most clean cooking interventions are externally funded but internally hollow. Monitoring and evaluation systems are built to validate donor-defined targets—not to detect failure or respond to reality. Feedback loops are outsourced to NGOs or short-term consultants and by the time a program learns what went wrong, the funding window has closed, the team has dissolved, and the next round of planning has begun. In the absence of embedded learning mechanisms, failure is not something to fix—it is something to stage around.

This pattern of delayed recognition leads to a deeper pathology: ghosting governance. The ministry hosts the launch. The donor funds the training. The NGO delivers the stoves. The consultants evaluate the uptake. But no one owns the outcome. And when failure arrives, it is everyone’s problem—and therefore no one’s.

What makes this cycle so persistent is that it mimics success. The project has outputs. The budget is spent. The pilot program is documented. The numbers are reported. But what remains is a trail of unused stoves, fatigued communities, and abandoned intent. The stove, in this context, becomes a symbol not of health or modernity, but of bureaucratic theater—where distribution replaces delivery, and visibility stands in for value

If clean cooking is to escape this trap, the system must be restructured to hold implementers, funders, and policymakers accountable for abandonment—not to obscure it.  Until then, the problem is not that clean cooking interventions fail. The problem is that they are never designed to succeed.

If clean cooking is where institutions fail quietly, school kitchens are where that failure becomes visible. In Addis Ababa, thousands of women cook for public school children each day using traditional biomass fuels under conditions of extreme indoor air pollution. The problem is not the absence of alternatives. Clean technologies exist. What’s missing is the institutional architecture to deliver them reliably, adaptively, and at scale.

This school-based initiative is not a conventional pilot that tests tools under ideal conditions. It is a delivery system designed to function under the pressures that typically cause institutional failure—fragmented mandates, behavioral complexity, and bureaucratic inertia. Each element of the proposed model is drawn from the anatomy of escape and assembled not as theory, but as practice: a working system that reveals whether clean cooking can be delivered, maintained, and improved in real-time, inside public school kitchens.

The pilot must begin with a sharply defined mandate: deliver measurable clean cooking outcomes in a small number of Addis Ababa public schools. The use case is concrete—reducing PM2.5 exposure and fuel consumption in school feeding programs that currently rely on polluting biomass. The outcome is equally clear: sustained stove usage over time, verified through sensors and user reports.

This is not a symbolic project. It is deliberately narrow because narrowness allows precision. The schools are known. The cooks are already employed. The food is already being prepared. What is being introduced is a new delivery system: one that replaces guesswork with feedback, one that installs a stove only if it will be used, maintained, and improved based on what the users actually experience.

By anchoring the pilot in school kitchens, the program avoids the ambiguity that has plagued past efforts. It does not rely on household diffusion, ambiguous adoption, or general awareness campaigns. It begins where cooking already happens, every day, in visible institutional settings. This narrow scope is not a limitation—it is a design strength.

To function under real-world constraints, the school kitchen pilot must be run by a delivery unit with operational autonomy. Without it, even the best-designed intervention will stall—caught in procurement bottlenecks, politicized hiring, or delayed budget approvals. The implementing team must be embedded within a host institution but protected from its slowest moving parts.

This means delegated authority over staffing, contracting, and day-to-day execution. Stove procurement cannot be held up by central tender boards. Maintenance teams must be hired based on performance, not placement rules. Feedback systems must be deployed without waiting for donor committee sign-off.

This is not about institutional independence. It is about responsive capacity: the ability to learn from what’s happening inside school kitchens and act immediately. If a stove malfunctions, if a cook reverts to biomass, if usage drops—someone must have the authority and resources to respond. Without that, the pilot becomes just another distribution scheme wrapped in reform language.

Autonomy without protection is just exposure. For the school kitchen pilot to survive its own usefulness, it must be politically shielded from interference, inertia, and quiet sabotage. That shielding cannot come from donor contracts or ministry memos. It must come from a clear, visible signal that this pilot matters to someone with the authority to protect it.

Political prioritization, in this context, means executive anchoring. The clean cooking pilot should be housed within a delivery unit linked to the Prime Minister’s Office or Mayoral office—not buried in a line ministry with conflicting mandates. This anchor does not need to micromanage. It needs to create space: space to hire quickly, procure directly, and iterate based on feedback without triggering institutional resistance.

This kind of shielding does more than prevent obstruction. It enables responsiveness. When cooks abandon a stove or air quality data shows regression, the team needs cover to adjust without delay or blame. And if the pilot generates visible gains—lower PM2.5 readings, happier cooks, cost savings—those results become political assets. Prioritization earns room for autonomy, and successful feedback loops feed back into political capital.

Reform does not scale because it succeeds. It scales because it is protected long enough to succeed.

With political protection in place and operational autonomy secured, feedback can finally do what it’s supposed to: guide action. In most clean cooking programs, failure is invisible until it is irreversible. That cannot happen here. The school kitchen pilot must be equipped to feel what is happening in real time—and respond.

Each participating school must have usage monitoring in place: sensor data to track whether the stove is used, logs for malfunctions and maintenance, and direct input from cooks. These signals must flow directly into the delivery unit—not as quarterly reports, but as daily operational information. If PM2.5 levels don’t drop, if cooks switch back to firewood, if stoves sit idle—these are not evaluation findings. They are triggers for immediate redesign.

Just as important is who feels the feedback. Monitoring cannot be outsourced to consultants. It must be internal to the team implementing the pilot, with dedicated staff responsible for translating data and user feedback into technical and operational adjustments. This is not about accountability to donors. It is about responsiveness to users.

If the stove breaks, the system must feel it. If the user disengages, the team must know why. Feedback is not a validation exercise. It is the nervous system of a living pilot.

For the school kitchen pilot to function as a learning structure—not just a spending mechanism—its financing must be calibrated to support both risk and realism. That means combining early, flexible capital with real-world constraints. Too much insulation, and the pilot becomes irrelevant. Too little, and it collapses before it can adapt.

Donors should underwrite the initial risk: covering the cost of stove deployment, training, monitoring equipment, and responsive maintenance. But continued funding should not be unconditional. Disbursements should be tied to usage verification, air quality improvements, and documented iterations based on feedback. Carbon revenues, where available, should be escrowed and released only when stoves are used consistently and emission reductions are confirmed.

This structure disciplines the system without disabling it. It protects the pilot from early failure, but embeds incentives for adaptation. Most importantly, it shifts the funding logic from “did you deliver the input?” to “did the system respond and improve?” In past programs, financing has followed effort. In this pilot, it must follow learning.

What holds this system together is not just its components—but how they interact and who drives them. A functional delivery structure is not just defined by what it does—it’s defined by who is inside it, and how its parts reinforce one another. The pilot must be staffed by actors with skin in the game and the authority to act: local manufacturers empowered to iterate, cooks who provide ground truth, funders committed to learning rather than showcasing, and technical staff who can translate data into design changes. Each element of the system reinforces the others. A narrow mandate makes autonomy feasible. Autonomy becomes effective when politically shielded. Shielding creates the space for feedback to be absorbed rather than ignored. And feedback, when acted on, generates results that justify continued prioritization and funding. This is not a conventional project team. It is a coordinated test of whether clean cooking can be delivered in real time, under real constraints, with real consequences.

No design survives contact with reality. That’s not a flaw in the school kitchen pilot—it’s the point. This pilot is not meant to prove success under ideal conditions. It is built to surface failure early—under the same pressures that typically cause clean cooking efforts to collapse. The goal is not perfection, but preemption.

Some failures will be technical. Stoves may break down. Sensors may stop reporting. Heat distribution may be uneven. If the pilot is working as intended, these are not setbacks—they are signals. But signals only matter if the system can act. Maintenance must be locally available. Spare parts must be stocked. Feedback must flow to actors with the authority and tools to respond.

Other failures will be operational. Fuel may arrive late. The price may rise unexpectedly. A key machine may break down, stalling production. These risks are not outliers—they are the norm. That means mitigation must be built in: backup delivery plans, cost buffers, redundancy in production, and rapid-response teams that can resolve issues without triggering paralysis.

Still others will be organizational. Procurement may stall. Coordination may fray. Decision rights may blur. Autonomy—so essential for responsiveness—may begin to erode. History shows this doesn’t happen all at once. It begins subtly: budget approvals rerouted, staff appointments delayed, procurement re-centralized. To guard against this, shielding must be formalized in advance—anchored by a directive from the Prime Minister’s Officeor Mayoral office that activates when rollout begins. This directive should include a sunset clause: the delivery unit is not permanent, but a temporary structure designed to operate under extraordinary conditions. Its legitimacy derives from its limited mandate. In short, organizational failure is not a surprise—it is a known risk that must be preempted by structural insulation, not post-hoc correction.

The point of a stress test is not to admire failure, but to reveal what kind of system can recover from it. The anatomy of escape must be tested not only for its capacity to learn—but for its ability to act under pressure. That is the only kind of pilot worth running. Not one that assumes success, but one that demonstrates whether adaptive governance is possible when success is at risk.

What Ethiopia needs is not another demonstration project, but a demonstration of possibility. Clean cooking has been studied, piloted, and subsidized for decades—yet its institutional architecture remains broken. The lesson is not that the problem is intractable. It is that past solutions were designed for conditions that did not exist.

This pilot begins from the opposite premise: that failure is not a deviation from plan, but the condition under which any viable plan must function. That insight is not limited to clean cooking. The structure proposed here—bounded, shielded, feedback-driven—is not a blueprint for stoves. It is a test of whether the state can learn under pressure.

If clean cooking is where Ethiopia’s governance learns to adapt, it will not just solve a neglected problem. It will have pioneered a method to approach others like it: domains such as forestry, forestry, urban development, or solid waste management—where incentives are misaligned, actors are fragmented, and systems have long operated without accountability. This is not a model to be scaled. It is a capacity to be built.

TsegayeNega is a professor emeritus at Carleton College in the United States and founder and CEO of Anega Energies Manufacturing.

Contributed by Tsegaye Nega