Insight, Industrial Policy
Rwanda is designing industrial parks around shared energy, water and by-products. Namibia is aligning concessional capital across an industrial ecosystem rather than one plant at a time. Both are right, and the model they are borrowing was never designed. It accreted over twenty years.
Rwanda is using a World Bank-supported programme to explore how Korea's eco-industrial park model could be built into its expanding industrial zones (as reported), connecting neighbouring firms through shared systems for energy, waste heat, wastewater and by-products, so that one company's waste stream becomes another's input.
Namibia is developing a Sectoral Transformation Investment Plan that could unlock up to 250 million dollars of Climate Investment Funds capital (as reported), with priorities including shared industrial infrastructure, mineral beneficiation, green hydrogen and ammonia, low-carbon manufacturing and participation by smaller firms.
These are the same move at two different levels. Both take the industrial ecosystem rather than the individual plant as the unit of intervention. Both are a considerable advance on installing solar panels factory by factory, and both are the right instinct.
The difficulty is in the model they are borrowing, and it is worth being precise about it, because the failure mode is well documented and entirely avoidable.
The term industrial symbiosis was coined to describe the network of by-product exchanges in Kalundborg, Denmark. Marian Chertow's literature and taxonomy of the field established it as a research area, and her later paper on uncovering industrial symbiosis contains the finding that matters here.
Kalundborg was not planned. The exchanges arose through private initiative rather than government planning, and developed over a period of about twenty years. Firms discovered, one bilateral deal at a time, that a neighbour's waste had value, usually because a manager knew another manager.
Chertow's 2007 assessment of what happened next is the sentence every industrial park programme should read. Since 1989, efforts to replicate deliberately what was largely self-organising in Kalundborg have followed many paths, some with much success and some with very little. The deliberate replication of an accreted system has a mixed record, and the mixed part is not usually a technical failure.
The follow-on work with John Ehrenfeld on organising self-organising systems frames the problem honestly in its title. You are trying to design something whose defining property was that nobody designed it.
None of which is an argument against Rwanda's approach. It is an argument about what has to be supplied deliberately to substitute for twenty years and a small town's worth of personal acquaintance.
The engineering is the easy part. A pipe carrying waste heat between two plants is not difficult. What the pipe creates is a set of obligations that a factory manager has good reason to resist.
A dependency on somebody else's production schedule. If your process input is your neighbour's by-product, then your output depends on their shift pattern, their maintenance shutdown, their order book and their commercial decisions. You have converted a purchased input with many suppliers into a single-source input with one, and the supplier is not in your business.
A dependency on their survival. Firms close. When the supplier of your steam or your feedstock exits, you are holding an asset specified for an input that no longer exists. This is asset specificity in the transaction cost sense, and it is exactly the condition under which firms prefer ownership or long contracts to spot exchange.
Exposure to their quality control. A by-product stream is not a specified product. Its composition varies with what the producing plant was doing that week, and the receiving plant absorbs that variance.
Shared liability and shared reputation. Effluent, emissions and safety incidents in a connected system are harder to attribute, and the connection itself becomes a channel for regulatory and reputational risk.
Read that list and the twenty years at Kalundborg stop looking like inefficiency. They look like the time it takes to build enough trust and enough operating history for a manager to accept those exposures.
Which identifies what a designed park has to supply, and it is not infrastructure. It is the contractual and institutional substitute for trust: supply agreements with defined quality, availability commitments, force majeure terms, penalty and termination clauses, arbitration that does not require going to court, and a fallback so that a plant whose symbiotic input fails can buy on the market at a known cost.
A park that builds the pipes and leaves the contracts to the tenants has built the cheap half.
Before any of that, a firm has to know what its neighbours have.
This is genuinely non-trivial and it is where several eco-industrial park programmes stall. Waste stream data is commercially sensitive, because it reveals throughput, yield, process choices and sometimes compliance history. It is also frequently not measured, because nobody has needed to know the composition of something they pay to dispose of.
So the first instrument a zone needs is a by-product inventory that firms will actually complete, which requires a trusted intermediary holding the data and a rule about what it may disclose. That is an institution, it costs very little, and it is the precondition for every exchange that follows.
Rwanda's programme has an advantage worth naming here. Designing this into zones that are still expanding is far cheaper than retrofitting it, and the zone authority is a natural candidate for the intermediary role because it already holds a relationship with every tenant. The window is the design phase, and it closes when the tenants are in.
Namibia's plan operates on the constraint rather than the project, which is the more sophisticated instrument and the harder one to evaluate.
Financing a hydrogen plant is a transaction with a counterparty, a term sheet and a completion date. Financing the shared constraints across an industrial ecosystem, power, water, port capacity, skills, local suppliers and community agreements, is a programme whose success is only visible in what other people subsequently build.
Two consequences follow.
Attribution becomes genuinely difficult, and that is not a reason to avoid it. If shared infrastructure works, the evidence is investments that would not otherwise have happened. That requires a counterfactual, and the honest version is comparing against similar sites without the shared investment rather than asserting the difference. It is the same problem as any programme whose product is an enabling condition, and it is answerable with the right design chosen at the start.
Sequencing determines whether it works at all. Shared infrastructure sized for an industrial ecosystem that does not yet exist is a stranded asset. Sized for the anchor tenant alone, it forecloses the ecosystem. The judgement between those two is the whole design decision, and it is made once, early, usually with insufficient information.
The inclusion of smaller firms in the priorities is the part most likely to be quietly dropped. Shared infrastructure is easiest to justify on the anchor's demand, and access terms for smaller users tend to be settled after the anchor's contract is signed, at which point the anchor has no reason to accept anything that raises its own cost. This is the third party access question arriving in an industrial zone, and the answer has the same shape: published terms, decided before commitment, or it will not happen.
Eco-industrial park programmes report connections built, firms located and megawatts shared. Those measure construction.
Exchanges still operating at year five, and why the failed ones failed. Symbiotic links break, and the reason is almost never the pipe. A register of terminated exchanges with causes would be the most useful document in this field and no programme publishes one.
The share of exchanges between firms with no prior relationship. This is the direct test of whether the zone is manufacturing the connections or merely hosting ones that would have formed anyway.
Whether smaller tenants access shared infrastructure on the same terms as the anchor. Published tariffs, or their absence, answers this immediately.
What happens to a plant when its symbiotic supplier stops. Whether a fallback exists, whether it was contracted, and what it costs. A zone in which no tenant can answer this has distributed a risk nobody has priced.
The Lab works on this in local manufacturing and climate and ecosystems, and on what a shared industrial system means for the firms and communities around it.
If you are designing a zone whose value depends on exchanges between firms that have never dealt with each other, tell us what you need to know.
This is an independent insight piece by Transitions Lab. For the Lab's applied work, see Local Manufacturing & Supply Chains. See also The Ban Is Not the Policy on third party access as the difference between industrial policy and a transfer, Own the Battery, Rent the Shopfront on asset specificity and where the boundary of a firm falls, Capability Is the Slow Part on why the institutional layer takes a decade, and The Survey Is the First Act of the Mine on third-party access rules that decide whether an information asset builds an ecosystem or concentrates one. To discuss a study, see Contact.