The model, developed by Gavin Melles of Swinburne’s School of Architecture and Design, is unusual in the granularity of its economics. It separates rigid plastics—bottles and containers—from flexible plastics such as films, pouches and laminates, and assigns each stream its true processing cost. Mechanical recycling of rigid plastics costs roughly $800 per tonne, covering collection transport, sorting at materials recovery facilities, washing, grinding and extrusion, and the resulting recyclate sells for $800 to $900 per tonne. That is marginally viable economics, enough to attract private investment when supported by container deposit schemes. Flexible plastics are another universe entirely. Because they are typically multi-layer structures of three to seven incompatible polymer layers bonded with adhesives and printed inks, they cannot be mechanically separated. The only mature pathway is solvent-based dissolution followed by membrane filtration, which costs about $19,500 per tonne—against a recyclate value of just $600 to $800 per tonne. Every tonne processed through this route loses roughly $18,700.

That order-of-magnitude cost gap is not a temporary market inefficiency, the model argues, but a consequence of physics. The functional properties of flexible packaging—barrier performance, food protection, shelf-life extension—derive from the laminate structure itself, not from its constituent polymers. Solvent-based membrane separation can purify individual polymer streams to food-contact quality, achieving 85 to 95 percent recovery, but the separated materials command only commodity pricing because the value was in the structure that has now been destroyed. Even cutting membrane costs in half, to $9,750 per tonne, would still leave losses of nearly $9,000 per tonne. For membrane processing to reach cost parity with elimination, which the model prices at $130 per tonne, costs would need to fall by 92 to 95 percent—faster than any documented environmental technology learning trajectory, including solar photovoltaics and aviation biofuels.

The model’s architecture captures this reality through eight interconnected feedback loops, three reinforcing and five balancing. The reinforcing loops describe how rigid plastic recycling can grow profitably: food-grade recyclate from clean feedstock commands a 30 percent premium, scale economies reduce costs 15 percent for every doubling of cumulative capacity, and container deposit schemes improve feedstock quality so dramatically that effective membrane recovery rises from 51 percent on contaminated kerbside material to 85 percent on clean deposit-collected material. The balancing loops describe the constraints. The most consequential, labeled the diminishing returns loop, operates on the low-hanging-fruit principle: as recovery targets rise, the profitable rigid materials are recovered first, and progressively more of the loss-making flexible materials must be processed, escalating total system costs exponentially.

The numbers behind that escalation are stark. At 40 percent recovery, processing around 160,000 tonnes of flexible plastics annually generates $3 billion in losses, with a marginal cost of roughly $15 million per percentage point. At 60 percent, losses reach $6.5 billion and the marginal cost climbs to $25 million per point. Attempting 70 percent would require processing 562,000 tonnes of flexible plastics, generating $10.5 billion in annual losses at a marginal cost of $78 million per percentage point—a 3.1-fold escalation in the final ten points. The model calculates that even with extended producer responsibility fees of $300 per tonne generating $612 million annually, plus $200 million in public funding, the fiscal pressure of full universal processing would exceed 12 times available revenue. No subsidy, efficiency gain or political will can overcome that arithmetic.

Historical validation gives the model credibility. Calibrated on Australian Plastics Flows and Fates data from 2015 to 2020 and tested on unseen data from 2021 to 2024, it achieved a mean absolute percentage error of 5.8 percent, within the 8 percent threshold. Rigid plastics tracked actual recovery within 3.2 percent error and flexible plastics within 6.8 percent. More strikingly, the model reproduced the November 2022 collapse of REDcycle—the voluntary soft plastics program that operated 1,700 collection points but secretly stockpiled material when processing proved unviable, ultimately landfilling 11,000 tonnes upon bankruptcy—as an emergent phenomenon, predicting the timing within three months and the magnitude within 13 percent, without any exogenous shock representing the event. The collapse emerged from the model’s fiscal pressure loop: at $18,700 per tonne in losses, any voluntary program exceeding roughly 4,000 tonnes per year generated losses exceeding total available public funding of about $80 million annually.

Five scenarios simulated to 2040 quantify the strategic choice. Business as usual, continuing the voluntary approach, achieves only 33 percent recovery at $320 million annually. Moderate policy reaches 48 percent at $480 million. The universal processing scenario—full extended producer responsibility, complete container deposit coverage, maximum $200 million membrane investment and a 70 percent target—catastrophically fails, achieving only 57 percent despite $1.18 billion in annual spending, with fiscal pressure reaching 2.4 and membrane plant utilization collapsing to 35 percent as contaminated flexible feedstock overwhelms facilities. The optimal strategy, which the model calls 80/15/5, eliminates 80 percent of unnecessary flexible formats such as shopping bags and excessive overwrap, redesigns 15 percent to mono-material alternatives, and processes only the 5 percent that are genuinely essential—medical sterile packaging, pharmaceutical blister packs—through membrane technology. It achieves 68 percent recovery at $510 million annually: 2.3 times cheaper than universal processing with 11 percentage points higher recovery.

The elimination economics are the model’s most viral finding. Banning unnecessary flexible packaging costs $130 per tonne, covering industry reformulation, alternative packaging development, supply chain adaptation and enforcement—cheaper even than landfilling, which runs $150 to $250 per tonne in Australian gate fees, and roughly 150 times cheaper than membrane processing. Eliminating 641,000 tonnes per year of unnecessary formats saves $12.4 billion annually in avoided processing losses. Lifecycle assessment cited in the study suggests elimination of formats like bread bags generates 85 percent lower greenhouse gas emissions, 90 percent lower water consumption and 95 percent lower toxic releases than membrane recycling. International precedent supports feasibility: the European Union’s Single-Use Plastics Directive achieved 70 to 90 percent reductions in targeted formats within three years, Australian bag bans cut shopping bag volumes 77 percent in four years, and Rwanda, Kenya and Bangladesh have achieved near-complete elimination of plastic bags across diverse economic contexts.

The model also indicts voluntary governance. Eight years of voluntary commitments produced just four percentage points of improvement despite the $690 million investment. The Australian Packaging Covenant involved more than 500 organizations across 12 working groups producing 47 technical specifications with annual adjustments—process complexity that raised investment risk premiums without creating funding certainty. By contrast, New South Wales’s binding container deposit scheme, a legal requirement with penalties, lifted beverage container recovery from 44 to 72 percent within two years on a $50 million infrastructure investment, achieving 14-fold better cost-effectiveness than national voluntary efforts. The study concludes that binding regulation should lead circular economy policy, with voluntary initiatives as supplements, and recommends extended producer responsibility fees of $300 to $400 per tonne generating $720 million to $960 million annually, allocated with 40 percent to processing subsidies, 25 percent to elimination programs and 15 percent to collection infrastructure.

The broader lesson extends well beyond Australia. The rigid–flexible divide reflects universal physics: multi-layer structures cannot be mechanically separated while preserving value, creating order-of-magnitude cost differences that persist across jurisdictions. Diminishing returns above material-specific thresholds apply to any mixed waste system, with the threshold varying by composition—potentially 65 to 70 percent where rigid plastics dominate, but as low as 50 to 55 percent where flexibles do. Australia’s low population density of 3.3 people per square kilometer raises collection transport costs to $150 to $200 per tonne, making elimination even more attractive in low-density contexts than in Europe. The study’s transferable prescriptions are blunt: do not treat all plastics homogeneously, consider elimination before processing for materials with negative economics, avoid voluntary approaches that achieve two to three times slower progress than binding mandates, and set targets based on material composition rather than arbitrary percentages. For a world that has recycled only 9 percent of the 10 billion tonnes of plastic produced since 1950, the message is that some packaging is worth recycling, some is worth redesigning, and most flexible formats are worth banning outright.

Subject of Research: System dynamics modeling reveals elimination of flexible plastic packaging beats membrane recycling in Australia

Article Title: System dynamics modeling of membrane technology deployment in Australia

Article References: Melles, G. (2026). System dynamics modeling of membrane technology deployment in Australia. Frontiers in Membrane Science and Technology, 5, Article 1765275. https://doi.org/10.3389/frmst.2026.1765275

Image Credits: AI Generated

DOI: 10.3389/frmst.2026.1765275

Keywords: plastic packaging, membrane technology, circular economy, extended producer responsibility, system dynamics, recycling economics, Australia, System, dynamics, modeling, membrane, technology