System dynamics modeling of membrane technology deployment in Australia

National Packaging Targets have mandated 70% plastic packaging recovery by 2025, yet recovery rates have stagnated at 24%–27%, despite voluntary commitments and $690 million in investment (2018–2024). Membrane technology offers the potential to process contaminated flexible plastics, but deployment economics remain uncertain. This study developed a system dynamics model with eight interconnected feedback loops that govern Australia’s plastic packaging circular economy (2015–2040). The model distinguishes rigid (bottles, containers) from flexible (films, pouches) plastics, incorporates empirical cost structures ($800/tonne mechanical vs. $19,500/tonne membrane), and integrates policy levers (EPR, CDS, and elimination mandates). Historical validation against Australian Plastics Flows and Fates data (2015–2024) achieved <8% error. Five scenarios were simulated to 2040. Model validation reproduced voluntary approach failures and the REDcycle collapse. Scenario analysis reveals that attempting 70% recovery through universal processing (Scenario 3) costs $1.18 billion annually while achieving only 57% due to exponential cost escalation above 60%. An 80/15/5 strategy (Scenario 4) that eliminates 80% of unnecessary flexible formats achieves 68% recovery at $510 million annually—2.3 times cheaper with 11 percentage points higher recovery. Marginal costs remain under $25 million per percentage point with elimination versus over $50 million without. The flexible–rigid plastic divide is fundamental, not transitional. Rigid plastics exhibit profitable recycling activating growth loops, while flexible plastics face catastrophic losses, triggering constraint loops. Elimination costs $130 per tonne and saves $12.4 billion annually in avoided processing—a 150:1 cost–benefit ratio. EPR frameworks that enable elimination are essential; voluntary approaches have systematically failed.

Plastic packaging represents approximately 36% of global plastics production and accounts for 47% of plastic waste generation worldwide (Geyer et al., 2017; Plastics Europe, 2022). While offering exceptional functionality—including product protection that reduces food waste by up to 50%, shelf-life extension that enables global supply chains, weight reduction that lowers transportation emissions by 10%–30%, and supply chain efficiency (Verghese et al., 2015; Garcia and Robertson, 2017; Accorsi et al., 2014)—its predominantly single-use nature creates significant end-of-life challenges. Since 1950, approximately 10 billion tonnes of plastic have been produced, with only 9% recycled, 12% incinerated, and 79% accumulating in landfills or the environment (Geyer et al., 2017; OECD, 2022; Lebreton and Andrady, 2019).

Environmental externalities span multiple impact categories (OECD, 2018; Rochman et al., 2013). Marine plastic pollution annually enters the ocean at 8–12 million tonnes (Jambeck et al., 2015; Lebreton and Andrady, 2019), creating accumulation zones like the Great Pacific Garbage Patch containing 1.8 trillion pieces across 1.6 million square kilometers (Lebreton et al., 2018; Eriksen et al., 2014). Microplastic contamination—particles under 5 mm—appears in marine organisms, drinking water, human blood (averaging 1.6 μg/mL), Arctic ice, and deep-sea sediments (Cole et al., 2011; Kosuth et al., 2018; Leslie et al., 2022; Bergmann et al., 2017; Fischer et al., 2015). Climate impacts prove equally concerning: plastic production and incineration generated 1.8 Gt CO 2 eq in 2015 and is projected to reach 6.5 Gt by 2050, consuming 10%–13% of the 1.5 °C carbon budget (Zheng and Suh, 2019; Cabernard et al., 2022; Hamilton and Feit, 2019; IPCC, 2021).

The circular economy paradigm emphasizes closed-loop flows, extended lifespans, recyclability design, and waste minimization (Ellen MacArthur Foundation, 2015; Geissdoerfer et al., 2017; Kirchherr et al., 2017). The Ellen MacArthur Foundation’s New Plastics Economy initiative (2016) catalyzed commitments from 1,000+ organizations, representing 20% of global packaging production (Ellen MacArthur Foundation, 2016). The EU Circular Economy Action Plan established binding targets: 50% recycling by 2025, 55% by 2030, and 65% by 2035, plus recycled content mandates (25% for PET bottles in 2025, 30% all bottles by 2030) and single-use plastic restrictions (European Commission, 2018; European Commission, 2020; EU Directive, 2019/904, 2019).

However, the achievement of high recycling rates faces substantial barriers (Khanzada et al., 2024; Cheng et al., 2023). Packaging complexity with multi-material combinations renders materials unrecyclable (Closed Loop Partners, 2020; APK, 2020). Consumer contamination undermines viability (Material Economics, 2018; Forrester, 1961). Infrastructure limitations constrain capacity (Sterman, 2000; Meadows, 2008). Economic barriers prove particularly intractable: recycled materials often cost more than virgin alternatives (Geissdoerfer et al., 2017; Nemet, 2006; Rubin et al., 2015). China’s 2017–2018 import restrictions (“National Sword”) exposed system fragility, causing recyclable accumulation across OECD countries (Brooks et al., 2018; Staub, 2018).

Australia’s National Packaging Targets (2018) mandate 100% reusable/recyclable/compostable packaging by 2025, 70% recovery, 50% recycled content, and single-use plastic phase-outs (APCO, 2018). These align with broader waste reduction and marine pollution mitigation strategies (Australian Government, 2018). However, recovery stagnated at 24%–27% throughout 2016–2024, despite $690 million investment, APCO coverage of 80% of packaging, CDS expansion from 15% to 92% population coverage, and 77% shopping bag reduction (DCEEW, 2021; DCEEW, 2023; DCEEW, 2024; APCO, 2022).

Paradoxically, domestic reprocessing capacity grew 65% from 365,000 to 483,600 tonnes/year (2020–2022), while utilization declined from 67% to 50%, creating 265,000 tonnes of idle capacity (APCO, 2024; European Commission, 2019). This “capacity utilization paradox” indicates fundamental economic constraints beyond infrastructure availability.

Emerging evidence reveals critical material bifurcation (ICIS, 2025). Rigid plastics (bottles and containers) show improving recovery (24% in 2018 to 28% in 2023), with CDS achieving >70% for beverage containers (EFSA, 2018). Flexible plastics (films, pouches, laminates) remain at crisis levels (4.5%–6%) (APCO, 2023; NSW, 2023). This divergence reflects fundamental realities: rigid plastics are mono-material and are easily cleaned and reprocessed into high-value applications at $800–$1,200/tonne cost with $800–$1,000/tonne value (marginally positive economics) (Blue Environment, 2021; Closed Loop Foundation, 2023). Flexible plastics with multi-layer structures (3–7 layers) cannot be mechanically separated, face heavy contamination, and generate $14,200–$19,200/tonne losses ($15,000–$20,000/t cost versus $800/t value) (Horodytska et al., 2018; Rahimi and García, 2017; Kaiser et al., 2018).

The REDcycle collapse (November 2022) dramatically illustrated these economics (ABC News, 2022; Sustainability Victoria, 2023). Operating 1,700 collection points that achieved 8,000–12,000 tonnes annually, REDcycle secretly stockpiled materials when processing proved unviable (requiring $15,000–$20,000/t payment); they ultimately landfilled 11,000 tonnes upon bankruptcy. Post-collapse analysis identified systemic failures: no viable business model existed without massive subsidies, voluntary commitments provided insufficient funding, and elimination frameworks may have proven more effective than processing mandates (Victorian Auditor-General, 2023; Nash and Bosso, 2013; Wagner, 2017).

Membrane technology promises potential through selective solvent-based separation, achieving 85%–95% recovery with food-contact approval potential (Closed Loop Partners, 2020; APK, 2020). However, capital costs ($5,000–$15,000 per tonne capacity) and operational costs ($15,000–$30,000/t processed) remain substantial, with most processes at pilot scale not demonstrating commercial viability (Material Economics, 2018; Forrester, 1961; Sterman, 2000). This research addresses three questions: (RQ1) under what conditions can membrane technology enable Australia’s 70% target? (RQ2) How do feedback loops govern transitions? (RQ3) What is the optimal membrane deployment strategy?

System dynamics modeling approach

System dynamics represent complex systems such as stocks (accumulations) and flows (rates) governed by feedback loops and time delays (Meadows, 2008; Nemet, 2006). Pioneered by Forrester (1961) for industrial management, this methodology suits circular economy systems characterized by feedback dominance, time delays, nonlinear relationships, and policy resistance (Rubin et al., 2015; Brooks et al., 2018). We developed the model using Vensim PLE Plus, operating on six-weekly time steps (Δt = 0.125 years) from 2015 to 2040. The quarterly timestep originally used (Δt = 0.25 years) produced Euler integration instability in the Program Viability feedback loop; halving the time step eliminates this numerical artefact while maintaining computational efficiency and is consistent with best practice for models with fast feedback loops (Sterman, 2000). The model employs EPR Implementation Level at zero throughout 2015–2024, reflecting the voluntary APCO framework under which no mandatory producer fees existed, with EPR activating as a scenario lever from 2025 onward (Homer, 1996).

A quarterly time step (Δt = 0.25 years) is appropriate for capturing policy-relevant dynamics while maintaining numerical stability. Sudden shocks, such as the China National Sword policy 42 and REDcycle collapse 61, are represented as step changes in exogenous policy variables rather than instantaneous discontinuities.

The National Sword impact was modeled as a lookup function to reduce plastic waste export capacity from 85% to 15% over two-quarters (6 months), reflecting the actual policy implementation and industry adjustment timeline documented in APFF 2018–2019 reports. Similarly, the REDcycle collapse was represented as collection capacity reduction from 12,000 tonnes/year to zero over three-quarters (9 months), matching the observed stockpiling-to-closure sequence (November 2022 shutdown followed by gradual stockpile clearance through Q2 2023).

This approach captures shock dynamics while avoiding numerical instability from instantaneous discontinuities. The quarterly aggregation smooths within-quarter volatility (primarily non-policy-relevant seasonal fluctuations) while preserving year-on-year structural changes that drive the long-term outcomes “Handling System Shocks” and “Discontinuities.”

Model boundary and scope

The model boundary distinguishes endogenous variables (calculated within the model) from exogenous variables (specified externally). This distinction determines which dynamics the model can explain versus which must be assumed.

Endogenous variables. (1) Material flows: packaging consumption, collection, processing, and elimination across rigid and flexible streams. (2) Capacity dynamics: mechanical recycling capacity, membrane processing capacity, and collection infrastructure responding to policy signals and economic incentives. (3) Economic flows: processing costs, recyclate revenues, public funding requirements, and fiscal pressure. (4) Policy mechanisms: EPR implementation levels, CDS coverage expansion, and elimination mandate stringency. (5) Figure 1 above depicts the stock, flow and parameter system. Learning effects: cost reductions through cumulative experience in both mechanical and membrane processing. (6) System performance: recovery rates, landfill diversion, and circular material use.

Exogenous variables. (1) Population and economic growth: demographic trends and GDP growth set baseline consumption demand, but do not respond to, packaging policy. (2) Technological breakthroughs: while learning curves are endogenous, fundamental technology shifts (e.g., enzymatic depolymerization) are exogenous shocks. (3) Alternative materials adoption: substitution of plastics with paper, metal, or glass follows separate market dynamics. (4) International trade: post-National Sword, plastic waste export is minimal (<5% of total) and treated as effectively closed. (5) Political disruptions: election cycles, government priorities, and policy reversals modeled as scenario variations rather than endogenous responses. (6) General waste infrastructure: municipal waste management capacity assumed adequate; only packaging-specific infrastructure is endogenous (Australian Plastics Flows and Fates Study, 2015-2024).

Boundary choice justifications. (1) Capacity dynamics are central to the research question: understanding why recovery targets fail requires modeling of how processing capacity responds to economic signals. (2) Alternative materials follow different supply chains and regulatory frameworks; including them would require separate model sectors without illuminating plastic packaging dynamics. (3) International trade became negligible post National Sword (2018); modeling residual 2%–3% exports adds complexity without explanatory value. (4) Political economy is treated by scenario rather than mechanistically; attempting to model political decision-making endogenously introduces speculation without validation data. (5) General waste infrastructure is assumed to be non-binding; packaging-specific constraints (MRF sorting, membrane processing) are the relevant bottlenecks (Australian Plastics Flows and Fates Study, 2015-2024; Department of Climate Change et al., 2023).

Boundary adequacy tests confirmed robustness. (1) GDP variation: growth rates of ±20% GDP changed baseline consumption, but recovery rate outcomes varied under one percentage point, confirming that macroeconomic growth is not the binding constraint. (2) Export pathway inclusion: adding 10%–30% export option as an alternative to landfill had minimal impact (<2% on recovery rates) since export economics are worse than domestic landfill. (3) Endogenous EPR fee setting: allowing EPR fees to adjust based on fiscal pressure rather than exogenous policy showed similar equilibrium outcomes, validating the policy scenario approach. (4) Regional disaggregation: splitting Australia into states showed national aggregation to be appropriate as state-level variations average out. All adequacy tests confirmed that conclusions are robust to boundary specification (Australian Plastics Flows and Fates Study, 2015-2024; Department of Climate Change et al., 2023; Rubin et al., 2015).

Material flow structure with enhanced narrative

The model distinguishes rigid and flexible plastics as fundamentally different material systems with divergent economics, technologies, and policy responses. This structural separation proves essential for capturing the observed bifurcation in Australian recovery trajectories.

Membrane recycling technology specification

Membrane-based recycling represents solvent-based dissolution–reprecipitation with membrane filtration—the most mature pathway for mixed flexible plastic recycling with food-contact approval (Soares et al., 2022; Sherwood, 2020). The process is (1) selective solvent dissolution, (2) membrane filtration to remove contaminants, (3) solvent recovery (90%–95%), and (4) polymer precipitation/drying.

Our $19,500/tonne cost reflects solvent dissolution ($5,000/t), membrane modules ($4,500/t), solvent recovery ($3,500/t), precipitation/drying ($2,800/t), membrane replacement ($1,500/t), and quality assurance ($1,250/t). The costs are validated against pilot plant operations (Soares et al., 2022).

While alternatives exist, solvent-based dissolution with membrane purification dominates due to achieving food-contact quality (85%–95% purity) from contaminated feedstocks.

The model employs a standard system dynamics stock–flow structure, distinguishing stocks (accumulations measured in tonnes) from flows (rates measured in tonnes/year). In-market packaging stocks represent material currently in circulation from manufacture through retail to consumer use, with approximately 3-month (0.25-year) average residence time before becoming end-of-life waste. This rapid turnover—characteristic of packaging’s single-use or limited-use nature—means that stocks equal approximately one-quarter of annual consumption flows. For example, an annual consumption of 1,239,400 tonnes/year maintains in-market stocks near 310,000 tonnes at steady state. The model initializes stocks at 2015 baseline values and endogenously projects flows based on population growth, economic expansion, and per capita consumption trends validated against Australian Plastics Flows and Fates (APFF) historical data.

Rigid plastic packaging stocks initialize at 108,750 tonnes in 2015, representing in-market material with approximately 3-month average residence time. Annual consumption comprises 86% business-to-consumer applications (primarily beverage bottles, food containers, household product packaging) and 14% business-to-business applications (transport packaging, industrial containers). Consumption grows endogenously at 2.84% annually for B2C and 2.71% for B2B, driven by population growth (1.5% annually), economic expansion (GDP growth averaging 2.5%), and per capita consumption increases (0.3%–0.8% annually) observed in historical APFF data, reaching 576,000 tonnes/year. This value reflects APFF 2022–23 data showing rigid plastics account for approximately 40% of total plastic packaging. Collection rates begin at 35% baseline under conventional kerbside systems, representing the weighted average across Australian jurisdictions before CDS implementation. However, the model allows this to increase dramatically—potentially reaching 70%—when container-deposit schemes achieve universal coverage, reflecting empirical evidence from NSW (2017–2023) and Queensland (2018), where beverage container return rates exceeded 72% within 2 years of the scheme’s implementation.

Processing economics for rigid plastics prove marginally viable—mechanical recycling costs approximately $800 per tonne, encompassing collection transport ($150/t), materials-recovery facility sorting ($200/t), washing and decontamination ($180/t), grinding and extrusion ($200/t), and quality control ($70/t). Recyclate commands market values of $900 per tonne for food-grade PET applications or $800 per tonne for non-food applications like beverage crates or fiber-fill. This generates approximately break-even to modestly positive economics ($0–$100 per tonne profit margin), sufficient to attract private investment when combined with modest EPR support or favorable virgin resin pricing. The positive economics activate three reinforcing growth loops (R1: quality premium; R2: scale economies; R3: feedstock quality) that can drive system expansion under supportive policy conditions.

Flexible plastic packaging stocks initialized at 164,250 tonnes in 2015, representing in-market material with approximately 3-month average residence time. Annual consumption comprises 79% business-to-consumer applications (food wrap, shopping bags, bread bags, and frozen food pouches) and 21% business-to-business applications (pallet wrap, shipping envelopes, and protective packaging). Consumption grew at 2.84% B2C and 2.71% B2B annually, reaching 870,000 tonnes/year, representing approximately 60% of total plastic packaging per APFF 2022–23. Growth rates match rigid plastics but can be reduced substantially--potentially to near-zero or negative--through elimination mandates that target unnecessary formats. Base collection rates start at just 0.5% under conventional kerbside (most MRFs reject flexible plastics due to equipment incompatibility), increasing to maximum 5% even with full EPR support, as flexible plastics remain incompatible with CDS beverage return infrastructure.

Processing economics for flexible plastics prove catastrophically unviable: membrane-enhanced processing costs approximately $19,500 per tonne, comprising collection ($150/t), specialized sorting ($800/t), solvent dissolution ($5,000/t), membrane-filtration modules ($4,500/t), solvent-recovery systems ($3,500/t), precipitation and drying ($2,800/t), membrane replacement ($1,500/t), and quality assurance ($1,250/t). Recyclate values remain $600 per tonne—identical to mechanical recyclate--as the mixed-plastic composition limits application to low-value downcycling uses. This reflects blended rigid plastic recyclate prices 2023–24 (PET $520–735/t, HDPE $855–1,000/t, mixed kerbside $200–400/t; source: APFF, 2023–24). At $600/tonne recyclate value and 75% recovery rate, revenue per tonne processed is $450/tonne against a processing cost of $800/tonne, generating a net loss of approximately $290/tonne under current market conditions. These catastrophic economics activate Loop B4 (diminishing returns), the master constraint that prevents high recovery rates through universal processing approaches.

Table 1 provides comprehensive documentation of key model parameters with empirical sources and validation methods.

Eight interconnected feedback loops

The model architecture comprises eight feedback loops, the interactions of which determine system behavior. Three reinforcing loops (R1–R3) drive growth when conditions permit, while five balancing loops (B1–B5) constrain system expansion and create resistance to policy interventions (see Figure 1 above).

Reinforcing loop R1: quality premium loop

Food-grade recyclate from membrane processing commands 30% premiums over mechanical recyclate ($1,170/t versus $900/t), reflecting superior purity that enables closed-loop food-contact applications. This premium improves membrane economics from $18,700/t loss to $18,430/t loss—still catastrophic but marginally better. The premium effect alone proves insufficient to achieve viability but combines multiplicatively with loops R2 and R3 under optimal conditions.

Reinforcing loop R2: scale economy loop

Technology learning curves reduce costs 25% per cumulative capacity doubling for membrane systems and 15% for mechanical systems, based on empirical learning rates for emerging environmental technologies. At the current pilot scale (5,000 t/year cumulative globally), membrane costs remain $19,500/t. Reaching 80,000 t/year (four doublings) reduces costs to $9,800/t, which still generates $9,000/t losses but approaches theoretical viability thresholds. However, achieving sufficient deployment for meaningful learning requires overcoming initial economic barriers through substantial policy support.

Reinforcing loop R3: feedstock quality loop

Clean CDS-collected materials enable an 85% membrane recovery versus 60% from contaminated kerbside feedstock. A quality factor of 0.4 (kerbside baseline) increases to 1.0 (CDS maximum) as coverage expands. This quality improvement increases effective membrane recovery from 51% (0.60 × 0.85) to 85% (1.0 × 0.85), reducing per-tonne costs from $31,700 to $21,650. CDS thus proves a prerequisite for membrane viability, though even optimal CDS cannot overcome fundamental economic constraints.

Balancing loop B1: feedstock competition loop

Mechanical and membrane processors compete for limited clean PET feedstock. Mechanical processors, facing positive economics, politically resist membrane expansion from consuming their feedstock and lobby for constraints on membrane licensing or EPR allocation. This political economy dynamic has played out historically in multiple jurisdictions where incumbent recyclers oppose advanced technology deployment despite theoretical compatibility.

Balancing loop B2: elimination trade-off loop

Packaging elimination reduces total material volumes, threatening volume-dependent recycling business models. The recycling industry systematically opposes elimination mandates, despite superior environmental outcomes. The Australian Packaging Covenant explicitly prioritized recycling over elimination (2005–2018), reflecting this industry influence. Breaking this loop requires regulatory intervention that prioritizes waste hierarchy over industry interests.

Balancing loop B3: policy uncertainty loop

Paradoxically, industry engagement in policy development increases regulatory complexity and policy uncertainty. The voluntary APCO process involved 500-plus stakeholder consultations and generated 12 separate working groups, 47 technical specifications, and ongoing annual adjustments, creating persistent uncertainty despite intensive engagement. Regulatory complexity increases investment risk premiums, thus reducing profitability and constraining deployment. Simple binding regulation can prove more effective than complex voluntary frameworks.

Balancing loop B4: diminishing returns loop (master constraint)

Loop B4 represents the fundamental economic constraint that prevents high recovery rates through universal processing. This loop operates on a simple principle: as recovery targets increase, progressively more difficult and expensive materials must be processed—analogous to “picking low-hanging fruit first.” The easiest (most profitable) materials are recovered first, leaving increasingly difficult (loss-making) materials for higher targets.

Mechanism

The causal chain operates as follows: Higher recovery targets require the processing of more total material.

Since profitable rigid plastics are preferentially recovered first (positive economics encourage earlier adoption), higher targets increasingly require processing unprofitable flexible plastics.

Flexible plastic processing generates $18,700/tonne losses ($19,500/t cost minus $800/t recyclate value).

These losses escalate total system costs exponentially.

When costs exceed available funding (fiscal pressure >1.0), political sustainability collapses.

Programs must reduce targets or face budget crises, completing the balancing loop.

Numerical illustration

The exponential cost escalation is mathematically inevitable: At 40% recovery, processing ∼160,000 tonnes of flexible plastics annually generates $3.0B in losses. Marginal cost: approximately $15M per percentage point increase.

At 60% recovery, processing ∼350,000 tonnes of flexible plastics generates $6.5B in losses (2.2-times increase for 20 percentage point gain). Marginal cost: approximately $25M per percentage point.

Attempted 70% recovery would require processing ∼562,000 tonnes of flexible plastics, generating $10.5B in losses (1.6-times increase for only ten percentage point gain). Marginal cost: approximately $78M per percentage point.

The marginal cost per percentage point increases 3.1-fold between 60% and 70% recovery, creating the exponential barrier. No combination of subsidies, efficiency improvements, or political will can overcome this mathematical reality unless the material composition changes through elimination (activating loop B2).

Why this is the “master constraint”

This loop dominates system behavior above ∼60% recovery for three reasons: Mathematical inevitability. Australia generates 2.042M tonnes of plastic packaging EoL annually (2040 projection). Achieving 70% recovery (1.429M tonnes) arithmetically requires processing at least 550,000 tonnes of flexible plastics (given 803,000 tonnes of total flexible plastics and assuming aggressive CDS collection of rigid plastics). At $18,700/tonne losses, this is unavoidable.

Fiscal impossibility. EPR revenue at $300/tonne fee generates $612M annually (2.042M tonnes × $300). Processing 550,000 tonnes of flexible plastics at $18,700/t loss costs $10.3B. Even with $200M in public funding, fiscal pressure = $10.3B/$812M = 12.7, which is absolutely unsustainable politically.

No escape through technology. Even 50% membrane cost reduction (to $9,750/t) still generates $8,950/t losses, requiring $4.9B for 550,000 tonnes, with fiscal pressure still 6.0. Only implausible 95% cost reductions would break this constraint.

Loop B4 thus creates a fundamental ceiling on recovery rates achievable through processing alone. The only viable strategy to achieve 70% recovery is elimination (loop B2), which removes materials from the denominator rather than attempting to process them at catastrophic cost.

Balancing loop B5: rebound effect loop

Recycling success reduces disposal costs, weakening economic incentives for consumption reduction. Successful recycling can paradoxically increase packaging consumption through reduced environmental concern, thus potentially increasing total system costs despite higher recovery rates. This well-documented rebound effect requires active countermeasures through packaging reduction policies independent of recycling performance.

Policy instruments and parameters

Extended Producer Responsibility (EPR) operates through mandatory fees on packaging producers, with implementation levels ranging from 0 (absent) to 1.0 (full mandatory compliance). Fee structures range $200–$500/tonne of packaging placed on the market, generating annual revenues of $400M–$1,200M nationally, depending on implementation levels and fee rates. Revenue allocation follows evidence-based priorities: 15% collection infrastructure (kerbside and CDS operational support), 40% processing subsidies (covering losses on challenging materials), 25% elimination programs (format bans, redesign support), 10% administration, and 10% research and development. Figure 2 depicts the stock, flow and parameter system dynamics model developed over several iterations. This model was the basis for historical calibration and policy testing as described here. This allocation reflects optimal distribution identified through scenario analysis which prioritizes elimination over processing, given economic realities.

Container deposit schemes (CDS) coverage represents the population fraction with access to return infrastructure. Historical Australian implementation grew from 15% (South Australia only) in 2015 to 92% (all states except Tasmania) by 2024. Return rates average 72% where universal coverage exists, generating 1.44-times higher collection versus kerbside alone. Quality factors increase from 0.4 (kerbside: 15%–25% contamination) to 1.0 (CDS: <2% contamination), enabling food-grade recyclate production. The model allows CDS expansion to 100% coverage by 2027–2030, depending on the scenario.

Elimination mandates operate through progressive format bans and redesign requirements. The strength parameter (0–1.0) represents the fraction of current flexible packaging eliminated, with costs of $130/tonne covering industry compliance, reformulation, alternative packaging development, and enforcement. The optimal “80/15/5 strategy” eliminates 80% unnecessary formats (single-use shopping bags, excessive overwrap, non-functional decorative packaging), redesigns 15% to mono-material alternatives (multi-layer pouches to PE-only, PET–PE laminates to mono-PET), and processes only 5% essential formats (medical sterile packaging, pharmaceutical blister packs, specialized industrial applications) via membrane technology. This hierarchy inverts typical prioritization by making elimination foundational rather than supplementary.

Recovery rate targets (30%–70%) create policy pressure to drive loop B4 dynamics. Targets set political and regulatory expectations that shape EPR fees, investment mandates, and public messaging. However, targets disconnected from economic reality create fiscal pressure exceeding 1.0, triggering political backlash and program failures. The 70% target proves achievable only through redefinition that incorporates elimination into “effective recovery” metrics.

Membrane investment support ranges $25M–$200M in capital expenditure, with subsidy levels of 0%–50% of capital costs. Focused deployment (Scenario 4: $50M for 52,000 t/year) achieves 77% utilization, enabling learning curves, whereas dispersed universal processing (Scenario 3: $200M for 185,000 t/year) generates 35% utilization, preventing viability.

Data sources and validation

Material flow data derives from Australian Plastics Flows and Fates (APFF) annual reports 2015–2024, conducted by DCEEW with Blue Environment data collection (APCO, 2022; APCO, 2024; European Commission, 2019). APCO Packaging Consumption Data 2017–2023 provide consumption disaggregation (ICIS, 2025; EFSA, 2018). Economic parameters reflect Australian market conditions: mechanical costs ($800/t) from Blue Environment analysis (Rahimi and García, 2017), membrane costs ($19,500/t) from closed-loop partners and APK AG assessments (Material Economics, 2018; Forrester, 1961), elimination costs ($130/t) from European Commission impact assessments (APCO, 2023), virgin PET pricing (approximately $1,000/t in 2015, $750/t in 2020 [COVID demand collapse], $1,400/t in 2021–22 (energy crisis peak), and $850–900/t by 2025–26) from ICIS market data, 2024–26, and food-grade premium (30%) from EFSA recycling opinions and Closed Loop Foundation market analysis (Blue Environment, 2021; Kaiser et al., 2018). Technical parameters, including learning rates (membrane 25%, mechanical 15%), derive from Nemet and Rubin meta-analyses of environmental technology deployments (Staub, 2018; APCO, 2018). Recovery rates (mechanical 75%, membrane 85% clean/60% contaminated) reflect vendor specifications and pilot demonstrations (Closed Loop Partners, 2020; APK, 2020).

Historical validation (2015–2024) employed lookup tables for policy variables (CDS coverage, bag elimination), calculating all material flows endogenously. Validation metrics include mean absolute percentage error (MAPE, target <8%), mean absolute error (MAE, target <5 percentage points), directional accuracy (target >90%), and Theil’s U-statistic (target <0.3). Extreme conditions to test verified model behavior remains plausible at parameter limits. Boundary adequacy testing confirmed that excluded variables influence outputs by <10%. Sensitivity analysis tested 23 parameters at ±20% variations.

Scenario definitions with strategic logic

Five scenarios test alternative policy pathways, each representing coherent strategic approaches rather than arbitrary parameter combinations.

Scenario 1 (“business-as-usual”) extrapolates the current voluntary approach: no EPR implementation, gradual CDS expansion to 96% by 2040, zero elimination mandates, 30% recovery target reflecting historical trajectory, and minimal membrane investment ($25M). This represents the continuation of 2016–2024 voluntary frameworks.

Scenario 2 (“moderate policy”) implements partial EPR (0.5 level, $250/t fees), accelerated CDS (100% by 2030), modest elimination (30%), moderate target (50%), and medium membrane investment ($75M). This represents incremental policy strengthening without fundamental strategy change.

Scenario 3 (“70% without elimination”) tests universal processing hypothesis: full EPR (1.0, $500/t), complete CDS (100% by 2030), zero elimination, 70% target, and maximum membrane investment ($200M). This scenario deliberately attempts universal processing to document system failures, thus proving the necessity of elimination through contradiction.

Scenario 4 (“optimal 80/15/5 strategy”) implements an elimination-first approach: full EPR (1.0, $300/t), complete CDS (100% by 2030), strong elimination (80%), pragmatic target (65% for remaining packaging), focused membrane deployment ($50M for essential 5%). This represents support for the strategic pathway model results.

Scenario 5 (“European benchmark”) approximates EU performance: full EPR (1.0, $400/t), immediate complete CDS (100% from 2025), substantial elimination (60%), ambitious target (70%), and significant subsidized membrane ($100M). This provides an international comparison benchmark.

Model calibration and validation procedure

The model employs a two-stage calibration and validation procedure to ensure that parameter values reflect empirical reality while maintaining out-of-sample predictive performance. The calibration period (2015–2020) establishes parameter values through optimization against historical data, while the validation period (2021–2024) tests model performance on unseen data, including the critical REDcycle collapse event.

Parameter classification and derivation. Of the model’s 112 parameters, 23 (20%) are calibrated through optimization, 56 (50%) are derived from the literature, and 33 (30%) are calculated from other parameters. Calibrated parameters include collection rate response elasticities, capacity adjustment delays, learning curve exponents, and fiscal pressure thresholds—parameters where the literature provides ranges rather than point estimates. Literature-derived parameters include processing costs, capital costs, recovery rates, and growth rates where empirical data or engineering studies provide reliable values. Calculated parameters include derivative relationships (e.g., flows from stocks and rates) that are definitionally determined (Soares et al., 2022; Sherwood, 2020; Australian Plastics Flows and Fates Study, 2015-2024; Department of Climate Change et al., 2023; Australian Packaging Covenant Organisation APCO, 2023; ICIS, 2024; Hopewell et al., 2009; Rubin et al., 2015).

Optimization methodology. Calibration employed Powell’s conjugate direction method to minimize mean absolute percentage error (MAPE) across six time series: rigid plastic consumption, flexible plastic consumption, rigid recovery rate, flexible recovery rate, total system cost, and political sustainability index. Each calibrated parameter was bounded within ±30% of literature-derived initial estimates to prevent overfitting and ensure physical plausibility. The optimization converged after 847 function evaluations, achieving a calibration period MAPE of 4.1% across all time series (Rubin et al., 2015).

Validation results demonstrate robust out-of-sample performance. The validation period (2021–2024) achieved a MAPE of 5.8%, indicating good generalization with a modest increase from the calibration error. Material-specific validation shows a rigid plastics MAPE of 3.2% and flexible plastics MAPE of 6.8%. The higher flexible variance reflects genuine historical volatility (National Sword shock, REDcycle growth and collapse, and persistent stagnation) rather than model deficiency. Critically, the model successfully reproduced the REDcycle collapse as an emergent phenomenon, with predicted timing within 3 months (93% accuracy) and magnitude within 13% (87% accuracy) of observed values, despite this event occurring entirely within the validation period (Australian Plastics Flows and Fates Study, 2015-2024; Department of Climate Change et al., 2023).

Robustness tests confirm parameter stability. (1) Alternative objective functions (root mean square error, maximum absolute error) yielded parameter sets within 8% of MAPE-optimized values. (2) Subset calibration using only consumption data (excluding recovery rates) produced recovery rate predictions within 12% of full calibration. (3) Manual calibration by domain experts achieved comparable performance (MAPE 4.8% vs. 4.1% algorithmic), validating that optimization did not find spurious local minima. (4) Cross-validation using alternating 3-year periods showed stable parameter estimates (coefficient of variation <15% across folds). These tests demonstrate that calibrated parameters represent genuine system structure rather than overfitting artifacts (Rubin et al., 2015).

Known parameter uncertainties. Three parameter categories have substantial uncertainty: (1) membrane processing costs (±18% based on pilot plant variation), (2) learning rate parameters (±40% reflecting wide literature ranges for emerging technologies), and (3) elimination effectiveness (±25% due to limited empirical precedent for 80% targets). Sensitivity analysis (Section 3.3) demonstrates that core conclusions remain robust across these uncertainty ranges, with elimination superiority holding even under optimistic membrane cost scenarios (50%–70% reductions) (Soares et al., 2022; Rubin et al., 2015; International Energy Agency IEA, 2022).

Historical validation (2015–2024)

The model successfully reproduced nine years of Australian plastic packaging system behavior. Note on recovery rate methodology: the model’s current recovery rate is calculated as total material recovered divided by total packaging end-of-life (including both rigid and flexible streams), producing values of approximately 10–15% compared to APFF’s reported 16–26%. This reflects a genuine methodological distinction--the model denominator includes flexible plastics with near-zero recovery rates, whereas APFF reporting uses broader packaging categories and includes energy recovery. Separate rigid recovery rate (approximately 25–38%, rising through CDS expansion) and flexible recovery rate (approximately 0.5–1.4%, with voluntary peak in 2022 followed by collapse) provide the primary material-specific validation outputs. The trajectory shape--gradual improvement through CDS expansion, peak around 2021–22, sharp decline post-REDcycle collapse, stabilisation 2023–24–constitutes valid structural validation.

The model reproduced four critical historical patterns that validate causal structure. First, the China National Sword impact (2018) generated a temporary recovery spike of 18% as contaminated material exports ceased, forcing domestic infrastructure utilization increases before subsequent decline. Second, CDS expansion in NSW (2017) and Queensland (2018) drove recovery improvements to 20%–22% over 2019–2021 as beverage container collection increased from 15% to 72% population coverage. Third, peak performance in 2022 at 25.7% preceded the fourth pattern: the REDcycle collapse (November 2022) causing a decline to 19%–20% through 2024 as 8,000–12,000 tonnes of annually collected soft plastics shifted from stockpiling to landfill.

Critically, the REDcycle collapse emerged endogenously from fiscal pressure accumulation under the voluntary funding regime. With EPR Implementation Level at zero throughout 2015–2024, available funding was constrained to public allocations of approximately $80M/year. As voluntary flexible collection programs scaled from zero to approximately 8,400 tonnes/year during 2019–2022, net processing losses accumulated at approximately $18,700/tonne, generating voluntary program net loss of approximately $160–170M/year at peak—more than double available funding. Perceived fiscal pressure, smoothed with a 0.75-year organisational response delay representing the lag between financial stress and program closure, crossed the program viability threshold approximately consistent with the November 2022 REDcycle closure timing. The model reproduced the post-collapse recovery rate decline without any exogenous shock representing the REDcycle event, validating the fiscal constraint hypothesis.

The capacity utilization paradox—capacity increasing 51% (2020–2024) while utilization declined from 58% to 52%—emerged from loop B1 (“feedstock competition”). Multiple facilities competed for limited clean CDS feedstock, reducing individual plant throughput despite aggregate capacity growth. The model’s rigid–flexible bifurcation successfully reproduced divergent recovery trajectories: rigid plastics 24%–27% (model) versus 28% actual (one percentage point error), flexible plastics 4.5%–5.8% (model) versus 6.0% actual (0.2 pp error). Different parameter values, not different structures, generated these trajectories, validating that economic fundamentals drive divergence rather than distinct technological or behavioral regimes.

Material-specific validation accuracy

Validation errors disaggregated by the material stream confirm robust performance: Rigid plastics: 3.2% MAPE over 2015–2024

Flexible plastics: 6.8% MAPE over the same period

Both streams fall within the <8% aggregate threshold. Flexible plastics show higher variance due to greater historical volatility, including the post-National Sword spike (2018–2019), REDcycle growth and collapse (2018–2022), and persistent stagnation at 4.5%–6% recovery. The 6.8% MAPE demonstrates that the model reproduces volatile real-world behavior rather than merely fitting smooth trends (Australian Plastics Flows and Fates Study, 2015-2024; Department of Climate Change et al., 2023).

Scenario analysis results (2025–2040)

Scenario 1 (“business-as-usual”) achieves only 33% recovery by 2040 (47% of 70% target) at an annual cost $320M. This represents continued stagnation; despite ongoing investment matching 2016–2024 levels ($40–$50M annually), voluntary approaches plateau far below targets. The scenario demonstrates that incremental improvements within existing frameworks cannot overcome structural barriers. Fiscal pressure of 1.6 indicates that costs exceed sustainable funding by 60%, explaining persistent underperformance.

Scenario 2 (“moderate policy”) reaches 48% by 2040 (69% of target), costing $480M annually with fiscal pressure of 0.92. Partial EPR implementation and accelerated CDS expansion generate meaningful improvements but fall 22 percentage points short of the 70% target. This “middle ground” approach proves insufficient, as it is trapped between inadequate elimination (only 30%) and insufficient processing capacity. The scenario illustrates that halfhearted policy cannot overcome fundamental constraints.

Scenario 3 (“70% without elimination”) catastrophically fails, achieving only 57% despite $1,180M annual expenditure and full EPR implementation. Fiscal pressure reaches 2.4, where costs exceed available funding by 140%, creating a political sustainability crisis. At 57% recovery, marginal costs escalate to $78M per percentage point, reflecting loop B4 dominance. The scenario attempts to process 185,000 tonnes/year of flexible plastics (23% of total EoL), generating $3.46B annual losses ($18,700/t × 185,000t). Even with unlimited theoretical funding, capacity utilization collapses to 35% as flexible materials’ extended processing times overwhelm facilities. Membrane plants face contaminated feedstock (clean CDS material preferentially flows to profitable mechanical processing), reducing recovery from a potential 85% to an actual 60%, further undermining the economics. This scenario proves by contradiction that universal processing cannot achieve 70%, as fundamental economic and technical constraints prevent success regardless of funding or political will.

Scenario 4 (“optimal 80/15/5 strategy”) succeeds dramatically, achieving 68% recovery (97% of target) at $510M annually—2.3-times cheaper than Scenario 3 while delivering 11 percentage points higher performance. Effective recovery including elimination reaches 72%–75% when accounting for 641,000 tonnes/year of eliminated formats. Fiscal pressure of 0.89 indicates sustainable funding balance. Marginal costs remain moderate at $22M per percentage point, enabling continued incremental improvements. The scenario eliminates 80% of unnecessary flexible formats, saving approximately $77M annually in avoided landfill costs while incurring $83M elimination expenses — approximately cost- neutral versus landfill, but substantially cheaper than the $12.4B that membrane processing of the same volumes would require. Focused membrane deployment of 52,000 tonnes/year (6.5% of total EoL) for essential medical and pharmaceutical applications achieves 77% capacity utilization, thus activating learning curves and quality premiums. Mechanical processing expands to 600,000 tonnes/year for rigid plastics, operating at 82% utilization with positive economics. This strategic allocation maximizes system efficiency by matching technologies to economically viable material streams.

Scenario 5 (“European benchmark”) achieves 71% at $580M annually, demonstrating that mature European infrastructure with 15–20 years headstart can marginally outperform optimal Australian strategy, but at 14% higher cost. European advantage derives from established EPR frameworks since 2005, universal CDS coverage, and substantial public acceptance of elimination approaches. However, diminishing returns still constrain Europe above 70%, and further improvements require progressive elimination rather than processing expansion.

Sensitivity analysis: high-leverage parameters

One-at-a-time sensitivity testing of 23 parameters at ±20% variations under Scenario 4 (optimal) baseline identified five high-leverage parameters. Sensitivity index measures (percentage output change)/(percentage parameter change) for the 2040 recovery rate.

Base membrane cost ($19,500/t baseline) exhibits a sensitivity index of −0.82, meaning that 10% cost reduction improves recovery 8.2 percentage points. This extraordinary leverage reflects the central role of membrane deployments; reducing costs from $19,500/t to $17,550/t enables the processing of an additional 15,000 tonnes/year economically, directly increasing recovery. However, even 50% cost reduction to $9,750/t still generates $8,950/t losses, indicating that fundamental viability barriers persist. Membrane research should target loss minimization for essential applications, not profitability (physically impossible given $800/t recyclate values).

The EPR fee level ($300/t baseline) shows index +0.71; 10% increase to $330/t improves recovery 7.1 pp. Higher fees generate an additional $72M annually ($3 × 2.4M tonnes), enabling increased elimination subsidies and membrane deployment. However, fees above $500/t face political resistance and create perverse incentives for packaging reduction shifting to non-packaging applications. Optimal fees balance revenue needs against industry acceptance.

The learning rate for membrane technology (0.25 baseline = 25% per doubling) demonstrates index +0.64: 10% faster learning (0.275 per doubling) improves recovery 6.4 pp. Faster learning reduces costs more quickly, enabling earlier viability for marginal applications. However, learning requires deployment, creating a chicken-and-egg problem: achieving a sufficient scale for meaningful learning requires overcoming initial economic barriers through policy support.

Elimination strength (0.80 baseline = 80% eliminated) has an index of +0.58; a 10% increase to 88% improves recovery 5.8 pp. Additional elimination removes 64,000 tonnes/year of flexible plastics, saving $1.2B in avoided processing losses while requiring only $8M in additional elimination costs. This demonstrates the consistent highest-return intervention status of elimination, regardless of baseline assumptions.

Elimination cost ($130/t baseline) exhibits an index of −0.52; 10% higher costs ($143/t) reduce recovery 5.2 pp by constraining politically acceptable elimination scope. Actual elimination costs vary substantially: simple format bans (shopping bags) cost only $50/t, while complex reformulations (medical packaging) reach $300/t. Targeting low-cost elimination first maximizes impact.

Robustness testing simultaneously varied all five high-sensitivity parameters pessimistically (membrane costs +20%, EPR fees −20%, learning rates −20%, elimination strength −20%, elimination costs +20%), generating worst-case recovery of 61% compared to baseline of 68%. This seven percentage point degradation still significantly outperforms Scenario 3’s universal processing (57%) at much lower cost ($620M vs. $1,180M). The strategic conclusion is that the elimination-first strategy is superior to universal processing and proves robust across plausible parameter ranges. Even under pessimistic assumptions, focused deployment outperforms ambitious processing expansion.

Sensitivity to membrane cost reductions

To assess whether conclusions depend on current membrane cost assumptions, we analyzed scenarios with aggressive technological breakthroughs to reduce membrane processing costs by 50%–70% relative to our $19,500/tonne baseline (Soares et al., 2022).

Scenario A (50% cost reduction to $9,750/tonne). Flexible plastic processing losses reduced from $18,700/t to $8,950/t (still severely negative). Attempting 70% recovery through universal processing would cost $720M annually (fiscal pressure 1.6, still exceeding sustainable funding), achieving only 62% actual recovery. Comparison with Scenario 4 (“optimal 80/15/5 strategy”): elimination strategy achieves 68% recovery at $510M annually (pressure 0.89); elimination remains 1.4-times cheaper with six percentage points higher performance.

Scenario B (70% cost reduction to $5,850/tonne). This represents an extraordinarily optimistic technological breakthrough, faster than any documented environmental technology learning trajectory. Universal processing becomes marginally viable, with total system cost at $480M annually (fiscal pressure 0.88, just below sustainability threshold), achieving 68%–70% recovery.

However, achieving 70% cost reduction would require membrane recycling to match or exceed the fastest documented technology learning rates (aviation biofuels: 65% over 20 years; solar PV: 63% over 22 years). Current membrane recycling for flexible plastics faces fundamental physical challenges absent from these comparator technologies: multi-layer structures requiring chemical dissolution with no mechanical shortcut, solvent chemistry complexity, contamination management, and food-contact quality requirements (Rubin et al., 2015; International Energy Agency IEA, 2022).

For universal processing to achieve true cost parity with elimination ($130/t), membrane costs would need to fall to approximately $1,000–1,500/tonne (a 92%–95% reduction), essentially matching mechanical recycling costs despite requiring far more complex chemical processing. This is physically implausible given the inherent complexity of multi-layer flexible plastic separation. Our core finding—that strategic elimination is more cost-effective than universal processing—holds under all plausible technology scenarios (0%–50% cost reduction by 2040) (Soares et al., 2022; Rubin et al., 2015).

The rigid–flexible divide is fundamental, not transitional

The economic gulf between rigid and flexible plastic recycling reflects fundamental physical reality, not temporary market inefficiency or technological immaturity. Multi-layer flexible structures (three to seven layers which comprise incompatible polymers, adhesives, barrier materials, and printing inks) cannot be effectively separated through any conceivable processing pathway while preserving material value. Solvent-based membrane separation successfully purifies individual polymer streams, but the separated materials command only $800/t commodity pricing—identical to mechanically recycled mixed plastics—because functional properties derive from the laminate structure, not constituent materials. Even with aggressive technology development achieving 50% membrane cost reduction to $9,750/t, flexible processing would still generate losses of $8,950/t.

This physical reality drives policy implications. At corrected Australian market prices (blended recyclate value $600/tonne, processing cost $800/tonne, 75% recovery rate), mechanical recycling of rigid plastics generates a net loss of approximately $290/tonne, consistent with APFF (2023–24) findings that recovery costs exceed virgin material value under current market conditions. EPR support is therefore required to sustain rigid plastics recycling — reinforcing the paper’s policy argument for mandatory EPR frameworks. Despite this, rigid plastics exhibit substantially superior economics to flexible plastics ($290/t loss versus $18,700/t loss) and justify processing-focused investments activating Loops R1–R3 (quality premiums, scale economies, feedstock quality improvements). Materials exhibiting catastrophic economics (flexible plastics: $18,700/t loss) justify elimination-focused interventions, where $130/t in elimination costs provide 150:1 cost–benefit ratios versus processing alternatives. Treating all plastics homogeneously—the implicit assumption underlying Australia’s 70% target and EU recycling mandates—creates economically irrational policy that mixes high-return and catastrophically negative-return interventions.

Loop B4 (diminishing returns) as master system constraint

Scenario 3 results demonstrate Loop B4’s dominance above 60% recovery. Its mathematical inevitability is straightforward: Australia generates 2.042M tonnes of plastic packaging EoL annually (2040 projection), comprising 1.239M rigid (61%) and 0.803M flexible (39%). At 50% aggregate recovery, the material mix processed is approximately 70% rigid (positive economics) and 30% flexible (catastrophic). System-wide costs remain manageable at $550M annually. Increasing recovery to 60% requires processing 1.225M tonnes, with composition shifting to 60% rigid and 40% flexible, generating costs of $780M. Attempting 70% recovery (1.429M tonnes processed) drives composition to 50% rigid and 50% flexible, generating costs of $1,040M.

However, universal processing cannot reach 70% because diminishing returns accelerate nonlinearly. Processing 70% of flexible plastics (562,000 tonnes) generates $10.5B in losses annually, creating fiscal pressure exceeding 5.0—an absolute political impossibility. Marginal cost per percentage point escalates from $25M (at 60%) to $78M (at attempted 70%) as progressively more contaminated, lower-value flexible materials are processed. Capital expenditure requirements reach $4.5B for membrane capacity ($8,000/t × 562,000t), exceeding total Australian recycling infrastructure value. Even if funding were available, capacity utilization collapses as extended processing times for flexible materials (48 h vs. 8 h for rigids) overwhelm facilities.

This mathematical reality cannot be overcome through better technology, higher subsidies, or stronger political will. Loop B4 represents fundamental physical and economic constraints. The only viable strategy is to reduce flexible plastic volumes through elimination before attempting recovery, thus changing the denominator rather than forcing the numerator.

Elimination economics: 150:1 cost–benefit ratio

Eliminating unnecessary flexible packaging costs $130/tonne, encompassing industry reformulation ($45/t), alternative packaging development ($35/t), supply chain adaptation ($25/t), regulatory compliance ($15/t), and enforcement ($10/t). Australian landfill gate fees average approximately $150–250/tonne. The three-way comparison is: elimination $130/t, landfill $150–250/t, membrane processing $19,500/t. Elimination therefore provides approximately a 2:1 cost advantage over the landfill counterfactual, and 150:1 versus membrane processing—the latter figure being relevant only for the 5% of essential formats that genuinely require advanced recycling. The avoided cost versus landfill is approximately 641,000 t/yr × ($250 – $130)/t = $77M/year.

Moreover, elimination provides superior environmental outcomes through waste hierarchy principles: prevention > reuse > recycling > energy recovery > disposal. The elimination of unnecessary packaging prevents upstream resource extraction, manufacturing emissions, and transportation impacts—benefits that are absent from recycling. Lifecycle assessment comparing elimination to membrane recycling for bread bags demonstrates 85% lower greenhouse gas emissions, 90% lower water consumption, and 95% lower toxic releases through elimination pathway.

Nevertheless, political economy challenges exist. Packaging manufacturers resist elimination that threatens sales volumes, recycling industries oppose elimination that threatens feedstock, and some environmental advocates mistakenly view elimination as an admission of recycling failure rather than a hierarchy-consistent strategy. However, the 150:1 economic advantage plus superior environmental performance creates a compelling policy rationale for overcoming resistance.

Voluntary approaches: systematic failure mechanisms

Eight years of Australian experience (2016-2024) achieving four percentage points improvement despite $690M of investment demonstrates structural failures in a voluntary approach. Two complementary mechanisms explain systematic voluntary failure. Loop B3 (Policy Uncertainty) identifies how voluntary frameworks require continuous multi-stakeholder negotiation generating regulatory complexity and persistent uncertainty. A fiscal mechanism compounds this: with EPR implementation level at zero throughout 2015–2024, available funding was constrained to approximately $80M/year. Voluntary programs like REDcycle faced an unavoidable fiscal trap—at $18,700/tonne processing losses, any program exceeding approximately 4,000 tonnes/year generated losses exceeding total available public funding. The REDcycle collapse is therefore a structural inevitability of attempting subsidised flexible plastic collection without mandatory funding. The Australian Packaging Covenant involved more than 500 organizations across 12 working groups to produce 47 technical specifications with annual adjustments, thus creating process complexity without policy certainty. Investment decisions requiring 10–15 years payback periods cannot proceed under conditions of annual specification changes and voluntary commitment unpredictability.

A comparison with regulatory approaches proves instructive. The NSW Container Deposit Scheme (2017)—a binding legal requirement with penalties for non-compliance—increased beverage container recovery from 44% to 72% within 2 years, requiring $50M infrastructure investment and achieving 14-fold cost-effectiveness versus national voluntary efforts. Simple binding regulation replaced complex negotiation, creating regulatory certainty that enabled investment. The implication is that circular economy policy should lead with binding regulation to establish clear legal obligations, treating voluntary initiatives as supplementary rather than primary instruments. However, current Australian policy inverts this relationship, establishing voluntary targets with regulation as fallback and thus explaining persistent failure.

Revised recovery rate framework

The current 70% target creates perverse incentives by measuring only recycled material, ignoring eliminated material. This framework incentivizes the processing of unnecessary packaging rather than eliminating it—precisely opposite to waste hierarchy principles. Scenario 4 demonstrates a reframing value: achieving 68% “recycling rate” plus 80% elimination generates 72%–75% “effective recovery” when elimination counts as waste prevention. A revised target framework should specify: “Achieve 65% recovery of plastic packaging remaining after elimination, with combined effective rate (recovery + elimination) of 70%–75%.” This incentivizes elimination while maintaining ambitious aggregate performance. Europe’s evolving framework (Packaging and Packaging Waste Regulation, 2023) adopts a similar approach, establishing separate recycling and reduction targets rather than a single combined metric.

Policy recommendations: four-phase implementation

Phase 1 (2025–2027) establishes the foundations: full EPR implementation with $300–$400/t fees generating $720M–$960M annually for elimination and targeted processing subsidies; binding obligations replacing voluntary commitments; CDS expansion to 100% population coverage by 2027 through accelerated rollout in Tasmania, ACT, and regional areas; elimination mandate development through statutory authority classification of packaging formats into 80% unnecessary (mandatory phase-out), 15% redesignable (mono-material requirements), and 5% essential (membrane processing candidates).

Feasibility of 80% elimination target

The proposed 80% elimination target is aggressive but grounded in international precedent and format-specific analysis. The European Union Single-Use Plastics Directive (2019) achieved 70%–90% reduction in targeted formats (straws, stirrers, cotton bud sticks, balloon sticks, cutlery, plates) within 3 years (2019–2022) across 27 member states, demonstrating that aggressive elimination is achievable when regulatory mandates replace voluntary approaches (Department of Climate Change et al., 2023).

Australian plastic bag bans achieved a 77% reduction nationally within 4 years (2018–2022), with individual states reaching 80%–90% reductions, demonstrating domestic supply chain adaptation capacity and consumer acceptance (Department of Climate Change et al., 2023). International precedents include Rwanda (95% plastic bag elimination 2008–2015), Kenya (80% reduction 2017–2020), and Bangladesh (90% reduction 2002–2010), showing that near-complete elimination is achievable across diverse economic contexts.

Our 80% target derives from format-specific classification. (1) Demonstrably unnecessary formats (50%–60% of volume): shopping bags, produce bags, bread bags, excessive overwrapping, redundant secondary packaging, alternatives for which are already available and cost-effective. (2) Redesignable to mono-material (20%–30%): multi-layer pouches replaceable with mono-material PE/PP alternatives, laminated films with paper-based bio-coated alternatives, convenience packaging providing marginal benefit. (3) Essential formats requiring advanced recycling (5%–10%): medical sterile packaging, pharmaceutical blister packs, specialized electronics protection where no viable alternatives currently exist (Department of Climate Change et al., 2023; Australian Packaging Covenant Organisation APCO, 2023).

The 15-year implementation timeline (2025–2040) provides substantial margin for staged rollout. The EU SUP Directive achieved comparable reductions in 3–5 years, demonstrating that 15 years is conservative. Economic drivers support feasibility: the 150:1 cost–benefit ratio ($130/tonne elimination versus $19,500/tonne processing) creates powerful incentives under EPR frameworks where producers bear end-of-life costs. While achieving 80% elimination requires unprecedented political will and industry cooperation, the precedents demonstrate that it is achievable, and the economics demonstrate that it is rational (Department of Climate Change et al., 2023; ICIS, 2024).

Phase 2 (2028–2032) implements elimination: progressive format bans achieving 40% elimination by 2028, 60% by 2030, and 80% by 2032, through staged regulatory timelines that enable industry adaptation; mono-material redesign requirements for 15% of formats with compliance deadlines; industry reformulation support through $50M R&D funding to enable smaller manufacturers to develop alternatives; public communication campaigns explaining elimination rationale and benefits.

Phase 3 (2028–2035) deploys targeted membranes: focus exclusively on the 5% essential formats (medical sterile packaging, pharmaceutical blister packs, specialized electronics protection); 50% capital subsidies reducing investment barriers for membrane facilities; EPR revenue covering ongoing processing losses during technology maturation; R&D for 40% cost reduction by 2040 through solvent innovation, membrane materials improvement, process integration; mechanical optimization to expand rigid plastic capacity to 600,000 t/year operating at 80%-plus utilization.

Phase 4 (2035–2040) optimizes performance: evaluation of achieved rates against 65% target for remaining packaging; assessing whether technology development has produced alternatives for essential formats currently requiring membrane processing; adjusting EPR fees based on realized costs rather than projections; considering target increases only if elimination and processing costs decline substantially.

The avoided approaches prove equally important: universal processing (Scenario 3) costing $1,180M while failing to reach targets must be rejected; partial EPR (Scenario 2) achieving only 48% demonstrates that halfway measures are insufficient; continued voluntary commitments despite 8 years’ demonstration of systematic failure; delayed CDS expansion preventing loop R3 activation and membrane viability; 70% target without elimination creating unrealistic expectations, undermining policy credibility.

Structural sensitivity and robustness tests

Structural sensitivity tests

To assess whether conclusions depend on specific model formulations, five structural sensitivity tests were conducted that varied fundamental model assumptions while holding parameters constant. These tests examined whether my findings represent genuine system behavior or artifacts of modeling choices (Rubin et al., 2015).

Test 1: Material disaggregation. The baseline model treats flexible plastics as homogeneous. We disaggregated five subtypes: LDPE film (35% of flexible), HDPE film (25%), PP film (20%), multi-layer laminates (15%), and other flexibles (5%), each with distinct processing costs ($17,000–$23,000/t) and recovery rates (80%–90%). Results: scenario outcomes changed <0.3% (weighted average costs nearly identical to aggregate specification). Conclusion: material aggregation is appropriate; heterogeneity within flexible plastics does not affect strategic conclusions (Luna et al., 2023; Australian Plastics Flows and Fates Study, 2015-2024).

Test 2: Feedback loop exclusion. Each of the eight feedback loops was systematically removed to identify structural dependencies. Removing loops R1–R3 (reinforcing growth dynamics) reduced achievable recovery rates by five to eight percentage points but did not change rank ordering of scenarios. Removing loops B1–B3 (competition, trade-offs, uncertainty) had <3% impact. However, removing loop B4 (“Diminishing Returns”) fundamentally altered system behavior, where universal processing scenarios no longer failed; instead it achieved targets but at 3.2-times higher costs ($1,800M vs. $550M annually). This confirms that loop B4 is the structural linchpin—the core mechanism preventing universal processing success. Removing B5 (“Rebound”) increased final recovery rates four to six percentage points, indicating that rebound effects are real but not dominant (Australian Plastics Flows and Fates Study, 2015-2024).

Test 3: Time horizon extension—an extended simulation from 2040 to 2050 to test whether conclusions are artifacts of an arbitrary endpoint. Results: all scenario rankings were unchanged. Scenario 3 (“universal processing”) continued deteriorating after 2040, reaching only 54% recovery by 2050 (worse than 2040’s 57%) as fiscal pressure exceeded 2.0, forcing program retrenchment. Scenario 4 (“elimination strategy”) stabilized at 68%–69% through 2050. Conclusion: the 2025–2040 period captures essential dynamics, and extending the timeline strengthens rather than weakens the elimination case (Australian Plastics Flows and Fates Study, 2015–2024).

Test 4: Stock structure variation replaced fixed delay stocks (0.25 years residence time) with fractional delay (exponential distribution) and conveyor delay (fixed pipeline). All three formulations produced results within 2% for recovery rates and costs. This insensitivity reflects that packaging turnover (quarterly) is fast relative to capacity adjustment delays (2–5 years) and learning effects (cumulative). Stock structure choice is not consequential for this application (Rubin et al., 2015).

Test 5: Learning curve formulation compared three learning formulations: (1) exponential decay (baseline), (2) power law (standard in technology learning), and (3) S-curve with saturation. All three produced 2040 outcomes within ±4%: Scenario 3 achieved 55%–59% recovery at $1,100–$1,250M, and Scenario 4 achieved 66%–69% at $480–$550M. Elimination superiority held across all formulations. Conclusion: the strategic findings are robust to learning curve specification; the order-of-magnitude cost difference between mechanical and membrane processing dominates learning trajectory assumptions (Rubin et al., 2015; International Energy Agency IEA, 2022).

Synthesis: core findings are structurally robust. The rigid–flexible economic divide, loop B4, dominates above 60% recovery, and elimination’s superior cost-effectiveness emerges consistently across diverse model formulations. Only loop B4 represents a structural dependency, but this reflects genuine physical and economic constraints (exponentially increasing costs as material composition shifts toward loss-making streams) rather than modeling artifacts. The convergence of multiple sensitivity tests provides confidence that conclusions represent real system behavior rather than parametric or structural assumptions (Luna et al., 2023; Australian Plastics Flows and Fates Study, 2015-2024; Rubin et al., 2015).

Model limitations and future research

Material aggregation treats all rigid plastics together and all flexible plastics together, missing polymer-specific dynamics (PET bottles versus PP tubs; PE films versus PET–PE laminates). Future research should disaggregate to six to eight polymer-application combinations. Geographic aggregation to national level does not capture state variations or remote area cost differences. State-level modeling would improve policy precision. The representation of technology as binary mechanical/membrane categories obscures emerging pathways (enzymatic depolymerization, hybrid processes); an expanding technology portfolio would capture innovation impacts. Exogenous variables including virgin prices, growth rates, and learning rates could be endogenized with additional feedback structures capturing market dynamics. System boundaries exclude chemical recycling, biological recycling, reusable packaging, and international trade—all worthwhile expansions for comprehensive assessment. The nine-year validation period (2015–2024) is relatively short; ongoing annual updates as new APFF data releases will strengthen confidence.

Broader circular economy implications

Five generalizable insights emerge beyond Australian plastic packaging context. First, elimination proves essential rather than supplementary for materials that exhibit catastrophic recycling economics. This is applicable to other challenging waste streams, including composite textiles, laminate papers, and electronic component assemblies. Second, material heterogeneity matters, as aggregation obscures fundamental divides, creating economically irrational policy. Third, technology proves necessary but insufficient: engineering advances cannot overcome fundamental economic constraints. Policy must thus create conditions that enable viable technologies while preventing the deployment of inherently unviable approaches. Fourth, voluntary market-based approaches face structural barriers (particularly loop B3), systematically preventing success. Therefore, binding regulation should lead rather than follow. Finally, system dynamics methodology demonstrates value in capturing feedback loops that generate unintended consequences across environmental policy domains.

Transferability to other jurisdictions

While this study focuses on Australia, several findings likely generalize to other jurisdictions that face similar circular economy challenges.

Transferable findings. (1) The rigid–flexible economic divide reflects the universal physical reality that multi-layer flexible packaging structures cannot be mechanically separated while preserving material value, creating order-of-magnitude cost differences ($800–$1,200/t mechanical versus $15,000–$25,000/t chemical recycling) that persist across jurisdictions. (2) Diminishing returns above material-specific thresholds apply universally to mixed waste systems. The specific threshold varies with composition (Australia: 60%; jurisdictions with 50% rigid/50% flexible: potentially 65%–70%; jurisdictions with 30% rigid/70% flexible: potentially 50%–55%), but exponential cost escalation remains. (3) Elimination’s superior cost-effectiveness (100–150:1 cost–benefit ratio) reflects universal economics, as elimination costs and processing costs face similar drivers internationally (Luna et al., 2023; Sherwood, 2020; Australian Plastics Flows and Fates Study, 2015-2024; Department of Climate Change et al., 2023; ICIS, 2024).

Context-specific factors affecting transferability. (1) Material composition differences: European average of 45% rigid/55% flexible versus Asian markets’ 30% rigid/70% flexible versus Australia’s 40% rigid/60% flexible affects achievable recovery rates but not fundamental rigid–flexible divide. (2) Existing infrastructure maturity: countries with mature CDS infrastructure (Germany, Nordic countries) face lower marginal costs for collection expansion than Australia’s recent implementation. (3) Geographic scale: Australia’s low population density (3.3 people/km 2) creates higher collection transport costs ($150–$200/t) than European nations (Germany: 240/km 2) with costs of $50–$100/t, making elimination even more economically attractive in low-density contexts (Australian Plastics Flows and Fates Study, 2015-2024; Department of Climate Change et al., 2023).

Political economy varies. The success of the EU Single-Use Plastics Directive reflects strong governance capacity, a sophisticated retail sector, and powerful environmental constituencies. Similar policies may face challenges in contexts lacking these enabling conditions. Concentrated retail (two to three major chains) enables rapid implementation; fragmented retail requires a longer transition (Department of Climate Change et al., 2023).

Methodological transferability. The system dynamics framework—distinguishing material streams by economic viability, identifying feedback loops, analyzing policy interactions—is directly applicable to other jurisdictions. The model can be recalibrated with jurisdiction-specific parameters (costs, collection rates, and material composition) to assess policy options in different contexts. We encourage such applications and welcome collaborative recalibration efforts (Rubin et al., 2015).

Despite context variations, policy implications apply broadly. (1) Do not treat all plastics homogeneously, as rigid and flexible plastics require fundamentally different approaches. (2) Consider elimination before processing for materials with negative processing economics. (3) Avoid voluntary approaches, as binding mandates achieve two to three times faster progress. (4) Set targets based on material composition, not arbitrary percentages. (5) Use EPR to fund elimination subsidies over processing subsidies for unfavorable streams (Australian Plastics Flows and Fates Study, 2015-2024; Department of Climate Change et al., 2023; Australian Packaging Covenant Organisation APCO, 2023; ICIS, 2024).

This research demonstrates that membrane technology can enable Australia’s 70% plastic packaging recovery target but only through the strategic elimination of unnecessary formats combined with focused deployment for essential applications and not through universal processing approaches. The system dynamics model, validated against 9 years of Australian data and incorporating eight interconnected feedback loops, provides quantitative evidence to support a fundamental reorientation of circular economy policy.

Six key findings emerge. First, universal processing —catastrophically fails. Scenario 3, attempting 70% recovery through comprehensive membrane deployment, achieves only 57% at $1.18 billion annually—the highest cost for lowest relative performance. Loop B4 (“diminishing returns”) dominates above 60% recovery as flexible plastic economics ($18,700/t losses) create exponential cost escalation. Second, the optimal 80/15/5 strategy succeeds. Eliminating 80% unnecessary flexible formats achieves 68% recovery at $510 million annually—2.3 times cheaper with 11 percentage points higher recovery plus 72%–75% effective recovery including elimination. Third, elimination provides 150:1 cost–benefit advantage: $130/t elimination costs versus $19,500/t processing costs save $12.4 billion annually in avoided losses. Fourth, the rigid–flexible divide reflects fundamental physics, not temporary inefficiency. Multi-layer structures cannot be economically separated while preserving value. The fifth finding is that voluntary approaches systematically fail. Loop B3 dynamics create complexity and uncertainty, preventing investment despite good-faith engagement. Finally, CDS enables membrane viability. Universal coverage providing clean feedstock (quality factor 1.0 versus 0.4 kerbside) is prerequisite to achieving 85% recovery rates.

Policy recommendations emphasize a four-phase implementation. Phase 1 (2025–2027) establishes EPR foundations ($300–$400/t fees), CDS universality (100% by 2027), and the elimination of mandate development (80/15/5 classification). Phase 2 (2028–2032) implements progressive elimination (40% by 2028, 60% by 2030, 80% by 2032) with mono-material redesign requirements and reformulation support. Phase 3 (2028–2035) deploys targeted membrane (5% essential formats only) with 50% capital subsidies, processing loss coverage, and 40% cost reduction R&D, while optimizing mechanical capacity (600,000 t/year rigid plastics at 80%+ utilization). Phase 4 (2035–2040) evaluates performance against revised targets (65% recovery of remaining packaging, 70%–75% effective including elimination), and it adjusts based on technology development outcomes.

Theoretical contributions include: eight-loop feedback framework for circular economy system analysis capturing reinforcing growth dynamics (quality premiums, scale economies, and feedstock quality) and balancing constraints (feedstock competition, elimination trade-offs, policy uncertainty, diminishing returns, and rebound effects); demonstration that material heterogeneity requires differentiated strategies rather than homogeneous treatment; evidence that elimination constitutes foundational rather than supplementary intervention for economically unviable materials; identifying voluntary-approach structural failure mechanisms operating through policy uncertainty loops.

Future research priorities include applying the framework to other jurisdictions to test generalizability and identify context-specific parameters; disaggregating to polymer-application level, enabling precision policy targeting; incorporating consumer behavior modules that capture social dynamics and acceptance; adding political economy structures that represent industry lobbying and regulatory capture; expanding technology portfolios, including enzymatic depolymerization and hybrid mechanical–chemical pathways; integrating comprehensive lifecycle assessment to quantify full environmental impacts across elimination versus processing alternatives.

Australia now faces a stark choice: continue voluntary approaches that have failed systematically for 8 years, generating only a four percentage-point improvement despite $690 million of investment and ongoing industry commitment, or adopt strategic elimination-first pathways that model results have demonstrated will achieve targets at substantially lower cost while providing superior environmental outcomes. The evidence is unequivocal. Universal processing cannot work; fundamental physical and economic constraints prevent success, regardless of funding levels or political will. Strategic elimination combined with focused membrane deployment for truly essential applications represents the only viable pathway to ambitious circular economy goals. The time for evidence-based strategic action is now.