Behavioral Model Links Waste Bank Retention to Landfill Pressure

A study in Balikpapan, Indonesia, found that household satisfaction with Waste Bank services was more strongly associated with continued depositing than attitudes, perceived control, or peer influence were with current participation. Researchers linked a cross-sectional survey of 298 households to a 20-year system dynamics simulation, finding that service-centered interventions produced much larger modeled participation gains than recruitment alone. Even the strongest integrated scenario, however, diverted only 3.61% of municipal waste and delayed the modeled landfill-capacity zero crossing by about 3.23 days. The results point to a clear scale limitation: making community recycling more reliable can sustain participation, but it cannot by itself relieve the city’s wider waste and landfill pressures.

The research was an exploratory behavioral–system dynamics study rather than a field trial or a historically validated forecast. It combined two logistic-regression models with a stock-and-flow simulation of household participation, recyclable waste diversion, municipal landfill inflow, and remaining landfill capacity. The investigators surveyed households in Kelurahan Sepinggan from September to October 2025, then translated estimated engagement and retention probabilities into transitions among non-participants, active participants, and dormant participants. The simulation used 2025 as its base year and compared five policy scenarios over 20 years, with 1,000 Monte Carlo runs per scenario to test whether the ranking of policy outcomes remained stable under simultaneous uncertainty.

The survey used non-probability convenience sampling among residents accessible to enumerators in the Waste Bank service area. Of 332 completed questionnaires, 34 were excluded because respondents lived outside Sepinggan, leaving 298 analytical cases. Among them, 122 households, or 40.94%, reported current active Waste Bank membership, while 176 were classified as not currently active. Within the active subgroup, 59 households deposited recyclable waste at least once a month, whereas 63 deposited less frequently and were treated as operationally dormant for the model. This last category was not an observed dropout: the cross-sectional survey did not follow households over time, so it could not directly measure the process of becoming inactive or later returning.

The engagement model examined whether current membership could be distinguished by attitude, perceived behavioral control, peer influence, and an exploratory formal normative index. It performed poorly. Its apparent area under the receiver-operating-characteristic curve was 0.561, with a bootstrap 95% confidence interval of 0.493 to 0.631, and repeated cross-validation produced a mean AUC of 0.513. Averaged out-of-fold predictions yielded an AUC of 0.491, indicating essentially no discrimination in that validation assessment. Attitude, perceived behavioral control, and peer influence were not significantly associated with active engagement. The formal normative index had a negative conditional association with membership, with an odds ratio of 0.691 and a 95% confidence interval of 0.530 to 0.901, but the investigators treated this as exploratory because the construct was respecified after the original normative indicators failed a unidimensionality assessment.

The evidence for continued participation was stronger. Among the 122 active households, a one-standard-deviation increase in service satisfaction was associated with 2.71 times the odds of depositing at least monthly, with a 95% confidence interval of 1.524 to 4.816 and p less than 0.001. This model had an apparent AUC of 0.715, a bootstrap 95% confidence interval of 0.633 to 0.795, and a mean repeated-cross-validation AUC of 0.716, although its estimates remained based on a relatively small subgroup. Because the survey was cross-sectional, the association does not prove that improving service will cause more frequent depositing or that satisfaction alone determines retention. It nevertheless provided the study’s clearest behavioral signal and supported separating recruitment from the maintenance of regular use.

The simulation translated those associations into a closed population of 21,630 residents, preserving transfers among non-participants, active participants, and dormant participants. Behavioral states adjusted gradually toward intervention-defined targets, while the proportion of active participants generated visibility that fed back into peer influence and perceived behavioral control. These feedback links were theoretical assumptions built into the model, not effects estimated from longitudinal observations. Cross-sectional probabilities were also converted into annual flows through a scaling parameter that was not independently estimated. Five scenarios tested existing conditions, recruitment alone, service retention alone, moderate integrated intervention, and a high-intensity exploratory package, with normalized inputs intended to represent relative policy strength rather than budgets, population coverage, or guaranteed treatment effects.

In the deterministic simulations, baseline conditions ended with 945 active participants, 3,755 tons of annual waste reduction, and a landfill capacity balance of negative 3,279,660 tons. The zero crossing—the point at which modeled capacity is exhausted—occurred in year 3.059. Recruitment alone produced only 1,018 active participants and delayed the crossing by 0.21 day. Service improvement was far more influential, producing 1,672 active participants and a delay of 1.31 days. The moderate integrated strategy generated 1,782 active participants, an 88.52% increase over baseline, while raising all intervention targets from 0.3 to 0.8 added only 12 more participants. That pattern indicates diminishing incremental returns in the model, but it does not establish cost-effectiveness because costs and dose-response relationships were not estimated.

Uncertainty analysis preserved the broad policy ordering while exposing substantial overlap in landfill outcomes. Median year-20 active participation rose from 945 at baseline to 1,008 with recruitment, 1,662 with service improvement, 1,730 with moderate integration, and 1,792 with high-intensity integration. The corresponding median share of the modeled population was only 8.28% under the strongest scenario. Median landfill balances ranged from negative 3.741 million tons at baseline to negative 3.593 million tons under the high-intensity package, but the central 95% simulation regions overlapped extensively. Median zero crossings ranged from 2.993 to 3.042 years, and the boundary trajectories crossed zero before year 4. These are conditional simulation regions based on assigned distributions, not confidence intervals for real future depletion dates, and none of the behavioral scenarios prevented early exhaustion under the modeled conditions.

The study’s most consequential result is the disconnect between program-level success and city-scale impact. The high-intensity integrated scenario nearly doubled active participation relative to baseline, yet its annual waste reduction reached only 7,128 tons, or 3.61% of municipal waste generation. Its year-20 capacity deficit improved by about 1.60%, and the landfill zero crossing moved by roughly three days. The findings therefore support treating Waste Banks as one element of a broader diversion system involving wider household coverage, organic-waste management, recycling capacity, treatment infrastructure, and measures addressing growth in municipal waste. They also favor operational measures such as dependable collection schedules, clear communication, accurate transaction records, convenient depositing arrangements, and follow-up with dormant participants, while emphasizing that the modeled intervention values are not directly translatable into budgets or coverage targets.

Several limitations constrain interpretation. The convenience sample came from one urban area and was not stratified by membership status, so it cannot be assumed to represent households across Balikpapan. Household survey results were scaled to a population model built from resident counts, creating an additional aggregation assumption. Adjustment times, engagement conversion, intervention intensity, population scaling, waste parameters, and initial landfill capacity were estimated, calibrated, or assigned rather than learned from a longitudinal record of participation and waste flows. No independent historical calibration or validation was possible, and formal institutional ethics approval was not obtained before data collection, although trained enumerators explained the study, confidentiality arrangements, voluntary participation, and the right to withdraw, and respondents provided affirmative consent. The model is consequently best used to compare mechanisms and scales, not to issue an exact forecast of landfill life or a proven implementation package. Its defensible conclusion is narrower but still useful: service satisfaction was the strongest measured signal for sustained depositing, and behavioral improvement alone was too small to alter the projected timing of Balikpapan’s landfill exhaustion materially.

Subject of Research: Cross-sectional household behavior linked to a 20-year urban waste-management system dynamics model in Balikpapan, Indonesia

Article Title: Integrating the Theory of Planned Behavior and System Dynamics to Evaluate Urban Waste Management Policies

Article References: Setiadi, E. P., & Ho, H.-C. (2026). Integrating the Theory of Planned Behavior and System Dynamics to Evaluate Urban Waste Management Policies. Cleaner Engineering and Technology, Article 101342. https://doi.org/10.1016/j.clet.2026.101342

Image Credits: AI Generated

DOI: 10.1016/j.clet.2026.101342

Keywords: Waste Banks, system dynamics, Theory of Planned Behavior, household recycling, landfill capacity, service satisfaction, behavioral modeling, urban waste, Monte Carlo uncertainty, Balikpapan, Integrating, Theory