Delay-consistent adaptive incentive governance for circular beverage packaging systems: distinguishing structural and algorithmic limits in PET and returnable glass recovery in Mexico

Mexico's 2030 target of 80% PET-container recovery under the ANNEP agreement remains 16 percentage points above the current national recovery rate, with returnable glass already operating near its infrastructure asymptote. Whether adaptive incentive governance can close this gap depends on distinguishing the mechanism sources of policy performance: the adaptive feedback timing itself, the average incentive intensity at which policies operate, or the underlying behavioral response of the recovery system. This paper investigates the distinction using a delay-consistent adaptive proportional–integral controller embedded in a system-dynamics closed-loop supply chain calibrated against four independent empirical sources for the Mexican beverage-container context. Four policies are compared in a 216-cell balanced factorial design: a no-incentive baseline, a constant deposit-refund, an adaptive proportional–integral controller, and a quarterly threshold rule. A new stratified matched-budget identification experiment decomposes policy performance into three analytically distinct levels — algorithmic effects, operating-point effects, and structural constraints — by comparing the adaptive controller against an equivalent-intensity fixed benchmark across baseline, demand-dip, and demand-whiplash regimes. The matched-budget analysis indicates that the adaptive controller and the equivalent-intensity fixed policy converge to essentially identical steady-state return rates (approximately 0.759) under all three demand regimes; the previously observed adaptive advantage is attributable to operating-point effects rather than to feedback timing. This identification supports interpreting the 2.1-percentage-point gap between the observed ceiling and the 0.78 composite target as structural rather than algorithmic in origin: the ceiling reflects the saturating Michaelis–Menten behavioral response of the recovery system, not a limitation of the control algorithm. Closing this gap requires policy instruments capable of expanding the behavioral elasticity envelope itself — reverse-vending densification, extended producer responsibility, informal-sector formalization, and consumer information campaigns — rather than stronger deposit-refund incentives alone. • Matched-budget identification experiment isolates algorithmic from operating-point effects in deposit-refund governance. • Adaptive controller and equivalent-intensity fixed policy converge to identical return rates across three demand regimes. • The 0.759 composite return-rate ceiling is structural, not algorithmic — reflecting behavioral elasticity limits. • Structural ceiling stops recovery 2.1 percentage points below Mexico's 0.78 composite target under current elasticities. • Closing the gap requires elasticity-envelope expansion, not stronger deposit-refund incentives.