Get started
Work

Accounting State Space as a Foundation for Economic Agent-Based Modeling: An Input–Output Benchmark and Inventory Extension

Year: 2026
Type: preprint
Abstract: This paper proposes Accounting-State Agent-Based Modeling (AS-ABM), a framework in which each agent maintains a double-entry bookkeeping state space formalized through Exchange Algebra and interacts with other agents through stage-ordered transactions. Because double-entry consistency constraints are inherently entity-specific, the agent-based architecture enforces them by construction at the transaction level rather than through ex post aggregate checks. As a minimal demonstration, we construct an input–output (IO) transaction network using synthetic input-coefficient matrices, treating the IO coefficients as controlled experimental parameters. We show that the framework exactly reproduces the Leontief multiplier-effect benchmark under unconstrained conditions, thereby establishing the equivalence of our dynamic cumulative-difference measure with the static multiplier. When inventory dynamics and input constraints are introduced, the cumulative multiplier becomes inventory-dependent: binding input constraints suppress propagation, and larger safety stocks relax these constraints so that the multiplier recovers toward—but does not exceed—the unconstrained benchmark. The replenishment rule generates overshoot–undershoot dynamics mechanically related to the bullwhip effect; however, across 99 independently generated IO structures these dynamics are highly seed-dependent, and we find no statistically significant amplification of fluctuations with system size. The results suggest that accounting state space provides a workable foundation for constructing dynamic economic models that are simultaneously accounting-consistent, mechanism-based, and extensible.
DOI: 10.21203/rs.3.rs-8485050/v2
Author: Kaya Akagi
Language: English
Open Access status: green
OpenAlex ID: W7204537820
Cites:
Cited by:
Related to: 10