M1 Endogenous demand

Topic

The model analyzes automation via artificial intelligence by simultaneously considering firms' substitution decisions and their consequences for the economy as a whole. Each sector consists of various roles or occupations, characterized by an initial workforce size and the fraction of workers potentially replaceable by AI. Since the workforce size represents actual people, the maximum number of replaceable workers is necessarily treated as an integer, with any fractional results from calculations being discarded. The economic attractiveness of automating a given role is determined by comparing the cost of AI implementation against the avoidable wage costs, thereby yielding an investment payback period for each replaced worker.

Unlike a framework based on a fixed investment criterion, each sector's tolerance regarding the payback period evolves throughout the simulation. As long as automation remains economically advantageous for the sector, this threshold progressively loosens. Consequently, roles yielding the greatest savings or the fastest investment recovery are automated first, while those offering less economic benefit enter the process only once the sector has sufficiently increased its willingness to automate. The sequence thus emerges from the economic conditions themselves, without the need to externally establish a ranking of occupations.

This process also incorporates an endogenous brake. The economic outcome achieved by a sector undergoing automation is continuously compared against the outcome that would have been realized had nothing changed. As long as the transformation remains advantageous, it can continue to progress. If the net result falls below that baseline scenario, the threshold stops loosening, and the sector's automation halts. This brake is considered absorbing, meaning the process does not spontaneously restart in subsequent periods.

Cumulative adoption translates into an effective, whole number of replaced workers. In each period, the number of newly displaced individuals is identified, the remaining workforce is updated, and calculations are performed for both ongoing wage payments and the costs associated with AI usage. These calculations are performed individually for the various sectors, maintaining the relationship between occupational structure, employment, and costs.

Finally, the model incorporates the interaction between these sectoral decisions and the economy as a whole. Labor income generated across all sectors determines aggregate consumption capacity andbased on structural sharesthe demand directed toward each sector. Consequently, a reduction in employment in one sector does not merely affect its own activity; by lowering total labor income, it also reduces the demand received by other sectors.

This closes the loop linking automation, employment, income, demand, and business outcomes. For each sector, the results obtained with AI can be compared against those of a scenario where the original structure remains unchanged. At the same time, aggregating the results makes it possible to observe the impact on the entire system through indicators such as total employment, the proportion of workers displaced by automation, and an estimate of the economic output generated.

ID:5049