• 15 October 2025
    • preprint
    • Published by HAL in HAL
Abstract
This study presents a decision-support tool designed for the regulation of food fraud, based on on hybridizing of four approaches: i) an economic model inspired by the Chinese milk crisis, opposing a food firm and a public authority. The model integrates the main hypotheses of food fraud mechanisms, considering organoleptic standards, production and compliance costs, the firm's ability to substitute the required nutrient with a less expensive chemical substance, as well as the cost of acquiring that substance. It also considers the probability of a health crisis occurring, the financial penalty stipulated by regulations, the overall health risk, and the expected profits based on the firm's different strategic choices; ii) A sequential game where the authority first chooses the nutritional standard and the fine applicable in case of a health crisis to minimize health risk, and then the firm determines its optimal compliance level to maximize its expected profit. The game is solved by backward induction; iii) the DCA method to locally solve step 2 of the game, and iv) The ML model (CaatBoost) to estimate the firm's best response function for solving step 1 of the game. The experimental results of the tool demonstrate that the organoleptic standard primarily drives firm's compliance. Strict nutritional standards are effective to reduce health risk only when paired with high fines. Optimal regulation is adaptive: when production capacity, prices, or organoleptic standards are low, maximum standards and fines are required; otherwise,relaxing the standard best limits health risks.
All Related Versions

This publication has 19 references indexed in Scilit: