This article (purchase required) uses data mining of nearly 72,000 official food inspections from China from 2018 – 2023. It tests the hypothesis that data manipulation by local food inspection agencies has led to an overall underestimate of food fraud and food safety incidents in China.
The authors examined the distribution of non-compliant samples near the qualified standard value using exceedance multiples. To quantify the extent of data manipulation, they used an exhaustive algorithm to construct counterfactual estimates.
They report an abnormal distribution of unqualified samples near standard value, indicating potential data manipulation. Robustness tests supported this inference.
They conclude that over 11% of unqualified (failed) samples may have been adjusted to qualified status during 2018–2023, with higher manipulation rates in eastern regions than in central and western regions. The manipulation rate of unqualified samples across 25 sample provinces ranged from 8.13% to 16.30%.
Comments