Discovering if mexican compensatory duties should be extended beyond China via a data science approach

Abstract

In international trade, when suspicion exist of dumping or subsidy, countries may draw upon compensatory duties. Such duties are intended to alleviate the harm caused to domestic economy. In Mexican case, compensatory duties are aimed at products imported from specific nations since their prices are too low that are considered as menacing for local producers. In most of the cases, compensatory duties are aimed at China. Although Mexican authorities set these duties after a thoughtful process, this research tries to prove that these taxes fall short of scope and are focused on Chinese products only. Every year in Mexico take place around 10 millions of imports and exports, and Mexican government, via the Tax Administration Service (SAT), make the information regarding these operations available to be consulted. That information is analyzed following a data science approach to suggest whether compensatory duties should be extended to other countries or not. By relying on statistical methods and machine learning classification algorithms, the import unitary prices of Chinese products and other countries are studied. If a statistical difference can be estimated or the algorithms can distinguish from charged and non-charged prices, then the data would support the existence of a compensatory duty fixed on China. On the contrary, then the justification of a compensatory duty fixed on China would be at least questionable. It can be advanced that statistical approach, for some products, found counter-intuitive price behaviors supported by classification algorithms, and that not all algorithms are suitable for this matter. Such findings suggest that, for products with counter-intuitive price behaviors, compensatory duties fixed on China are questionable.