从产品级出口篮子的角度理解中国202425年的前沿（英）

Understanding China’s2024-25 Frontloading from Prepared by Jason Lu and Dimitre Milkov WP/26/13 IMF Working Papers describe research inprogress by the author(s) and are published toelicit comments and to encourage debate.The views expressed in IMF Working Papers arethose of the author(s) and do not necessarily 2026JAN IMFWorking PaperResearch Department Understanding China’s 2024–25 Frontloading from the Lens of Product-Level Export Baskets Prepared byJason Lu and Dimitre Milkov* Authorized for distribution byRafaelPortillo IMF Working Papersdescribe research in progress by the author(s) and are published to elicitcomments and to encourage debate.The views expressed in IMF Working Papers are those of the ABSTRACT:A striking feature of US-China trade tensions in mid-2025 is China’s acceleration of exports to theUS ahead of new tariff increases, a phenomenon we term export frontloading. To understand how this wasachieved, we develop a factor model analytical framework to characterize China’s product-level exports, acrosstime and destinations, according to a set of latent export baskets. Applying this to data from China’s GeneralAdministration of Customs, we document the channels behind the 2024-25 episode and compare them with the2018 US-China trade tensions. Our analysis points to broad-based adjustments across multiple dimensionsina manner not observed in 2018: (i) shipments to the US accelerated in the second half of 2024,possibly RECOMMENDED CITATION:Lu, Jason, and Dimitre Milkov. 2026. “Understanding China’s 2024–25 ExportFrontloading from the Lens of Product-Level Export Baskets.” IMF Working Paper. WORKING PAPERS Understanding China’s 2024–25Frontloading from the Lens of Prepared byJason Lu and Dimitre Milkov Understanding China’s 2024–25 Frontloading from the Lens of Jason Lu∗Dimitre MilkovFirst draft: July 22, 2025This version: January 21, 2026 Abstract A striking feature of US-China trade tensions in mid-2025 is China’s acceleration of exportsto the US ahead of new tariff increases, a phenomenon we term export frontloading. To under-stand how this was achieved, we develop a factor model analytical framework to characterizeChina’s product-level exports, across time and destinations, according to a set of latent exportbaskets. Applying this to data from China’s General Administration of Customs, we documentthe channels behind the 2024-25 episode and compare them with the 2018 US-China trade ten-sions. Our analysis points to broad-based adjustments across multiple dimensions in a mannernot observed in 2018:(i) shipments to the US accelerated in the second half of 2024, possi- Keywords:Export Frontloading, Trade Tariffs, Production Relocation, Intertemporal Realloca- 1Introduction China’s exports to the United States (US) have faced a renewed wave of tariff increases since early2025, with the possibility of these increases anticipated by firms and other market participants. Inprinciple, higher tariffs reduce the competitiveness of Chinese goods in the US market, leading to a these adverse effects can be temporarily mitigated by export frontloading to the US before the Export frontloading was evident in aggregate trade data showing a marked increase in Chineseexports to the US during the final months of 2024. Nominal export flows in December 2024 were6.6 billion USD higher than a year earlier, an increase of 15.6 percent, and a notable acceleration The objective of this paper is to understand how frontloading was achieved by identifying thechannels that facilitated this adjustment over multiple months, and to contrast these dynamics withthose observed during the 2018 US-China trade tensions. In particular, our goal is to disentangle the At the core of our analysis is the novel application of factor models to identify latent exportbaskets at the product level to capture the dynamics of China’s export flows over time and acrossdestinations. At a practical level, the factor model provides a means of dimensionality reduction, Specifically, our framework represents the panel of export flows across product categories anddestinations at each point in time using a factor model, where each factor corresponds to a latentproduct-level export basket.For each factor, the loadings describe the composition of the corre- sponding export basket, that is, the specific weights across all product categories in the basket, Our representation of the export flows thus far consists of a few latent export baskets, eachwith specific weights across product categories, and the value of each basket that each destinationreceives in every period. At any given point in time, the value of exports from a specific basket toa particular destination can be alternatively expressed as the product of two components: (i) the Leveraging this representation, we decompose the dynamics in each basket’s exports acrossdestinations over time into two components: changes in the total amount of exports of a specificbasket to