Key Takeaways
- • Li Yuankun, an Assistant Research Fellow at CASS's Institute of World Economics and Politics, argues that the OECD's MAGIC database inflates China's measured subsidy gap by using China's Loan Prime Rate (LPR) as the benchmark for below-market borrowing (BMB).
- • In a benchmark-sensitivity test on Chinese A-share listed firms from 2019 to 2024, the LPR-based BMB measure was 16.68 times the China government-bond-based estimate and 5.65 times the SHIBOR-based estimate in the Top-147 large-firm sample.
- • The OECD's 2026 MAGIC report finds Chinese firms received three to eight times more government support than OECD-based firms between 2005 and 2024, with BMB driving much of the gap.
- • The OECD attributes almost 60 per cent of Chinese firms' global market-share gains to subsidies, a model-based figure that Li argues is also sensitive to the benchmark choice.
- • Li calls for transparent benchmark rates, sample boundaries and sensitivity tests before subsidy estimates inform trade measures.
Summary
Chinese economists and trade scholars are challenging the methodology of the OECD's MAGIC Database of Industrial Subsidies, released on 1 June 2026, which estimates government support to large manufacturing firms worldwide. According to a VoxChina.org article originally published on 1 July 2026, Li Yuankun of the Chinese Academy of Social Sciences focuses on the interest-rate benchmark used to estimate below-market borrowing. Other critics include Zhu He and Guo Kai of the China Finance 40 Forum (CF40), who questioned the OECD's subsidy-centric explanation of China's emerging industries, and Cui Fan of the University of International Business and Economics, who criticized the database's methodology and research agenda.
The OECD's 2026 report separates industrial support into government grants, corporate income-tax concessions and below-market borrowing. It reports that Chinese firms received three to eight times more support than OECD-based firms between 2005 and 2024, and that almost 60 per cent of Chinese firms' global market-share gains can be explained by subsidies. The OECD states that its results do not prejudge WTO or countervailing-duty assessments. Li argues that the headline comparison depends heavily on BMB, which is a constructed estimate comparing a firm's borrowing cost with a counterfactual benchmark. Grants and tax concessions are closer to observable fiscal support.
Li's central concern is comparability. The OECD's 2024 industrial-subsidy work compares China's benchmark lending rates, which refer to the LPR after 2019, with interbank rates for OECD banks. Li argues the LPR is a lending-rate reference compiled by the National Interbank Funding Center, not an interbank funding cost or low-risk sovereign benchmark. He cites People's Bank of China data showing that in March 2026, 48.81 per cent of new renminbi loans were priced below the one-year LPR of 3.0 per cent, suggesting that treating below-LPR pricing as below-market borrowing is too broad a signal.
Because the OECD's firm-level inputs are not public, Li conducted a narrower test holding firms, debt proxy and simplified formula constant while changing only the base rate, using Chinese A-share listed firms from 2019 to 2024. In the Top-147 large-firm sample, the LPR-based BMB estimate was 16.68 times the government-bond-based estimate and 5.65 times the SHIBOR-based estimate, rising to 17.89 and 7.38 times respectively in 2024 alone. Li acknowledges these figures do not replicate the OECD's full method, which includes credit-risk spreads, guarantee adjustments and non-listed firms, but argues the base-rate choice alone can magnify measured BMB several-fold.
Li also notes that MAGIC covers only three firm-level instruments, while industrial policy also includes procurement rules, tariffs, local-content requirements and regulatory preferences. He argues the almost-60-per-cent market-share attribution is a model-based estimate that should not be treated as a standalone fact. He concludes that benchmark rates, sample boundaries and sensitivity tests should be made transparent before subsidy estimates shape trade measures.
The OECD's 2026 report separates industrial support into government grants, corporate income-tax concessions and below-market borrowing. It reports that Chinese firms received three to eight times more support than OECD-based firms between 2005 and 2024, and that almost 60 per cent of Chinese firms' global market-share gains can be explained by subsidies. The OECD states that its results do not prejudge WTO or countervailing-duty assessments. Li argues that the headline comparison depends heavily on BMB, which is a constructed estimate comparing a firm's borrowing cost with a counterfactual benchmark. Grants and tax concessions are closer to observable fiscal support.
Li's central concern is comparability. The OECD's 2024 industrial-subsidy work compares China's benchmark lending rates, which refer to the LPR after 2019, with interbank rates for OECD banks. Li argues the LPR is a lending-rate reference compiled by the National Interbank Funding Center, not an interbank funding cost or low-risk sovereign benchmark. He cites People's Bank of China data showing that in March 2026, 48.81 per cent of new renminbi loans were priced below the one-year LPR of 3.0 per cent, suggesting that treating below-LPR pricing as below-market borrowing is too broad a signal.
Because the OECD's firm-level inputs are not public, Li conducted a narrower test holding firms, debt proxy and simplified formula constant while changing only the base rate, using Chinese A-share listed firms from 2019 to 2024. In the Top-147 large-firm sample, the LPR-based BMB estimate was 16.68 times the government-bond-based estimate and 5.65 times the SHIBOR-based estimate, rising to 17.89 and 7.38 times respectively in 2024 alone. Li acknowledges these figures do not replicate the OECD's full method, which includes credit-risk spreads, guarantee adjustments and non-listed firms, but argues the base-rate choice alone can magnify measured BMB several-fold.
Li also notes that MAGIC covers only three firm-level instruments, while industrial policy also includes procurement rules, tariffs, local-content requirements and regulatory preferences. He argues the almost-60-per-cent market-share attribution is a model-based estimate that should not be treated as a standalone fact. He concludes that benchmark rates, sample boundaries and sensitivity tests should be made transparent before subsidy estimates shape trade measures.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.