Replication Package for "The Tariff Tax Falls Hardest on the Middle: Effective Tariff Burdens on Lower-Middle-Income Economies under the U.S. Trade Wars, 2017–2026"

This replication package accompanies the article “The Tariff Tax Falls Hardest on the Middle: Effective Tariff Burdens on Lower-Middle-Income Economies under the U.S. Trade Wars, 2017–2026” by Ismet Ismatullah (Universitas Muhammadiyah Sukabumi). The article measures the effective U.S. tariff rate (calculated duties divided by customs value) for 234 trading partners over 116 months, from January 2017 to August 2026. It compares the rate across World Bank income groups through the 2018–2019 and 2025–2026 trade-war waves and the February 2026 Supreme Court ruling on IEEPA tariffs. Contents Replication_Tariff_Tax.ipynb is the reference implementation, a Python/Jupyter notebook. It reproduces Tables 2–8 and Figures 1–3 of the article and Appendix Tables A2–A13 from the panel workbook. The notebook ends with an automatic check of 24 key estimates against the published values, and exports all tables (Excel) and figures (PNG). Panel_Tarif_Efektif_2017_2026.xlsx is the processed partner-month panel: 234 partners × 116 months, 27,144 rows. Its sheets are: Panel_Negara_Bulan: imports, calculated duties, the effective rate, the World Bank income group in force each month, and the period indicators Post2018, Post2025 and PostRuling. Ringkasan_Kelompok, Ringkasan_Negara and Tren_Bulanan: summary tables built from formulas. CPI_Bulanan and CPI_Metadata: ten U.S. CPI series, not used in the article. Ref_Negara: country reference list and policy-date parameters. Reklasifikasi: the 51 partners that change income group. Catatan_Data: data notes. Sheet names and notes are in Indonesian. A guide is on the “Panduan” sheet. Statistical_Appendix.pdf / .docx covers: Variable definitions and summary statistics. Imports, duties and effective rates by income group and year. Reclassifications, additional hump-shape specifications, and difference-in-differences with the 2018 wave. Event-study coefficients and parallel-trends tests. The decomposition method, the 2025 cross-section, low-income preference episodes, and ASEAN partners. Replication notes (Tables A1–A15). tariff_tax_analysis.do is a Stata version of the same specifications, for Stata users (Stata 15+, with reghdfe, ftools and estout). It also contains two optional sections that run once additional files are supplied: GNI per capita (WDI) and HS chapters 50–67 duties (USITC DataWeb). The Python notebook is the verified reference implementation. The do-file has not been run in Stata by the author. Data_Mentah_HSSC.zip holds the raw source downloads: USITC DataWeb (2017–2022 and 2023–2026), FRED and BLS CPI series, and World Bank OGHIST. Data sources U.S. International Trade Commission, DataWeb: imports for consumption, customs value and calculated duties by country, monthly. Retrieved 13 and 15 October 2026. https://dataweb.usitc.gov World Bank, Historical Country Income Classifications (OGHIST), FY89–FY27. Retrieved 8 October 2026. U.S. Bureau of Labor Statistics CPI series, via FRED (https://fred.stlouisfed.org) and BLS (https://data.bls.gov). Retrieved 13–15 October 2026. Processing notes Partner-months with zero imports (733 rows) keep a value of zero and a missing rate, not a rate of zero. Income groups follow the World Bank fiscal year (July–June). Twenty-six territories and economies lack a classification in some or all months. Gaza Strip and West Bank both carry the ISO3 code PSE, so use the country name as the unique key. BLS did not publish CPI data for October 2025. Those values are imputed as the mean of September and November 2025. The processed panel was checked cell by cell against the raw DataWeb downloads, and the income groups were checked against OGHIST. How to replicate (Google Colab, free, no installation) Download Replication_Tariff_Tax.ipynb and Panel_Tarif_Efektif_2017_2026.xlsx from this record. Open https://colab.research.google.com, choose Upload, and select the notebook. Select Runtime → Run all. When the “Choose Files” button appears, select the panel workbook. If Colab asks to restart the session after installing packages, restart and select Run all again. The run takes about 2–4 minutes. Step 12 should report “24 of 24 checks OK”. All tables and figures are downloaded as replication_output.zip. The notebook also runs in any local Jupyter environment with Python 3.10 or later. Place the workbook in the same folder as the notebook. The estimation package is pinned to pyfixest 0.60.0 so that results do not change with future package versions. Estimates are weighted least squares with fixed effects and standard errors clustered by partner. Clustered standard errors computed with other software, including the Stata do-file, may differ in the second decimal because of small-sample corrections. CitationPlease cite the article, and this package using its DOI.