Ten papers, reproduced with visible SQL
Every analytical page on TradeWeave is a replication in spirit. This gallery makes it explicit: pick a well-known trade-economics paper, reproduce its headline statistic or figure using the Parquet data on this site, and show the numerical comparison side by side. Each page carries the original citation, the query, the figure, the numbers, and an honest note on why any divergence exists.
Reading these replications
Three things about the input side shape every card below. The workbench uses annual BACI at the HS6 level, and pulls gravity regressors from the public CEPII Gravity release. Annual aggregation washes out the monthly or quarterly variation that several of the original papers exploited (Amiti, Redding & Weinstein 2019 work at the monthly tariff cut; Autor-Dorn-Hanson 2013 runs on decadal Census commuting-zone panels built from Longitudinal Business Database plants). HS6 is the finest resolution that BACI reconciles, so we do not see the tariff-line heterogeneity that HS8 or HS10 national schedules expose, and we cannot follow individual firms the way Melitz-style papers do (Bernard et al. 2003 AER uses the Linked Longitudinal Firm Trade Transactions Database). The CEPII Gravity file gives bilateral distance, contiguity, common language, colonial history, and RTA dummies, but not the proprietary firm-level or customs-level micro that Anderson-van Wincoop (2003), Eaton-Kortum (2002) and Caliendo-Parro (2015) used for their structural estimations. The replications below therefore reproduce the qualitative shape and the order of magnitudeof the headline result; they do not, and cannot, reproduce point estimates to two decimal places. Where the gap is meaningful (ADH most obviously), the card says so.
Cards
Hidalgo & Hausmann, The Building Blocks of Economic Complexity
Schott, The Relative Sophistication of Chinese Exports
Autor, Dorn & Hanson, The China Syndrome: Local Labor Market Effects of Import Competition
Silva & Tenreyro, The Log of Gravity
Feenstra, New Product Varieties and the Measurement of International Prices
Eaton & Kortum, Technology, Geography, and Trade
Caliendo & Parro, Estimates of the Trade and Welfare Effects of NAFTA
Melitz, The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity
Anderson & van Wincoop, Gravity with Gravitas (the US-Canada border puzzle)
Hausmann, Hwang & Rodrik, What You Export Matters
Comparison table
Side-by-side of each paper’s published headline coefficient and the workbench replication estimate, at a glance. The verdict column is deliberately coarse: matchesmeans the sign and magnitude fall within what the original sample variation would admit; close means the qualitative shape survives with a material but defensible numerical gap; approximation flags cases where BACI / public-CEPII-Gravity cannot reproduce the proprietary-micro specification, and the page is best read as a sanity check rather than a reproduction.
| Paper | Published coefficient / claim | Workbench estimate | Verdict |
|---|---|---|---|
| Hidalgo & Hausmann PNAS 106(26): 10570–10575 | r(ECI, log GDPpc) ≈ 0.78 (cross-section of 128 countries, 1985 baseline basket) | r = 0.758 on 201 country-rows, 2022 cross-section, slope ln(GDPpc) on ECI = 1.09 | matches |
| Schott Economic Policy 23(53): 5–49 | Chinese export basket moving closer to OECD basket 1990–2005 (sophistication catch-up) | Finger-Kreinin(CHN, OECD-38) over HS6 rises 0.334 (1995) → 0.479 (2024), monotone | close |
| Autor, Dorn & Hanson American Economic Review 103(6): 2121–2168 | US commuting-zone ΔIPW → 0.55 percentage-points of working-age employment lost per $1,000 per worker of rising Chinese exposure 1990–2007 | Not ΔIPW — BACI-feasible proxy: China’s share of country j manufactures imports, OECD-38, 2000 → 2016. USA 9.4% → 25.0%, average 4.1% → 13.4%. | approximation |
| Silva & Tenreyro Review of Economics and Statistics 88(4): 641–658 | OLS distance elasticity ≈ −1.0 to −1.3 (log-linearised Anderson-van Wincoop gravity); PPML distance elasticity roughly half that in absolute value | 2020 BACI-CEPII: OLS ln(X) = −1.20·ln(dist) + controls (n=23,795). PPML: −0.50·ln(dist). Direction and size gap reproduce Silva-Tenreyro. | matches |
| Feenstra American Economic Review 84(1): 157–177 | New varieties enter the trade basket continuously; standard price indices understate true welfare-adjusted prices because they ignore new-good entry. | Variety proxy = distinct (country, HS6) pairs with positive exports. 1995: 369,973 → 2024: 505,935 pairs (+36.7%). Peak 542,622 in 2015. | close |
| Eaton & Kortum Econometrica 70(5): 1741–1779 | Distance-bin coefficients in a Ricardian gravity: −3.10 (0-375 mi), −1.76, −1.09, −0.98, −0.52, 0 (>6000 mi ref). Implied overall distance elasticity ≈ −1.75. | 2020 CEPII Gravity × BACI: raw mean log-trade gaps vs >6000-mi reference are monotone and sharply declining, consistent with the EK pattern (no country fixed effects absorbed). | approximation |
| Caliendo & Parro Review of Economic Studies 82(1): 1–44 | Sectoral trade elasticities (θ) span 0.39 (Other transport) to 51.08 (Petroleum) across 20 ISIC Rev.3 sectors (Table 1). | HS6 Armington |σ| spans ~0.10 to 131.8 across 4,566 products; median 7.89, IQR 5.15–12.88. Section V (minerals) highest, Section XII (footwear) lowest. | close |
| Melitz Econometrica 71(6): 1695–1725 | Firm heterogeneity: productive firms export, unproductive serve domestic only. Trade liberalization reallocates share to the productive tail. | Proxy: across-country variance of log export share in top-100 HS6. Rose from 13.78 (1995) to 22.22 (2024), +61% — consistent with selection getting stronger. | approximation |
| Anderson & van Wincoop American Economic Review 93(1): 170–192 | McCallum (1995) 22× border effect; Anderson-van Wincoop MR correction reduces to ~10.7× (provinces) / 1.5× (states). | Raw size-normalized USA↔CAN intensity ≈ 5.3×-6.2× in 2019 BACI. ~30-36× vs world median pair. MR decomposition flagged but NOT executed (needs intranational trade). | approximation |
| Hausmann, Hwang & Rodrik Journal of Economic Growth 12(1): 1–25 | EXPY (income content of export basket) predicts 10-year growth. Table 2 coefficient on log(EXPY) ≈ +0.05, conditional on initial income. | 1995→2015 cross-section, 185 countries: β on ln(EXPY_1995) = +0.0268 controlling for ln(GDPpc_1995), R² = 0.34. Positive, conditional, about half HHR. | close |
Replicated papers by publication decade
cite
@misc{hossen_2026_figure-1,
author = {Md Deluair Hossen},
title = {Replicated papers by publication decade},
year = {2026},
howpublished = {TradeWeave Workbench},
url = {https://tradeweave.org},
note = {Figure: Figure 1}
}What “replication” means here
Replication in applied economics runs along a spectrum. At one end: re-run the original authors’ code on the original data and obtain identical point estimates (the Chang & Li 2022 sense). At the other end: retell the qualitative claim with fresh data and a similar identification strategy and check whether the sign and rough magnitude survive. These pages live much closer to the second end, because we are using only the public Parquet tables on this site: CEPII BACI, CEPII Gravity, WDI. No proprietary firm-level data, no Census micro-data, no IRS administrative links.
That constraint is honest rather than shameful. The point is to show that you can stand up a rough version of The Log of Gravity in twenty lines of SQL plus seventy lines of iteratively-reweighted least squares; that Hidalgo-Hausmann’s ECI-income correlation is visible in 2022 with the same strength it had in 1985; that Schott’s Chinese-sophistication trajectory continued through 2024. Where our specification diverges sharply from the original paper (ADH, most obviously), the page says so on the first line of the prose, not as a footnote.
What is next
The gallery now covers the canonical gravity tradition (Anderson-van Wincoop 2003, Silva-Tenreyro 2006, Eaton-Kortum 2002), the heterogeneous-firm and variety tradition (Melitz 2003, Feenstra 1994), the complexity/export-quality tradition (Hidalgo-Hausmann 2009, Hausmann-Hwang-Rodrik 2007, Schott 2008), the sectoral-elasticities tradition (Caliendo-Parro 2015), and one causal labour-economics paper (Autor-Dorn-Hanson 2013). Broda-Weinstein (2006) new-variety welfare bias is feasible with HS6 unit-value variance inside each source country and is the next natural addition. Kehoe-Ruhl (2013) extensive-margin growth accounting is also on the short list. The gallery will grow.
Citations on each detail page include DOI, volume, issue, pages, and a BibTeX block. Queries are verbatim — copy, paste, run them yourself at /sql.