Production data at HS6 level from Solleder, Silvy, and Olarreaga (2024)

Description

This dataset contains the data from the paper "Protection for sale without aggregation bias" (Solleder, Silvy, and Olarreaga, 2024).

Production data indicated by "Estimates" in the source field of the dataset were estimated using the Regression-Enhanced Random Forest (RERF) algorithm developed by Zhang et al. (2017). It allows us to predict production data trained and tested on the Prodcom dataset, containing mainly European countries, for which production data is available at a high level of disaggregation with a correspondence to the HS six-digit classification. For more information on the procedure, please take a look at Solleder, Silvy, and Olarreaga (2024).

Production data indicated by "FAO" in the source field of the dataset is sourced from FAO and has been converted by the authors to 6-digit HS codes.

The dta file has been created with STATA 17. The csv file is a comma-separated value file. The separator is ',', and the first row is variable names. The content is the same in both files. Variables are:

• country: ISO 3166 3-character country codes, string; 
• year: years (1999-2015), numeric;
• commoditycode: product  6-digit HS codes in HS revision 1992 (H0), numeric;
• production: production in country - year in 1000 USD, numeric;
• source: "Estimates" / "FAO" (see above) 

References

Solleder, J, F Silvy and M Olarreaga, 2024. Protection for sale without aggregation bias, CEPR Discussion Paper No. 19418. CEPR Press, Paris & London. https://cepr.org/publications/dp19418

Zhang, Haozhe, Dan Nettleton and Zhengyuan Zhu, 2017. Regression-Enhanced Random Forests. JSM Proceedings, Section on Statistical Learning and Data Science, Alexandria, 636-647. https://doi.org/10.48550/arXiv.1904.10416