Global Antidumping Database (Bown, 2016)

This repository shows how to clean the Global Antidumping Database (GAD) and do some summary statistics. Comments on better coding or errors are welcomed. Contact: ianhe2019@ou.edu.

The GAD was published by Chad P. Bown and collects detailed information on the national use of antidumping (AD) investigations in 40 countries and European Union. Fortunately, the database is freely available; unfortunately, Bown claimed that he no longer updates the data. The version I cleaned is version 6.0, updated in June 2016 and containing data through 2015Q4.

Introduction to the AD Investigation

According to the WTO, before successfully imposing an AD duty, the importing country must affirm the existence of two phenomena:

1. the export firm is actually dumping its products;
2. the dumpling behavior is causing injury to the market in the importing country.

The process of investigating the dumping and injury is called antidumping investigation, which is the responsibility of one or two government agencies. A single-track AD investigative process refers to a system in which there is only one government agency carrying out the dumping and injury investigations at the same time, and a dual-track process is a system in which two independent government agencies handle the AD dumping and injury investigations. Most countries follows a single track procedure; the typical countries that follow a dual-track procedure are China and the US.

Introduction to the Database

• The raw data for each AD user country is saved in a Microsoft Excel 97-2003 Worksheet file, within which are typically four spreadsheets:
  • A "Master" sheet collecting case-level data on important dates and decision outcomes. For example, in the US, an AD case usually has 7 important dates, including initiation date, ITC preliminary determination date, DOC preliminary determination date, ITC final determination date, DOC final determination date, AD imposition date, and revoke date. An illustration of time frame for US AD investigations can be found here.
  • A "Products" sheet collecting case-product data on the Harmonized System (HS) product codes listed in the AD investigation petition. The digits of HS code vary across cases and user countries; for example, China's AD cases are usually at 6-digit HS level and the US AD cases are usually at 10-digit HS level.
  • A "Domestic Firms" sheet collecting a set of data on domestic firms that submitted petitions to request the AD investigations, and a "Foreign Firms" sheet collecting a set of data on foreign firms subject to the AD investigation. Here I will skip these two sheets because my focus is on the country-product-level statistics of the AD investigations.

Cleaning the Database

I used Stata to clean up and construct a product-case level dataset. The procedure consists of two steps:

1. Withdraw data from the raw database (in xls format).
2. Clean up data to include only information on AD cases against China. This is just an example; you can follow the coding to easily contruct a dataset with cases against other countries.

I don't clean or reconstruct the dataset when I am withdrawing data; this is the most efficient way to deal with a big dataset, in my opinion.

Note on Withdrawal

• I save 34 raw data files in a folder named "GAD". To run my coding (here), you need change the file path in the do file.
• I use the foreach loop to withdraw data from Excel sheets; however, data on Peru (PER) and Philippines (PHL) cannot be withdrawn succefully if they are included in the loop [I get an error message: r(198)]. That's why I withdraw their data independently.
• I use merge m:1 to merge data from Products and Master sheets. "Products" is in memory (called master file in Stata language) and "Master" is the using file. Note that an AD case may investigate multiple products from the same foreign country.
• The key date variables are stored in "dd/mm/yyyy" or "mm/dd/yyyy" format in the raw database. They include
  • Date of initiation.
  • Dates of preliminary dumping and injury decisions.
  • Dates of final dumping and injury decisions.
  • Dates of imposition of final AD measure.
  • Date of revocation of AD measure.

Note on Construction of Dataset

• My first step in constructing a dataset is always drop those useless variables; especially in a big dataset, this action could reduce Stata running time a lot.
• It's important to clean the country/area names and drop observations without HS code information. These two variables are usually the key variables for merging datasets. Note that the most disaggregated level of HS code we can use to merge datasets is 6-digit; this is why I generate a variable hs06 to save this information.
• Although information on final AD measure imposed is stored in a variable named f_ad_duty, data files for some countries/regions (e.g., Australia, Canada, and Taiwan) save this information in variables with different names.
• The GAD contains data usually from 1980s to 2010s. Some research may not use all years; for example, in my coding I only select the data between 2000 and 2009. In trade research, it's rare to use data with a daily frequency, so I write codes for converting the date variable format from "date" to "year-month" and "year".
• I use a very rude method to drop the repetitive observations (at country-product-initial-year level). I recommend you should find a more reseasonable way to drop those repetitive observations, depending on your research questions.

Summary Statistics

In the part of summary statistics, I contruct two figures and two LaTeX tables:

• A figure presenting number of products over time. The first panel shows number of products under AD investigations, and the number is counted at region-product level and in the year of initiation. The second panel shows number of products receiving final AD duties, and yhe number is counted at region-product level and in the year of imposition.
• A figure presenting number of products across industries. I use 2-digit HS code to proxy for industries.
• A table reporting number of products receiving AD investigations and final AD duties, respectively, across regions. The rate of products receiving affirmative decisions is also reported.
• A table reporting the average length of AD investigations across regions. The type of AD investigative process (single-track or dual-track) is also reported.

Note that I use an external Stata package to export the LaTeX tables: dataout. You can download it by running the following code.

ssc install dataout, replace

A disadvantage of the dataout command is that it automatically adds some LaTeX codes such as \documentclass[]{article} in the preamble (and as of now I do not find a way to prevent this default action). Usually, when you include it in your LaTeX file, please delete all codes except the tabular environment; otherwise, your pdf file may not be produced successfully.