# Replication materials for: Intermediation in domestic value added trade within the China-centred and United States-centred subnetworks, 2017–2023

## Overview

This repository contains the data and materials used to reproduce the analyses reported in the article:

**Intermediation in domestic value added trade within the China-centred and United States-centred subnetworks, 2017–2023**

The repository includes:
(1) the original source files downloaded from the Asian Development Bank (ADB) globalization portal,
(2) author-prepared input files used for importing the data into Stata,
(3) Stata code used to process the data,
(4) intermediate and merged output files produced in Stata,
(5) UCINET input files, UCINET network files, and documentation of the UCINET stage of the analysis,
and (6) supplementary result files used for rank-stability and intermediation-concentration analysis.

======================================================================
1. SOURCE DATA
======================================================================

The original data used in this study were downloaded from the Asian Development Bank (ADB) globalization portal:

https://kidb.adb.org/globalization/current

Source files:
- ADB-MRIO-2017_September 2024.xlsx
- ADB-MRIO-2023_September 2024.xlsx

Date of download:
- 21 July 2025

Data scope:
- 73 entities in total: 72 countries and one aggregate entity labeled ROW (Rest of the World)
- 35 sectors per entity
- 5 categories of final use

The original ADB files are stored in:
01_raw_adb/

======================================================================
2. REPOSITORY STRUCTURE
======================================================================

01_raw_adb/
Contains the original ADB source files used in the analysis.

02_input_for_stata/
Contains the author-prepared input files used to import the data into Stata:
- adb2017_country74.csv
- adb2023_country74.csv
- adb_countrylist74.csv

03_stata_code/
Contains the Stata .do files used to process the data.

Full-year execution scripts:
- 01_run_icio_2017.do
- 02_run_icio_2023.do

Grouped processing scripts:
- group1_2017.do to group8_2017.do
- group1_2023.do to group8_2023.do

Append/merge scripts:
- append_groups_2017.do
- append_groups_2023.do

Software metadata:
- software_versions.txt

04_intermediate_outputs/
Contains intermediate grouped Stata outputs and merged Stata output files:
- va_2017_group1.xlsx to va_2017_group8.xlsx
- va_2023_group1.xlsx to va_2023_group8.xlsx
- va_2017_all.dta
- va_2017_all.xlsx
- va_2023_all.dta
- va_2023_all.xlsx

05_ucinet/
Contains UCINET input files, UCINET network files, and documentation:
- va_origin_exporter_dest_2017.xlsx
- va_origin_exporter_dest_2023.xlsx
- PRC_2017.##d
- PRC_2017.##h
- PRC_2023.##d
- PRC_2023.##h
- USA_2017.##d
- USA_2017.##h
- USA_2023.##d
- USA_2023.##h
- README_UCINET.txt

06_supplementary_results/
Contains supplementary input and output files used for additional article-level calculations:
- rank_stability_input.csv
- rank_stability_2017_2023.xlsx
- intermediation_concentration_input.csv
- intermediation_concentration_results.xlsx

docs/
Contains additional repository documentation:
- DATA_INVENTORY.txt
- LICENSE_NOTE.txt
- CITATION.txt

======================================================================
3. PREPARATION OF INPUT FILES FOR STATA
======================================================================

The original ADB matrix data were converted into numeric input files suitable for use with the Stata package and command structure applied in this study.

Original matrix structure:
- 73 entities x 35 sectors = 2,555 rows
- 73 entities x 35 sectors = 2,555 production columns
- 73 entities x 5 final-use categories = 365 final-use columns
- total number of columns = 2,920

Formatting rules for the prepared matrix files:
- file format: CSV
- separator: comma
- encoding: UTF-8 without BOM
- content: numeric values only
- no country names, sector names, headers, or totals inside the matrix body

Preparation steps:
1. Numeric values were copied from the original Excel matrix into a new worksheet.
2. The worksheet was saved as a CSV file.
3. Because the local Excel settings used semicolons as separators, the separators were converted from semicolons (;) to commas (,).
4. Any trailing empty row was removed.
5. The file was saved in UTF-8 encoding without BOM.
6. A separate one-column country/entity list file was prepared for import into Stata. The file contains one country or entity code per row, including ROW and the technical placeholder XXX.

Prepared input files:
- adb2017_country74.csv
- adb2023_country74.csv
- adb_countrylist74.csv

These files are stored in:
02_input_for_stata/

======================================================================
4. TECHNICAL ADJUSTMENT FOR STATA COMPATIBILITY
======================================================================

The Stata command icio_load did not recognize the 73rd entity (ROW) correctly.

To solve this technical problem, an additional artificial placeholder entity labeled XXX was added.

This adjustment involved:
- adding 35 rows of zeros,
- adding 35 production columns of zeros,
- adding 5 final-use columns of zeros,
- adding XXX as the final entry in the entity list.

Final working structure used in Stata:
- 74 entities x 35 sectors = 2,590 rows
- 74 entities x 35 sectors = 2,590 production columns
- 74 entities x 5 final-use categories = 370 final-use columns
- total number of columns = 2,960

Important note:
XXX is a purely technical placeholder introduced only for software compatibility.
It does not represent a real country or economy and contains only zero values.

======================================================================
5. STATA WORKFLOW
======================================================================

The data were processed in Stata/MP 16.0.

The import relied on the command:
icio_load, iciotable(user, userpath("...") tablename("...") countrylistname("..."))

Two alternative execution paths are provided.
Path A: full-year execution
The scripts 01_run_icio_2017.do and 02_run_icio_2023.do run the complete Stata workflow for each year. This path reproduces the full calculation directly, but it is computationally time-consuming.

Path B: grouped execution
For practical reasons, the empirical workflow used by the authors divided the full set of entities into eight processing groups for each year. The grouped scripts are:
- group1_2017.do to group8_2017.do
- group1_2023.do to group8_2023.do

The group outputs were then merged using:
- append_groups_2017.do
- append_groups_2023.do

This grouped path was used to reduce computational burden and manage export limitations.

This produced the merged files:
- va_2017_all.dta
- va_2017_all.xlsx
- va_2023_all.dta
- va_2023_all.xlsx

These outputs are stored in:
04_intermediate_outputs/

======================================================================
6. UCINET WORKFLOW
======================================================================

The Stata outputs were subsequently used in UCINET for network analysis.

The UCINET stage used Excel edge-list files prepared from the Stata outputs:
- va_origin_exporter_dest_2017.xlsx
- va_origin_exporter_dest_2023.xlsx

These files are stored in:
05_ucinet/

The edge lists were imported into UCINET and converted into weighted, directed adjacency matrices for four networks:
- PRC_2017
- PRC_2023
- USA_2017
- USA_2023

The UCINET network files are stored as paired .##h and .##d files in:
05_ucinet/

The UCINET stage was used to compute:
- flow betweenness centrality,
- continuous core–periphery scores,
- Bonacich beta centrality.

Detailed documentation of the UCINET import procedure, settings, network files, and measures is available in:
05_ucinet/README_UCINET.txt

======================================================================
7. SUPPLEMENTARY ARTICLE-LEVEL RESULTS
======================================================================

Additional calculations were performed on the UCINET-derived network measures to support the article-level interpretation.

These supplementary calculations include:
(1) rank-stability analysis between 2017 and 2023,
and (2) intermediation-concentration analysis.

The supplementary input and output files are stored in:
06_supplementary_results/

Files included:
- rank_stability_input.csv
- rank_stability_2017_2023.xlsx
- intermediation_concentration_input.csv
- intermediation_concentration_results.xlsx

The rank-stability analysis compares countries’ positions between 2017 and 2023 using Spearman rank correlation and Kendall’s tau-b for selected network measures.

The intermediation-concentration analysis summarizes the concentration of normalized flow betweenness values using TOP10 share, upper-quartile share, median nFlowBet, and the Herfindahl-Hirschman Index.

Further methodological details are provided in:
05_ucinet/README_UCINET.txt

======================================================================
8. REPRODUCIBILITY NOTE
======================================================================

This repository documents the empirical workflow used in the study and provides the files necessary to reproduce the Stata-based data preparation stage, the UCINET-based network analysis stage, and the supplementary article-level calculations.

The workflow includes:
- preparation of ADB MRIO input files for Stata,
- generation of DVA flow outputs in Stata,
- preparation and import of UCINET edge-list files,
- construction of weighted directed UCINET networks,
- calculation of selected network measures,
- supplementary rank-stability and intermediation-concentration calculations.

Users should consult README_UCINET.txt for detailed documentation of the UCINET import procedure, settings, and network measures.

======================================================================
9. SOFTWARE
======================================================================

The analysis used:
- Stata/MP 16.0
- UCINET 6.804

Further software details are provided in:
03_stata_code/software_versions.txt

======================================================================
10. CITATION AND SOURCE ACKNOWLEDGMENT
======================================================================

Users of this repository should cite:
(1) the associated article,
and
(2) the original ADB data source.

Please consult:
docs/CITATION.txt
docs/LICENSE_NOTE.txt

======================================================================
11. CONTACT
======================================================================

For questions concerning these replication materials, please contact:

dr Aleksandra Nacewska-Twardowska
aleksandra.nacewska-twardowska@uni.lodz.pl
Department of International Business and Trade, Institute of International Economics, 
Faculty of Economics and Sociology, University of Lodz 
POW street 3/5, 90-255 Lodz, Poland