Researchers’ institutional mobility: bibliometric evidence on academic inbreeding and internationalization

paper by: Vít Macháček, Martin Srholec, Márcia R. Ferreira, Nicolas Robinson-Garcia, Rodrigo Costas

see the paper

This repository contains:

1. Public dataset
2. Source codes for generating the data

Public dataset

Dataset is available in ./data/dataset_public.csv. It contains institutional estimates on both the aggregate level and individual disciplines

The following columns are available in the CSV file:

• organization_name - Organization name by GRID
• discipline_name - Discipline name (FoR - division level)
• grid_id - ID from the GRID database
• discipline_id - Discipline ID
• number_researchers - number of researchers identified on the organization in 2018
• pct_insiders - percentage share of insiders (0-100)
• pct_domesticoutsiders - percentage share of domestic outsiders
• pct_foreignoutsiders - percentage share of foreign outsiders
• grid_country_code - organization country (by GRID)
• established_year - organizations' year of establishment (by GRID)
• country_name
• geo_region - geographic region
• latitude - GPS coordinates from GRID
• longitude - GPS coordinates from GRID
• continent

Source codes for generating the dataset

Prerequisities:

• Access to the Dimensions dump in SQL-like database. Use your DB query editor to run SQL procedures and store the DB_CONNECTION_STRING in the .env file in the root directory, so that Python can access it.

Dataset Generation procedure:

1. Generate publication-author-discipline grid table (SQL):

• Excluding publications from 2019 and later.
• Result table dbo.dim_mob_pub
• Main logic in ./code/01_create_dim_mob_pub.sql
• The computation lasts approx. 30 minutes

Creates a table where each row is a unique author-paper-discipline combination.

column	description
pub_id	publication_id
researcher_id	Resarcher ID after disambiguation (main ID)
for_division_id	Discipline identifier from Fields Of Research (level 2 - Divisions)
author_id	Author ID (before author disambiguation; not used)
grid_id	Institution ID from GRID database
country_code	Organization country
pub_year	Publication year
N	The order of researchers' publication (sorted by pub_year)
T	Years from the first researchers' publication

SQL code to generate the table is here. Replace the {{dimensions_database}} and {{target_database}} placeholders. Works in MSSQL dialect.

2. Generate a researcher-level table (Python)

Table where each row is a unique researcher-2018_institution-discipline combination. I.e. multiply affiliated researchers in 2018 will be represented mutliple times.

• Result table dbo.dim_mob_researchers
• Main logic in ./code/create_dim_mob_researchers.ipynb notebook
• This procedure took a bit longer, but it was over-night

Procedure

1. Select all organizations with more than 100 researchers in 2018 (from table organization) - 5902 organizations with unique grid_id

2. Load discipline data from table for_division

3. Iterate over all organizations and all disciplines:

   a. select all publications of researchers active on the organization in 2018 (see SQL query in Python code) from the dim_mob_pub

   b. Group by researcher_id and apply function getResearcherMobility on all

   c. writes results into SQL dim_mob_researchers

Mind that if researcher published in multiple disciplines in 2018 he will be calculated multiple times

Following attributes are available in researchers' table:

column	description
grid_id	Grid ID of 2018 affiliated university
country_code	2018 affiliated university country
for_division_id	Discipline ID (if applicable)
researcher_id	ID of a single researcher
N	number of researchers' paper (in given field; if applicable)
T	Years since first publication in 2018
sameInst	Is grid_id listed in the researchers' starting papers?
sameCountry	Is country_code listed in researchers' starting papers?

Jupyter Notebook to generate researchers.

3. Scaling-up to organizations and disciplines

• Result in dbo.dim_mob_organizations

• Main logic in ./code/create_dim_mob_organization.sql

• <30 seconds

• Two queries:

  a) Average indicator per organization and field of research

  b) Average indicator per organization (for == 'All') - Beware researchers assigned to multiple discipline are counted mutliple times!

column	description
grid_id	GRID institution ID
for_division_id	Field of Research ID
organization_name	GRID institution name
for_division	discipline name
researchers	Number of researchers in given GRID and FOR in 2018
ResSameInst	Share of researchers starting their career on the same institution (affiliated to the same institution in T<2)
ResSameCountry	Share of researchers starting their career on the same country (was affiliated to the same country in T<2)
ResOtherInstSameCountry	Share of researchers starting in the same country, but on different institution (ResSameCountry - ResSameInst)
ResOtherCountry	Share of researchers starting in other countries (1 - (ResOtherInstSameCountry + ResSameInst))
country_code	GRID institution country
latitude	GRID institution latitude
longitude	GRID institution longitude

4. Filtering Leiden-ranking universities

The GRID universities were semi-automatically matched with Leiden Ranking universities. The final dataset only includes succesfully matched universities.