Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Differentiating Emigration from Return Migration of Scholars Using Name-Based Nationality Detection Models

Faeze Ghorbanpour, Thiago Zordan Malaguth, Aliakbar Akbaritabar

TL;DR

This paper tackles left-censoring in migration research by predicting scholars' nationalities from their full names using machine learning. It builds a large, three-level taxonomy of nationalities (12/30/175 classes) trained on 2.6 million name–nationality pairs derived from Wikipedia and evaluates multiple character-based models, with ByT5 delivering the best overall performance. Applying the model to 8.2 million Scopus author records reveals that nationality-based measures shift interpretation of emigration versus return migration, notably reclassifying a substantial portion of USA-origin emigration to China as returns when name-origin nationality is used (79% vs 41% by academic origin), and affecting perceived destination patterns. The authors validate robustness with out-of-domain tests and publicly release code and data to enable replication, underscoring the method's potential to reduce left-censoring in digital-trace migration studies while acknowledging limitations related to name similarity and generational name changes.

Abstract

Most web and digital trace data do not include information about an individual's nationality due to privacy concerns. The lack of data on nationality can create challenges for migration research. It can lead to a left-censoring issue since we are uncertain about the migrant's country of origin. Once we observe an emigration event, if we know the nationality, we can differentiate it from return migration. We propose methods to detect the nationality with the least available data, i.e., full names. We use the detected nationality in comparison with the country of academic origin, which is a common approach in studying the migration of researchers. We gathered 2.6 million unique name-nationality pairs from Wikipedia and categorized them into families of nationalities with three granularity levels to use as our training data. Using a character-based machine learning model, we achieved a weighted F1 score of 84% for the broadest and 67% for the most granular, country-level categorization. In our empirical study, we used the trained and tested model to assign nationality to 8+ million scholars' full names in Scopus data. Our results show that using the country of first publication as a proxy for nationality underestimates the size of return flows, especially for countries with a more diverse academic workforce, such as the USA, Australia, and Canada. We found that around 48% of emigration from the USA was return migration once we used the country of name origin, in contrast to 33% based on academic origin. In the most recent period, 79% of scholars whose affiliation has consistently changed from the USA to China, and are considered emigrants, have Chinese names in contrast to 41% with a Chinese academic origin. Our proposed methods for addressing left-censoring issues are beneficial for other research that uses digital trace data to study migration.

Differentiating Emigration from Return Migration of Scholars Using Name-Based Nationality Detection Models

TL;DR

This paper tackles left-censoring in migration research by predicting scholars' nationalities from their full names using machine learning. It builds a large, three-level taxonomy of nationalities (12/30/175 classes) trained on 2.6 million name–nationality pairs derived from Wikipedia and evaluates multiple character-based models, with ByT5 delivering the best overall performance. Applying the model to 8.2 million Scopus author records reveals that nationality-based measures shift interpretation of emigration versus return migration, notably reclassifying a substantial portion of USA-origin emigration to China as returns when name-origin nationality is used (79% vs 41% by academic origin), and affecting perceived destination patterns. The authors validate robustness with out-of-domain tests and publicly release code and data to enable replication, underscoring the method's potential to reduce left-censoring in digital-trace migration studies while acknowledging limitations related to name similarity and generational name changes.

Abstract

Most web and digital trace data do not include information about an individual's nationality due to privacy concerns. The lack of data on nationality can create challenges for migration research. It can lead to a left-censoring issue since we are uncertain about the migrant's country of origin. Once we observe an emigration event, if we know the nationality, we can differentiate it from return migration. We propose methods to detect the nationality with the least available data, i.e., full names. We use the detected nationality in comparison with the country of academic origin, which is a common approach in studying the migration of researchers. We gathered 2.6 million unique name-nationality pairs from Wikipedia and categorized them into families of nationalities with three granularity levels to use as our training data. Using a character-based machine learning model, we achieved a weighted F1 score of 84% for the broadest and 67% for the most granular, country-level categorization. In our empirical study, we used the trained and tested model to assign nationality to 8+ million scholars' full names in Scopus data. Our results show that using the country of first publication as a proxy for nationality underestimates the size of return flows, especially for countries with a more diverse academic workforce, such as the USA, Australia, and Canada. We found that around 48% of emigration from the USA was return migration once we used the country of name origin, in contrast to 33% based on academic origin. In the most recent period, 79% of scholars whose affiliation has consistently changed from the USA to China, and are considered emigrants, have Chinese names in contrast to 41% with a Chinese academic origin. Our proposed methods for addressing left-censoring issues are beneficial for other research that uses digital trace data to study migration.
Paper Structure (15 sections, 9 figures, 9 tables)

This paper contains 15 sections, 9 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Data and Methods
  4. Results
  5. Discussion and Conclusions
  6. Availability of Model Code and Training Data
  7. Acknowledgements
  8. Ethics Checklist
  9. Data Preprocessing
  10. Rationale for Country Inclusion and Exclusion
  11. Training Details
  12. Computational Resources
  13. Test Datasets
  14. Error Analysis
  15. Assessment Per Class

Figures (9)

  • Figure 1: Distribution of nationality classes in the training dataset. Each column sums up to 100% and represents one of the three classification levels, from the left (broader) to the right (more detailed). All classes are shown, but only those bigger than 1.5% of the total are labeled.
  • Figure 2: Composition of scholars affiliated with countries (names printed above panels) included in (top row) and excluded from the training data (bottom row) based on the country assigned using ML model (Level 2 regions, printed inside tree map).
  • Figure 3: Top 5 countries with the highest proportion of scholars among those affiliated to institutions in the United States divided into emigration (left bars) and return migration (right bars) based on academic origin (top bars) versus the country of ML-name origin (bottom bars).
  • Figure 4: The proportion of return migration based on academic origin (top panel) versus the country of ML-name origin (bottom panel) divided by female scholars (diamonds) and male scholars (circles), relative to the size of the population of scholars (point size).
  • Figure 5: Distribution of nationality classes in the Athletes (top) and IUSSP (bottom) out-of-domain test datasets. Each column sums up to 100% and represents one of the three classification levels, from the left (broader) to the right (more detailed). All classes are shown, but only those bigger than 2% of the total are labeled.
  • ...and 4 more figures