Digital Divide and Sectoral Employment in Developing Economies
Description
This repository contains the data and replication materials for the study “Digital Divide and Structural Transformation: Evidence from Sectoral Employment in Developing and Emerging Economies.” The empirical dataset is a strongly balanced country-year panel covering 96 developing and emerging economies over the period 2007–2022. The dataset includes sectoral employment shares in industry, agriculture, and services; sectoral value-added indicators; urbanization; trade openness; gross capital formation; human capital; government effectiveness; electricity access; and the ICT variables used to construct the digital-development measure. The composite ICT index is constructed using fixed broadband subscriptions, Internet use, fixed telephone subscriptions, and mobile-cellular subscriptions. The main digital-divide indicator measures each country’s normalized annual distance from the ICT frontier within the study sample. The repository includes the final harmonized panel dataset, the constructed ICT index, the distance-to-frontier digital-divide indicator, variables used in the income-group heterogeneity and instrumental-variable analyses, and Stata do-files required to reproduce the fixed-effects estimates, channel analysis, clustered-bootstrap indirect effects, heterogeneity analysis, instrumental-variable estimates, and robustness checks. The underlying indicators were compiled from publicly available international databases, including the World Development Indicators, the International Telecommunication Union, the UNCTAD Productive Capacities Index, the Penn World Table, and the Worldwide Governance Indicators. Users of the replication package should cite the original data providers where appropriate.
Files
Steps to reproduce
The dataset was assembled from publicly available international databases and harmonized into a strongly balanced country-year panel covering 96 developing and emerging economies from 2007 to 2022. Countries were classified as low-, lower-middle-, or upper-middle-income economies using the World Bank FY2026 income classification. The underlying indicators were collected from the World Development Indicators, the International Telecommunication Union, the UNCTAD Productive Capacities Index, the Penn World Table, and the Worldwide Governance Indicators. Country names and identifiers were standardized before the datasets were merged by country and year. The final database includes sectoral employment shares, sectoral value-added shares, urbanization, trade openness, gross capital formation, human capital, government effectiveness, electricity access, and ICT indicators. The main ICT index was constructed from four variables: fixed broadband subscriptions per 100 people, individuals using the Internet as a percentage of the population, fixed telephone subscriptions per 100 people, and mobile-cellular subscriptions per 100 people. Each indicator was standardized using pooled country-year observations. Principal Component Analysis was then applied to the covariance matrix of the standardized indicators, and the first principal component was retained as the composite ICT index. The resulting component scores were rescaled to the 0–1 interval using min–max normalization. The distance-to-frontier digital-divide indicator was calculated separately for each year. The country with the highest normalized ICT-index value in a given year was treated as the annual frontier. An alternative distance-to-frontier indicator was constructed from the ICT sub-index of the UNCTAD Productive Capacities Index and used for robustness analysis. The instrumental variable was created by interacting each country’s baseline electricity-access rate in 2007 with a leave-one-out global ICT diffusion trend. The empirical analysis was conducted in Stata. The replication do-files reproduce the principal-component analysis, index construction, fixed-effects regressions, specifications with year fixed effects, Driscoll–Kraay standard errors, lagged digital-divide models, income-group heterogeneity analysis, two-step channel regressions, country-clustered bootstrap indirect effects, instrumental-variable estimates, and robustness checks. The repository includes the final panel dataset, constructed indicators, and the Stata code required to reproduce the main tables and empirical results.
Institutions
- University of World Economy and DiplomacyTashkent, Tashkent