Foreign-Demand Shocks and the Dynamic Labour-Market Visibility Gap: A Cross-Country Panel Dataset, 1995–2022

This dataset accompanies the article “Visible First, Distributive Later: Foreign-Demand Shocks and the Evolution of the Labour-Market Visibility Gap.” It provides an unbalanced country-year panel covering 68 economies over 1995–2022, with a maximum of 1,904 observations. The data integrate OECD Trade in Employment statistics, ILOSTAT labour-market indicators, and World Bank macroeconomic and institutional variables. The core exposure variable is a partner-weighted foreign-demand innovation constructed from lagged bilateral employment weights and partner-country real GDP growth deviations from the annual cross-country mean. Its adverse-shock transformation assigns larger positive values to deeper negative foreign-demand innovations. The dataset also includes a foreign-demand employment lock-in measure combining the share of domestic employment sustained by foreign final demand with the concentration of that demand across partner economies. Labour-market outcomes include unemployment, labour income share, working hours, labour-force participation, and broad labour underutilisation where available. The principal burden measures are oriented so that positive values indicate deterioration. Visible unemployment burden is defined as the standardised annual change in unemployment, while labour-income-share loss is the standardised negative annual change in labour’s income share. Their difference forms the static visibility gap in contemporaneous analysis. Horizon-specific cumulative versions for h = 0, 1, 2, and 3 constitute the dynamic visibility path, allowing the relative incidence of unemployment and distributive loss to be traced over time. Additional variables include inflation, trade exposure, manufacturing share, GDP per capita, government consumption, subsidies and transfers, union density, collective-bargaining coverage, income-group classifications, country identifiers, and year indicators. Labour-market outcomes are not imputed; consequently, sample size varies across variables. Unemployment changes are available for 1,835 observations, the labour-income-share measure and core visibility gap for 1,223, working-hours changes for 1,155, and participation changes for 1,835. The dataset is designed for replication of two-way fixed-effects models, panel local projections, institutional heterogeneity tests, and robustness analyses. Because OECD Trade in Employment coverage is weighted toward high-income, particularly European, economies, the sample should not be interpreted as globally representative.