On the Network Topology of Unemployment Transmission in the European Union: A Quantile and Frequency Approach

Unemployment policy is national, but unemployment dynamics may be connected across European Union (EU) member states. A penalised quantile vector autoregression of monthly EU27 unemployment rates, 2000–2026, is decomposed into fifteen connectedness networks over five quantiles and three frequency bands, reported as point estimates. At every quantile, the share of connectedness attributed to contemporaneous correlation falls from the shortest to the longest band, and the share attributed to lagged dynamics is larger in the tails than at the median. Upper-tail connectedness exceeds the median’s against fitted benchmark systems in which the two coincide, an exploratory departure. In an exploratory second stage, edge weights are associated with sectoral similarity and bilateral trade. Simulations set these statuses: bootstrap intervals under-cover connectedness levels, and an edge screen against a no-spillover null does not control false discoveries. Separately, a rank-score test of each direct lag-one coefficient, a narrower object than an edge, examined by simulation without established control of false discoveries, supports one link on the panel, at low power.