Spatial aggregation bias in wage curve and NAWRU estimation

I argue in this paper that the estimation of wage curves and NAWRUs at the country level suffers from spatial aggregation bias. Using European data for the years 2000-2017, I find steeper country level wage curves and higher NAWRUs, compared to estimating at the underlying regional level. The distribution of regional unemployment rates within countries over time is not mean-scaled. Regions with low unemployment rates are the main drivers of changes in aggregate unemployment. The steepness of a log-linear wage curve in regions with low unemployment dominates at the aggregate (country) level, overestimating wage pressure. Lagged wages are important in explaining wage growth, together with unemployment. This suggests that a wage curve fits the data better than the assumption of a NAWRU or long run natural rate of unemployment. With regional wage curves, spatial aggregation bias can produce aggregate data that resembles such a natural rate of unemployment, however.