Nowcasting. Developing the sources and methods to improve high-frequency labour market forecasting

Forecasting unemployment dynamics is a significant concern for policymakers and welfare institutions. Timely decisions on passive and active labor market (LM) policies are increasingly seen as key for their efficiency and cost effectiveness. The effectiveness of policies can benefit from the high-frequency update of forecasts on the LM, as this might reduce the typical decision and implementation lags of policies. Unfortunately, official information about the LM is released with a substantial delay (about two and a half months for quarterly LM statistics; about one month for aggregate monthly provisional data for employment and unemployment). We implement a nowcasting approach to real-time LM forecasting as a straightforward solution to official information lag problems. In economics, nowcasting tools are becoming increasingly popular in the real-time prediction of macroeconomic aggregates such as industrial production, gross domestic product (GDP) and inflation. We merge real-time data on hiring and firing with high- and standard-frequency official data in a Vector Auto-Regressive (VAR) model to provide high-frequency forecasts of specific segments of employment. By merging official data releases with a high-dimensional set of high-frequency data, we contribute to the estimation of the current LM dynamics and the understanding of their underlying developments in the real-time.