August 11, 2020

Forecasting wage growth and price inflation in the Netherlands with a BVAR model

This background document describes a Bayesian Vector Autoregression (BVAR) model to support the CPB short-term forecasts of wage growth and price inflation.

BVAR models are already used at the CPB to support the forecasts of GDP and unemployment. We have now also developed a BVAR model for HICP inflation and hourly wage growth. We tested and optimized the model to produce the most accurate forecast possible. The model is automized for regular use during the forecasting rounds. 

Our analyses show that the BVAR forecast of HICP inflation results in a smaller prediction error than the forecast for hourly wage growth. This means the BVAR produces a more accurate forecast of HICP inflation than of hourly wage growth. We also find that for the 2016- 2019 period the forecast error of the BVAR for HICP inflation is compared to the prediction error of the CPB forecast. Moreover, the BVAR for hourly wage growth has a smaller forecast error than the CPB forecast. We therefore conclude that the BVAR is a useful additional tool to the existing modelling suite at the CPB for forecasting wage growth and price inflation.


Jurriaan Paans