Abstract:
We show that payment system data can help predict economic activity in Denmark by employing mixed-data sampling (MIDAS) regression methods to forecast quarterly macroeconomic variables using high-frequency data. Among a set of frequently used predictors of economic activity, payment system data delivers some of the best nowcasts and one-quarter-ahead forecasts of GDP. Forecast combinations that blend monthly payment system data with other high-frequency predictors also score high in terms of their nowcasting performance. However,the forecasting performance of the payment system data deteriorates during the first half of 2020, as changes in the payment behavior recorded in the payment system data are not large enough to explain the sharp drop in economic activity during the first wave of the covid-19 pandemic.