Nowcasting Danish Private Consumption: Estimates in Real-Time

Abstract

This thesis investigates how to estimate the current state (combined with the recent past and near future) of the quarterly Danish private consumption. The method is called nowcasting and enables the possibility to use information from the latest release of high-frequency indicators relating to the target variable. The selection of real-time monthly indicators is based on quantitative hard data, qualitative soft data and Google Trends data. The underlying statistical framework suggests a dynamic factor model to capture the collinearity in the monthly indicators. The thesis presents this method thoroughly before estimating and evaluating the method on the private consumption. Several in-sample and out-of-sample exercises is conducted to evaluate the nowcast and to assess the different indicators impact. I find evidence that the nowcast estimates the current state of the private consumption to a better extend than the benchmark presented in the thesis. Out-of-sample exercises showed that combinations of different indicator groups tend to beat models that only rely on a single group. The combination of hard and soft data exploits the timeliness and information best, whereas Google Trends data does not necessarily bring any new valuable information to a full model. I also find strong evidence that the nowcast model is able to predict turning points during the Covid-19 pandemic, when more information becomes available during a quarter.