Abstract

Nowcasting combines 'now' and 'forecasting' and refers to predicting the present or the very near future before true data becomes available. Given that it is a real-time analysis technique that aims to provide current and immediate insights into economic conditions, weather, public health trends, or social behaviors, this method is of high practical importance.

This paper proposes a method to produce preliminary but timely predictions of the present (nowcasts) using the Dutch 'Road Freight and Transport Survey' (RFTS) as an example. The sample is drawn quarterly, and the owner of a sampled license plate has to report trips and loadings for one randomly assigned week. The sample size is chosen to make yearly estimates of sufficient quality. Quarterly, weekly, or even daily estimates can be made, but the sample sizes will decrease, and standard errors will increase beyond reason. To produce estimates at a higher than yearly frequency with reasonable precision, we propose to include in a structural time series model (STM) both design-based quarterly estimates from the past, and official economic indicators, such as freight transport by water, rail and air; consumer, producer and business confidence; gross value added; turnover; or international trade.

This paper proposes a multivariate STM to nowcast quarterly indicators for the RFTS, using auxiliary series observed at a monthly frequency. Direct quarterly sample estimates for the RFTS from 2013 to 2021 for the STM are considered. The STM improves the accuracy in two ways. First, by using sample information from previous reference periods, and secondly, by using additional information from the auxiliary series. The auxiliary series are observed at a monthly frequency and, therefore, more timely available than the target series of the RFTS. In this way, more precise and timely nowcasts for the quarterly RFTS figures can be made. We will consider several relevant target variables for the publication of official statistics, such as the weight of the transported load, the vehicle kilometers, or the number of loaded trips. We will demonstrate how much our proposed method can improve the accuracy of quarterly survey estimates compared with direct estimates. Therefore, nowcasting is an important real-time analysis and decision-making method in economic and social monitoring or crisis management.