Tracking Weekly Activity using New Data Sources

Policymakers increasingly rely on real-time measures of economic activity to inform decisions, yet official statistics are typically available only with substantial delay. This paper develops a methodology to extract information from high-frequency indicators in order to produce weekly estimates of official monthly statistics in real time. Building on real-time tracking and nowcasting models, our approach addresses two challenges inherent in alternative indicators: the absence of seasonal adjustment and the prevalence of outliers. We incorporate seasonal components directly into the model, allowing the seasonal structure of low-frequency data to inform high frequency proxies, and employ fat-tailed distributions to mitigate the influence of large, infrequent shocks. Applying our methodology to UK data, we track retail sales, monthly GDP, and vacancies using proxies such as debit card spending (Revolut) and online job advertisements. Our results show that weekly estimates improve real-time prediction of official releases and highlight the usefulness of high-frequency alternative data.