The exploitation of novel data streams (for example from card payments) will allow rapid and real-time model estimates to be made available to stakeholders. As the quantity of data available for Office for National Statistics (ONS) to process increases, there needs to be comparable development in the methods of exploiting this data. The project aims to create a model to 'nowcast' consumer spending by category and geography.

Explaining the science

This project will exploit the recent advances in the signature methodology using controlled differential equation models, which give a generic approach to efficient use of multimodal and irregularly sampled (or missing) time series data. This approach has been shown to be very effective in related problems for astronomy, medicine, finance and human-computer interaction, and is well suited to dealing with the complex systems under consideration.

Project aims

Challenge 1 focuses on specific economic applications. The challenge is to produce high-frequency near to real-time estimates of consumer spending and consumer income at very detailed geographic level. This requires combining detailed consumer expenditure data with other data sources on economic activity and developing the best possible near to real-time model estimates for high dimensional outcomes.  

A related challenge is the production of high-frequency measures of inflation, in particular the CPI. Real-time spending data and large-scale data on the prices of goods are available from ONS. The challenge is to link these two types of data, in order to develop improved price indices, estimate demand, and evaluate the impacts of tax, health and other policies on household welfare and spending.
Challenge 2 seeks to understand how disparate, irregularly sampled time series can be understood and exploited in an official statistics context. The project aims to build robust and widely applicable methods for nowcasting fine-grained economic data, using passive or indirect observations of related variables.  


By creating a signature model for Nowcasting, the project aims to provide new economic indicators to assist policy makers with:

  • High-frequency near to real-time consumer income forecasts at very detailed geographic level.
  • Improved high-frequency measures of CPI
  • Analysis of impacts on spending of exogenous factors like COVID-19 lockdown policies, opening or closure of shops, new roads or rail links, or changes in tax structure

Researchers and collaborators

Contact info

Tony Zemaitis
[email protected]