This project considers methods and data sources to nowcast variables more detailed than headline GDP, including household consumption, capital investment, household production, and public service quality indices. The nowcasting of these variables pose different challenges and may require different approaches to nowcasting GDP. We consider a wide range of data sources, including monthly industry output data, ONS ‘Real Time Indicators’, and a range of other seemingly related (or unrelated) variables. We test a range of statistical methods and models, and evaluate them using out-of-sample prediction performance.
Many nowcasting and forecasting models are focused on headline or aggregate measures, such as GDP for the whole economy or a specific region. These models can use a range of macroeconomic indicators with general relevance for the economy. Less attention has been paid to nowcasting variables below the aggregate, such as the expenditure components of GDP (namely household consumption and capital investment) or ‘Beyond GDP’ variables, such as household production and public service quality indices. This project considers nowcasting for these applications.
Household consumption and capital investment collectively account for two-thirds of GDP in the UK, so are important variables for estimating quarterly GDP. Data on government spending and trade are more readily available at quarterly frequency. Modelling consumption and investment may require the use of more detailed variables given their specific nature.
Similarly, nowcasts of household production and public service quality indices may benefit from unique or unusual data sources or methods, given their particular concepts. These variables support “Beyond GDP” work at ONS, and are not currently available at quarterly frequency.
For nowcasts of GDP consumption and investment, we employ mixed-data sampling models as they allow us to exploit the information of monthly and weekly predictors to predict the quarterly series. Our first approach revisits the MIDAS model of Galvao and Lopresto (2020), which incorporates mixed-frequency methods to extract information from monthly official and unofficial data sources in order to predict quarterly GDP. We also augment this model with a real-time database of monthly estimates of GDP(O) sectoral GVA. Finally, we consider the suite of faster indicators at different frequencies, which the ONS now publishes, and add these to the model. We will then examine which approaches perform better in forecasting consumption and investment in an out-of-sample exercise.
For nowcasting household production and public service quality indices, our nowcasting approach will be based on the existing measurement methodologies. We will use both a top-only approach (focusing exclusively on the aggregate variable), and a bottom-up approach (making a nowcast for each subcomponent of the aggregate variable and aggregating). In both cases, we will consider a large panel of seemingly related (and also unrelated) variables and use statistical techniques to select relevant variables. The performance of each model is measured in an out-of-sample cross-validation exercise.
The research is ongoing, and findings will be shared when available.
The project will offer alternative estimates of consumption and investment growth available at the time the first estimate of these expenditure components is published by the ONS. These could be used by ONS or external researchers to better understand short-term estimates of these components in official statistics. It will also contribute to understanding of the predictive power of ONS ‘Faster Indicators’, and the literature concerned with modelling components of GDP.
Similarly, the nowcasts of household production and public service quality indices will offer ONS more timely estimates of variables which are currently unavailable at quarterly frequency, for use in “Beyond GDP” research and official statistics.
Galvão, A.B. and Macqueen, R. ‘Nowcasting UK Consumption and Investment with Monthly Output Components and Real-time Indicators’ Economic Statistics Centre of Excellence: The Next Five Years, poster exhibition, 12 December 2022, One Birdcage Walk, London.