Timely and high frequency regional macroeconomic indicators are essential for effective decision making and for supporting growth across the UK. By combining national and regional statistics produced by ONS with data produced by the devolved administrations of Scotland and Northern Ireland, our newly developed nowcasting model produces estimates of quarterly economic growth for all UK regions to approximately the same timetable as the release of UK GVA data. Our estimates are being used by policymakers at sub-national level and enable new analysis and understanding of the pattern of historical regional growth.
There is a clear demand for better devolved, regional and local statistics in terms of quality, coverage, timeliness, frequency and granularity. Historically, the ONS produced regional output, specifically Gross Value Added (GVA), data at the annual frequency and with a significant release delay. Since September 2019, the ONS has released country and regional output data at the quarterly frequency around 6 months after the end of the quarter. Existing short-term indicators already exist for Northern Ireland and Wales, and Scotland has, for some time, produced its own quarterly GVA estimate.
Complementing this work, we have developed a nowcasting model that produces timely quarterly regional estimates of GVA. Our estimates are conditioned on both the historical annual data for the regions and the quarterly data for the UK. Since November 2018, our regional estimates have been released shortly after ONS first estimates of quarterly GVA for the UK. We have also produced historical quarterly estimates of regional GVA growth dating back to 1970. We are working with ONS to incorporate their new Regional Short-Term Indicators into our econometric models and are providing training for ONS staff to run these models.
We use Bayesian Vector Autoregressive (BVAR) models, widely used to produce timelier and higher frequency estimates of a range of economic variables like economic growth. In broad terms, nowcasting methods using these models seek to exploit the mixed-frequency nature of the data and accommodate the differing publication timetables of those indicator variables chosen for their putative ability to explain the within-year or quarterly variable of interest, such as quarterly economic growth. A range of different econometric methods have been used to ‘nowcast’ macroeconomic variables. We use two main models: a stacked VAR model and a state space VAR model. In order to make these models suitable for our application, we have explored different econometric innovations to meet our needs and, in particular, to impose the ‘cross-sectional’ constraint that the quarterly regional data add up to the observed UK data. We also explored the use of ‘machine learning’ to enable more efficient computation and deliver more accurate nowcasts of regional growth.
Our research has demonstrated that empirical macroeconomic methods are able to produce higher-frequency and more timely estimates of regional economic growth. This meets a key demand from stakeholders for more timely information on the performance of regional economies. Our methods also provide consistent historical regional growth data back to 1970 that enables current regional growth to be set into an appropriate historical context, helping inform policy and decision making at the regional level.
We are publishing nowcasts as regular research outputs. These have generated much media coverage and have been cited by many sources in the private sector and by politicians in the UK Parliament and elsewhere. We are working closely with ONS to integrate their newly produced quarterly country and regional GDP estimates into our model and to develop their ability to use these methods, enabling new ONS statistical outputs.
Project Papers, Presentations and Data Outputs
Gary Koop, Stuart McIntyre, James Mitchell and Aubrey Poon (2021) ‘Regional Nowcasting in the UK’ ESCoE Conference on Economic Measurement 2021, Poster Exhibition, 11-13 May 2021. Poster Presentation
McIntyre, S., Mitchell, J. and Poon, A. ‘Nowcasting ‘True Monthly US GDP during the Pandemic‘ ESCoE Conference on Economic Measurement 2021, Contributed Session B: Nowcasting, 11-13 May 2021 (presentation at 45:00)
Koop, G., McIntyre, S., Mitchell, J., & Poon, A. ‘Reconciled Estimates of Monthly GDP in the US‘ ESCoE Discussion Paper Series, ESCoE DP 2020-16, 26 Nov 2020
Koop, G., McIntyre, S., Mitchell, J., & Poon, A. (2020). Reconciled Estimates and Nowcasts of Regional Output in the UK. National Institute Economic Review, 253, R44-R59. doi:10.1017/nie.2020.29
Koop, G., McIntyre, S. and Mitchell, J. (2020) “UK regional nowcasting using a mixed frequency vector auto-regressive model with entropic tilting” Journal of the Royal Statistical Society, Statistics in Society Series A, Vol 183, Issue 1, Wiley https://doi.org/10.1111/rssa.12491
Koop, G., McIntyre, S., Mitchell, J., Poon, A. (2020) Regional output growth in the United Kingdom: More timely and higher frequency estimates from 1970, Journal of Applied Econometrics Vol 35, Issue 2, Wiley https://doi.org/10.1002/jae.2748
Koop, G., McIntyre, S., Mitchell, J., & Poon, A. ‘Regional Output Growth in the United Kingdom: More Timely and Higher Frequency Estimates, 1970-2017‘ ESCoE Discussion Paper Series, ESCoE DP 2018-14, 20 Nov 2018
Koop, G., McIntyre, S. and Mitchell, J. ‘UK Regional Nowcasting using a Mixed Frequency Vector Autoregressive Model‘ ESCoE Discussion Paper Series, ESCoE DP 2018-07, 12 Jun 2018
For a full list of the quarterly updates visit here.