A talk presented at the Royal Statistical Society Conference, September 2018
GDP growth is a key metric in considering the economic status of the UK. The ONS produces quarterly growth figures as preliminary estimates that are typically released several weeks after the end of the quarter from data that does not cover the whole quarter. This delay, and later revisions, present issues for policy makers, investors and businesses as the statistics they use to make informed decisions are not current. Therefore, it would be advantageous to produce early indicators of GDP growth that can reliably detect changes in advance of the GDP figure publication. Here we investigate the potential of using traffic flow data to provide early indication of GDP change. We take annual average daily flow figures and show that they lead annual GDP figures by one year using cross correlations, a trend that is consistent across different vehicle types. And that the previous year’s traffic flow correlate with annual GDP growth (Pearson’s R = 0.76, p < 0.01, n = 12). This shows that traffic flow is a potential early indicator of GDP growth and compares well with similar work carried out by Stats Netherlands using road sensor data (Killan, Ros, "Road Traffic Correlations with Economic Variables: The Big Data Perspective, 2017). Results from ongoing work will also be included. This uses traffic counts, recorded from camera positions on Motorways and Major A-roads from Highways England that provides traffic flow data at fifteen-minute intervals by site . This data contains counts separated by vehicle size so different types of vehicle (Car, HGVs & Coaches etc.) making it an ideal data-set for further investigating the potential of traffic flow as an early indicator for GDP growth as well as what types of vehicle and at what times and locations (e.g. cars at rush hour on weekdays in the South East) might be the best indicator of economic activity.
The slides are available here; https://onsbigdata.github.io/RSS-2018/#/
The presentation mostly concentrates on models predicting GDP from daily traffic flow data. It finds that while it is possible to replicate the approach carried out by other National Statistics Institutes (references in presentation), this approach does not work with a longer time series that includes the 2007/8 financial crisis and subsequent recession in the UK.
Please email [email protected] if you have any questions or comments.