Are UK’s September and October output figures set to disappoint?

Important note: The chart above, and what follows, are not a prediction, projection, forecast, or anything like that. They are simply an illustration of the factors that one model has picked out as good historical predictors of the UK’s monthly gross value added (GVA). The number of variables used is large; the number of actual datapoints is minuscule (monthly data for January 2020 to August 2021); and the illustration is based on a black box deep learning model that has simply trained itself to best predict each historical GVA figure. The model therefore has little explanatory power and, I suspect, predictive power. However, it is an interesting and different way of looking at what might have happened in September and October 2021.

Update on 11th November 2021: The actual GDP outturn figures (first estimates, subject to revision) for September 2021 are now out from ONS. Turns out the model’s September projection wasn’t that accurate. [We’ll see about October, but it might similarly be quite wrong.] According to ONS, real output in September was 0.5% higher than in August, whereas the model had suggested this to be 0.3% lower than August 2021. A key factor which the model would have difficulty of picking up is to do with public sector output and how it is measured, and the effect of the coronavirus testing, tracing and vaccination programme.

I have previously suggested that the Google mobility indices are decent predictors of output levels and that it would be interesting to feed all that data, as well as data about various COVID-19 related developments, into a machine learning model to see what it spits out. Well, here are some of the results of one such model (which, full disclosure, I have not even evaluated — just looked at and presented the results “out of curiosity”; I also admit to significant subjective biases in choosing which variables “made most sense” to include in the model and to highlight in the chart).

Caveats aside, the result is interesting. Taken at face value, the model is suggesting that far from a continued bounce back, we might in fact see a contraction in output in September and October, relative to August levels. That feels (and I’m intentionally using that word) quite unlikely, but it must be the case that there is enormous uncertainty about this; not least because the model is predicting real GDP growth and inflation has been high and volatie; and because there are so many factors on both the supply and demand side of the economy that work in opposite directions, sometimes somewhat mysteriously through second and third-order effects.

If there is any truth in the model’s predictions, the explanations for such a negative outlook will be around what I have previously coined “pandemic fundamentals”: the spread of coronavirus infections and the physical, psychological and policy implications of it. Since August, the UK has experienced further spikes in new infections and COVID-related deaths have been edging up. So far, this hasn’t resulted in further significant restrictions on households or businesses, but may well be impacting their confidence. What has happened is that several economic support measures — such as the furlough scheme and the £20 uplift to universal credit — have come to an end; both bound to have some impact on at least some households’ ability to spend.

The mobility indices also tell an interesting story. On one hand, one has to take them with a grain of salt — not least because they are very dependent on holidays and which weekdays happen to fall into which month. On the other hand, those are precisely the factors that will influence output, so maybe it can be slightly less of a concern in the interpretation of the data. [Note, however, that that does depend on which measure of output we are talking about. I suspect that “seasonally adjusted” figures will in fact not quite reflect those same fluctuations, and it is the seasonally adjusted information that will most likely be used to make policy judgements.]

For October (up to and including the 25th), both the “Grocery and pharmacy” and “Retail and recreation” mobility indices are somewhat down from their August levels. Fascinatingly, “Residential mobility” (which has tended to be negatively correlated with GDP) remained almost as high in September and October as it did in the previous 3 months; “Workplace mobility” (which has tended to be positively correlated with GDP) remained almost as low as in June; and “Transit station mobility” was only slightly higher than in the summer months. [Note: I’m about to do another blog about working patterns during the week, using this data, so will will provide link here when available.]

So we shall see. I believe September GDP figures will be out on the 13th of October, so I will come back to this blog then to see how badly wrong the machine learning model was…

--

--

--

I help complex organisations make the right strategic decisions through innovative, insightful and incisive analysis and recommendations.

Love podcasts or audiobooks? Learn on the go with our new app.

Recommended from Medium

9 Must-have skills you need to become a Data Scientist, updated

Pandas DataFrameGroupBy

Logistics in the times of COVID

COVID Comorbidity Tweets and Data Science (Part I)

Accelerate : Become a Data Engineer

Setup, Delivery & Punchline!

August 2020 monthly Machine Learning Reading List | by Durgesh Samariya

Basic sales prediction for Sales Managers & Marketers using Regression Analysis

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Tera Allas

Tera Allas

I help complex organisations make the right strategic decisions through innovative, insightful and incisive analysis and recommendations.

More from Medium

Gold Rush: Mineral Exploration with Machine Learning

Color scale map of the United States showing radiometric data.

Dimensionality Reduction

Predicting fuel efficiency of cars.

Survey: Semiconductors Supply Chain