Q&A: How can data science help investors map the route out of lockdown?
Q&A: How can data science help investors map the route out of lockdown?
The Covid-19 crisis has brought data into our everyday lives like never before. Where once we might have checked our news apps for the football scores, we’re now looking for the latest on the "R rate", showing the level of new infections, and wondering what it means in terms of a return to normal life.
Data science has been an established part of the investor's toolkit at Schroders since the Data Insights Unit (DIU) was established in 2014. The team, headed by Mark Ainsworth, provides a research service to help Schroders’ investors extract actionable insight from alternative and big data, enabling them to make better investment decisions.
Mark, can you give an example of the kind of alternative data you use?
“Alternative data is the name for the data sets that aren’t normally part of a fundamental investor's toolkit, such as market data or data from company reports. This data can be useful but often exists in a form that’s too large to access via usual methods such as Excel, or is messy and unstructured.”
“The use of apps or websites is an example of this. Our team is able to see how popular any app or website in the world is. This can give us an insight into the kind of information the company itself will have on how its business is doing. Over the years, this has been used hundreds and hundreds times in investment research.
“The Covid-19 crisis presented a situation where we suddenly found a different use for that kind of data.”
And what kind of use was that?
“Take Google Maps, or their Chinese equivalent. These apps can give us a real-time picture of how a population’s behaviour is changing in response to the lockdowns.
“People tend to use mapping apps when they're out travelling in the car or on foot. The data tells us how much people's behaviour changed as a result of lockdowns, but also how it started returning to normal again as restrictions began to be lifted towards the end of March.
“Obviously, Wuhan was the original epicentre of the outbreak, but in fact the whole of China was locked down. The total drop in usage of mapping apps across the entire country was more than 35% and is only very slowly returning to normal. We’ve seen that it's not been an instant recovery but more of a gentle recovery.”
Can you give us other examples?
“Another example of alternative data is websites. Any website in the world can be a source of data but that data can be complex to display and interpret.
“One of the things we've done is to build a web scraper that extrapolates data on traffic congestion from one particular traffic company. We've turned it into a dashboard that our investors can access.
“Take Beijing and Wuhan as examples again. In March in Beijing, there were few instances of the virus spreading and people were behaving fairly normally. Our dashboards told us that morning rush hour was roughly normal.
“However, during the middle of the day and at weekends, Beijing's streets were not congested. There were no traffic jams and that was starkly different from normal times. Wuhan was in full lockdown, so traffic movements there were way below normal levels. But we could see that Beijing was also very restrained.
“This is an example of the kind of insights we can provide about what’s happening with the virus. It’s something we can look out for in any country of the world and will be very relevant as lockdowns start to lift.”
And what about data on the spread of the virus itself?
“We created our own epidemiological model to track the spread of the virus in different countries in March as lockdown was being introduced in continental Europe but not in the UK. The model takes into account the demographic profile of different countries and how susceptible the population is likely to be to the virus.
“It’s been clear that the official cases vastly undercount the actual number of people who have had Covid-19, possibly by as much as a factor of 50. This is one of the first messages we delivered to our investment colleagues back in early March: pay a lot more attention to the mortality statistics than the number of positive cases, because the mortality figures will be a lot more reliable.
“We were also able to use the data on the spread of the virus to see which countries are starting to bring it under control. We can then start to draw conclusions about what's really going on and the likely actions that governments will take, i.e. easing lockdowns or introducing tighter measures. The decisions that governments make then obviously affect financial markets and how companies operate.”
There are a lot of news reports about the speed of its spread and levels of immunity in the population - are you able to use that data?
“A big part of being a data scientist isn’t just about obtaining data but interpreting it.
“For example, there have been a couple of big studies where the researchers took blood samples from over a thousand people in Santa Clara County in California and in parts of Florida and they tested it for evidence of antibodies. The presence of antibodies is the key indicator that people have had this virus in the past and they now have immunity.
“The estimates from that county in California suggested that about 2% to 3% of the population had already developed immunity. Now, that would mean that the number of people who'd got it was actually 50 times higher than the official case count. It would also imply that the disease is much less dangerous than people thought.
“Clearly, if this were the case it would have big implications for public policy and the re-opening of the economy.
“But we did some digging and we found that the particular test used in the study has a false positive rate. What this means is that if you gave this test to 100 people who had never had this virus then, on average, about two of them would come up positive by mistake. So it's entirely possible that the study got this result by accident. It could be that there's nobody in Santa Clara County who has immunity.
“This work in digging into the detail of the test and interpreting the findings means we could give important guidance to our investors, advising them to ignore this particular study and be very wary of other similar ones.
“This is where our expertise in handling large data sets and spotting any problems with them can become really crucial. We continue to watch the data and distil it down into our overall judgements about the virus: what makes it spread faster or slower; what makes it more or less dangerous, etc. And our colleagues can then use these judgements to make more discerning investment decisions.”
This article is issued by Cazenove Capital which is part of the Schroder Group and a trading name of Schroder & Co. Limited, 1 London Wall Place, London EC2Y 5AU. Authorised by the Prudential Regulation Authority and regulated by the Financial Conduct Authority and the Prudential Regulation Authority. Nothing in this document should be deemed to constitute the provision of financial, investment or other professional advice in any way. Past performance is not a guide to future performance. The value of an investment and the income from it may go down as well as up and investors may not get back the amount originally invested. This document may include forward-looking statements that are based upon our current opinions, expectations and projections. We undertake no obligation to update or revise any forward-looking statements. Actual results could differ materially from those anticipated in the forward-looking statements. All data contained within this document is sourced from Cazenove Capital unless otherwise stated.