In flagrant disregard of government advice, many American citizens took to the streets in April to claim that draconian lockdown measures impinged on their civil liberties, and more prominently, their businesses. To combat the financial peril of small businesses, the UK government will have to borrow an unprecedented 38 per cent of the year’s GDP if social distancing measures are in place until the end of 2020, according to the Resolution Foundation. As uncertainty rages, it’s vital to understand how economies recover from crises.
‘We thought it would be interesting to see the different types of data available to study small businesses in the wake of a natural hazard event,’ says Filippo Simini, senior lecturer at the Faculty of Engineering, University of Bristol. The team’s methods were remarkably simple: they retrieved data from publicly available Facebook business pages, adjusted for unusually high social activity, and then compared posting schedules before and after the natural hazard event.
By doing so they accurately estimated business recovery time from three natural hazard events; the Gorkha Earthquake in Kathmandu in 2015, which affected 8.1 million Nepalese citizens; Hurricane Maria in San Juan, which caused significant infrastructural damage and affected the entirety of the Puerto Rican population; and the Chiapas Earthquake in Juchitán de Zaragoza, the second strongest earthquake in Mexico’s history. All estimated recovery times were validated using text analyses, surveys, official reports and scientific publications – which all pointed to similar estimates.
‘Our next step is to try to evaluate the recovery of small businesses after the Covid-19 lockdown measures,’ says Simini. ‘This technique can be done in real-time, meaning you could collect data at any time; not just during natural hazard events, but theoretically during periods of economic crisis,’ he adds.
In developing countries, this method could prove cheaper and more scalable than traditional field surveys of business recovery. It is also privacy compliant, and has no geographic limitation, though Simini caveats that the results are only as good as the data – you may miss businesses that are less active on Facebook.
In developed countries, government evaluation of long-term business data is more sophisticated – however, the real-time nature of this new technique could provide dynamic support for economic monitoring. ‘We do want to be very cautious and test the methodology first to see how it works. Hopefully in the future this could have a global application as a monitoring system to spot early warning signals of unusual economic activity.’
As data powerhouses, many have already called upon Facebook and Twitter to use their data to assist during humanitarian crises, with some success. As data powerhouses, many have called upon Facebook and Twitter to use their data to assist during humanitarian crises. Recently, Facebook have started to look at how their data can be used to support non-governmental and humanitarian organisations in understanding the impacts of natural hazard events. They also have their ‘Data for Good’ campaign, which uses privacy-preserved data to aid humanitarian issues. Through the COVID-19 pandemic, Facebook are producing High Resolution Population Density Maps, Disease Prevention Maps and Social Connectedness measures to help governments and non-profits track the pandemic and the effectiveness of prevention initiatives.
Nevertheless, Simini adds that while social media data is starting to be released for humanitarian good, we must remain aware of privacy and quality issues. Such data sets only capture a sample of the population meaning social-media driven analyses should support existing models, not replace them.