The rapid global spread of COVID-19 has demanded that governments respond at breakneck speed to limit the human and economic costs of the crisis. And the scale of government response is unprecedented - a staggering 11 trillion USD in fiscal support provided by 55 countries in a matter of months – as is the demand for timely data to ensure that resources are targeted where they are most needed.
Having accurate and timely data has become the foundation of a resilient and effective government, and national statistical systems are central to this. These systems have been placed under extreme pressure during the crisis. Temporary office closures and disruptions to operations and surveys have jeopardised dozens of statistical publications. Increased demand for new and more timely data on the impacts of the pandemic have stretched available resources. And pragmatic decisions have had to be made to reprioritise and redeploy staff and resources as well as to innovate to deliver new statistical products.
The impact of COVID-19 on national statistical systems has been profound, and it will continue to have enduring impacts on how we collect, produce, and disseminate data for the foreseeable future.
National Statistical Offices (NSOs) that have fared the best during the pandemic are those that have invested in modernization and have the infrastructure and production processes in place to enable statistical work to continue remotely. For example, investments made in digital infrastructure, remote access work environments, and online collaboration tools have paid off big dividends during the crisis. For other NSOs, the crisis has led them to fast-track investments and plans to upgrade infrastructure and shift to remote work and online survey collection methods. In addition, NSOs that have been able to innovate and tap alternative administrative and big data sources have been more successful in meeting increased demand for timely data products.
From high-income to low-income countries, NSOs are harnessing new and innovative ways to assist governments to understand and address the social and economic fallout from the pandemic, including developing new types of surveys, employing new data sources and analytical methods, and engaging in new data partnerships.
With the suspension of face-to-face household interviewing, NSOs quickly adapted household surveys and field operations to focus on phone and online data collection. Staff have been retrained and redeployed to provide additional capacity to these functions. New rapid response surveys of households and businesses have been rolled out by mobile phone or online format in many countries with results published within weeks or even days. These have provided near real-time data to governments on key economic and social signals, including employment, income, financial stress and business activity.
For instance, household pulse surveys in the U.S. have monitored spending patterns, food security, mental health and educational disruption, while in the U.K. and Australia rapid surveys have monitored everything from alcohol consumption and exercise to the share of unpaid care and domestic work. These rapid surveys are effectively a new line of business for many NSOs – delivered much faster, but with smaller sample sizes, lower statistical quality and less depth of analysis. Such surveys do not meet the ‘gold standard’ often lauded by NSOs and statisticians, but they can reduce production time by over one year. As such, they should become a new tool in the statistician’s arsenal moving forward, providing timely insights from households and businesses for government and community use.
Rapid mobile surveys have also been common during COVID-19 in developing countries, focusing on households and businesses and often undertaken with support from partners and donors. For example, members of the Intersecretariat Working Group on Household Surveys have supported 480 COVID-19 surveys to date. However, with the flood of new surveys, there is a risk of poor-quality results which provide little information needed to inform policy. To ensure quality, the World Bank and others have developed guidance for undertaking rapid mobile surveys. Good practices for ensuring rigour include the use of an existing representative sampling frame, as has been deployed in recent rapid mobile surveys in South Africa and Bangladesh.
New Data Sources and Methods
NSOs and experts have also tapped administrative and big data sources and applied new methods to derive real-time measures of the economy. For example, the UK’s Office of National Statistics (ONS) is combining rapid response surveys with a wide range of real-time datasets to produce faster indicators of economic activity, including online job advertisements, price changes for high-demand products, energy performance certificates, foot and road traffic, and shipping data. The broad powers of the ONS to request data from other agencies has enabled access to new administrative data sources which are then linked to census and other datasets.
Additionally, the Australian Bureau of Statistics (ABS) has partnered with the government’s taxation office to obtain near real-time insights on the change in payroll jobs and hours worked based on payroll accounting software. ABS has since received 351 million Single Touch Payroll transactions and used these to generate 2.7 billion daily observations. ABS has also approached other partners, including supermarkets, banks and other private organisations to extend the use of private sector data assets, and 19 new statistical products have been launched to date.
Meanwhile, Ghana’s Statistical Service has partnered with Vodafone and Flowminder Foundation to produce rapid mobility estimates using anonymised and aggregated mobile phone data to support government interventions against COVID-19, and Colombia’s NSO, DANE, is monitoring prices of basic necessities at five-day intervals to address price manipulation of basic products and medicines.
The demand for information on the current state of the economy has also led to the advancement of methods for nowcasting estimates of GDP. These ‘flash estimates’ use large datasets (e.g. industrial production, sentiment indexes, interest rates, and traffic volumes) combined with big data analytics using statistical regression and machine learning algorithms to provide more timely estimates of GDP. Nowcasting and modelling methods have also been used by the International Labour Organization to model global losses in working hours as a result of COVID-19.
The COVID-19 pandemic has shown that national statistical systems can innovate in a time of crisis. The appetite for timely data is unlikely to subside anytime soon, and the collection and production of statistics may never fully return to business as usual. Important lessons we should draw from the crisis include the immense value of statistical modernization and digital infrastructure for ensuring business continuity, the considerable value in tapping administrative data and other big data sources for timely statistics, and the importance of building capacity within national statistical workforces to work with new data sources and embrace new methods.
It is critical that we don’t forgo traditional methods and standard surveys, however supplementing them with additional rapid surveys, administrative data and new data sources is the right way to go, and requires modern, agile and innovative NSOs.