key: cord-0018751-bb5f71nm
authors: Allin, Paul; Hand, David J.
title: Building back better needs better use of statistics
date: 2021-03-26
journal: Signif (Oxf)
DOI: 10.1111/1740-9713.01514
sha: 3a9a1b7765430032a090e7efed6020a0564aa3f9
doc_id: 18751
cord_uid: bb5f71nm

Paul Allin and David J. Hand call for official statistics to take centre stage

N ever have so many statistics been quoted so widely. The Covid-19 pandemic is being tracked though statistics on population health and wellbeing and on the state of health and care services, the economy, and livelihoods. Statistics from a variety of sources are pored over by politicians, policy-makers, commentators, business leaders and the public. For many, statistics provide points of reference in a torrent of opinion, claim and counterclaim about what is going on and how to manage the pandemic and its impact.

We also hear many calls that it is time to reset how we live our lives, to build back better, and to adopt a new narrative for societal progress. This chimes with a longrunning desire to go beyond tracking gross domestic product (GDP) as the single indicator of growth, moving to wider measures of sustainable wellbeing using data from a diversity of sources. What appears to be missing is a system for delivering statistics of the required quality, including timeliness, relevance, reliability, and coherence. Such a system must allow for a number of factors to be considered together, and balances drawn, rather than focusing on the measure of one factor at a time.

Official statistics systems have a crucial role in this. The vision that official statistics should be indispensable in democratic society is clearly set out in the United Nations' (UN) fundamental principles of national official statistics (bit.ly/3pb33wL), but this vision needs to be unpacked and acted upon. For example, GDP is probably one of the most widely mentioned of the thousands of official statistics. It is indispensable to economists. However, recent UK research shows that public understanding about GDP is very limited. GDP may be in the news, but people see it as an example of economic jargon, inaccessible and not important to their everyday lives (bit.ly/2LGnhZH).

Moreover, as we see in Covid-19 statistics, official statistics are not the only source. Statistics are also published by independent, peer-reviewed, research centres, including scientific modelling and projections of the possible course of the pandemic. We have become familiar with R, the reproduction number (average number of secondary infections produced by a single infected person) in this way. At least in the UK, and we suspect elsewhere, the figures that grab most attention are daily counts of identified new cases of Covid-19, hospital admissions, and deaths of people who have tested positive with the virus. These are invariably described as government figures or statistics, which in everyday terms they are. However, they are not formally official statistics unless they have been so specified. With that designation comes the promise of meeting high standards of production, analysis and presentation and, crucially, the requirement to show how these have been met. Of course, the administrative data from which counts are derived are primarily and hugely valuable for managing essential public services.

So, how to deliver on the UN vision -and get official statistics widely used -in an increasingly complex data ecosystem? As we see it, the goal is for statistics to be used OpiniOn SIGNIFICANCE widely. Who provides a specific set of statistics is not important, only the quality of those statistics, especially their practical utility. As for what widespread use means, it is certainly not just within government. The key concept here is of statistics aimed at serving the public good. As well as informing government, statistics should also inform business and planners in all areas, and help citizens make effective decisions in their lives. This is also about helping to hold government to account. The totality of all these uses makes statistics "part of the lifeblood of democratic debate" (bit.ly/3a8uV8I).

One potential sticking point is a lack of understanding of how and where statistics are used. Much is documented about concepts and methods used in the production of statistics; much less about why statistics are produced and who is using them. The predominant approach in official statistics seems to be for producers to work most closely with their stakeholders in government departments and central banks. That is but part of serving the public good.

We should also recognise that, while statistics and other evidence can inform and illuminate policy options, politicians make policy decisions on more than evidence. Similarly, we as individuals make decisions, including whether or not to change our behaviour, based at most in part on the information we have. Other factors, such as beliefs and attitudes, also play a role. In short, reason does not prevail in human life. Improving society by following the evidence is tempered by "the roles played by custom, habit, passion and the imagination", 1 even where democracy calls for decisive, transparent and informed decisions. This does not devalue the use of statistics but, rather, stresses their importance in informing our decisions.

On the basis of our research, and our experience of UK and other European statistical systems, we conclude that official statistics systems can be more effective and increase their public value. We acknowledge that there is much under way that should be built on. Technical and methodological standards are crucial in delivering official statistics but are not sufficient. More attention should be given to those aspects of quality that address the fitness for purpose of statistical products and outputs. More trust has to be earned by the producers of official statistics. Producers of statistics should work with others to turn indicators and statistics into messages and accurate narratives that can be understood, trusted, and used by people and businesses as well as by government. National statistics offices must engage more with the public, businesses, and civil society, as the UK statistical service is now committed to do.

The Covid-19 pandemic is a testing time for statistics. It is showing that single sources struggle to provide statistics that are timely, relevant and reliable. Timeliness is at a premium when political decisions are urgently needed, so administrative datasuch as counts of vaccinations, cases and deaths -are inevitably highlighted. Such counts can be accompanied with guidance (see bit.ly/2LMyvMs, for example), and the challenges of using administrative data for statistical purposes are documented, 2 but administrative data still need to be handled with care. As the Welsh first minister observed, "If you are testing more people you will identify more positive cases" (bit.ly/3qfN7CM) -especially pertinent when testing regimes have been greatly expanded. Thus, even reported cases per 100,000 of the population are not reliable estimates of rates of infection. For those we need to wait around a week in the UK, for the official survey statistics on infection rates (bit.ly/3jMAYTG), during which time the narrative has progressed as further administrative data counts are published.

We conclude that official statistics systems should be broadened. They should generate quality public statistics that can be used to help get answers to the many urgent questions about society and how we can sustainably improve our lives and livelihoods. National statistical agencies would be at the heart of such systems, perhaps importing more data from other sources, but also providing leadership, and curating and quality-assuring sets of statistics drawn from official and other organisations.

We are heartened that others have the same kind of approach in mind, like Julia Lane's recent call in Significance for "a new, more democratic, public data infrastructure" in the USA. 3 Similarly, MacFeely and Nastava Paul Allin is a visiting professor at Imperial College London, following a career in the UK government statistical service.

are concerned by the huge challenges faced in many countries in compiling indicators to assess progress towards the UN's Sustainable Development Goals. They propose that official statistics switch to "a mixed business model: one combining the manufacture of official statistics with the franchising of production under license", with accreditation based on the fundamental principles for official statistics. 4 Above all else, public statistics need to feature even more in debate and decisionmaking, so that we are not constrained to considering progress -or other aspects of our societies -simply in terms of the indicators we have, as opposed to those we would prefer. This is not about waiting for perfect statistical systems. As Gus O'Donnell, former UK cabinet secretary, puts it: "Of course, measurement is hard, but roughly measuring the right concepts is a better way to make policy choices than using more precise measures of the wrong concepts" (on.ft.com/3d1pLNF). Delivering a major change in direction and helping us all go beyond GDP and other established measures will be challenging for official statistics systems -but it is essential.

The Infidel and the Professor

Statistical challenges of administrative and transaction data

After Covid-19, the US statistical system needs to change

You say you want a [data] revolution": A proposal to use unofficial statistics for the SDG Global Indicator Framework

Hand is emeritus professor of mathematics and a senior research investigator at Imperial College London, where he previously chaired the statistics section