Closed: Crises, politics and statistics (Closed)

I put below my contribution to the discussion. For easier reading, I have broken it down in three sections.

DATA CURRENTLY AVAILABLE
I share Walter’s view that, in the peak of the crisis, data that had to be hastily used to monitor evolution of cases (infected, dead, recovered) had enormous shortcomings. I refer to my LSS article were some of them are mentioned. But I would just stress again the impossibility to draw international comparisons. Just looking at the data, one cannot explain easily why some countries start to ease lockdown, while others maintain strict confinement restrictions.

Yet, despite so many concerns about currently available figures, these have been used, and more interestingly, have been useful for extremely difficult decisions. Perhaps statisticians should reflect on whether to take more on board the definition of quality as “fit for purpose”. That sometimes, timeliness is an overwhelming criterium for usefulness.

I also agree with Steve MacFeely’s concern about the limited utility of some Covid platforms posted by statistical organizations, sometimes gathering whatever found in their drawers. But, though I have not carried out any exhaustive study, I could pick up a couple of examples on the positive side. One is the data on displacements at municipal level (district level for large cities) that daily publishes INE Spain. These could be very valuable to assess the impact of lockdown measures as they are gradually lifted. Another is the very timely data from death registers published by INSEE and ISTAT, also at municipal level. With only a few days delay, it provides very interesting insight into pockets of non-infected persons, just to mention a single specific use.

DATA MISSING AND NEEDED NOW
As Dennis Trewin mentions for the case of Australia, many countries are launching epidemiological surveys to follow evolution of main categories of individuals with respect to the disease. My two countries (Luxembourg and Spain) are launching them now. Again, this is a challenge for official statisticians due to the complexity of the operation which has to rely mainly on non-statistical expertise. Are the roles played by each actor well defined? Taking the importance of the information gathered to inform deconfinement measures and learn more about the disease immunity and contagion capacity, this is an enterprise that shouldn’t be allowed to fail.

Real time data on health care system capacities and its response levels are very important to divert cases to less congested facilities. These are not statistics as such, but perhaps statisticians with their expertise in communicating large amounts of information could play a role in building the processes needed hereon. This is in line with my remarks in the next, and last, section.
I refer to my article that mentions some other datasets that will be needed. https://luxstat.wordpress.com/opinions/

TRANSFORMING THE STATISTICAL PROFESSION
I fully agree with Walter on his analysis of the role of statistics on an emerging, yet to be defined, society.

Current crisis has caught official statistics, as many other professions, unprepared; thus, unable to provide rapidly the answers authorities needed to decide on a solid basis. Hence, policies have been supported by those data available, notwithstanding their shortcomings, to take decisions of unprecedented importance.

I believe it is now the time for statisticians to play a key role. For this purpose, they will be obliged to question methodologies and principles so far unquestionable, but that nowadays do not have the speed to adapt, amongst other features, that is expected. I think I share on this point Jmdurr’s remarks.

In parallel it corresponds to institutions to take a significant step towards a digitalization of their processes that, apart from increasing their efficiency, will allow for a very rapid production of the quantitative information essential in situations like the one we are suffering today.

And to society as a whole to acknowledge that increasingly it will have to rely on uncomplete information, whose quality is not solidly stated, perhaps not even measurable.

In my view, and from a purely statistical perspective, an opportunity has opened now and it should be seized. It would be regrettable that all the collective knowledge potential and management capacity enshrined in our profession, is not put to the service of society in times when it is so much needed. But to that end, programmes, processes and methods will have to be reinvented (I mention as very simple instance electronic capture of digitised data as the sole information source). Most industrial and institutional sectors will have to launch gigantic transformations in order to survive or remain relevant; we cannot lag behind.

My thanks to Walter for provoking this very timely, interesting and essential discussion. I had intended to contribute to it since the moment I read it.
In the meantime, I have written an article on the issue that was published early this week by the Luxembourg Statistical Society (LSS), in fact, it is the English version – not a translation – of a text in Spanish that I wrote on 13th April and posted on Linkedin a couple of days after.
You can see the article on the website of the LSS:

https://luxstat.wordpress.com/opinions/

Later today I will comment on some aspects of Walter’s paper and the ensuing discussion. Parts of my remarks will be drawn from the article; others will be new thoughts after a few days of further reflection.

The need for estimates from official statistics during a global crisis

The health crisis that broke out at the beginning of March simultaneously in most European countries created a completely new situation from the perspective of identifying data sources that would allow to assess its short-term impact on the economy. Official statistical quality data, calculated on the basis of well-defined methodologies, using traditional sources, are published almost two months after the end of the reference period, without diminishing, under normal economic conditions, the usefulness of the statistical results.
What happens when unexpected events occur that suddenly change the behavior of the economy and of the whole society? National statistical offices are required to seek the best compromise between quality and dissemination speed of statistical data and information.
The American economist of Romanian origin Nicholas Georgescu - Roegen, the founder of bioeconomics, compared the economy with a human body, in The Entropy Law and the Economic Process, using the principle of analogy. We could, therefore, compare the current crisis in the economy with a serious accident on a highway. The medical analyzes of a person involved in an accident collected one day before its occurrence are only partially useful at the time of the accident, since, at the beginning of the accident, primary but summary investigations will be performed, using tools specific to emergency interventions, and later, in specialized medical units, exhaustive assessments will be carried out using sophisticated medical instruments.
Similar, in assessing the effects of the impact of the crisis, the official statistics will have, by specific methods, to provide preliminary data, although not very detailed and which are not characterized by the rigor specific to the official statistics, but which are obtained in a very short time and with a tolerable margin of error.
In order to meet these challenges, national statistical institutes and international bodies have rapidly created instruments to measure the contraction of economic activity, when traditional channels for collecting primary data have been affected. An eloquent example France, INSEE carrying out the first ad - hoc research to estimate the effects of this health crisis on the economic environment.
INS Romania has started its own ad-hoc researches, in order to measure the impact of the health crisis on the economy (https://insse.ro/cms/ro/covid-19-impactul-economic).
The first, carried out in the first days after the outbreak of the pandemic, aimed at assessing the impact of the health crisis on the economy starting from applying a questionnaire to the statisticians of the Territorial Statistics Directorates. Over 800 respondents gave a first picture on the size of the economic catastrophe that was about to occur in March and which is expected to continue in April.
The second ad-hoc research aimed at evaluating the size of the activity volume reduction in the economy and the reduction of employees number. Qualitative data collected from a representative sample (over 8000 companies) were used, at national level, but also by activity sectors, but also other quantitative statistical data. It was estimated a reduction in the volume of activity in the economy compared to a month with normal activity with 30% in March and 40% in April, respectively.
The third research led to the estimation regarding the reduction size of the foreign trade activity by categories of companies grouped according to their size, the number of employees, the form of property, etc. The research was carried out on a representative sample of more than 1700 companies covering at national level about 70% of the foreign trade activity of the last month.

The prolonged existence of COVID-19 and the actions to manage it nationally, regionally and internationally provide national statistical offices with challenges that only two months ago were unforeseeable. Statistical offices have much to gain from working together and sharing experiences and ideas. Few countries will have the capacity to meet all such demands, or even anticipate what they might be. What has changed fundamentally is that at a macro-level, the health of the population and the economic capability of countries have become intertwined on a scale that is far outside what our information systems, institutional and managerial capability, supply chains and connectivity have been designed for. The immediate action taken to close down economic activity and limit personal interactions is unsustainable. Lockdowns are likely to have increased inequities through differential access to even basic necessities.
Much of our current statistical capability is anchored in information sources which enable regular comparisons of levels of activity, over common periods of time and between sectors. Such measures are usually designed with a fixed frequency and report with a lag. Coherence is usually strong within types of statistics (e.g. economic statistics, demographic measures), but much less strong between types. The questions we need to address now are not those we have planned for. We need more frequent, more immediately available information, but cannot compromise on representativity, quality measures or coherence. We need sources to be more granular, to be able to pinpoint emerging concentrations of need, and measure extra-ordinary changes to rates of change almost in real time. There will be new demands placed on information about health services, government economic leadership, social cohesion and community solidarity, and the reach of redistributive programmes. The statistical infrastructure and methodological expertise in statistical offices has become an even more vital national resource, and it is important that there is a good understanding of where it needs to focus beyond enabling the large array of existing sources to continue in the face of COVID-19. We are already seeing how new information types have become critical to shaping government decisions. There seems to be at least eight elements of the statistical implications of COVID-19:
1. Managing effectively the medical consequences of the COVID-19 pandemic.
2. Ensuring that the means exist to meet the needs of service organisations of all forms for access to basic population statistics.
3. Providing statistical methods for COVID-19 surveillance testing once the quarantine periods wind down.
4. Ascertaining and meeting the information needs for staging the removal of lockdowns.
5. Monitoring the effect across the population of limiting the forms of service delivery.
6. Rethinking the scope, frequency, timeliness and granularity of public statistics.
7. Reassessing statistical priorities and rethinking priority areas for rapid statistical innovation in the face of possible fiscal restraint.
8. Strengthening the connections with expert users in government, community and international organisations.

Len Cook New Zealand
16 April, 2020

As a more lengthy contribution to this discussion platform, following an invitation by Pieter Everaers, please find following an op-ed published in Canada’s Globe and Mail April 13, 2020: https://www.theglobeandmail.com/opinion/article-during-the-pandemic-why…. (This newspaper is a bit like a cross between the New York Times and the Wall Street Journal for Canada.)

In order to make this op-ed more accessible for an international audience, I’ve interspersed additional text to provide some further context and amplification. The text of the original op-ed is indicated for clarity.
To start, up to the week prior to its publication, there was increasing public pressure for the federal and provincial governments to be more open about their projections of the COVID-19 epidemic curves, especially the numbers of cases, deaths, and hospitalizations, and how these trends were likely to affect hospital capacity including ICU beds and ventilators. This pressure in Canada increased as the UK projections from the Imperial College group were disseminated, and increased even more as those projections plus the ones from the IHME group in Seattle induced a dramatic change in the US federal approach.

So finally, in the preceding week, the federal government and several provinces provided some such data. However, from a statistical and epidemiological perspective, these data and projections were very limited, and they generally did not extend far enough into the future to inform Canadians when the stringent physical distancing and lock-downs could end.
Further, I had good reason to believe that one of the impediments was the widespread lack of coherent and timely data. Part of my personal knowledge stems from having been responsible for Statistics Canada’s health statistics program from before 1990 to my retirement in 2009. In this role, I had struggled continually to improve Canada’s health data infrastructure, but repeatedly faced blockages. I was also responsible for a group of truly excellent microsimulation modelers, where among others we had built models of SARS and H1N1.

(op-ed) Canadians are finally beginning to see projections of COVID-19 cases, deaths and needs for intensive-care units from various provinces and the federal government. We are also starting to see simulations (https://www.cmaj.ca/content/cmaj/early/2020/04/08/cmaj.200476.full.pdf) that look beyond the next month or two when, hopefully, epidemic curves are clearly flattening.

The simulations cited in the previous paragraph were done by one of Canada’s leading infectious disease epidemiologists In an interview on TV the week before my op-ed was published (https://www.tvo.org/video/ep-18-ontarios-oncoming-covid-19-toll?), he made extremely critical comments about Ontario’s modeling, and complained strongly about his inability to access the needed data. In general, it is likely that university-based modelers in Canada have stronger analytical capacity than staff within government ministries of health.

(op-ed) Canada’s national data-collection capacity will be critical for the next stage of the pandemic, when relaxing of the stringent physical-distancing measures can begin. Yet our data-collection infrastructure is proving woefully inadequate.

(op-ed) To be effective, an extraordinary and co-ordinated national effort is required, with much more extensive testing and real-time standardized reporting of results, from local to provincial to federal agencies. These data on the tests will be much more powerful for managing the pandemic if they also include pre-existing diseases and risk factors such as smoking.
In Canada, the provinces have the bulk of the responsibility for delivering health care. They also each have their own approaches to data collection. As a result, it is extraordinarily difficult to assemble coherent national-level data. Further, data collections are typically siloed. But for sensible management of the pandemic, especially in the upcoming “relaxation phase”, it will be critical to have not only real-time coherent data but also multivariate longitudinal data.

For example, there are widespread indications that susceptibility to more severe breathing problems is associated with various comorbidities. But there are insufficient data to understand better just which comorbidities are most important – is it the heart disease or diabetes themselves, or possibly the drugs patients have been taking to control those diseases that are the true risk factors. It is also important, for the more sophisticated kinds of modeling, to have better data on the distributions of times between events like admission to the emergency department, being put on a ventilator, and then recovering or dying. The need is for longitudinally linked microdata covering the gamut of patients’ health care encounters, not siloed or aggregated data.

(op-ed) These kinds of data flows are obviously feasible with current computing and communication technologies. Indeed, they were feasible 20 years ago when the federal government created the Canada Health Infoway corporation and provided it with billions of dollars. One of its missions was to work with the provinces to develop interoperable real-time “outbreak detection” systems.

(op-ed) Had these systems been in place even as late as last year, Canada would not have wasted critical weeks and months in reacting to COVID-19. And if these systems were in place now, we could manage relaxing the current lockdown phase with “smart quarantine” (https://www.nytimes.com/2020/04/07/opinion/coronavirus-smart-quarantine…) and reap the major benefits of returning the economy to normalcy at a faster rate.

Health Infoway was created with the mandate from the outset to work with the provinces to create essentially a standardized pan-Canadian inter-operable electronic medical or electronic health record (EMR or EHR). However, the main mechanism they had to influence the provinces was a 50%, and in some cases a 75% cash subsidy for the software development. Further, given Infoway’s judgment that their focus had to be on patient care, and that they had to be very careful not to raise concerns among the leadership of the medical profession, they continually refused to include in their work anything that made reference to “health system uses” of EMR or EHR data.

(op-ed) So why do we still not have this real-time standardized data-reporting capacity?

(op-ed) One blockage is the constitutional conflict over jurisdiction; the provinces claim almost exclusive jurisdiction over health care. The federal government also plays a substantial role, spending billions on health research and fiscal transfers to the provinces and regulating drugs and devices – on top of the billions given to Infoway – but it has been too timid (https://www.cmaj.ca/content/179/1/9) to use all its powers much beyond ineffectual cajoling.
It is unclear why, at the very highest levels of the federal government, there has been such reticence to use the powers it does have under the constitution, including exclusive jurisdiction over “statistics”, to be more forceful in compelling the provinces to establish the requisite data systems. One possible opportunity in the tragedy of this pandemic is that the very real felt issues with current data flows will finally lead to more effective action.

(op-ed) Another blockage is fear of transparency. It has taken strong public pressure for governments to begin providing even limited epidemic-curve projections on which their policies are based.
Provincial ministers of health are understandably leery of providing the kinds of detailed data which, when carefully analyzed by others, can be used to produce embarrassing information. Further, there is a long history of strong, effective, but behind-the-scenes resistance to providing high quality data to others by the medical profession.

(op-ed) Of course, we need to ensure patients’ sensitive health data remain confidential except as needed in their circle of care. However, as the Council of Canadian Academies noted in its 2015 report (https://cca-reports.ca/reports/accessing-health-and-health-related-data…), data custodians too often use privacy concerns to block access, stymieing major benefits of health research and, in the current emergency, support for both smart quarantine and much better modelling and projections.

In a phrase, Canada has long suffered from a “privacy chill”. This is complicated in the past few years by the very real and growing concerns about the sometimes awful behaviour of the huge private social media corporations. It is essential for NSOs and the bona fide academic research community to make a clear distinction between the public good benefits they can produce with highly sensitive and confidential patient data, and the private profit oriented motivations of the far more powerful social media corporations.

(op-ed) What can we do about these completely unacceptable blockages? There are several places to start.

(op-ed) The Canadian Medical Association (CMA) can offer strong leadership by supporting real-time interoperable data not only for their own interests and individual patient care, but also for broader health-system uses, not least for epidemic detection and management.

The CMA released a study in February 2020 (https://www.cma.ca/new-report-roadmap-improve-virtual-care-canada) where they call for essentially a real-time inter-operable EMR system that works across all of Canada. However, most of the focus in this detailed study is on aspects of interest to doctors themselves such as billing and liability, with the balance on how this will improve patient care. There is basically nothing on the potential for “health system use” of the resulting data flows.

(op-ed) The private-sector vendors of electronic medical-record systems can immediately cease their profit-capturing data blockages and allow their software to interoperate in real-time with those of other vendors and government systems.
In a number of these systems, the only way to export a patient’s data is as a pdf file, which is completely useless from a statistical perspective. It is obviously in the self-interest of these software vendors to make it as difficult as possible for a provincial government or a doctor’s office to purchase and migrate to a competitor’s software. From its inception, Infoway was supposed to prevent this kind of vendor behaviour. Provinces have the power to force doctors to use only EMR software that does provide inter-operability, though in general they have not done so.

(op-ed) Provincial governments can agree quickly on more in-depth and uniform data standards for hospitals, labs and physicians so that, along with the federal government, they can quickly and unambiguously assemble these data, especially virus-testing results.

(op-ed) Privacy commissioners need to alleviate the excessive concerns over privacy around health data, to rise above responding only to complaints, and to make it clear that – especially in this emergency situation – they support essential data flows, provided that basic privacy protections are in place.

As far as I can tell, Canada’s privacy commissioners have been totally silent, unlike in New Zealand, for example: https://www.nzherald.co.nz/business/news/article.cfm?c_id=3&objectid=12….

(op-ed) The Public Health Agency of Canada and the provinces can open up their data beyond a few pages to the energy and creativity of Canada’s excellent university-based health researchers and modellers, and support the CIHR-funded pan-Canadian network (https://dash.hdrn.ca/).

In many areas related to the pandemic, there is an explosion of innovation and creativity around the world, from the search for a vaccine to fabricating ventilators to devices for testing for antibodies. There is an analogous potential for epidemiological data analysis and modeling, but it is being stymied in Canada by the very poor quality and extremely limited data being made available.

(op-ed) In turn, Statistics Canada can expedite a virtual form of its Research Data Centres so that bona fide health researchers can access much higher-quality data with appropriate privacy protections.
Of course, much of the most powerful data for these kinds of analyses will be patient-level longitudinal microdata, whose confidentiality must be protected. Statistics Canada has a network of university-based Research Data Centres (RDCs) within which certified researchers with certified projects can access such data. But they have all been closed as part of the lock-down. Compared to the Netherlands NSO, Statistics Canada has been a real laggard in developing virtual RDC data access. In order to harness and improve dramatically the extent and quality of all kinds of pandemic-related statistical analysis, Statistics Canada could be moving much more aggressively to provide virtual data access, though perhaps in the first instance to a more limited group of bona-fide researchers.

(op-ed) The federal government must assert its leadership and authority, using its constitutional powers, to set critical national standards and enforce the collection, sharing and use of public-health data – and finally bring Canada into the 21st century of critical data infrastructure.

I’m hoping, in the climate of the current pandemic, when all sorts of unprecedented public policy initiatives that would be unthinkable in more normal times are being implemented, that the ideas sketched in this op-ed can be acted upon. However, these ideas have been around in Canada for decades, and have not been acted upon up to now. We need to start by understanding why, including the various blockages and vested interests. Assuring the data and analytical infrastructure for managing the pandemic, and over the longer term for maintaining ongoing pandemic preparedness, is a vital role for official statistics.

(This work is licensed under a Creative Commons Attribution-NoDerivatives 4.0 International License.
This means that you are free to reprint this article for any non-profit or for-profit purpose, so long as no changes are made, and proper attribution is provided. Please credit the author and QUOI Media Group when you reprint this content. And if you let us know that you’ve used it, we’ll happily share it widely on our social media channels: quoi@quoimedia.com.)

Do the European Centre for Disease Prevention and Control COVID-19 daily tables count as "official"? https://www.ecdc.europa.eu/en/publications-data/download-todays-data-ge…
They are invaluable for comparing epidemic curves across countries.
But even these data must be used with care as disease detection regimes are not identical everywhere. Dennis Trewin makes good points about the "unknowns"

It seems to me that the diagnosis an some ideas on the role of official statistics in the epidemia or pandemia can be found in my paper "The role of official statistics in fragile countries, regions and branches of economy" published in the SJIAOS Vol. 33, No. 4, pp. 833 - 849. I take the liberty to recommend this paper to the managers of statistical offices (national and international).

Thank you for the nice discussion. There are many stimulating questions, and I will respond to a few that I am particularly interested in as a young sociology researcher.

First, about the "social and political utility" of official statistics. It is indeed clear that new statics are of primary importance for political decision-making and for informing the general public. Also, let us remember that there is the covid-19 pandemic, but also the pandemic of fear, which is at least as important, if not perhaps even more important. It is the main object of study in certain research circles in sociology, anthropology, social psychology, and even among our historian friends. So, the dissemination of official statistics contributes to reduce the globalized phenomenon of the fear generated by this pandemic while directly preserving the psychological health of citizens at a time when a large part of the world's population is under various forms of health confinement. And it seems that "fake news" circulates faster than "real information", especially in crisis times.

This brings us to the second point I would like to discuss, the new requirements in terms of " rapidity " and innovation in the dissemination of statistical analysis. Today, small, very up-to-date analyses are sometimes more important than large studies that provide relevant but belated insights. More specifically, to provide useful knowledge to decision-makers and citizens, official statistics are called upon to be as reactive as possible with regard to the actuality and real evolution of social, economic, institutional and political dynamics around this covid-19 pandemic. But all scientific knowledge is a scientific construction that takes time. Moreover, on a point already made in previous opinions, this is indeed the area of expertise, precisely because expertise is certainly less in-depth knowledge than scientific research, but it is rapidly produced. Unfortunately, expertise often stifles official statistics, but it cannot replace them, especially at times of crisis when the latter need projections of various scenarios. "Forecasting" is what can be of most help to decision-makers today. And , improving administrative statistics is a very important future challenge in this way.

Thirdly, I think it is time to ask new questions at a time when this pandemic is revealing the complex degree of hyper-connectivity of our societies. Perhaps we already knew that our world is a "globalized village"?!? Today we are internalizing the meaning of this notion. Isn't the interdependence of the targets of the SDGs ultimately a much more complex interdependence than we imagined? This despite statistical models and despite the tools for visualizing this mesh of connections. Thus, your question in relation to The Achievement of the 2030 Agenda, "will it be more difficult after this pandemic? "is a more difficult question than it seems.

I can of course only support this comment. This is the most concise and up-to-date example of what statistical offices can contribute by providing their services (besides compiling their own statistical products).
In the longer term, however, we should ask ourselves whether this will change the orientation of our industry to the effect that we will increasingly become a service provider. Such services should generally have to do with our strengths and experience: For example, we could help manage (and certify) the quality of statistics, even if they are produced outside our own factory walls.

In these moments of the unexpected Covid-19 crisis, the statistical community, as mentioned by John & Michel, has been reactive at different international, regional and national scales.

The NSOs are suddenly faced with major and urgent challenges: Ensuring the continuity of their work in the confinement circumstances, communicating with their users about the adaptations introduced, responding to new urgent requests related to the crisis and its impact on economies and societies as well as revising more frequently the macroeconomic frameworks with all the difficulties linked to the availability of data.

Given all those challenges, the use of technology has been crucial to ensure telework, communicate with partners, conduct surveys (telephone and internet), e-manage administrative procedures, e-learning, etc. ....The focus is to respond quickly to the needs of the moment and prepare for the next phases.

That said, these challenges have imposed themselves all over the world, but their magnitude is not the same. For countries whose production of statistics is mainly based on face-to-face surveys, the challenge is much greater.

I think that this crisis has shaken the world in its different spheres and the same goes for the official statistical community. This crisis would boost the implementation of the digitalization processes in developping countries, the massive use of administrative data and the use of new methods and data sources, e-learning platforms, etc.

The crisis is certainly cruel but timely to drive change. More than ever, the statistical community needs to collaborate and put partnership at the heart of its concerns.

Thanks a lot to Mr. Walter Radermacher for opening this discussion and IAOS and John Pullinger to host.

NSOs priority is the health of their workers.

I do agree with Dennis Trewin on "Using Random Sampling to Learn About COVID-19 Known Unknowns".

If you don't do any blood tests then you don't have high cholesterol. The cases of COVID-19 depend on how many persons each country tested and how they were tested. We cannot take decision on this variable.

In my opinion, NSOs have to lead random samples, jointly with national epidemiology authority, to test the COVID19 and to produce trustable statistics. Currently, we do not have trustable statistics on COVID19. Without data on the COVID-19, we cannot possibly understand how the pandemic is progressing locally.

About the next phase? we should be clear:

We face an unprecedented situation, and we will have unprecedented data and unprecedented quality data.

The covid-19 crisis is a real challenge for National Statistical Offices (NSO). Their usual tools (surveys, administrative sources) are not performing as usual: their staff is mainly working from home and business managers are struggling to adapt to the loss of activity and thus less available to respond to surveys. On the other hand, the need for timely data is critical in a time when the economic and social situations change abruptly.

But this is also an opportunity to develop the use of new sources and new methodologies, yet experimental, but of paramount interest for governments and economic stakeholders to monitor the effects of the crisis and prepare mitigation policiy responses.

As John noted, International organisations play their role in gathering how NSO adapt their methodologies to produce relevant outputs. For example, the UNSD launched the "Response of statistical systems to COVID-19" website (https://covid-19-response.unstatshub.org/) to provide a space for the global statistical community to share guidance, actions, tools and best practices to ensure the operational continuity of data programmes by National Statistical Offices, and to address issues of open and timely access to critical data needed by governments and all sectors of society to respond to the global COVID-19 crisis.

Similarly, the ILO collect information on how countries adapt the Labour Force Survey to the crisis (https://ilostat.ilo.org/topics/covid-19/covid-19-impact-on-labour-marke…)

Many countries develop new techniques to provide real time information. For example, the French statistical institute, Insee, now updates every two weeks its economic outlook using alternative sources such as credit card data, electricical consumption, railway trips...to estimate the impact on the GDP and household consumption (https://insee.fr/fr/statistiques/4478214?sommaire=4473296)

The benefit could be to weaken the traditional reservation towards methodologies or sources that don't fully comply with the highest professional standards. It forces us to be more pro active to society needs, thus giving a high visiblity to the utility of official statistics, in a time where it is sometimes challenged. I believe that at the end of the crisis, things will not be back to business as usual. We will be more open to innovative and experimental methodologies, but at the same time stakeholders will be more demanding to fresh and timely data.

The response of the official statistics community around the world to the Covid-19 crisis has been extraordinary and impressive. Connections with those charged with making urgent and life saving decisions have been deepened, new service offerings have been developed and implemented, innovative ways of gathering and analysing data have been put into effect and partnerships strengthened at local, national and global levels. It has been inspirational to witness, all the more so given the personal and organisational challenges being faced. Resources I have found valuable include those created by the United Nations Statistics Division ( https://covid-19-response.unstatshub.org/ )and PARIS21 ( https://paris21.org/news-center/news/Our-Response-to-COVID-19 ).
Without good quality, timely and trustworthy official statistics, decision makers are flying blind in their response and the public is left ever more confused and anxious about what is going on in the world around them. At present colleagues are working flat out to deal with the current situation and there is no time to think ahead. However, we will have an opportunity to learn from the response to Covid-19 to establish new capabilities. Strong statistical capability will help our countries and the world chart a course, not just through Covid-19, but through the economic, social and environmental challenges that lie beyond.
I think that Walter Radermacher is right to stress the opportunity provided by the Sustainable Development Goals to provide a framework (and also a timetable) for action. Indeed, only a few days ago. the Deputy Secretary General of the United Nations, Amina Mohammed issued an encouragement to the data community to do just that ( www.data4sdgs.org/news/un-deputy-secretary-general-participates-data4sd… ). She stressed the role of the Global Partnership for Sustainable Development Data, highlighting that it is through partnership that we will be effective in our response.
In this 200th anniversary year of Florence Nightingale, the first female member of the Royal Statistical Society, we can be encouraged by the transformation her statistical work heralded. As a nurse, she was on the front line putting her life at risk, as are all the health and care workers and others providing us all with vital services today. As a statistician she worked in partnership to ensure that the voice of evidence stood out to challenge misconceptions and provide a sure guide for decision making. She can be a model for us now across the statistical community as we “step up, step forward and step on the gas” to mobilise the power of data to help our countries and the world as a whole make the decisions that will get us through the current crisis and towards a brighter future beyond.

Using Random Sampling to Learn About Covid-19 Known Unknowns

Walter’s discussion opener outlines the important tole of official statistics. However, he does not mention what I believe is the most important contribution – support for a national survey to better understand what is happening with the population at large in terms of COVID-19 infection. The COVID-19 tests to date have focussed on those with symptoms or judged to be at risk, clearly not a random sample. This has been appropriate in the earlier stages particularly when the availability of tests is limited. However, we need different data to understand a possible second wave and when to lift restrictions.

NSOs have strong skills (or should have) in survey design, questionnaire design and how to manage uncertainties in the collected data (eg non-response). They also have good sampling frames which are not necessarily available to other organisations.

The tests would have to be done by health professionals so I would not expect the national statistical offices to be involved in the data collection itself.

The focus should be on the data that is needed but we haven’t got – the known unknowns? There are several categories: people who haven’t had the virus, those who have the virus but are asymptomatic, those who have the symptoms but have not been tested, those who are actually ill, those who have had the virus and recovered, and potentially those who have recovered and then lapsed.

In plain language, Governments all over the world have been flying blind when it comes to the level and nature of undetected COVID-19 in the community. These data are essential in enabling Governments to gain a better understanding what is driving community transmission and the possibility of a second wave.

With the actions taken to date by Governments, there is definite evidence of slowing the trend. There have been massive economic, welfare and social cohesion costs to achieve this. A plan to return to some form of normality will need to guard against a significant second wave and the inherent pressures on health systems. These plans will no doubt be informed by complex models developed by epidemiologists to forecast the evolution of the pandemic. And the quality of the forecasts will depend critically on the quality of the data supplied to the models.

It is relatively straightforward to implement a random sample and testing selected persons to identify which categories they fall into. It would need to include children not just adults.

By randomly testing people every week, we can capture invaluable information. We can build up estimates of the proportions of the population in each category, and how these proportions are changing with time. We can use demographic information about people sampled to gain insight about differential risks of different groups according to age, gender and socio-economic factors.
This work would have to be done in close association with epidemiologists.

Although, the ABS is not yet undertaking a survey of the type described above, they have done a number of other things to support COVID-19 policy analysis such as a survey of business activity and expectations, financial and employment impacts of households, and early release of parts of key publications. They are intimately involved in the national co-ordination effort.

Official statistics should be on board to help in crisis management

I agree. But in a useful and coordinated way. At the moment there's a scramble to put out information. Every agency has some sort of Covid platform, including my own. Are just adding to the noise I wonder?

An important question, I think, emerging from the pandemic is why were Google allowed to publish detailed statistics, harvested from their applications and platforms, on population movements, but NSOs have been prevented by Data Protection Commissioners from accessing and using similar data?

On March 27 the UK Information Commissioner’s Office noted that “data protection law enables the data sharing in the public interest and provides the safeguards for data that the public would expect”. Should we infer that official statistics are not compiled in the ‘public interest’?

How do we address this? Social license is elastic - it arguably contracts when times are good, but loosens in times of crises. It appears to be quite loose at the moment. Should official statisticians now try and make the case, that we too should have access to important secondary data sources?

Much to ponder...

Thanks for opening this discussion. Timely and provocative. As usual with Walter's commentaries, there is so much to react to. So I will probably reply in installments.

Now epidemiology is most important, in the next phase other statistics will become relevant again! Yes, I think this is certainly true - we see evidence of it already.

The Secretary-General of WHO has been clear, a key action guiding the path to containing Covid-19 is to 'measure, measure, measure'. I hope policy makers don't forget this message. Not only is it a key action guiding the path to containing Covid-19, it will also make other paths more clear too, not least, economic recovery.

It is has been interesting to watch the re-emergence of 'expertise' during this crisis. Again, I hope statistics as a profession, will be afforded the same respect, in the months and years ahead. But how do we as a profession ensure this is the case?

More to follow...