Governing-by-the numbers – Résumé after one and a half years

2.1The value chain and the co-production of statistics and society

In the opening contribution (‘Governing-by-the-numbers Reflections on the future of official statistics in a digital and globalised society’) [1], a distinction between data and facts was introduced, in order to better consider and name different phases of the value chain from raw material to knowledge with relevance for decisions and public discourse.

Figure 1.

The statistical value chain.

2.2Discussion platforms and other main topics of the journal in this period

3.1Data policy – One ring to rule them all?

Digitisation of society, the economy, the education and health sectors has been a high priority in national and international politics for some time, so for example in the European Union: “The success of Europe’s digital transformation will depend on establishing effective rules to ensure trustworthy technologies, and to give businesses the confidence and means to digitise.”55

Besides the business sector, it is also about supporting “decision-makers at the local, state and national levels as they accelerate the responsible re-use and opening of data for the benefit of society and the equitable spread of economic opportunity.”66“Rooted in the notion of an information commons, it begins from the premise that data collected from the public, often using public funds or publicly funded infrastructure, should also belong to the public or, at the very least, be made broadly accessible to those pursuing public-interest goals.” For the objective of Open Data, eight key principles are developed that record what belongs to a successful policy, starting with the data via insights to the actions [25].

The World Bank has dedicated its World Development Report 2021 [26] to the topic ‘Data for Better Lives’ with the aim of providing a blueprint on how to harness the power of data for development. When data is seen as force for the public good,77 the view narrows however from the very broad panorama of data with its countless pathways of possible uses to the concrete reality of (official) statistics (which as such is still complex).

This illustrates that there is a relationship between a data policy on the one hand and a statistics policy on the other: “The blueprint for building an integrated national data system is a national data strategy, which is a country’s plan for capturing greater economic and social value from data in line with the principles of a social contract for data. By contrast, the blueprint for building a national statistical system is a national statistical development strategy, which focuses on official statistics [26].” Clarifying the relationship between data and statistics is strategically important for the (official) statistics community if it does not want to run the risk of its specific concerns being lost in the general effort to promote and regulate data markets.

In January 2017, six months after the Brexit referendum and at the beginning of Donald Trump’s presidency, William Davies published a widely acclaimed article “How statistics lost their power – and why we should fear what comes next” [27]. In it, he expresses his concerns that nothing less than the end of a statistical era has arrived, with serious consequences for public discourse, trust in experts as well as politics, and with options for populist politicians to use this for their purposes. With ubiquitous amounts of data and almost infinite possibilities of use, informational ecosystems are fundamentally changing; statistical logic is being replaced by data logic. “With the authority of statistics waning, and nothing stepping into the public sphere to replace it, people can live in whatever imagined community they feel most aligned to and willing to believe in. Where statistics can be used to correct faulty claims about the economy or society or population, in an age of data analytics there are few mechanisms to prevent people from giving way to their instinctive reactions or emotional prejudices [27].”

After a year in a state of pandemic emergency, it is time to return to the discussion initiated by William Davies. It is necessary to ask the fundamental question, whether we need official statistics as the backbone of democratic decision-making, and if so, what their tasks are and how they should be financed and anchored in the political system. The public infrastructure that provides society, politics and the economy with elementary facts is official statistics. Facts on the basis of which momentous political decisions are made are based on international methodological standards (e.g., consumer price index for Central Banks’ monetary policy). Official statistics must perform their tasks with great efficiency and continuity; they work with long lead times, industrial production lines, international standards and democratically decided work programmes. In this way, internationally, nationally, and temporally consistent indicators of high quality are provided.

Let us use the example of transport: data strategies aim at promoting and regulating individual (data) mobility. In addition, there is public rail transport with data and statistics, which must also be made fit for the future. This requires investments, because new areas are to be established on which modern high-speed trains are to run. Individual data use alone can be inefficient and ineffective. In the 1960s, we thought that promoting individual transport was the best option. Today we know that this one-sidedness has led us into congested cities and roads because rail expansion was not pushed with enough verve. If the infrastructure of public statistics is not modernised, geared to new technologies (high-speed statistics) and new terrains are not opened up (COVID-19, biodiversity, …), there will be parallel infrastructures in both the public and private sectors that will develop their own standards. Or, more precisely: We will have an outdated, unattractive public (data) railway with multiple rail widths (partly public, partly private) and incompatible industry standards; a setback for trust, transparency, and public discourse.

To prevent such a situation, the integration of a country’s different producers under one roof into a well-coordinated statistical system is crucial. Roles must be assigned, responsibilities defined, so that citizens can rely on the highest quality standards being met. To make it easy for users to obtain information, a certificate should be introduced that provides trustworthy information about the quality profile of an information product. Certification requires a neutral and trustworthy institution that sets and verifies quality standards.

In connection with the World Development Report 2021, the World Bank has also designed a new Statistical Performance Index, being is of interest in this context,88 which identifies five key pillars regarding statistical data usage and production in modern societies (see Fig. 2).

This approach provides a structure that can inspire the further development of existing frameworks of international standards and quality codes.

In this context, it will be crucial for statistical strategies on the one hand to clarify the commonalities with general data strategies, but on the other hand also to emphasise the special features and political importance of the public infrastructure called official statistics.99

All too often, an attitude that focuses purely on the use of data is presented as a passe-partout for all processes, industries and value chains, without their specific characteristics actually being adequately represented in this way. Against this background, language and terminology used gain special importance, so that it seems appropriate to avoid ambiguities in the use of ‘data’ and ‘statistics’. It may also be worth arguing against simplistic narratives of data use for all sorts of purposes (‘data-driven’ decisions, policies, etc.), which generate false expectations that are detrimental to statistics.

3.2The scientific approach: STEM1010 or SHAPE1111?

In 1924 John Maynard Keynes sketched the “ideal type” of an economist in the following way: “The master-economist must possess a rare combination of gifts. He must be mathematician, historian, statesman, philosopher – in some degree. He must understand symbols and speak in words. He must contemplate the particular in terms of the general and touch abstract and concrete in the same flight of thought. He must study the present in the light of the past for the purposes of the future.”1212

This well-known quotation leads us to the question of what actually qualifies a statistician, and even more specifically a professional in official statistics. The answer to this question is unfortunately all too often truncated. This begins with the interpretation of what characterises statistics as a science. Statistics is seen as a part of the natural sciences (STEM), or more precisely of mathematics,1313 and scientific attention is primarily focused on inferential, analytic statistics.1414 In contrast, descriptive statistics with its methodological questions is in the background of scientific interest. An excerpt from an article on “science and statistics: the scientific basis of the census” in this journal may be suitable to illustrate this point of view: There “seems to be a general consensus about the object of statistics as a science and nature of the statistical method. Any empirical science observes physical objects (people, animals, celestial objects, rocks, etc.) and their characteristics at a clearly delimited location and specific time [30].”

Figure 2.

The pillars and dimensions that construct the new SPI.1515

This reduction to physical objects and their ‘measurement’ does not, of course, meet the typical target of observation in (official) statistics, as Theodore Porter [31], Heinz Grohmann [32] or Alain Desrosières [33] have clearly pointed out in their works. Imagine, for example, approaching the well-known indicators of statistics, such as GDP or the Consumer Price Index, with such an inadequate scientific toolkit. Desrosières’ work is closely linked to the scientific movement of the so-called “economics of convention” (in French ‘économie des conventions’) [34]. Here, the name of this movement already contains the core of the message that statistics (at least in the field of social or economic phenomena) has to deal with facts that cannot be measured directly, such as a tree, but which first require a qualitative description (or as Grohmann put it “theoretisches Konstrukt”) and subsequently its translation into a corresponding statistical concept. The aim here is not to give preference to either a deductive or an inductive approach, but to iterate between these with corresponding learning cycles, as Box already expressed it in his 1976 reflection on statistics and science [35]. Included in the qualitative description of phenomena, may one like it or not, are conventions, questions of design, which are decisive for the later result of the statistics. Statistics are products of a complex sequence of process steps, some of which are not possible without interaction with the users of statistics and their questions and information needs. To make this clear in terminology, Desrosières distinguishes between ‘measurement’ (physical objects) and ‘quantification’ (social objects).

The choice of an adequate scientific approach, based on appropriate epistemology, has manifold consequences for the practice of statistics (both in research and application). In conclusion: only with a scientific approach that does justice to the complexity of statistics and without cutting it down to specific disciplines (and their perspective of knowledge), can the questions and problems that are essential for the quality of statistics be addressed. Returning to the question raised in the title, it can therefore be concluded that the scientific roots of statistics are based in natural sciences (STEM), but that the professional profile of the statistician also includes scientific components from the social sciences (SHAPE).1616

3.3Misuse and misconduct, trust and integrity

The term ‘statistics’ has the same linguistic roots as ‘state’. Since the Enlightenment, statistics has been closely married to the nation state, to democracies of various kinds and, unfortunately, also to dictatorships. Because of this nature of the relationship with politics, it is by no means a novelty for official statistics to have to face attempts to influence (and even manipulate) the results. In principle, it can be observed that this risk for influencing or manipulating increases with proximity to and relevance for political decisions (in accordance with ‘Goodhart’s Law’).1717

In the transition that followed the fall of the Berlin Wall in 1989, official statistics of centrally planned economies also had to be transformed to become systems of market-oriented democracies. Against this background, it was necessary to take an analytical look at statistical governance in these states before the fall of communism1818 in order to codify guidelines in professional and scientific standards on this basis. In 1991 the Conference of European Statisticians developed and adopted the Fundamental Principles of Official Statistics;1919 in 1994, W. Seltzer presented an overview on “various ways in which the professional integrity and credibility of statistical agencies can be undermined” [38].

Although much has been achieved since then in terms of countering harmful influence, the issue itself has not been resolved: Misuse of statistics is an abuse of power [39]; manipulation of official statistics is corruption [40].

Over the past three decades, work to establish an independent statistical infrastructure in the public sector has been achieved mainly by making the principles of good statistics the norm in the public sector through laws and codes enshrined in them. However, this approach to statistical governance is now reaching its limits (or generates increasing marginal costs), for various reasons.

As a conclusion, it will have to be considered whether and to what extent a complementary approach should be developed alongside the existing one, in which the aim is to assess the integrity and independence of statistics from a global perspective, comparable to the World Press Freedom Index,2020 the Academic Freedom Index2121 or the OECD Public Integrity Maturity Model.2222 In this way, a broader peer community and regular resonance with the public would be generated, a novel institutional design responding to “the wider societal conditions of confidence and trust [41].”

This diagnosis and the therapeutic route to be taken are well in line with an OECD approach to improving integrity in the public sector in general. In essence, the integrity and independence of statistics should be further strengthened, which means

Of course, it is of eminent importance in such a project that the highest possible quality standards of statistics are met. In this respect, it should be investigated with which of the already existing global platforms mentioned above2323 a fruitful cooperation and partnership can be entered into.

3.4Innovativeness, relevance and strength

Given the urgency and stress of the current pandemic situation, it seems natural to focus the statistical issues of COVID on the decisions that are about to be taken. At this point, however, a broader perspective and longer horizon is advocated. Why does a more strategic and long-term view seem so essential right now, when we hardly have time for the most dramatic questions?

There are two important arguments for this:

Firstly, the most pressing risks identified by the World Economic Forum in the Global Risk Report 20212424 lists a number of risks, which are perceived as present dangers or existential threats, with infectious diseases, climate action failure and biodiversity loss as the risks with the highest values in terms of their likelihood of occurrence and impact. Similar to this assessment from the economy is that of the scientific community, which call for action to reach net-zero emissions, reverse declines in biodiversity, and improve data-sharing to prepare for future health emergencies.2525 The question is therefore to what extent it is possible to design the measures for exits from COVID in such a way that they do not take us back to a supposed normality ex ante, thereby missing the opportunity to make this normality more compatible with the global strategy of sustainable development.2626 This is of course not a statistical debate but a scientific or political one.

Secondly, it must be borne in mind that more substantial changes, extensions or accelerations of official statistics cannot be realised overnight. Official statistics are a kind of ocean liner whose course can only be changed with considerable advance planning and preparation. It is therefore time to look far ahead now in order to be able to adapt to the information needs of the future, two to five years from now.

3.5Literacy

For some years now, a resistance to evidence-based governance has been growing; scepticism towards all forms of experts does not stop at scientists or statisticians. Coupled with a lack of statistical literacy and the impression of being at the mercy of the representatives of a supposed technocratic regime, the counter-position is forming in which the existence of neutral facts is negated or relativised. The reduction of social and economic questions to numerical aggregates and averages no longer seems acceptable, unless the results come from one’s own calculations and correspond to the ‘truths’ that demagogues deliver about what is going on in society.

The fact that initiatives to improve data literacy are gaining momentum, supported not only by business but also by politics and science, is very welcome. Data literacy serves to promote maturity in a modern digitalised world and is important for all people – not just specialists. This education, like other education, is about several competence dimensions: Knowledge, skills, and values [43].

However, a broad, balanced, and situational approach is rarely found in practice. Rather, the focus seems to be on teaching technical skills of data science, mathematics, and IT, reminiscent of the do-it-yourself wave of the 1970s, in which screwing, repairing, and constructing by anyone was propagated, sometimes even in cases where a good craftsman would have done the job better and cheaper than an amateur. For the citizen, the entrepreneur, the teacher, the student who wants to understand and apply the indicators of public statistics, sophisticated skills of data science (e.g., own analysis of raw data, knowledge of algorithms) are just as irrelevant as in-depth knowledge in the mathematical field of stochastics.

Citizens should know enough about the informational product and its properties to be able to assess its quality regarding personal application goals and questions. This requires basic mathematical knowledge as well as experience in dealing with quantitative information; knowledge of descriptive statistics and its application in the processes of economic and social statistics is required. What the consumer price index says (or doesn’t say) about inflation should be taught in school and adult education; everyone should understand the indicators of sustainable development. For advanced users, microdata are also available [3] as ‘public use files’ to experiment with their own statistical evaluations and gain experience.

The question for official statistics is whether and to what extent they can and want to be involved in this matter. However, with all sympathy for improving general literacy, the answer will depend on what free capacities and resources are available for that purpose.