A Fast-Evolving Landscape For Official Statistics: How To Respond To The Challenges?

Whire robot

In April 2022, during its annual Conference, the IAOS has formally established the “Krakow Working Group” whose aim is to contribute to the collective reflection on the challenges that our “datafied” societies pose to Official Statistics.

SJIAOS is launching a virtual discussion on the issues raised in the editorial. The main points arising from this discussion will constitute useful inputs to the deliberations of the Group.

The Krakow Working Group will report on advancement of its work at the IAOS Conference in Zambia in April 2023 and the ISI Congress in Ottawa in July 2023. It is expected that the Group will be able to present recommendations to be widely disseminated in the second half of 2023.

Extra SJIAOS discussion: Krakow Working Group

Today, we live in a digitised world where data has become a global commodity that is produced, shared and traded by a myriad of actors around the world. Sometimes called the data revolution, this trend is facilitated by technological innovations such as the Internet of Things or Artificial Intelligence. It reflects an increased demand for, and use of data in the public debate and a strong appetite for data-driven decisions by governments, businesses and citizenry at large.

In this complex new data ecosystem, competition for timely and relevant data for decision-making is strong. Official Statistics are now only one actor among many, albeit an important one. New actors, such as private producers or data scientists are sometimes perceived as more agile, more responsive and more innovative.

Do you feel that Official Statistics have lost their quasi-monopoly position? How do you think Official Statistics should position themselves in this new competitive environment?

To remain relevant in the face of increased competition, Official Statistics need to ensure that they continue to deliver on their key mission to inform public policies and public debates. They have great assets, such as strong legal frameworks, strong methodological standards, strong reputation for the quality and the dissemination of the data produced and strong ethical principles. But they have to address weaknesses such as a lack of flexibility, too defensive strategies vis-à-vis other actors and poor communication strategies.

Methodologists in National Statistical Offices should continue to reflect on how to exploit new data sources and invest in data science techniques to turn administrative and big data into worthy statistics in order to respond to policy-makers’ and citizens’ demands for more timely and granular information, while respecting fundamental professional and ethical principles about privacy and confidentiality. They should continue to innovate through data labs for experimental statistics.

In your view, what are Official Statistics main assets? And how should they address their weaknesses in the face of increased competition from other actors in the data ecosystem?

We live in a world where social media bubbles, “alternative facts”, emotions and beliefs often prevail over facts. Official Statisticians should combat misuse and misinformation, and not be afraid to speak-out in case of false information, outright manipulation or errors of interpretation. Official Statistics should also contribute to statistical literacy of both data producers and users. In order to reach out to a wider audience, they should also expand their social media culture and be more present where evidence-based opinions are formed. For this, they should build partnerships and collaborations with other actors, in particular the media and fact-checkers.

A lot of innovation in data visualisation and communication has taken place in NSOs during the Covid-19 pandemics related not only to established statistics but also to experimental ones. The increased presence in the media during the pandemic through blogs, data viz, articles has reinforced Official Statistics’ reputation as a reliable, trusted source of information. It is important to pursue these initiatives and invest in new ones over time.

Do you agree that Official Statistics should improve their dissemination and communication strategies, including by being more present in all types of media? If so, how?

In the new data environment, not all players have the same approach to the notions independence, integrity, quality, transparency or privacy in relation to Official Statistics. The presence and active engagement of more and different actors requires to review existing governance arrangements. Unlike for Official Statistics, which are typically governed by statistical laws that formally set out their mission, obligations, means and interactions with other stakeholders, at present, there are no formal institutional frameworks governing the data ecosystem at national and international levels.

Over the past couple of years, some countries and international organisations have adopted laws, regulations and guidance related to data governance. These mainly deal with issues of personal data privacy, cross-border exchange of data, or data sharing among private actors. They generally do not cover statistics nor the inter-relation between statistics and data. This is a gap that will need to be filled and where Official Statistics should make their voice heard to express their needs and concerns. Discussions are taking place about new laws and regulations that would recognise Official Statistics as a key actor in the data ecosystem as a step forward. NSOs could play an important stewardship role in areas where they have the legitimacy to contribute, for example with respect to standards, classifications, methodologies or quality assurance. They should pro-actively engage and build partnerships with other public providers and private actors to offer their technical and legal services.

Do you agree that the emergence of new actors in the field of data and statistics requires new governance arrangements? How do you see NSOs playing a stewardship role?

Ensuring data quality in the new data ecosystem is essential for building and enhancing trust in statistics. The recourse to new data sources, including private ones, their linkages and integration with established sources implies to review the long-established quality frameworks that are commonly used in Official Statistics. Working with private and public actors, Official Statistics should take the lead in developing new or enhanced quality frameworks that take into account the nature of the new data used, the new processes put in place, as well as the new statistical products offered. These new frameworks should thus encompass the whole data value chain, from collection, to production, dissemination and use. They should respect  the fundamental/human rights frameworks at all stages of the processes. There should be full transparency about potential caveats and limitations of the new statistical products and clear guidance on their interpretation.

Tapping into new data sources, including citizen-generated and geospatial data, investing in data science skills, data labs, experimental statistics and data viz, combined with standard household surveys will contribute to provide more timely and granular information on people’s living conditions and on what matters in their lives, with the potential to reinforce trust in Official Statistics. But for this to happen, this information has to be made readily available and easily accessible. Official Statistics like other public data actors have therefore to make their data available, and more generally have to adopt open data policies by default, while fully respecting confidentiality.

Indeed, as more granular information may be derived from individual administrative records or other personal data, it is essential that citizens and companies have an absolute guaranty that safeguards are in place to ensure data privacy and confidentiality, and that there is no interference and full independence from public authorities or lobbies. Any breach, or even perceived breach, in these respects would greatly undermine trust in Official Statistics.

To create trustworthy statistical evidence for public debates and public policies, and engineer trust in the system, Official Statistics should also build solid alliances with other actors (producers, data scientists, statistical societies, the diversity of users and the media).

How in your view could Official Statistics strengthen their position as a trustworthy source of information? What types of alliances with other actors could Official Statistics build to enhance trust in the system?

A very important characteristics of Official Statistics which highly contributes to the trust the general public attaches to them, is their strong ethical underpinnings. Existing sources of ethics for statistics are many at both national and international levels. These include inter alia national constitutions, national statistical laws, the UN Fundamental Principles for Official Statistics, the European Statistics Code of Practice, the OECD Council Recommendation on Good Statistical Practice, the Principles Governing International Statistical Activities, and the International Statistical Institute Declaration on Professional Ethics, National professional associations’ codes and declarations.

These sources of ethics for statistics are do not necessarily apply to the new data environment where new actors and new practices have emerged. Official Statistics should be pro-active and take the lead to foster the adherence to strong ethical principles for data and statistics. They should advocate for the existing codes of ethics for Official Statistics.

A potentially controversial proposal would consist in initiating work with other relevant actors, including users, towards the creation of an International Forum to address issues around ethics but also misuse, governance and trust. The aim of such a Forum would be to develop and promote the effective use of a set of core values and of universal ethical data and statistics principles, based on existing frameworks and recent initiatives. Such a Forum should bring together the different actors in the data ecosystem.

The principles could cover regulations and arrangements encompassing professional independence, responsibilities and accountability, methodological excellence, protection of personal data, data accessibility, data sharing and exchange, data interoperability, etc., as well as prevention of abuse and misuse of data. For such a Forum to be effective, Official Statistics should promote the creation of an independent, ‘soft’ enforcement mechanism supported by the broad constituency to ensure that all participants in the International Forum are in compliance with the agreed principles. In a datafied society, it is institutional framework that should ensure the guaranties people need for the full respect of their fundamental rights. The Forum should aim at the emergence of such frameworks, which should be universally endorsed.

Do you support the creation of an International Forum involving all actors involved in the data ecosystem that would discuss and propose ethical principles, together with an independent enforcement mechanism? What are the main issues that such a Forum should consider?   



Comment on:
26/01/2023 12h41

I am retired from the US Federal Government.  Even if I were active, I would note that I am expressing my own views, and not those of any US Government agency.  That said, I did work for a statistical agency, and had some contact with or knowledge of others. 

There appears to be pressure on Government statistical agencies to provide ad hoc information, in the interest of addressing new issues, or at least that was the impression I had before I retired from the US Energy Information Administration (EIA).  That anecdotal experience may be somewhat consistent with the statement under the first questions above, in another context, that "...National Statistical Offices should continue to reflect on how ... to respond to policy-makers’ and citizens’ demands for more timely and granular information...."  Also, one large survey (or set of surveys) at the EIA was expanded permanently, some time ago, for more 'granularity,' for which I initiated a 'borrowing of strength' for a version of small area estimation.  But with pressure to expand operations to meet changing needs, though motivations are understandable, there also may come a point where there is danger of a loss of accuracy for standard statistics routinely collected, in the case of the EIA, to monitor energy markets, if care is not taken.  Resources are always limited.  

At the EIA, some long time surveys have come and gone or been modified, so I think changing needs have been considered there.  One needs to note, however, that it takes a lot of effort and experience to establish a good data stream, and ad hoc or new routine efforts should be undertaken with care. 

Also, statements about subject matter which assume too much accuracy can be misleading.  The foundations of Official Statistics should not be undermined, and that includes the need for measurement of accuracy, and/or a qualitative assessment, and/or at least some relevant information.     

The main function of a statistical agency is to provide reliable data which others may use.  That isn't the flashy part, but it's necessary.  That does not mean that innovation and initiative are not important, and needs do change.  But if Government statistical offices do not provide fundamental data, and everyone is flitting from one topic or data source du jour to another, then that could quickly become very destructive.  I just caution that there needs to be attention to the big picture, and as much stability as possible.  Everyone wants data to use for various projects.   Agencies may do some of their own analyses, but they should keep in mind that their reason for being is to collect usable data. 

Other questions in the discussion write-up above have more to do with data use, dissemination, sources and administration, but data quality could receive greater attention.  Sources are important with regard to accuracy,  but there are other considerations.  When aggregate data are reported by a statistical agency, do they always have metadata?  Relative standard errors are often relegated to an appendix.  It seems that attention to accuracy is likely already often too weak without letting the changing environment make accuracy an even lower priority.  Yes you need to pay attention to changing needs, and if data are not timely, they may not be very useful, but neither are they useful if the resulting discussions are just about noise.  

I think that National Statistical Offices should concentrate on the routinely needed data, to have as much consistency, and accuracy, as possible.  New players might be given more license to experiment.  Rather than NSOs lacking relevance by not 'keeping up,' they may lose relevance if they are no longer reliable for core data.  

Jim Knaub