‘We should trust statistics, but only good ones’: Interview with Paul Schreyer

SJIAOS: Thank you Paul for kindly accepting to be interviewed right before your retirement. How did your last CSSP go last June? Were there any specific outcomes of this meeting?

Paul Schreyer: The feedback from our Delegates was good! As in previous years, countries appreciated discussion of statistical policy questions – for example how Artificial Intelligence is used in the work of National Statistical Offices or how to define the limits of modelling and imputations that are acceptable for official statistics. The OECD has much to bring to the table, not least because CSSP Delegates can interact firsthand with other OECD policy committees. This allows members to identify current and future demand for statistics from policy makers but also to advise other parts of the OECD on their statistical and data projects. From an institutional perspective, the single most important outcome of the 2024 CSSP meeting was the very positive review of the statistical systems of Bulgaria and Croatia – an important step towards the accession of these countries to the OECD.

SJIAOS: Could you tell us about your career, especially at OECD, since you joined in 1988?

Paul Schreyer: As my entry date to the OECD tells you, I spent most of my professional life in this great organization, preceded only by a few years in academia in Innsbruck Austria and at the IfO Institute for Economic Research in Munich, Germany. The OECD has a lot to offer professionally, and I moved between different areas including science, technology and industry, environment and then statistics. In all domains, however, I dealt with data, indicators or quantitative analysis. Indeed, evidence-based analysis is a core value that the OECD wants to bring to its members and to the public. And as I have always enjoyed working with data and developing measurement tools, the OECD was an excellent match for me.

SJIAOS: What does your job as Chief Statistician and Director of Statistics and Data consist of?

Paul Schreyer: There were two main aspects to this role. One was providing leadership for the substantive work of the Statistics and Data Directorate that collects, harmonises and disseminates core macro-economic statistics such as National Accounts, Prices, or International Trade data. Together with several other international organisations, the Directorate also develops statistical standards, the current update of the System of National Accounts being a prominent example

The other aspect of the job is the role as OECD Chief Statistician properly speaking. Let me explain: OECD statistics are de-centralised in the sense that many important statistical activities are carried out in policy directorates, closely tied to the needs of the relevant policy community. For example, OECD health statistics and indicators are produced under the auspices of the OECD Health Committee, or environmental data under the guidance of the Environment Policy Committee. The Chief Statistician’s role is to make sure that these areas are provided with the right tools for their statistical work and that there is a minimum of coordination, harmonisation and a common statistical policy. Over the past few years, much effort has been devoted to modernise the OECD data management system and to manage access and use of new data sources. Lastly, the Chief Statistician also presents the Organisation towards the outside internationally in statistical matters.

SJIAOS: As IAOS relates to official statistics, could you explain to us what is the specificity of OECD statistics?

Paul Schreyer: Traditionally, and still to a large extent, OECD statistics have been sourced from national official statistics through transmission programmes with national statistical organisations. Increasingly, the OECD has also become a producer of data itself, either through large primary data collections (such as the PISA surveys) or by harnessing new data sources that have become available as part of the digital transformation (such as social media, or geospatial data). Whatever the source, the specificity of OECD data is threefold: first, they are demand-driven, i.e., they help addressing those policy themes that are most pertinent for our members; second, they strive for international comparability; third, they comply with minimum criteria for statistical quality. Living up to these characteristics requires constant work and attention but is essential for the Organisation’s reputation and usefulness.

SJIAOS: The “Recommendation of the OECD Council on Good Statistical Practice” applies to member states. How is it put in practice, and how does it impact NSOs in member states?

Paul Schreyer: The Recommendation is part of many ‘soft laws’ that the OECD has at its disposal Recommendations are not legally binding but they represent a political commitment to the principles they contain and entail an expectation that adherents, i.e., actors in the national statistical system will do their best to implement them. When the Recommendation was conceived in 2015, the CSSP also decided that adherence requires some form of review of countries’ statistical systems. These are mostly self-assessments that are made publicly available and, in some cases, specific reviews. For instance, a review of the Mexican statistical system against the Recommendation was carried out a few years ago which gave rise to practical recommendations and follow-up actions in Mexico. The Recommendation’s mere existence and the publicly available assessments create transparency and support for NSOs in complying with its key principles such as professional independence. I should add that it is also possible for non-members to adhere to the Recommendation and indeed, several have done so, for example Argentina or Peru.

SJIAOS: What guarantees the quality, reliability, neutrality and independence of OECD statistics, produced either by your directorate or by others?

Paul Schreyer: As a matter of principle, the OECD is held to its own Recommendation that I just described. In practice there are several mechanisms to ensure quality of OECD statistics. One is an internal quality review process that databases undergo periodically. Technology and well-defined processes also help: for instance, the Organisation’s internal data sourcing hub provides the tools for all data teams to document and protect source data in line with confidentiality requirements. The second, equally important mechanism is the filter that applies through the OECD Committee structure. Statistical working parties discuss methodologies, sign off on new databases and typically screen national figures when published by the OECD Finally, the analysis to which OECD economists and other analysts put the data creates an excellent feedback loop for constant quality improvement of OECD statistics.

SJIAOS: How do you work and coordinate with other international organizations, such as the UN, the IMF and Eurostat?

Paul Schreyer: We work closely and very well with other international organisations. Coordination occurs via mutual participation in meetings and regular bilateral contacts. For instance, as Chief Statistician I had regular teleconferences with the DirectorGeneral of Eurostat, of the UNSD, the World Bank and the IMF. Much coordination occurs also via joint projects and initiatives. These include common development of standards – I already mentioned the update of the SNA but there are other examples such as a handbook on measuring digital trade that we developed jointly with the IMF, the WTO and UNCTAD. Joint initiatives include also development of tools and capabilities and the SDMX consortium is a good example – as you will know, SDMX is an ISO standard designed to describe statistical data and metadata, normalise their exchange, and improve their efficient sharing across statistical organisations. Indeed, the data exchange between international organisations is increasingly SDMX based. This helps co-ordination.

Eurostat deserves special mention when it comes to co-ordination because most EU countries are also OECD Members. Whenever possible, the OECD collects the official data from its EU members through Eurostat which ensures data consistency and efficiency of transmission.

SJIAOS: You attended the last IAOS Conference in Mexico, held in May 2024. Could you give us your feelings about this conference, and how do you think it is useful for OECD and for the international statistical community?

Paul Schreyer: I found the Mexico conference both interesting and pleasant. From an OECD perspective, the IAOS provides a network that goes well beyond OECD countries and so adds to our possibilities to interact with official statisticians. It also blends nicely academic work and statistical production which means it is possible to find out about the latest developments in concepts and tools but also get a good sense for what is happening in statistical practice elsewhere, ‘on the ground’ so to say. Contents apart, the Mexican hosts should be commended for their organisation and hospitality.

SJIAOS: You were one of the rapporteurs of the Stiglitz-Sen-Fitoussi Commission in 2009; 15 years later, what has been the impact of this report?

Paul Schreyer: It would take at least another interview, and probably a long one, to provide a comprehensive answer. The short answer is that the report has had a significant impact on the discussion and practice of economic, social and environmental measurement. Many countries are now producing sets of indicators

that reflect the spirit of the report. The idea that our societies need to go beyond GDP in their measurement efforts is now well established. How much exactly is due to the report is hard to say but its visibility has been huge for two reasons: the prominent composition of the Commission including 4 Nobel laureates – Joe Stiglitz, Amartya Sen, Daniel Kahneman and Angus Deaton – and the fact that measurement was discussed at the highest political level – president Nicolas Sarkozy had put the Commission in place, provided financing and actively supported its recommendations while granting it full independence.

SJIAOS: At the time of leaving your job as Chief Statistician of the OECD, what are you the most proud of in your work and what do you see as the main challenges for the future?

Paul Schreyer: I hope I have contributed at least in some ways to the development of international statistical standards and methodology in national accounts, productivity measurement and prices. The OECD Smart Data Strategy – the Organisation’s master plan for its data management and capabilities – is another area to mention: I neither conceived it single-handedly nor did I manage it on a daily basis but I supported it with all the possibilities I had because I believe that people and data are the two most valuable assets that the OECD has. Looking ahead, the pace with which technology progresses and the speed with which events occur will continue to pose the main challenges for OECD data work. The Organisation needs to keep up with latest digital developments while preserving accuracy and credibility of its data and statistics.

SJIAOS: And a final question? Should we trust statistics?

Paul Schreyer: Yes, but only good ones.

Paul, thank you for your very detailed answers, and have a great retirement!

Interview conducted by Jean-Pierre Cling in June 2024