1 Introduction

It is a great pleasure to deliver the keynote of this year‘s External Statistics Conference hosted by the BIS‘s Irving Fisher Committee on Central Bank Statistics, the ECB and Narodowy Bank Polski.

A few weeks ago, I visited the Palazzo Pubblico (the town hall) in Siena, Tuscany. In the Sala della Pace (the Hall of Peace) you can find the famous fresco cycle of the Allegories of Good and Bad Government[1]. It is an Italian gothic masterpiece by Ambrogio Lorenzetti that depicts the effects of good and bad government on the city and the surrounding countryside. While bad government leads to war, disease and poverty, good government leads to peace, justice, a flourishing society and prosperity.

Though the fresco was created in 1338 and 1339, it is still today a powerful reminder that good or bad government is not an abstract idea, but something that has very concrete consequences for trust in institutions and for people’s lives – for prosperity and stability.

Standing in front of the Allegory of Good and Bad Government, I could not help thinking about our own responsibilities as statisticians. 

Good government depends on good decisions – and good decisions depend on good data. In an era of geo-economic fragmentation, when the world is becoming more complex, and more uncertain, the quality, timeliness, openness, and accessibility of external statistics are not a technical detail. They are a cornerstone of sound policy, international cooperation and, ultimately, of “good government” in the modern sense.

I would like to structure my keynote around three questions:

1. What do we take from the current environment – and what are the policy and statistical implications?

2. How can we provide relevant, objective data faster and in a more user‑friendly way?

3. How do reliable statistics contribute to democracy, and what responsibilities does this entail for us? 

2 The current environment: policy and statistical implications

Let me start with the first question: What do we take from the current environment – and what are the policy and statistical implications?

We are in an environment with exceptionally high levels of economic policy uncertainty worldwide. This has turned the focus towards geoeconomic fragmentation.

Geoeconomic fragmentation, as defined by den Besten, Di Casola and Habib, is a reduction in the degree of economic integration and multilateral cooperation, accompanied by a stronger alignment of trade and financial relations with geopolitical preferences[2].

And geopolitical risks and fragmentation have been the main drivers of many economic developments recently. 

Evidence of geoeconomic fragmentation has been discussed in earlier panels, and research by the Bundesbank confirms those findings. 

This research, for instance, finds that heightened geopolitical risks dampen imports, raise costs, and disrupt supply chains, thereby reinforcing the fragmentation of global commerce. 

In addition, Bundesbank studies indicate that hybrid threats place a drag on aggregate demand by amplifying uncertainty, weakening confidence, and tightening financial conditions.  

Furthermore, Bundesbank work on Germany’s external assets in the light of geoeconomic tensions finds that geopolitical factors already affect portfolio and other investment, while equity capital has so far been less responsive in a statistically robust way. In other words: portfolio investors and banks adjust more quickly than multinational enterprises with long‑term physical investments. These findings are detailed in the Bundesbank’s most recent monthly report[3].

Fragmentation is complex and varies across channels, instruments, and time frames. External statistics provide reliable data to quantify the effects.

A fundamental obstacle to a clear global picture is the existence of large and persistent bilateral asymmetries in external statistics. While one partner’s credits, exports, assets and balances should ideally mirror the other partner’s debits, imports, liabilities and balances, this is often not the case. Discrepancies harm the analytic potential of our data for policymakers and researchers.

A prominent example is the euro area’s current account vis‑à‑vis the United States, where discrepancies between the data reported by the European Central Bank (ECB) and by the US Bureau of Economic Analysis of around €66 billion were observed in 2025. According to ECB data, the imports of goods, services, and primary and secondary income from the United States were €57 billion higher than the corresponding exports. Data published by the U.S. Bureau of Economic Analysis (BEA) show the opposite direction: euro area exports to the United States were 9 billion higher than imports[4]. This is a politically relevant difference. 

Martin Schmitz presented a longer view on asymmetries at the IMF’s 13^th Statistical Forum in autumn last year[5].

Until 2020, all three main components – goods, services and, above all, primary income – contributed to large asymmetries, with ECB data pointing to higher euro area current account surpluses than those recorded in US statistics.

Since then, the picture has changed. While the asymmetry in goods has remained broadly stable in size and direction, the asymmetries in services and later also in primary income have reversed. In recent years, it is now mainly services – and to a lesser extent primary income – that drive the large asymmetries, but this time implying smaller euro area surpluses, or even larger deficits, in ECB data compared with BEA data.

Obviously, these asymmetries complicate global surveillance and policy analysis. They raise uncomfortable questions in the public debate: “Whose numbers are right?” and “What are the reasons behind these asymmetries?”

We suspect the main drivers of these asymmetries – not only in this special case – to be the following: First of all, compilers use different data sources and different estimation methods. Furthermore, complex multinational enterprise structures make it difficult to identify cross border trade between the institutional parts (units) of the corporation. And confidentiality constraints limit data comparisons of different statistical agencies. In some cases, asymmetries may also arise from different interpretations of international standards.

The implication is clear: if we want to understand all aspects of external trade properly, we need more consistent data.

This calls for a determined effort to reduce asymmetries. And much has already been done.

Europe provides an example with good work in the context of statistical coordination networks or what are known as asymmetry resolution mechanisms. Central banks and statistical offices are working together to clarify and reduce asymmetries. 

However, since the European legal framework for official statistics enables and supports such measures, they cannot be transferred to other parts of the world easily.

But work is also taking place internationally. For example, new databases at the OECD[6] and the IMF[7] allow an overview of bilateral asymmetries. They picture the top largest asymmetries or allow asymmetries to be viewed by country pair.

And there is promising work taking place at the BOPCOM’s Task Team on Global Asymmetries. This is a body mandated in autumn 2024 by the IMF’s BOPCOM, which brings together national experts from all over the world in addressing the most important asymmetries in external statistics. I’m looking forward to reading the task team’s recommendations.

I believe there are promising opportunities to push for greater ambition now. 

Many of our institutions have installed large case units (LCUs) that statistically analyse large multinational enterprise groups with numerous institutional units in several countries. These efforts aim to create consistent and coherent statistics that provide a clearer picture of cross-border activities.

The IMF’s Balance of Payments and International Investment Position Manual Version 7 (BPM7) White Cover (Pre-Edited) Version [8] – published in March 2025 – introduces breakdowns of nonfinancial and financial corporations based on control, offering valuable data to address and reduce asymmetries. 

While these initiatives are a step in the right direction, they are not sufficient to significantly lower asymmetry-levels.

Therefore, one concrete proposal is to include a dedicated recommendation on asymmetry reduction and its monitoring in the next G20 Data Gaps Initiative[9].

The G20 has developed a good framework for monitoring progress (traffic light system based on countries’ self-assessments). Fostering and implementing bilateral discussions between the respective statistical producers involved could make a great deal of difference.

To summarise my answer to the first question: external statistics offer a rich data source that shows the effects of geoeconomic fragmentation. At the same time, large asymmetries hinder country analyses and prevent us from answering questions related – for example – to tariff debates. Reducing these asymmetries is key. A G20 Data Gaps Initiative recommendation on this (together with its monitoring system based on self-assessments from jurisdictions) could be a driver for change.

3 Providing relevant, objective data faster and in a more user-friendly way

Let me turn to my second question: how can we provide relevant, objective data faster and in a more user‑friendly way? 

Take ChatGPT and its competitors as an example. These services offer low effort language-based entry points. Many of today’s students in schools and universities already turn to these services to explore topics. 

Of course, they cannot be compared to the high-quality data provided by official statistics – but they are much faster and easier to use. They may therefore become the standard entry point for users. How can we adapt?

With StatGPT,[10] a promising platform (developed by the IMF) is currently emerging that may have the potential to offer an even more flexible user experience, based on the high‑quality data provided by external statistics. It aims to combine the flexible natural language interaction current large language models offer with the rich high-quality data that is provided by official statistics using sound and precise information from SDMX[11]-definitions.

A key factor is easy finding and easy access. This is also the intention of international SDMX sponsor organisations: the Bank for International Settlements (BIS), the European Central Bank (ECB), Eurostat (the Statistical Office of the European Union), the International Monetary Fund (IMF), the Organisation for Economic Co-operation and Development (OECD), the United Nations Statistics Division (UNSD), and the World Bank. In their first statement on AI readiness[12] they said: “We will be working together with national and international partners, legal experts, and technology leaders to make official statistics discoverable, machine-readable, and globally usable, including by investigating a Global Trusted Data Commons and model context protocol (MCP) technologies.

We welcome the UN Statistical Commission’s leadership in calling for official statistics to be made AI-ready. We will complement and support the broader strategy led by the UN Statistical Commission, ensuring that official statistics remain a trusted global public good in the digital age. Ensuring that statistics are accessible, interoperable, and enriched with high-quality metadata is essential to safeguard trust and to strengthen the global statistical system.” Work is going on, and this is good news.

Also, smaller approaches like improving search functionality in websites and databases improve accessibility for users.

This is very helpful, but it does not make the provision or publication of the data faster.

During the pandemic, demand for high‑frequency, granular data exploded.[13] These changing demands are typical for every crisis.

During trade conflicts, users asked for more detailed geographical breakdowns and information on value chains. In the current environment, they want to understand exposures to specific countries, sectors and instruments.

At present, for the Eurosystem’s analyses, the effective US import tariff as well as counter-tariffs – such as China’s and the EU’s considerations on this subject – are important assumptions. These must be reliably available in a timely manner – for forecasting purposes, for example. Unlike other assumptions – such as the price of crude oil – the effective tariff rate is not an easy-to-obtain variable but is based on product and company data as well as almost daily evaluations of applicable laws and regulations combined with difficult legal classifications.

Not only central banks within the Eurosystem, but many research institutions and universities would benefit greatly from a centralised provision of appropriately prepared customs information as a public good.

If official statistics are too slow or too complex, alternative – often opaque – data sources fill the gap. Non‑traditional data providers – like current trade and tariff monitors – have become more visible. They can be useful, but they also raise questions about quality, coverage and comparability.

This is a challenge for us – but also an opportunity. It is a task for official statistics to respond.

We need to be faster, more flexible and more user‑oriented without compromising quality.

At the Bundesbank, we have taken several steps in this direction.

First, faster production. In securities statistics, for example, we have modernised our production systems and reduced processing times significantly – in some cases by around three‑quarters, from weeks to days.

Second, better products and services. We have strengthened user‑oriented structures and first‑level support for data requests. Our integrated data and analytics platform (IDA) serves as an internal “marketplace” for data and tools, making it easier for economists and other users to find and combine information.

Third, stronger data governance. Centralised financial market data management, bank‑wide data catalogues and simplified access procedures help us to use existing data more efficiently. Data strategies on openness, metadata, architecture, analytics and data literacy provide a framework for further progress.

Many of you are addressing similar issues and pursuing similar initiatives. 

At the international level, we face a related challenge: The speed and flexibility of our update process for our statistical standards. Growing complexity in manuals make life harder for compilers and users. Long time lags – often more than ten years from the emergence of new phenomena to their full integration into standards and implementation – mean that we are always catching up. And revision cycles become increasingly long and resource‑intensive. Until standards are available and implemented, only stand-alone indicators (from various sources) can be developed to provisionally fill the gap.

Take the emergence of crypto assets as an example:

• Bitcoin started in 2008 as a peer-to-peer electronic cash system

• 22 May 2010 “Bitcoin Pizza Day”: The first real-world retail transaction involving bitcoin

• In April 2017, Japan passed a law to accept bitcoin as a legal payment method. Russia announced that it would legalise the use of crypto assets  such as bitcoin

• 2019: BOPCOM published a clarification note on crypto assets[14] 

• With the BPM7 White Cover Version, a statistical standard was published in 2025 and will be implemented in 2030.

Taking 2020 as the year in which crypto assets were recognised by the relevant institutions, it took five years to publish a harmonised standard that is yet to be implemented. We therefore need a more flexible approach.

In my view, a stable core of concepts – such as economic ownership – is the bedrock of our work, while at the same time adding a more flexible layer for faster updates, for example through new breakdowns or supplementary items.

Rolling recommendations, especially via the G20 Data Gaps Initiative, can play an important role in quickly achieving more flexible adjustments to core concepts. Instead of “all‑or‑nothing” revision cycles every decade, we could integrate completed items into regular work, remove obsolete requirements and free capacity for new priorities – such as asymmetry reduction, climate‑related external data or digitalisation.

At the same time, we must preserve consistency between external statistics and national accounts and maintain long, comparable time series for research and policy. 

To summarise: In a fast‑changing world, official statistics must become faster, more flexible and more user‑friendly, while preserving conceptual consistency and long time series. Adapting international standards and G20 data gap recommendations more quickly and on a rolling basis is a proposition for responding to these needs.

4 Reliable statistics for democracy: our responsibilities

Let me now turn to my third question: How do reliable statistics contribute to democracy, and what responsibilities does this entail for us? 

To answer this question, I would like to cite a recent survey conducted in OECD countries on trust in public institutions and highlight three key findings[15]:

First, Reliable statistics are linked to greater trust in democracy. The OECD survey reveals a clear connection: people who trust official statistics are significantly more likely to trust their national government. 

Citizens in OECD countries who perceive official statistics as “always” or “often” reliable, are nearly four times more likely to trust the government compared to those who “rarely” or “never” trust statistics (59 % vs. 15 %). 

Second, democracy is transparent and accountable when statistics are reliable.

Fact-based decision-making is a key principle of democratic governance. At the same time, political decisions create new facts and statistics. Together, these elements serve as anchors for the legitimacy of democracy and public trust in government. 

The Bundesbank contributes to this transparency through initiatives like Open Data (e.g. via its statistics portal). 

In the same OECD survey, citizens who believe government promises can “always” or “often” be checked against official statistics are almost twice as likely to trust their national government (59 %) as those who think this is ‘rarely’ or ‘never’ possible (23 %).

A critical challenge is that the majority of citizens in both OECD countries and Germany perceive the ability to verify government actions through statistics as “rarely” or “never” possible. 

What responsibilities emerge from those findings?

First, clear documentation, intuitive visualisations and good storytelling make it easier for non‑experts to understand what the data show – and what they do not show. Honest communication of uncertainty and revisions is essential to build and maintain trust.

Second, highlighting the value of high-quality and independent official statistics is a continuous task. Our principles, commitments and codes of practice are strong arguments and statements from Madame Lagarde[16], Isabel Schnabel[17] and Joachim Nagel[18] provide good points of reference. Understanding the value of official statistics for “good government” has to be rooted deeply in our democratic societies.  

Third, responding when statistics are misused or misrepresented is important. This requires close cooperation between statisticians, economists and communication experts. 

To summarise: Independent official statistics are a public good and highly relevant for our democracy. Active and ongoing statistical communication is part of our responsibility. Our sound principles are an asset in meeting this responsibility.

5 Concluding remarks

Let me conclude.

Geoeconomic fragmentation is a defining challenge of our time. It affects trade, finance, production networks and, ultimately, welfare. The best response would be to prevent a lasting fragmentation and to strengthen rules‑based international cooperation. But as long as fragmentation pressures persist, we need to measure, understand and deal with them.

For the statistical community, this implies three key messages.

First, good data does exist for the analysis of geoeconomic fragmentation. To analyse it, we provide high‑quality, granular external statistics. However, asymmetries still limit our ability to tackle all related questions. A G20 Data Gaps Initiative recommendation on lowering asymmetries (together with its monitoring system based on self-commitments from jurisdictions) could be a driver.

Second, speed, flexibility and usability are key issues. In a fast‑changing world, official statistics must become faster, more flexible and more user‑friendly, while preserving conceptual consistency and long time series. We are all carrying out many steps in this regard. Rolling standards with a sound core and a dynamic G20 Data Gaps Initiative can support this.

Third, we must continue to actively communicate and protect the value of independent official statistics as a public good.

Going back to the frescos in Siena, independent statistics are a part of “good government”. That’s why I am delighted to see so many statisticians, economists, academics and representatives of national, European and international organisations gathered here today. This combination is quite special – and, I would say, exactly what we need in the current environment. In a fragmenting world, we will only be able to provide good answers if we all work together.

Footnotes:

1.  The Allegory of Good and Bad Government – Wikipedia https://en.wikipedia.org/wiki/The_Allegory_of_Good_and_Bad_Government, (accessed on 4 May 2026)
2. Den Besten, T., P. Di Casola and M. M. Habib (2023), Geopolitical fragmentation risks and international currencies, The international role of the euro, European Central Bank, June 2023. Online: https://www.ecb.europa.eu/press/other-publications/ire/article/html/ecb.ireart202306_01~11d437be4d.en.html (accessed on 15 May  2026)
3. Deutsche Bundesbank (2026), Germany’s external assets in the light of geoeconomic tensions, Monthly Report, April 2026, (accessed on 4 May 2026)
4. U.S. Bureau of Economic Analysis, "Table 1.3. U.S. International Transactions, Expanded Detail by Area and Country" and European Central Bank, Statistical Release, Euro area quarterly balance of payments and international investment position: fourth quarter of 2025 (accessed on 15 May 2026),Note that the BEA reports on the euro area 20 while the ECB already reports on the euro area 21 including Bulgaria.
5. Pastoris, Schmitz (2025). Mind the Gap: Analysing Asymmetries in the Bilateral Euro Area–US Current Account. IMF 13th Statistical Forum: Measuring Cross-Border Economic and Financial Linkages in a Dynamic World Online: https://www.imf.org/-/media/files/news/seminars/2025/13th-stats-forum/session-7-fausto-pastoris-martin-schmitz-mind-the-gap-analysing-asymmetries-in-the-bilateral-euro.pdf (accessed on 4 May 2026)
6. OECD Asymmetries Explorer for Trade in Services: https://oecd-main.shinyapps.io/OECD_TradeAsymmetriesExplorer/ (accessed on 4 May 2026)
7. IMF Asymmetries in Bilateral Direct Investment Positions at the IMF's compilers hub https://compilershub.imf.org/home (accessed on 4 May 2026)
8. International Monetary Fund. Balance of Payments and International Investment Position Manual. White Paper 7th ed. Washington, D.C.: International Monetary Fund, 2025 https://www.imf.org/-/media/files/data/statistics/bpm6/draft-bpm7-wcv.pdf (accessed on 4 May 2026)
9. G20 Data Gaps Initiative https://www.imf.org/en/news/seminars/conferences/g20-data-gaps-initiative (accessed on 4 May 2026)
10. James Tebrake, El Bachir Boukherouaa, Jeff Danforth, and Miss Nivashini Harikrishnan. "StatGPT: AI for Official Statistics", Departmental Papers 2026, 004 (2026), accessed on 4 May 2026, https://doi.org/10.5089/9798229036863.087
11. Statistical Data and Metadata eXchange, is an ISO standard designed to describe statistical data and metadata, normalise their exchange, and improve their efficient sharing across statistical and similar organisations. SDMX is sponsored by eight international organisations including the Bank for International Settlements (BIS), the European Central Bank (ECB), Eurostat (Statistical Office of the European Union), the International Labour Organization (ILO), the International Monetary Fund (IMF), the Organisation for Economic Cooperation and Development (OECD), the United Nations Statistical Division (UNSD), and the World Bank.
12. SDMX Sponsor Organisations’ Statement on AI Readiness, 14 December 2026, online: https://sdmx.org/about-sdmx/sdmx-sponsor-organizations-joint-statement-on-ai-readiness/ (accessed on 15 May 2026)
13. Biancotti, Rosolia, Veronese, Kirchner, Mouriaux. COVID-19 and Official Statistics:
14. Treatment of Crypto Assets in Macroeconomic Statistics, IMF, 2019,   https://www.imf.org/external/pubs/ft/bop/2019/pdf/Clarification0422.pdf (accessed on 15 May 2026)
15. OECD (2024), OECD Survey on Drivers of Trust in Public Institutions −⁠ 2024 Results: Building Trust in a Complex Policy Environment, OECD Publishing, Paris, https://doi.org/10.1787/9a20554b-en.
16. Monetary Policy Statement Press Conference, Christine Lagarde, President of the ECB, Frankfurt am Main, 17 April 2025, https://www.ecb.europa.eu/press/press_conference/monetary-policy-statement/2025/html/ecb.is250417~091c625eb6.en.html (accessed on 4 May 2026)
17. Isabel Schnabel, Don’t take it for granted: the value of high-quality data and statistics for the ECB’s policymaking, Blog post, 20 October 2020, https://www.ecb.europa.eu/press/blog/date/2020/html/ecb.blog201020~0f195b9ea2.en.html
18. Joachim Nagel, Staying the Course: How Central Bank Independence Guides Us Through Uncertainty, Speech at the Foreign Policy Association, New York, 20 October2025 https://www.bundesbank.de/en/press/speeches/staying-the-course-how-central-bank-independence-guides-us-through-uncertainty-968754 (accessed on 4 May 2026)