Introduction

There are significant differences in the way central bank mandates relate to climate change. Many Asian central banks seem open towards incorporating the fight against climate change in their policies. For example, the Reserve Bank of India (RBI) addressed environmental aspects in a note to its supervisees in 2007, followed by the Bangladesh Bank regulations making banks incorporate environmental dimensions in lending in 2011 (Durrani et al., 2020). In Europe, however, most central banks had initially been hesitant to consider climate change in their policies, given their missing or ambiguous legal mandates (Dikau & Volz, 2021). Parallels can be drawn with the asset management industry, where fiduciary duty, a legal obligation to consider only financially relevant factors in investment decisions, was interpreted as a need to focus on short-term returns (Clark & Hebb, 2005). Trustees started to incorporate longer term considerations such as climate change slowly and by imitating others (Woods, 2011).

The goal of this paper is to describe and account for the spatio-temporal process whereby central banks in Europe gradually started to engage with climate change risks. We do so by using the sample of 941 financial stability reports (FSRs) of 38 central banks in Europe, published between 2003 and 2022. We identify and analyse bigrams, i.e. pairs of words related to climate change that appear in the FSRs, and we apply a graph-analysis to study source references of the FSRs. These methods allow us to tackle three questions. Where and when did the recognition of climate change risks by European central banks start? How did it spread across Europe in terms of time and space? How did the harmonization of analytical approaches play out in a geographical sense? Addressing these questions is important. First, it helps us understand the evolution of sustainable finance as a process of innovation, grounded in the context of particular times and spaces. Second, our study represents a contribution to financial geography, as it uncovers the spatial production of finance with regard to central banks’ engagement with climate change risks.

We show that central bank mandate interpretations have started opening slowly towards climate change-related topics since the Paris Agreement of late 2015 (EP, 2018; MNB, 2022). This included the systematic measurement and tracking of climate risk exposures (BIS, 2021; EBA, 2022). Initially, however, the methodologies and data concerning climate risks were almost nonexistent, and their development required time-consuming research and data gathering. The discussion of climate risks in financial stability reports started in the Central Bank of the Netherlands (De Nederlandsche Bank, DNB) and the National Bank of Belgium (NBB). It subsequently spread across Europe, with various types of proximities playing a significant role in the diffusion process (Rutten, 2016; Simandan, 2016). Fairly quickly, however, the collection and publication of methodologies moved to the European Central Bank (ECB), which started coordinating the ways to measure the financial stability impact of climate change, and also began drafting Eurozone-wide analyses in a uniform manner. We expect that in the future, climate change related financial risk analysis may not only follow more similar methodologies across countries, but central banks may develop more or less harmonized macroprudential regulatory measures on the topic (ECB/ESRB, 2022).

This paper proceeds as follows. The next section discusses the literature on the geography of innovation in relation to climate change and central banks. Section 3 details our data and methods. The following sections present a three-part analysis of: the evolution of European central bank climate change risk considerations in FSR reports, the map of these considerations, and the networks of their source references. The last section concludes.

Climate change, central banks, and the geography of innovation

The 2008 global financial crisis made evident that central banks’ narrow focus on monetary policy, complemented by a separate micro- and macroprudential regulation and supervision does not grant sufficient attention to financial stability risks (Balls et al., 2018). The subsequent widening of central banks’ accountability created the public illusion that central banks are „solely responsible for managing the economy, restoring full employment, ensuring strong growth, preserving price stability and safeguarding financial stability” (Dall’Orto Mas et al., 2020: 23). Some central bank leaders, recognizing the long-term threat of climate change on financial stability, voiced their concerns on the tension between public pressure and the limiting nature of central bank mandates. In his famous speech at Lloyds, Mark Carney, who governed the Bank of Canada from 2008 to 2013, and the Bank of England from 2013 to 2020, stressed that the implications of climate change are “beyond the business cycle, the political cycle, and the horizon of technocratic authorities, like central banks, who are bound by their mandates” (Carney, 2015). As a result, both more relaxed mandate interpretations and new sustainability mandates were witnessed during the 2010s (e.g., MNB, 2022).

Initially, as Dikau and Volz (2021) showed, only a small minority (12%) of 135 central banks in their sample adopted explicit sustainability mandates, and majority did not even have implicit references to sustainability. Zooming in on Europe, Dikau and Volz (2021) showed that during the 2010s, most EU central bank mandates only indirectly referred to sustainability, by citing Article 3 of the Treaty on European Union. It states that the EU “[…] shall work for the sustainable development of Europe based on balanced economic growth and price stability, a highly competitive social market economy, aiming at full employment and social progress, and a high level of protection and improvement of the quality of the environment”.

According to both Dikau and Volz (2021) and Durrani et al. (2020) the growth in the adoption of climate change and sustainability in central bank mandates turned strong and evident after the Paris Agreement, signed on 12 December 2015 and entered into force on 4 November 2016. Its signatories agreed on country-level legally binding targets on CO2 emissions. Subsequently, the European Parliament (EP) reaffirmed its implications for the European Central Bank in its 2016 annual report, stating that “the ECB as an EU institution is bound by the Paris Agreement” (EP, 2018). This moment was a clear signal for most EU members: as the second anniversary of the Paris Agreement in 2017 drew closer, the Network for Greening the Financial System (NGFS) was created by eight central banks (French, German, Dutch, Swedish, English, Hong Kong, Singapore, and China). These central banks agreed to share best practices on climate change related issues a voluntary basis. By June 2023 NGFS had 127 central banks and financial regulatory agencies as its members (NGFS, 2023).

The pressure on central banks has also taken the form of public, media, and expert opinions, as the consequences of climate change have gradually arrived at the doorstep of central banks, including the ECB. During the summer of 2018, for example, drought led to low waters, and cargo ships were not able to travel on the Rhein, leading to fuel shortages in southern Germany (Süddeutsche Zeitung, 2018). Another factor was the criticism that the quantitative easing (QE) programmes of central banks, including the ECB have been overwhelmingly skewed towards high greenhouse gas emitters (Matikainen et al., 2017). With central banks becoming central to the allocation of growing financial resources, demands that they should explicitly include the climate agenda in their strategies became widespread. As a result of all these pressures, the ECB Strategy Review of 2021 laid out an action plan to include climate change considerations in the Bank’s monetary policy strategy, including regular climate stress testing of the Eurosystem balance sheet from 2023 onwards (ECB, 2021c).

Emerging studies are not unanimous in welcoming central banks’ engagement in the climate change policy. Cullen (2023) questioned whether central bank policies addressing climate change could ever be effective. Christophers (2017) criticised climate risk disclosures as a tool of neoliberal governance, based on false beliefs in markets efficiently incorporating such information. Our goal in this paper is not to judge such policies and their effectiveness, but to uncover the spatio-temporal dynamics of central bank engagement with climate change risks in terms of their impacts on financial stability. To this end we build on literature in economic and financial geography.

To start with, we understand finance in the light of literature on socio-technical systems (Geels, 2005) as a socio-technical regime, based on seven interconnected pillars of financial firms, regulation, policy, technology, market and user preferences, as well as the science and cultures of finance (Urban &Wójcik, 2019). The system is capable of innovation, constantly producing new financial services, but changing the system as a whole, and moving its socio-technical regime from finance as we know it to sustainable finance, is difficult, complex, and takes a long time. This is because the building blocks of finance as a socio-technical system are tightly interrelated in terms of functions even if they can vary across space in terms of forms. For example, the science of neoclassical finance with modern portfolio theory and efficient market hypothesis in the lead (see e.g. Clark & Wójcik, 2024) encourages market and user preferences for chasing short-term returns, which are in turn facilitated by short-term reporting (as part of regulation) and reflected in the culture of short-term performance incentives. What can facilitate such transition are changes in the broader environment of finance, e.g. the reality of climate change and its consequences, but what is necessary are also specific innovations. These can start as disjointed niche initiatives, but as they grow and link together, and put pressure on all building blocks of the existing socio-technical regime of finance, they may gradually challenge and replace it.

In the light of this definition, consider that central banking, a major building block of the current socio-technical regime of finance, has also been subject to innovation. Central bank mandates themselves are not set in stone but are the product of specific spatio-temporal contexts. The inflationary contexts of the 1970s and the 1980s in North America, Western Europe and beyond made the mandates of many central focus on price stability. However, in the wake of the global financial crisis of 2008, financial stability was elevated to a concept of similar merit, making regular financial stability reporting a widespread practice among central banks (Issing et al., 2009).

What can we expect the evolution and geography of central banks’ engagement with climate change, conceptualized as niche innovations, to look like? Innovations can come from the geographical and institutional centre of the existing regime or from its periphery. On the one hand, financial centres, where central banks typically reside, are places of financial expertise capable of generating niche innovations. Knox-Hayes (2009), for example, found that the first successful carbon markets were created where service providers with complementary skills were present, and where knowledge transfer was made easier by frequent interactions among market and other players, notably, in London and New York. On the other hand, inertia in the centre may be strong, and smaller, more peripheral, centres can stimulate innovation as they seek innovation to disrupt the status quo. Innovation may also be stimulated in more peripheral financial centres if they are exposed to changes in the environment of the existing regime, e.g. to disruptive impacts of climate change, such as flooding. Our first hypothesis is therefore to expect the first appearance of innovations related to climate change risk analysis in regions where climate threat is rampant, and not necessarily in leading financial centres.

Once a niche innovation emerges, its spread and impact depend on proximity. However, physical or “geographical proximity per se is neither a necessary nor a sufficient condition for learning to take place” (Boschma, 2005, 61). First, to learn from each other people need cognitive proximity – they must share the same knowledge base, as it helps them to communicate, understand, and process new knowledge successfully. However, when cognitive proximity (and overlap in knowledge) is too large, this limits potential for new knowledge, with too much silo thinking. Central bankers in Europe can certainly be expected to have much cognitive proximity, with overlapping knowledge, as most are likely to have background in economics (or law) and experience in the financial sector. One may rather question whether such a degree of cognitive proximity, with few geographers and environmental scientists working in central banks, does not stifle learning about the impacts of climate change.

Second, organisational proximity concerns relations within and among organisations, with loose arrangements and arms-length market transactions on one, and tight hierarchy within an organization on the other extreme (Boschma, 2005). Organizational proximity is inherent to the European System of Central Banks (ESCB), a platform for all central bankers in the EU, and more notably, to the European Central Bank (ECB). All EU nationals are eligible to work in and all EU countries have relations with the ECB. Despite that, the ECB represents a tighter community for Eurozone member countries. For example, the Governing Council, the main decision-making body of the ECB, consists of the governors of the Eurozone national central banks, in addition to its Executive Board members (Clark, 2015).

Third, social proximity refers to socially embedded relations among agents i.e. those involving trust based on friendship, kinship, and experience. While standardized knowledge can be exchanged without trust, the latter is required to exchange tacit knowledge. Social proximity may be sparked by organizational proximity, frequent work-related interactions at conferences, in expert groups of the ESCB and on the job at the ECB, just to name a few opportunities. These, put together, lead to the second hypothesis of this paper: we expect climate risk related innovation to spread quicker among Eurozone members than among other ESCB members thanks to the more closely-knit community of its actors.

Finally, institutional proximity concerns shared values, norms (informal institutions), laws and rules (formal institutions) at the macro-level. As such, it provides a stable environment and social cohesion conducive to interactive learning and innovation.

More specifically, the institutionalization of an innovation brings actors closer to each other and fosters future collaboration. One former example for this is provided by Bieri (2009), who described the geographical spread and the global acceptance of the bank risk regulatory framework, coordinated by a central institution, widely known for its location as the Basel-process. Our third hypothesis rests on this recognition, and provides evidence for the.

spatial institutionalization of an innovation. It is expected to gravitate towards its institutional centre, with a central institution collecting and aggregating the technical-institutional innovation. While the ultimate determinant of institutional proximity in European central banking is the EU, to a significant extent its forces are projected via its central element, the ECB, which we expect to move to the main stage in the spread of climate risk analysis with time.

Policy makers have been working to eliminate the impact of physical distance from the institutional geography of central banking within the EU: they have been creating opportunities to compensate for it with temporary physical proximity, as well as meetings and co-working enabled by the intensive use of digital technologies. Despite that, all types of distances, particularly institutional and organizational, rise dramatically as one moves from the Eurozone to other central banks of EU member states, and then to European central banks outside the EU. Admittedly, there are forces and opportunities of proximity, including joint workshops and technical assistance programmes to support countries waiting for EU accession (like with Ukraine and Moldova), and at international central bank organisations such as the Bank for International Settlements (Bieri, 2009). However, as this paper illustrates by analysing the usage patterns of climate change risk analytical techniques, these activities have still not eliminated the significance of various forms of distance.

Data and methodology

Financial stability reports

To understand the geography of climate change risk analysis, we looked at the FSRs of central banks in Europe. One may debate if FSRs are the right kind of publications to study the ways central banks discuss climate change risk. Admittedly, climate risks can be analyzed in a macroeconomic setting, which are not necessarily central to FSRs. Sovereigns are at risk of default as a result of climate change (Volz et al., 2020); and macro models with sectoral breakdowns can be powerful tools for macroeconomic analysis, examining the impact of climate change on GDP, inflation and other macroeconomic variables (CamEcon, 2021). The rising significance of financial stability in central bank mandates and modelling efforts after the crisis of 2008, however, tilted climate change analysis towards a financial stability perspective.

Financial stability, the ability of the state in which the “financial system […] is capable of withstanding shocks and the unravelling of financial imbalances” (ECB, 2016),Footnote 1 is assessed in FSRsFootnote 2 by most central banks. An FSR is a “vehicle to allow stakeholders to form a view about how effectively the central bank is undertaking its broader financial stability responsibilities” (Allen et al., 2004). It usually consists of’core’ and’non-core’ parts, whereby core-sections cover the same topics from issue to issue, and non-core parts include results from special and occasional studies (Cihák, 2006).

An FSR has a limited size, hence its contents an indication for the relative importance of matters, as seen by the team responsible for analysing financial stability. Admittedly, the implications of an urgent phenomenon may dominate the pages over long term but unfortunately serious but seemingly less urgent matters such as the climate crisis do not. In addition, these reports are written by a limited number of people within a limited timeframe. For instance, Brexit-related challenges or the energy crisis triggered by the war in Ukraine made several financial stability teams recalculate default probability estimations, which simply used the time and space from climate risk analysis.

The Task Force on Climate-related Financial Disclosures categorizes climate risks into physical risks, defined as the damage or complete destruction of physical assets stemming from climate change induced weather events, and transition risks, which accompany the transition to a low- or zero carbon economy. Both acute physical risks, which are event driven, and chronic risks, which result from long-term shifts in climate patterns, such as sea level rise (TCFD, 2017), may diminish or destroy a borrower firm or individual’s ability to repay its loan obligation on time (increase their ‘probability of default’ values), and increase the ‘loss given default’ values, both of which are core concepts in a typical analysis in FSRs within the Basel Risk Regulatory Framework (Aspachs et al., 2006). Similarly, transition risks, such as investments in low- or zero emission production techniques, house insulations, or even climate-related litigation costs affect future credit quality.

To add another argument for studying FSRs in the context of climate risk analysis by central banks, Pointner et al. (2019) summarize the relationship between further Basel III-risk types, typical concepts in financial stability reports, and physical risks. They concede that market risk (increased price volatility due to disasters), liquidity risk (sudden withdrawals, high demand for emergency loans), operational risk (destruction of banking infrastructure), reputational risk and systemic risk (rising correlation of defaults, underestimation of natural disasters) may all be results of sudden, acute physical disasters.

The data

We manually downloadedFootnote 3 941 FSRs of 38 central banks from Europe using the central banks’ publicly available websites (Table 1). We collected FSRs from the earliest date we found them online, until the end of 2022. While most of FSRs are available in English, there are exceptions. The Banque Centrale du Luxembourg publishes in French only (we translated their bigrams into English). Each of the 38 central banks publish one or two FSRs every year, or none,Footnote 4 and their length varies between a few and 200 pages.

Table 1 Countries included in the analysis
Full size table

We analysed bigrams and references in the FSRs. First, we scanned the FSRs for climate change related expressions and counted them, using a few dozen lines of code in R.Footnote 5 Since all FSRs come in a pdf format with ASCII coding, the data generated exhibit high accuracy. A few inaccuracies were detected: (1) with words added to graphs, (2) special characters used (such as the El Niño) or (3) where the words were separated (such as cli-mate). To ensure the veracity of our data we went through all of the FSRs of the most important central banks, checked other central banks where no climate-relevant information was indicated by the data; and read the pages where climate-relevant text was found. This latter was also crucial to obtain a general understanding of the spatio-temporal process.

Climate-relevant expressions are best captured as bigrams, or two consecutive words, such as ‘physical risks’, ‘hot house’ (a scenario defined by NGFS), ‘green bonds’ or ‘climate stress’. The use of its monogram versions, or single words, can appear in other contexts, such as ‘physical money’ or ‘economic climate’, and would distort the analysis. Bigram search is a widely accepted way of analyzing cultural tendencies in books (Michel et al., 2011) or in bibliographic databases and journal archives (Chumtong & Kaldewey, 2017). Financial geographers have also used bigram frequencies (Clark et al., 2015).

Relevant bigrams were identified using two sources. First, we used journal papers and central bank documents most often cited by FSRs’ climate change sections, which specifically focus on the modelling questions and financial implications of climate change.Footnote 6 These documents were selected either because they are the most often cited documents by FSRs or we found them relevant during the literature review. From these documents, we selected the top two thousand most frequent bigrams,Footnote 7 after removing stop words,Footnote 8 and expression stemmingFootnote 9 (removal of endings). Then we manually flagged climate change relevant items among the most frequent bigrams. The second source were the most relevant bigrams from the journal Nature Climate Change between 1990 and 2014, as identified and listed by Hamed et al. (2015). The reason for the use of a second source is to counter any criticism of a potential bias in the papers we used to learn relevant bigrams. Admittedly, there is some overlap between the authors of the climate-relevant FSR sections and the authors of these papers, who may use their own vocabulary in both. We counterbalance this potential effect by including bigrams from Nature Climate Change. The final bigram list is included in the Appendix, which shows bigrams analysed in the following sections.

In the next step, we classified each bigram in the learner set manually based on whether it related to physical risks (P), transition risks (T) or both (B). Finally, we counted the number of bigram appearances within the FSRs, and aggregated the results for each year.

The second part of the analysis involved the opening of all relevant FSRs, and manually collecting references from the sections discussing climate change. This way, a table was constructed with the FSR’s publication year, the title of publication, its year and institution of publication, containing a total of 587 references. We also added an additional variable to record if the publication is available from a central bank (not just FSRs but working papers and other publicationsFootnote 10), from a scientific journal or at any other institution.

Tracing climate change risk considerations over time

The results of the climate change related bigram search by central bank are presented in Fig. 1. It shows low occurrences up until 2018, when the first wave of attention to climate change risks is led by the Dutch and the Belgian National Banks, later followed by a more prominent presence of climate-related issues in other central banks’ FSRs.

Fig. 1
figure 1

Absolute number of climate change-related bigram occurrences in FSRs of selected central banks of Europe

Full size image

Until the mid-2010s, FSRs discussed natural disasters as deplorable events which are regrettably parts of life. The first larger bigram clustering is observable in the FSR of the Österreichische Nationalbank, Austria (ÖNB), in 2002. Their second FSR in that year discussed the financial consequences of the devastating floods of 2002. In a stark contrast to the way today’s FSRs capture climate change physical risks, the implications of the flood are discussed as a macroeconomic shock: “the flood is a typical example of a negative supply shock, as it damaged important infrastructural facilities, caused crop failures, destroyed production sites or forced temporary interruptions in production” (ÖNB FSR, 2002/2, p. 29). The report notes that negative supply shocks caused by natural disasters are often followed by positive demand shocks, triggered by the reconstruction. “Economic growth in Austria is going to decline by 0.12% in 2002 vis-a ‘-vis the baseline scenario, but is expected to be 0.2% to 0.3% higher in 2003 owing to investments”, without any reference to Basel-type concepts such as financial stability or default probabilities, and without the introduction of new innovations.

ECB FSRs mention climate related concepts (hurricanes in the Caribbean and on the Atlantic coast of the US) since 2004, the first year an FSR is available on their website. The likely reason behind this attention is that at the time, five of the ten largest reinsurers globally were European companies with considerable exposures in North America. Similarly, the increase in climate-related bigrams in 2007 was driven by a longer text box on the relationship between insurance sector and climate change (ECB FSR, June 2007, p. 127). It discussed increased expected losses due to “extreme weather events”, stating that “[i]nsurers and in particular reinsurers can transfer part of the risk associated with natural disasters to the capital markets using instruments such as weather derivatives and catastrophe bonds”. This seemed to solve the concerns related to financial stability, since “[t]he euro area insurance sector, and in particular the reinsurance segment, is increasingly prepared to handle possible future high-impact, albeit low-probability, events, or several closely spaced events affecting parts of the sector and individual insurers”.

These quotes illustrate the initial approach to climate change as an insurable risk. As such, the financial system was deemed capable of fending off temporary imbalances without interruptions in crucial services to households and businesses. As the December 2011 FSR stated the „occurrence of natural catastrophes also enables insurers and reinsurers to increase their prices”. Meanwhile “massive tornados that hit the South and Midwest of the United States in April and May, and Australian floods in January 2011 […] and the expected impact of La NiñaFootnote 11” (ECB FSR, Dec. 2011, 109) represented little challenge to the business-as-usual, since its effects were expected to be solved within the financial system.

The Dutch and the Belgian National Banks’ FSRs represent the most wordcounts in the first year with significant bigram mentions—2018. This turn was not without precedent: the DNB published an exploratory study on the exposure of the Netherlands to climate transition risks only in 2016 (Schotten et al., 2016), and as an explanatory summary document on transition and flood risks (DNB, 2017), without any consideration of further physical risk types. The attention to flood risk is certainly the product of the Dutch history of land reclamation from the sea and managing water-related risks, which is expected to worsen as a result of the warming climate.

The DNB published a real climate stress test one year later (Vermeulen et al., 2018), with parts of it cited in the following FSR (DNB FSR, autumn of 2018). The reason for their focus on transition risks and the exclusive attention to flood risk as physical climate risk sheds light on the difficulties involved in the calculations which is related to limited data availability. While assessing transition risks requires firm- and household level data on energy mix, intensity and some profit-and-loss calculations only, physical risk estimation requires accurate firm and household location data, sophisticated geographically sensitive models on physical risk exposures to several risk types and damage functions which explain the value loss when a natural catastrophe occurs.

The National Bank of Belgium (NBB) was also among the first to discuss climate change in length. It published a study on the macroeconomic impact of the fight against climate change already in 2011 (Bruggeman, 2011). This was followed by a long climate risk-related discussion in the FSR of 2018. The main difference to the approach of DNB was the NBB’s limited contribution to methodological innovations. The NBB FSR of 2018 focused on general economic implications, including a literature review and a brief discussion of financial stability related topics. In comparison, the DNB presented novel results using a new stress testing method, which was a significant a contribution to the climate risk assessment methodology (Vermeulen et al., 2018).

The DNB’s involvement confirms our first hypothesis, which stated that a climate change-related threat would ignite the analytical innovation on climate change risks.

Up until 2018, climate-related discussions in the FSRs other than the DNB’s accompanied the phenomena of climate change but did not bring real innovation to the table. The early involvement of the Belgian central bank in climate change risk assessment may be related to its physical proximity to the Dutch central bank, and likely interactions between the two, but also to the presence of EU institutions and environmental non-governmental organisations in Brussels.Footnote 12 In any case, its novel content in 2018 stayed behind the levels seen at the DNB.

Elsewhere, the FSR of the Bundesbank, Germany, referred to the rising number of natural catastrophes in 2005. Notably, the report indicated the limited ability of the insurance sector to distribute natural catastrophe risks on their own. Instead, it highlighted the need to use catastrophe bonds more intensively, involving other financial participants in physical risk sharing: “Hurricane Katrina recently highlighted the limits of insurability […] Catastrophe risks […]are transferred to the capital markets through the issuance of bonds. If the loss event occurs, the bondholders’ principal and interest, as a rule, pass to the issuer” (Bundesbank FSR 2005, 88). However, the FSRs of the Bundesbank remained relatively quiet about climate change until 2019. Its 2019 report subsequently described the analytical framework found in other publications, which could be used to measure the financial stability impact of climate change, but stopped short of providing figures or proposing novel approaches.

In a similar vein, the Banque de France first realized the impact of climate change on its reinsurers. It has been discussing the topic in its FSR since 2017 and 2018, followed by a reference to a coordinated work with the NGFS in 2019. However, it only reported on its first results in its June 2021 FSR. Admittedly, the June 2021 FSR summarizes the results of a stress test involving both transition and physical risks.

To analyze spatiotemporal patterns in a more granular manner, Table 2 contains the first year a bigram is mentioned three times or more in a country’s FSR. The top row contains the total number of climate-related bigram mentions for a particular country. The first figure column represents the frequency a specific bigram appears in all FSRs. Due to size limitations, only the top countries and top bigrams are displayed. We have determined the column order arbitrarily, to reflect total bigram mentions and temporal patterns.

Table 2 The total number of (stemmed) bigram mentions by country (top row), bigram mentions for the top bigrams (first column) and the first year an FSR mentions a bigram more than three times
Full size table

Table 2 demonstrates that in the year 2019, climate change-related writings appeared in the FSRs of the ECB, France, Austria, UK, Germany, Sweden, Romania, Slovenia, Portugal and Spain. While country-level developments are discussed later, Fig. 2 shows the temporal advantage of the Dutch, Belgian, French Central Banks and the ECB (we refer to them as the "central countries” to the spread of this innovation); as well as of the Eurozone. The green line in shows the share of these banks’ climate change-related bigram mentions in their FSRs out of all bigram mentions. It shows that in 2017 and 2018, these were the dominant central banks discussing climate change in their FSRs. Once the diffusion process reached other European countries, in 2019, other Eurozone central banks caught up the pace and joined the exchange (blue line). Subsequently, their relative share diminished as other non-Eurozone countries’ national banks started writing on the issue in their FSRs. Figure 2 and Table 2 reinforce our second hypothesis, namely, that after the 2017, the FSRs of Eurozone-countries, and among them, the central countries represented the majority of climate change-related bigram mentions.

Fig. 2
figure 2

The combined share of the ECB and the central banks of the Netherlands, Belgium, and France, and the share of the central banks of the Eurozone in all climate change-related bigrams in all European FSRs

Full size image

Putting the analytical framework to use not only required the development and application of new analytical techniques, but also investment in data collection efforts (as recognized by NGFS, 2022). In the ECB FSR, a special chapter addresses the data-related problem outright in the May 2019 issue. While on the transition risk side the uniform classification of sustainable assets was missing (the EU Taxonomy Regulation came into force in 2020, and its delegated acts in 2022 only), using a benchmark-level emissions intensity and a price for carbon-dioxide quotas is a simple task in terms of data requirements. In contrast, the ECB report refers to the existence of “measurement gaps” in “understanding of exposures of financial institutions to climate change-related risks” (ECB FSR May 2019 p 121). More specifically, this concerns the location of bank exposures (e.g., where the buildings financed by mortgages are), and their exposures to risks (how much will be damaged). As a result, the report reaches out to aggregated catastrophe loss figures from reinsurers. Two years later, the ECB first climate stress test relied on (still approximated) data provided by the startup Four Twenty SevenFootnote 13 (ECB, 2021b, c, 30), and modelled corporate exposures to physical climate risks in a simplified fashion (for instance, using headquarters as opposed to actual production sites), omitting residential mortgages entirely.

This problem was also visible on bigram-data: all central banks face the data and methodology related difficulties when assessing the impact of physical risks on bank balance sheets. In fact, only 18 percent of all bigrams are directly related to physical risks, whereas 33 percent cover transition risks (the rest covers general concepts related to both). On country level,Footnote 14 most first mover central banks (including NL, BE, ECB and FR) first wrote on transition risks, followed by discussions of physical risks.

In 2021, it was again the DNB that came up with a novel way to calculate exposures: it increased the granularity of its physical risk analysis significantly- unsurprisingly – related to flood risk in the Netherlands. We cite the following quote as a whole to illustrate the data-related pain points in such a modelling: “we have drawn together granular data from a range of individual sources to create a dataset representing over €700 billion in domestic real estate exposures. This includes both regular data that we gather as well as data from a special survey of insurers and pension funds. The picture that emerges is that the majority of real estate is located in parts of the Netherlands that may be affected by flooding. These include both areas which are not protected by flood defence systems (outside the dikes) and parts of the Netherlands that are protected against floods” (DNB FSR, Autumn 2021).

Mapping climate change risk considerations

To complement the analysis of bigrams over time we use a map showing the number of bigram occurrences between 2000 and 2022 in Fig. 3. The brightest coloured country on the map, with most occurrences, is Austria, where the Österreichische Nationalbank (ÖNB) follows a special structure in its FSRs. The non-core parts of the FSR are basically research papers, and as such, they offer significantly more text than the text boxes of the ECB or other banks. In 2019, ÖNB published a methodological overview of a Basel-conform climate change modelling (Pointner et al., 2019, in ÖNB FSR Dec. 2019), followed by a report on a climate transition stress test in November 2020 (Battiston et al., 2020, in ÖNB FSR, Nov. 2020), and a report on a carbon price shock in 2021 (Guth et al., 2021 in ÖNB FSR Nov 2021).

Fig. 3
figure 3

Number of bigrams in FSRs between 2000 (or earliest available) and 2022. Notes: the map excludes ECB FSRs; Bulgaria does not publish FSRs, Greece started publishing them only recently. Unfortunately, we could not find a shapefile which includes Kosovo

Full size image

The second brightest country on the map is Belgium, where the NBB continued to discuss climate change in its FSRs after 2018. However, similarly to their restrained attitude in 2018, their 2019 FSR was limited to the discussion of a survey carried out among commercial banks in Belgium. In 2020, it presented a scenario analysis of a “potential introduction of a minimum energy performance standard for residential real estate on the banking sector’s mortgage portfolio” (NBB FSR 2020, 147), which resembles to a kind of climate stress test on transition risks.

Another bright country in Fig. 3 is Norway, with the Central Bank of Norway first discussing climate change in its FSR of 2019. While the report states that “Norway’s exposure to direct physical climate change is limited”, and remains analyzed in a shallow manner only, a lengthy chapter is dedicated to the consequences of a potential drying in oil demand on the country’s financial stability (Norges Bank FSR, 2019, 47). This late consideration may be surprising for one of the largest oil-exporting country globally, where the Norges Bank itself runs an investment management fundFootnote 15 to invest oil- and gas-related revenues for future generations. The fund has decades-long track record in incorporating environmental (as well as social and governance-related) factors in its decision making (e.g., Clark & Monk, 2010; Reiche, 2008). It is therefore unlikely that the financial stability team of the Norges Bank had not been aware of the salience of the climate issue. Rather than lacking knowledge, relevant departments of Norges Bank probably did not see the point of analysing and presenting climate risk in an FSR.Footnote 16 This omission could have also been influenced by the social and organizational distance to the DNB and other innovating central banks.

The limited space dedicated to climate change in the FSRs of the Bank of England (BoE) may also be surprising, with a total of bigram mentions of around 240 (Fig. 3 and Table 2). While the BoE had contributed several innovations to climate risk analysis in the 2010s, its FSRs focused on other, seemingly more urgent matters such as Brexit.Footnote 17 The absence of top-down climate analyses is somewhat alleviated by the fact that the bottom-up stress test results, carried out using inputs from supervised banks and insurers, were published in May 2022 (Bank of England, 2022).

There are further surprises in Eastern Europe too. The FSRs of the Romanian Central Bank first referred to measures aiming at reducing greenhouse gas emissions already in 2011. This was a description of a programme financed jointly by the Romanian Government, the EU, IMF and the World Bank to improve the country’s energy efficiency. In 2019, Romania’s FSR mentioned climate change again. The special section of the FSR on climate started with discussing funds the EU had been spending on tackling climate change, followed by a high-level analysis showing the vulnerability of the financial system to climate risks in the country. Elsewhere, Serbia’s central bank addressed climate change in the FSRs of 2018 and of 2020. Besides the description of basic ideas and indicators, both reports mentioned credit lines provided by the European Bank for Reconstruction and Development (EBRD). In comparison, Slovenia, a Eurozone member, also dedicated considerable space to the transition risk of its financial sector in its FSR of 2019, admitting that the country had already met the 2030 emissions target in that year.

The sources of climate change risk considerations

The analysis of climate change-related citations in FSRs provides direct evidence on the influence a central bank has on others. Table 3 shows the gradual disappearance of national central banks from the top positions of influenceFootnote 18: references to papers and talks published the DNB and the Bank of England dominated FSRs in 2018, including DNB (2017), Schotten et al. (2016) and Vermeulen et al. (2018) for DNB, and speeches of Carney (2015) as well as Batten et al. (2016) for the Bank of England. DNB and BoE positions decline over time and disappeared from top five in 2021 and 2022. Once the community of central banks processed their innovations, approaches and methods to climate change risks underwent coordination, and international bodies grew to dominate the rankings of influence, led by the ECB, Network of Central Banks and Supervisors for Greening the Financial System (NGFS), and Intergovernmental Panel on Climate Change (IPCC). This is direct evidence in favour of our third hypothesis, stating that innovative knowledge accumulates at the ECB, from where it gets redistributed again.

Table 3 The most cited references on climate change in European central banks’ FSRs
Full size table

The directed graph of references in Fig. 4 zooms in on central bank publications only, visualizing our argument further. The arrows point from central bank FSRs towards other central banks which are referenced. The years represent the publication year of the FSRs. The graph visualizes the centrality of the DNB and the Bank of England in 2018, which only grew stronger in 2019. The pandemic year of 2020 generated a lower number of citations compared to the preceding year. Nevertheless, in that year the ECB became the most cited central bank, and its dominance expanded further in 2021. In 2022 the absolute number of references declined, but the ECB’s central location remained unchallenged.

Fig. 4
figure 4

Citation directions in FSR reports. Notes: Arrows point from the central bank that used a reference in their FSR towards the cited institution. The graph shows the number of links. Lines have the same weight irrespective of the number of citations

Full size image

Besides the centralization of citation patterns, the institutional agreement in methodological questions was underscored by the ECB itself: “a growing body of empirical evidence on climate-related risks to financial stability has now provided a robust analytical foundation for macroprudential policy considerations” (ECB/ESRB, 2022, 3). The focus has gradually shifted from discussing research to refinement questions, and to regulation. Both ECB/ESRB (2022) and the EBA (2022) papers started studying the ways climate risks can be incorporated into a macroprudential risk framework, implying that the regulators consider options on how higher or lower climate risks in bank portfolios should appear in capital allocation processes. More specifically, the EBA (2022, 12) states, “[t]he analysis demonstrates that the Pillar 1 framework already includes mechanisms that allow the inclusion of new types of risk drivers such as those related to environmental risks”. Such a regulation on climate risks and capital requirements would move the process of climate risk analysis even higher up the institutional ladder, to the EU legislation level, which would represent an additional centralization of an analytical method which initially started as an innovation outside the centre of the hierarchy.

Conclusions

Our goal was to examine the spatial and temporal process of financial innovation, as climate change risk analytical methods gradually conquered the pages of financial stability reports (FSRs) issued by European central banks. At least three major factors influenced the process: the evolving interpretation of central bank mandates as a result of public pressure, the recognition of the salience of climate change by central banks, and the spread and availability of know-how related to the analysis of physical and transition-related climate change risks.

Our results, based on identifying climate change related mentions and references within 941 FSRs of 38 central banks in Europe, confirmed our three hypotheses. First, we showed that the tipping point in climate risk analysis was the Paris Agreement of 2015, followed by the decisions of elected politicians, among them, MEPs of the European Parliament (EP, 2018), to explicitly instruct central banks to consider climate change in their policies. It was, however, a few central banks influenced less by a narrow interpretation of their mandates and exposed more to climate change threats, among them, the Dutch Central Bank (DNB), which developed methodological innovations and started publishing them.

Second, we showed that physical and organizational proximity mattered in the spread of climate risk analytical techniques. Countries in the proximity of the Netherlands, as well as Eurozone members, were among the first adopters, followed by other countries outside the Eurozone, and the EU. This observation underscored the expectation that distances and geographies matter and have a major role in determining the way European central banks are looking at climate risk exposures.

Third, our analysis of citations and references within FSRs has demonstrated that this successful innovation required the involvement of the centre, with the ECB gradually turning into an aggregator of analytical knowledge, and as a result, serving as a reference point for all central banks pursuing new research and analysis in this area. This centralization ensures the EU-wide acceptance of the results, catalyses communication and cross-country comparisons, as well as makes economies of scale possible. Moving forward, the role the ECB played in the organizational and social acceptance of the results of climate risk analysis may be a prerequisite for a uniform macroprudential treatment of climate change risks within the EU to be reflected in regulation.

To be sure, financial stability reports will not save us from the climate disaster and are only a reflection of whether and how central banks consider climate change risks in their analysis. Nevertheless, in order to make finance, and more specifically, banking an effective player in the arduous journey towards net zero, ideas to be implemented not only require the involvement of its individual actors, but should be catalysed by a flexible central player open to innovation. It can provide fruitful avenues for thought exchange by reducing the effects of geographical distance, and generate positive externalities. Further, such central players may impose regulatory frameworks with success once a mutually acceptable and implementable solution is found and supported by its members.