Measuring and mitigating systemic risks: how the forging of new alliances between central bank and academic economists legitimize the transnational macroprudential agenda

Abstract

After the great financial crisis of 2007–2009, central banks were handed a macroprudential mandate to contain systemic risks, a mandate seen as endangering their independence due to expected distributional conflicts. At the same time, depoliticization through scientific expertise was largely foreclosed, as systemic risk was a largely undefined concept. This paper focuses on how central banks dealt with this conundrum. It examines the scientific debate on systemic risk and macroprudential regulation post-crisis, focusing on the debate’s impact on final regulation. Employing author-topic-modeling on a unique dataset of 2397 published economic papers on the relevant topics, we detect the formation of a new alliance between central bankers and academic economists working jointly on developing systemic risk measures. Centered around a hinge of systemic risk contribution by individual banks, this new alliance expresses itself by incorporating the macroprudential concerns of practitioners into abstract market-based systemic risk measures. These measures develop incrementally, using and repurposing techniques from financial economics pre-crisis to legitimize and justify macroprudential interventions post-crisis. This alliance allows us to account for the incremental change witnessed post-crisis and point to its potential for long-term fundamental change.

Keywords:

• G-SIFI
• macroprudential regulation
• regulatory science
• central banks
• systemic risk
• epistemic alliances

Acknowledgments

The authors would like to thank Alexander Kentikelenis, Oliver Levingston and Leonard Seabrooke as well as three anonymous reviewers for helpful comments in completing this manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Regarding the importance of these overarching dogmatic ideas for policy possibilities (Best & Widmaier, 2006).

2 Variations of the terms were also used for the search.

3 Using only abstracts would have provided a too limited text corpus. A statistical analysis on possible differences between the content of the remaining abstracts and the abstracts of full documents found no statistically significant difference for our topics.

4 The process with which ATM constructs authors has two important implications for our study. First, we focus only on full texts instead of abstracts because attributing a topic to authors becomes difficult when the word count is too low in multi-authored publications (especially after removing non-essential words). Second, we use a cut-off threshold of 0.25 to distinguish authors as belonging and not belonging to a topic. If, for example, only 5% of an authors’ work is on a certain topic, one cannot reasonably claim that the author has worked on it.

5 Mixed authors either have more than one affiliation in one document or different affiliations in two different documents.

6 The advantage of cumulative sampling is the ability to place newer literature in the context of previous periods. By doing this, we can improve our tracing of the origins of concepts, which would be much harder if we analyzed the periods separately.

7 In 2003, Claudio Borio wrote a seminal paper on systemic risk and macroprudential regulation, which galvanized research on this topic (Thiemann, Aldegwy, et al., 2018).

8 There is a crucial period from the onset of the financial crises until the passing of Basel III in 2013, during which regulatory action was taken. Any systemic risk measure which could be the basis for policymaking had to be present in this period.

9 ATM is essentially a tool for automating coding for large quantities of text. As such, the task of the researcher is to determine the correct number of topics for the specific research question at hand. Although there are some quantitative measures to choose the “correct” number, most researchers rely on qualitative topic testing. This involves checking the internal consistency of topics and whether the consistent topics appear even if the number of topics is increased (Chang et al., 2009). For this study, one researcher incrementally increased the number of topics until the above criteria were met. A second researcher qualitatively checked whether the final number of topics aligned with a qualitative reading of the documents.

10 Econophysics sets out to discover the mathematical properties of tail events to arrive at power laws capable of predicting the probabilities of events in non-Gaussian distributions (Bisias et al., 2012).

11 This is remarkable considering that the entire sample only contains approximately 34% of pure practitioners.

12 CoVaR and EMS are among the top 10 words that characterize this topic, SRISK in place 22. Qualitative reading confirms the central position of these three risk measures.

13 The CoVaR paper by T. Adrian and M.K. Brunnermeier is published in the American Economic Review in Adrian and Brunnermeier (2016), the Expected Marginal Shortfall by Acharya, V. V., Pedersen, L. H., Philippon, T., & Richardson, M. in 2017 in The Review of Financial Studies (Acharya et al., 2017) as is the SRISK paper by Brownlees, C., & Engle, R. F. in the same special issue (Brownlees & Engle 2017). Both journals are among the most influential economic journals registered at RePEc.

14 A simple Google citation analysis of the topic also shows the three respective papers to be cited most of the papers in the sample, with 2386 for CoVaR, 2004 for the MES paper, and 1175 citations for the SRisk paper in its different versions (December 2019).

15 The MES measure already circulated in 2009, s. e.g. https://economics.mit.edu/files/4907

16 CoVar uses 25 years, which includes three crises and two up- and downswings vs. the much shorter time span of 5–10 years for the other two papers.

17 CoVar was publicized in the Geneva report, which suggested it as the methodology to classify financial institutions based on objective risk-spillover measures (Brunnermeier et al., 2009). The report has been an important document in the macroprudential ideational shift (Baker, 2013a), published by influential change agents who actively sought to persuade the transnational epistemic community of new macroprudential ideas.

Additional information

Notes on contributors

Matthias Thiemann

Matthias Thiemann is an assistant professor of European Public Policy. His work focuses on the structural and ideational factors that shaped financial regulation pre- and post-crisis, with a focus on shadow banking and macroprudential regulation. His work has been published inter alia in the American Journal of Sociology and the Journal of European Public Policy.

Carolina Raquel Melches

Carolina Raquél Melches is a research associate in the German Bundestag. She works in the field of financial and banking regulation, public finances and economic policy. Among others, she wrote articles on central bank digital currencies and the German debt brake. In 2018, she graduated summa cum laude in Public Policy and Economics at SciencesPo Paris.

Edin Ibrocevic

Edin Ibrocevic is a doctoral student at the Max-Planck-Institute for the Studies of Societies. His main focus lies on the impact of scientization of central banks on knowledge production, policymaking and the state-market boundary.