Recent reforms that followed the Great Financial Crisis, as the establishment of the Single Supervisory Mechanism in Europe and the Prudential Regulatory Authority in the UK, reflect the belief that the governance of banking supervision affects financial stability. However, while existing research identifies the pros and cons of having either a central bank or a separate agency responsible for microprudential banking supervision, the advantages of having this task shared by both institutions (shared supervision) have received considerably less attention.
In a recent Bank of England Working Paper, I fill this void by comparing empirically the impact of three supervisory governance models — supervision by the central bank, by an agency or by both of them — on bank non‑performing loans (as a share of total gross loans). To do so, I introduce a new database on the governance of microprudential banking supervision in 116 countries from 1970 to 2016. The findings of this work lead to two considerations. First, I find that shared supervision is associated with higher financial stability (proxied by non‑performing loans). Second, I show that this effect is statistically significant only in countries that score poorly in terms of corruption. Shared supervision is therefore associated with better financial stability outcomes in countries where regulatory capture is a concern. However, when the risk of capture is low, there is no significant difference in terms of non-performing loans between having a single supervisor or shared supervision. This is consistent with the hypothesis that supervisory governance affects financial stability by affecting the likelihood of supervisory capture. This short article summarises how I reach these conclusions.
Before doing so, it is important to better define shared supervision. In the paper I focus on those cases where microprudential banking supervision is a function shared by two different agencies. This means that I do not consider as shared those arrangements where supervision is assigned to two agencies but with different functions, such as the separation between prudential and conduct supervision (also known as ‘twin peaks’ model), or the sectoral separation of supervision across banking, insurances and securities supervisors. For example, I do not categorise the current setting in the United Kingdom as shared, since the Prudential Regulatory Authority is the only microprudential banking supervisor, while the Financial Conduct Authority supervises the conduct of the financial sector.
There are at least three reasons for which we should care about shared supervision. First, existing comparisons between supervision by a central bank and an agency, which overlook the case of shared supervision, led to mixed empirical results (e.g. Goodhart and Schoenmaker 1995; Dincer and Eichengreen 2013), leaving unclear which supervisory model works best for financial stability (Koetter, Roszbach and Spagnolo 2014). Second, even if the majority of countries has a central bank as sole supervisor (see page 10 of the paper), in other important economies, such as Germany and China, banking supervision is shared between the central bank and an agency (supervision in the United States is a special case and is excluded from the analysis; see the full paper for more details). Third, the theoretical literature argues that shared supervision prevents against the risk of capture of regulators and supervisors from the private sector. With two supervisors, banks would need to double their effort to take control over supervision (Laffont and Martimort 1999 and Boyer and Ponce 2012) compared to the costs of capturing a single institution, be it the central bank or another agency (by contrast, others argue that the presence of two supervisors might give rise to coordination problems that could slow down or delay actions aimed at reducing bank risk-taking: see: Barth et al., 2002; Wall and Eisenbeis, 2000; Briault, 1999). If this is true, the debate should not focus on whether supervision should be given to the central bank or an agency, but whether it should be given to the central bank and an agency.
A captured supervisor can be detrimental for financial stability, as it incentivises banks to take over more risks for two complementary reasons. First, a captured supervisor is more lenient in evaluating banks’ health. Second, banks may expect the captured supervisor to rescue them in case of distress regardless of their solvency, increasing banks’ moral hazard. We therefore expect non-performing loans (NPLs), which are a proxy for the health of the banking sector, to be higher when supervision is not shared. Descriptive evidence suggests that this might be the case. Chart 1 shows that NPLs are lower when supervision is shared rather than when it is in the hands of a single supervisor. Panel data regressions support this hypothesis: across the three governance models, the coefficient of shared supervision is the only one negative and significant, indicating that NPLs are lower under this arrangement.
Chart 1: Median NPLs under non-shared supervision (i.e. supervision by the central bank or an agency only) and shared supervision
There is however a second element that the theory on governance and supervisory capture highlights. Supervisory governance is not effective per se, but depending on the context to which it is applied. In other words, shared supervision should be associated with better financial stability outcomes only in those countries where there is a risk of supervisory capture. This means that in those countries where corruption is lower, shared supervision may have no impact on financial stability. To measure a country’s risk of capture, I use the inverse of the ‘control of corruption’ indicator from the Worldwide Governance Indicators database. This variable is an ideal indicator of the risk of capture of the supervisor since it measures ‘perceptions of the extent to which public power is exercised for private gain, including both petty and grand forms of corruption, as well as ‘capture’ of the state by elites and private interests’ (World Bank).
Chart 2 summarises this idea, showing that NPLs increase with the risk of capture in countries where supervision is not shared (left-panel), whereas this relationship is weaker where supervision is shared (right-panel).
Chart 2: The relationship between NPLs and shared supervision under different degrees of risk of supervisory capture
Note: The chart plots log NPLs (y-axis) against risk of capture (x-axis) in countries with non-shared (left-panel) and shared (right-panel) supervision. Each observation (dot) represents a combination of country and year in the sample. The orange lines show fitted values. The fitting lines show that, as the risk of capture increases, NPLs are higher in countries where supervision is not shared (positively sloped fitting line); on the other hand, when supervision is shared, the risk of capture does not affect the share of NPLs (horizontal fitting line). Risk of capture is the inverse of the variable ‘Control of Corruption’ of the Worldwide Governance Indicators database.
Chart 2 also shows that when the risk of capture is low NPLs (in logarithmic form) tend to be lower under non-shared supervision. This does not necessarily mean that shared supervision is undesirable when corruption is low, as the low degree of NPLs may be driven by a number of different factors which are not captured by the chart.
The results of panel data regressions provide evidence in support of this hypothesis. The interaction between shared supervision and risk of capture is negatively and significantly correlated with NPLs, even after controlling for a number of macroeconomic and financial factors as well as country and year-specific fixed effects.
It is important to stress that these preliminary results are based on regressions, and therefore do not identify the causal impact of governance on NPLs, but just a correlation. In this context, causality is difficult to establish as the implementation of a model of supervision may depend on a number of factors, including the conditions of the banking sector. For this reason, there could be a potential endogeneity bias in the regression estimates due to reversed causality. In this regard, however, it is worth stressing that the within-country variability of supervisory governance is very low, as countries historically did not reform their supervisory architecture often. It is therefore less likely that shifts in NPLs provoke a change in supervisory governance, suggesting that this bias might be smaller than expected.
In conclusion, this paper suggests that supervisory governance could have an impact only depending on the institutional setting in which they are implemented. Institutional factors, such as the risk of capture in a country, are able to influence the effectiveness of supervisory governance in keeping the banking system stable. If policymakers want to address reforms in the governance of banking supervision, they should be aware that the success of their efforts will be conditional on the existing political economy setting in which the reform is undertaken.
Nicolò Fraccaroli commenced this post when working in the Bank’s Macroprudential Strategy and Support Division.
If you want to get in touch, please email us at firstname.lastname@example.org or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.