Researchers have paid increasing attention to bank supervision, especially following the 2008 financial crisis (see Masciandaro and Quyntin 2013, and Mishkin 2001 for earlier work). Much of this new research has focused on supervisors’ incentives (Agarwal et al. 2014, Lucca et al. 2014, Carletti et al. 2015), but a number of other questions have received much less attention. Among these are:
- What does it take to supervise banks?
- What is the content of supervision?
- What does increased supervision do to bank outcomes?
In a new paper, we provide a theoretical framework for thinking about supervision and its relationship to regulation, and use hours data of Federal Reserve examiners to describe how supervisory efforts vary by bank size and risk, and to measure key trade-offs in allocating resources (Eisenbach et al. 2016).
Diverging objectives of banks and society help rationalise the need for banking policies. First, banks have limited liability with deposit insurance and so they don’t fully account for the risks they impose on depositors and other creditors who fund them. Second, banks can impose costs – externalities – on the financial system and the economy when they fail. Because banks don’t take these external costs into account, they engage in riskier activities than are socially optimal. Regulation and supervision can help realign banks’ risk-taking with the objectives of society as a whole.
The difference between regulation and supervision
We distinguish regulation from supervision by the type of information the two use about a bank – ‘hard’ or ‘soft.’ Regulators can only use hard information such as a bank’s business lines or the adequacy of its capital or liquidity. Supervisors, on the other hand, can also make use of softer information, like the quality and centrality of a bank’s risk management, and whether it appears to be side-lined in important decisions.
The distinction between soft and hard information is informed by results in Goldsmith-Pinkham et al. (2016). They provide a ‘look behind the curtain’ using computational linguistics techniques to parse the content of thousands of supervisory messages, or actions, sent by supervisors to banks, and then examine how the frequency of topics varies with key characteristics of the bank holding companies (BHCs), such as size and risk. Using a number of methods, they find that supervisory actions are a combination of soft and hard information; and importantly, supervision involves much more than mere compliance with regulations.
The reliance of supervision on soft information means that judgement must often inform supervisors’ decisions. This element is captured by allowing for false negatives and false positives in supervisors’ information, meaning that supervisors can observe a good signal although a bank is in trouble or a bad signal although a bank is fine.
As highlighted by the timeline below, once the hard regulations are set, a bank chooses a low-risk or high-risk action – for example, whether or not to engage in diligent risk management – and then later the risk is realised in an outcome, such as loan repayments or defaults. While regulations are set at the beginning, supervisors are active throughout, first monitoring to produce information (a signal) about the bank’s action, then intervening based on the signal.
In our setting, supervisor’s monitoring improves the reliability of the signal that supervisors act on, and curbs banks’ incentives to take excessive risk in the first place. The level of monitoring is optimally chosen taking both of those effects into account. Intervention can be flexibly chosen after the signal is observed. This flexibility is a big benefit of supervision compared with regulation, since it allows for an optimal response to the particular situation. However, commitment to harsher-than-optimal intervention ex post would have benefits on banks’ incentives ex ante.
In the model, supervisory monitoring and intervention lowers risk-taking. These assumptions are consistent with results of Hirtle et al. (2016), who study whether the intensity of supervision across BHCs affects the risk and performance of these firms. Using an assumption about asset-size rank within Federal Reserve districts to identify banks more likely to receive additional supervisory attention, they find that more supervisory attention is associated with lower risk, but not disproportionately lower returns or slower asset growth.
Supervision is costly and, even if more effort is beneficial in our framework, supervisory resources are limited. Supervisors face an economic trade-off and have to allocate their resources to where they are most valuable. To better understand this trade-off in practice, we analyse data of hours worked by Federal Reserve supervisors and their ratings of BHCs, combined with balance sheet information from regulatory filings.
In terms of how supervisory attention (hours measured in log terms) relates to bank size (assets measured in log terms), we intuitively expect supervisory attention to increase. But the extent to which this is true is not known – that is, comparing two banks, one twice the size of the other but otherwise similar, does the larger bank receive twice the attention, more than twice, or less than twice? Our regression estimates imply that if a bank is twice the size, on average and all else equal, it receives 62% more hours. So, supervisory attention increases less than proportionately with size.
Our model suggests two competing effects. On the one hand, a larger bank may warrant a disproportionately higher level of scrutiny because its failure would cause disproportionately more damage to the financial system and the economy. On the other hand, there may be economies of scale in supervision similar to the way a bank’s own headcount doesn’t necessarily double if it has twice the assets. Our 62% estimate is consistent with scale economies in supervision or a finding that larger banks actually do receive disproportionately more scrutiny, but the effect is dominated by the scale economies.
We next study how supervisory attention varies as a function of a bank’s riskiness as measured by its supervisory rating (on a scale of 1 to 5, with 1 being safest and 3 or higher indicating moderate to significant concerns). If supervisors are assigned to lower bank risk, as in our model, we would expect to see a positive relationship between supervisory hours and ratings, and this mechanism is confirmed in the data. For example, a 3-rated bank receives 66% more hours than a safer 1-rated bank. To put this in perspective, supervisory attention increases about the same amount when a bank’s rating goes from 1 to 3 as when its assets double.
We also study how supervisory resources are reallocated across banks, first by comparing supervision before and after the 2008 financial crisis. The post-2008 period has been characterised by the emergence of a number of enhanced supervisory frameworks for large banks. We estimate that large banks ($10 billion or more in assets) receive 65% more attention since 2008, while small banks receive 19% less – evidence of resource scarcity. We similarly find reallocation of supervisory resources within Federal Reserve districts when banks are in distress. In particular, resources are allocated away from small banks while the attention to large banks is unaffected.
In conclusion, the data provide support to the basic economic trade-offs – supervisory resources are allocated based both on the size and risk profile of an institution, but also on the overall resource constraints facing supervisors.
Figure 1. Limits and trade-offs in supervision
Disclaimer: The views expressed in this column are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System. Any errors or omissions are the responsibility of the authors.
Agarwal, S, D Lucca, A Seru and F Trebbi (2014) “Inconsistent regulators: Evidence from banking”, Quarterly Journal of Economics 129, 889-938.
Carletti, E, G Dell’Ariccia and R Marquez (2015) “Supervisory incentives in a banking union”, IMF unpublished manuscript.
Eisenbach, T M, D O Lucca and R M Townsend (2016) “The economics of bank supervision”, NBER Working Paper no w22201.
Goldsmith-Pinkham, P, B Hirtle and D O Lucca (2016) "Parsing the content of bank supervision".
Hirtle, B, A Kovner and M C Plosser (2016) "The impact of supervision on bank performance" .
Lucca, D, A Seru and F Trebbi (2014) "The revolving door and worker flows in banking regulation", Journal of Monetary Economics 65, 17-32.
Masciandaro, D and M Quintyn (2013) "The evolution of financial supervision: The continuing search for the Holy Grail", SUERF 50th Anniversary Volume Chapters: 263-318.