Interest Rate Risk – What Regulators Missed

I discussed the Federal Reserve’s supervision of liquidity risk at Silicon Valley Bank in an earlier post.  Examiners at least identified liquidity weaknesses at SVB, but supervisors were slow to translate findings into action or to even adjust ratings.  Supervision of interest rate risk was a more complete miss.

Interest rate risk (IRR) represents a bank’s exposure to changes in market interest rates.  There are two major forms of IRR:  earnings-at-risk (EAR) and economic value of equity (EVE).  EAR measures the sensitivity of net interest income to changes in market interest rates over a specified time horizon, typically one year.  EVE is the sensitivity of economic value (net present value of assets minus present value liabilities) to rate changes.  EVE presents a point-in-time snapshot of the balance sheet and is more indicative of the bank’s structural IRR.

A Supervisory Ratings Letter dated August 17, 2022, kept the Sensitivity to Market Risk rating at “2” or Satisfactory.  The report indicated that sensitivity “is adequately controlled with moderate potential that earnings performance or capital will be adversely affected.”  The report added that “management has established a moderately asset sensitive balance sheet, and the bank has benefited from the recent rate increases.”  Furthermore, “interest rate hedges implemented in 2021 have effectively mitigated the bank’s exposure to rising rates.” (Emphasis mine.)

Yikes.  SVB’s long-term MBS and CMO portfolio had increased nearly fivefold, from $14.4 billion in December 2019 to $84.5 billion in December 2021.  Most of the securities were held-to-maturity, providing SVB with little flexibility to adjust to rising rates.  By mid-2022, the consequences of the disastrous investment strategy were already becoming apparent.  The investment portfolio fell to $8.3 billion underwater by March 2022, falling further under to $14.5 billion by June.  The same letter’s Capital discussion also ignored the unrealized loss.

SVB had initiated interest rate swaps, which peaked at a face value of $12.8 billion in June 2021.  Swaps fell to $5.9 billion by March 2022 and were off the books by September.  As shown below, the bank’s EVE remained outside internal limits during most of this period.  EVE briefly moved within limits in April 2022, but swaps were already starting to wind down.  The respite appears to be the result of a deposit study and resulting change in assumptions.  Regulators should view attempts to solve problems though modeling or assumption changes as red flags. 

The Federal Reserve got around to issuing an MRA related to IRR in the November 15, 2022, CAMELS Examination Supervisory Letter.  However, the MRA focused on EAR rather than the more significant, if not existential threat to EVE.  The MRA identified some important weaknesses in risk management but glossed over the more important issue.  As a colleague of mine once remarked in another context, it was like deciding what sunglasses to wear at ground zero of a nuclear explosion.  A presentation to the Federal Reserve Board on February 14, 2023, gave a clearer presentation of SVB’s problems.  That presentation came just 24 days before SVB’s closure and by then it was too late.

The Federal Reserve’s supervision of IRR at SVB is hard to defend but somewhat easier to understand.  Interest rate risk is Banking 101, but it’s also a lot more abstract than credit risk or liquidity risk.  That is especially the case with EVE, which not just bankers but more than a few regulators view dismissively.  EAR is more relevant to day-to-day asset-liability management.  It’s also easier to compare EAR forecasts to actual results through back-testing.  However, EAR is also an incomplete measure.  It doesn’t matter for EAR purposes whether a portfolio has an average life of 13 months or 13 years.  EVE presents a better indication of exposure to tail events and potential costs to the FDIC.

From the late 1980s to a couple of months ago, credit risk caused most bank failures.  Bank runs or other liquidity events might push a bank over the edge.  The overall downward trend in interest rates for most of the past 35 years meant that IRR caused few failures.  The market experienced some rate volatility during that time.  However, exposed banks could usually ride out the storm.  The early 1980s and the early 2020s both experienced sharp and sustained rate movements.  Deposits have more recently become a less reliable source of funding.  That’s not just the case with SVB, where outflows were sudden and dramatic.  It was a slower bleed at First Republic, where liquidity facilities allowed the bank to replace its lost deposits.  But First Republic’s low yielding (2.89%) mortgage portfolio could not cover rising wholesale funding costs.

There are some specific regulatory measures of liquidity risk, such as the Liquidity Coverage Ratio.  That is not the case with IRR.  No regulation deals with IRR specifically except as part of broader safety and soundness requirements.  An interagency advisory identifies sound IRR practices.  However, that is guidance, not regulation.  There is a longstanding policy that examiners do not cite “violations” of guidance.  The five bank regulatory agencies reiterated this policy in 2018.  Regulators further noted their intent “to limit the use of numerical thresholds or other ‘bright lines’ in describing expectations in supervisory guidance.”  In other words, keep the guidance as squishy as possible. 

The Fed went a step further in 2021 issuing the “guidance on guidance” as a Final Rule.  At least the Fed did not go as far as the request by the American Bankers Association and Bank Policy Institute to limit MRAs, MOUs, and exam rating downgrades to instances where there is a violation of statute, regulation, or order.  The ABA/BPI approach would have made proactive supervision virtually impossible.

The Supervisory Ratings Letter still looks terrible, even with this additional context.  The November 2022 MRA looks a bit better, however.  MRAs in the IRR area usually emphasize process deficiencies rather than outsized exposures.  Although the MRA focused on the less important risk measure (EAR), at least some of the corrective actions could also make EVE assessments more reliable.  For example, the MRA’s requirement to analyze deposit mix shifts would likely affect EVE as wells are EAR.  Focusing on the MBS exposure would have been more relevant and addressing that exposure in 2021 would have been timelier.  Still, the MRA at least picked off some of the low hanging fruit.