Follow Topic Follow Contributor Share Feedback
<p><font color="#14ABB2">Probative Inquiry into the <br>2023 Financial Sector Stress Events</font></p><p><font size ="6"><font color="#455664">Executive Summary</font></font></p>

Probative Inquiry into the
2023 Financial Sector Stress Events

Executive Summary

by Starling Insights

Starling Insights Editorial Board

Sep 21, 2026

Feature

A Series of Unfortunate Events…

The evidence assembled and reviewed by Starling concerning lapses in the supervision of and the ultimate failure of SVB does not fully sustain the now established public narrative regarding the relevant circumstances leading up to, during, and following from the Financial Sector Stress Events of 2023. 

The Barr Report of April 2023 does not afford an adequately substantiated causal account of those events. Nor should it be expected to have done so. Assembled in a remarkably short period of time, while events were still fresh, the Barr Report was compiled almost exclusively on the basis of records then in the possession of the Federal Reserve. 

It is an impressive study that draws several conclusions that now appear beyond doubt: SVB assumed consequential financial risks; its risk governance capabilities struggled to keep pace with its growth; supervisors failed to ensure that its most consequential vulnerabilities were addressed in a timely manner; and the bank experienced a run that was extraordinary in its speed and scale. 

But the Barr Report advanced a causal narrative that has been interpreted as more definitive than the evidence assembled at the time permits. That established public narrative now encounters material conflicting and qualifying evidence curated through Starling’s ongoing Probative Inquiry.

Starling finds the central supervisory failure was not due to any absence of adequate information or recognition of the growing vulnerabilities to which SVB was exposed. Sufficient information existed across frontline Reserve Bank examination teams, Board supervisory portfolio leads, specialist Risk and Surveillance personnel, Legal and Policy advisers, and senior-most Board leadership. 

Nevertheless, due to long-standing uncertainties regarding where responsibility, authority, and accountability reside within the Federal Reserve System, supervisors lacked clarity regarding decision-rights and effective institutional mechanisms by which to determine whose assessment of the risks evident at SVB should govern the supervisory response. 

SVB struggled to meet supervisory demands because supervisors struggled to articulate them with sufficient clarity and timeliness.

This is but one factor among several that must be considered before a fully causal account can be asserted. As such, this Preliminary Report does not seek to replace the established public narrative with a new monocausal explanation. Rather, it aims to identify where that narrative is supported, where it overstates the available evidence, where subsequent learnings materially complicate that narrative, and where they contradict it fully. 

Our Inquiry also raises a question of continuing consequence: whether the organizational and supervisory weaknesses exposed by SVB’s failure have since been remedied, or whether systemic fragilities persist within the Federal Reserve’s practice of supervision. As a preliminary conclusion, the latter appears highly likely.

Detailed Discussion

The key findings of this Preliminary Report, based on the evidence included in the Appendix that accompanies it, are the following:

1.SVB failed in 2023 after suffering unrealized losses on its bond portfolio that wiped out its capital on a fair-value basis, more than 90% of its deposits were uninsured, meaning they were prone to run, and it had insufficient contingent-liquidity readiness.

This finding is based exclusively on public documents referenced in the footnotes hereto and supported by the findings reported by Charles River Associates (“CRA”) in the course of related analytics commissioned by Starling and reported here at Addendum Item 5.

SVB tripled in size from 2019 until 2021 largely because its depositor base of venture-capital-backed companies raised enormous amounts of capital during the Covid-19 pandemic and deposited most of the proceeds from those capital raises in SVB.1 Instead of using those deposits to fund loans, SVB chose to invest most of them in U.S. government securities and mortgage-backed securities guaranteed by Fannie Mae and Freddie Mac.2 This resulted in a securities portfolio that accounted for between 50% and 60% of its total assets from mid-2021 until December 31st 2022.3

SVB also gave depositors the option to have their deposits swept into third-party money market funds. Approximately $168 billion had been swept into money market funds at the end of 2022, showing that its depositors were ready and able to move funds away from SVB should they lose confidence in it.4 Finally, SVB did not have sufficient emergency liquidity facilities lined up in advance and appears to have been operationally unready to borrow large amounts of cash from the Federal Home Loan Banks (“FHLBs”) or the Federal Reserve’s Discount Window (“Discount Window”).5

SVB classified nearly 80% of its bonds as held-to-maturity (“HTM”) instead of available-for-sale (“AFS”).6 SVB’s HTM bonds had an average weighted duration that rose from 4.6 at June 30th 2021 to 6.2 at year-end 2022.7 Duration is a financial concept that measures the interest-rate sensitivity of interest-bearing debt securities. A bond portfolio with a duration of 6.2 means that for every 1% increase in interest rates the market value of the bond portfolio will decline by about 6.2%. Conversely, for every 1% decrease in interest rates, the market value of the bond portfolio will increase by about 6.2%.8

SVB hedged the interest-rate risk on its AFS securities until it allowed those hedges to lapse by the end of 2022,9 resulting in a substantial increase in the duration of its AFS securities.10 A hedge on a bond portfolio consists of a financial instrument whose value goes up when the value of a bond portfolio goes down. SVB did not hedge the interest-rate risk on its HTM bonds because hedging HTM bonds is generally considered to be inconsistent with their classification as HTM.11

SVB’s accountants may have required it to reclassify its HTM bonds as AFS bonds had SVB hedged its HTM securities. This would have made its reported income more volatile because generally accepted accounting principles (“GAAP”) require banks to reflect gains and losses on AFS bonds in their Other Comprehensive Income (“OCI”) during the period when they are incurred.12 Companies are not required to reflect gains and losses on HTM bonds in their OCI.13 Instead, they are reported as unrealized gains and losses.14

At year-end 2022, over 90% of SVB’s deposits were uninsured.15 Uninsured deposits are generally considered to be more prone to running at the first sign of trouble than insured deposits because they are not protected against losses unless the Systemic Risk Exception (“SRE”) to the Least-Cost Test is invoked. The Least-Cost Test requires the Federal Deposit Insurance Corporation (“FDIC”) to resolve an insolvent bank using the method that is least costly to the Deposit Insurance Fund (“DIF”).16 The Systemic Risk Exception to that test allows the FDIC to protect uninsured depositors against losses.17

During Q1 2022, SVB suffered unrealized losses on its securities portfolio of about $8.3 billion because of a rise in interest rates.18 That amounted to more than half its reported Common Equity Tier 1 (“CET1”) capital of $15.1 billion.19 By the end of Q3 2022, SVB had suffered accumulated unrealized losses on its securities portfolio of $18.7 billion because of a further rise in interest rates.20 That amounted to more than its reported CET1 capital of $17.0 billion,21 meaning it was insolvent on a fair-value basis.

The reason for these losses is that between year-end 2021 and March 10th 2023, the Federal Reserve raised the federal funds rate from near zero to over 4.5%.22 As noted above, for every 1% increase in interest rates, a bond portfolio with an average weighted duration of 6.2 would suffer losses equal to about 6.2% of the bond portfolio. Those losses would be recognized on AFS bonds and reflected in a firm’s reported OCI. This was not the case for SVB because it had opted out of reflecting Accumulated Other Comprehensive Income (“AOCI”) in its capital.23 Losses would be unrealized on HTM bonds and AFS losses would not be reflected in a firm’s reported income statement or capital.

Despite the fact that SVB was insolvent on a fair-value basis after Q3 2022, SVB’s depositors did not run until March 9th 2023. The run was unprecedented in its size and speed. SVB’s depositors withdrew $41 billion24 on March 9th and reportedly submitted orders to withdraw another $100 billion25 on March 10th. That amounted to 25% and 60% of its total deposits of $166 billion.26 The withdrawal orders for March 10th were not completed. Instead, SVB’s chartering authority — the California Department of Financial Protection and Innovation (“DFPI”) — revoked SVB’s charter, closed the bank, and appointed the FDIC as its receiver on Friday morning, March 10th.27

2.Fed supervisory staff knew, or should have known, by June 2022 at the latest or as early as March that SVB was highly vulnerable to massive losses, a run, and failure because of the size and duration of its bond portfolio, the likelihood that its uninsured depositors would run if they lost confidence in the bank, and its insufficient readiness to obtain emergency liquidity when needed.
  1.  

This finding is based primarily on SVB’s public call reports during 2022, the CRA analysis attached as Addendum Item 5 to this Preliminary Report, the OIG Report, and the summaries of the testimonies of [Board Staff 3] and [Board Staff 15] contained in the Appendix to this Preliminary Report.

SVB’s call reports during 2022 publicly disclosed that (1) its bond portfolio accounted for more than half its total assets during 2022, (2) it was either substantially undercapitalized or insolvent on a fair-value basis throughout 2022, and (3) approximately 90% of its deposits were uninsured throughout 2022 and thus prone to run.28

For example, its call reports for the first and second quarters of 2022 publicly disclosed that it had unrealized losses on its bond portfolio equal to more than half its reported CET1 capital at the end of those quarters because of rising interest rates.29 Its call reports for the third quarter and year-end 2022 publicly disclosed that it had unrealized losses on its bond portfolio that exceeded its reported CET1 capital at the end of those periods because of rising interest rates.30 That means it was substantially undercapitalized on a fair-value basis at the end of the first half of 2022 and insolvent on a fair-value basis at the end of the third quarter and year-end 2022.

In addition, SVB was insufficiently ready to borrow emergency liquidity from the FHLBs or the Discount Window.31 According to the researchers at CRA, the Federal Reserve supervisory staff knew about the “[m]aterial weaknesses in SVB’s contingent-liquidity readiness” because SVB “had repeatedly failed its own liquidity stress tests, had not completed planned expansion of repo capacity, had not conducted test transactions and had limited collateral pre-positioned at the discount window.”32

It is notable that the Barr Report did not specifically mention SVB’s call reports or identify that they should have put the supervisory staff on notice that SVB was substantially undercapitalized on a fair-value basis or highly vulnerable to a run by uninsured depositors as early as March 2022.33 But SVB’s call reports and insufficient contingent-liquidity readiness were known to, or observable by, Board S&R and the FRBSF supervisory staff and, thus, they either knew, or should have known, as early as March 2022 that SVB was highly vulnerable to massive losses, a run, and potential failure.

The OIG Report discussed a relevant S&R internal report issued in June 2022:34

In June 2022, the Board’s surveillance team issued a special topic report discussing the high level of unrealized losses on the investment securities at many banks as a result of interest rate increases. The surveillance team identified SVBFG as one of the institutions with the highest levels of unrealized losses and with a large portion of investments in its HTM portfolio.35 The report cautioned that those institutions with a high level of unrealized losses and a large share of HTM investment securities may need to realize sizable losses when they sell investment securities in their HTM portfolios in case of liquidity needs, such as a large deposit outflow. As of June 2022, the Board’s surveillance team also placed SVBFG on the Systemwide holding company watch list with a high adverse change probability warning, indicating the likelihood of a downgrade to its supervisory rating in the near future.

. . . .

After identifying SVB as one of the institutions most at risk, the Board surveillance team continued to warn of the effect of rising rates on banking conditions through surveillance reports available to System examiners and presentations and communications to supervision officials and staff.36

The Barr Report included a reference to the June Fed Surveillance report, but only in an appendix to the main report entitled “Additional Topics”:

The theme of higher rates was the focus of a special report on risks associated with unrealized losses on investment securities in June 2022. SVBFG was included in a list of banks with the highest ratios of unrealized losses relative to common equity tier 1 (CET1) capital and was larger than any bank ranked higher.37

This special topic report and watch list were provided to Board S&R and FRBSF supervisory staff in June 2022. They were also available to the supervisory staff between June 2022 and March 2023. Thus, the supervisory staff knew, or should have known, as early as June 2022 that SVB was highly vulnerable to massive losses and failure based on interest-rate risk and liquidity risk.

The Barr Report also stated that Board Risk and Surveillance staff warned the VCS and the full Board that SVB was highly vulnerable to losses based on interest-rate risk and liquidity risk in a presentation to the Board on February 14th 2023.38 Yet despite the availability of SVB’s public call reports during 2022, the June 2022 special topic report, and his participation in the meetings of the Federal Open Market Committee (“FOMC”) that raised interest rates aggressively after July 2022, then-VCS Barr testified at a hearing before the Senate Banking Committee in March 2023 that he did not learn “about the issues at Silicon Valley Bank with respect to interest rate risk” until the full Board meeting in mid-February 2023.39

In an interview with Starling, [Board Staff 15]40 argued that SVB’s failure resulted from a “toxic combination of many things,” beginning with its extraordinary growth. SVB doubled in size during 2020 and 2021. Historically, rapidly growing banks fail at a substantially higher rate than slower-growing institutions: “We’ve always known that fast growth was an indicia of potential future distress. But that bank grew super-fast.”

To this were added a remarkably high proportion of uninsured deposits, a depositor base concentrated in a small number of industries, and a very large bond portfolio carrying substantial duration risk and, from early 2022 until early 2023, significant unrealized losses. None of these facts was obscure. “All of this information was public,” [Board Staff 15] emphasized. “Every fact I’ve just mentioned — the rapid growth, the 90 plus percent uninsured deposits, the concentrated depositor base, the giant securities portfolio, the huge duration risk and mark-to-market losses in the book. Everybody knew that, and almost nobody seemed to be worried about it.”

The problem, in [Board Staff 15]’s telling, was not that supervisors were unaware of the underlying facts. Nor, in [Board Staff 15]’s view, was the problem principally a lack of information. Supervisors possessed information unavailable to the public, but [Board Staff 15] did not consider that additional information essential to recognizing the danger; public disclosures were sufficient to reveal the enormous duration risk, the potential for rapidly rising rates to produce severe unrealized losses, and the instability created by a deposit base that was overwhelmingly uninsured, [Board Staff 15] insisted. 

“You simply needed to be able to connect all those dots and see that this concatenation of facts could result in a really, really rapid sudden demise of that firm,” [Board Staff 15] said. “Silicon Valley Bank, everybody knew, ex ante, there was a lot of interest-rate risk going on there,” [Board Staff 15] explained, “but no one knew if the amount of interest-rate risk that was going on there was exceeding some risk tolerance threshold, because I don’t think we really had adequately set one.”

[Board Staff 3] reinforced the views shared with Starling by [Board Staff 15]. In the 2022 timeframe, [Board Staff 3] explained, the Board’s Risk function (“Risk”) sat within Board S&R but was intended to look horizontally across the Federal Reserve’s supervisory portfolios, to identify emerging threats, compare their significance across institutions, and bring concerns to the attention of the portfolios responsible for individual banks. 

[Board Staff 3] recalled that the Board’s Surveillance staff, which was headed by the same Board officer who headed the Board’s Risk function, had identified interest-rate risk as an emerging trend by the first quarter of 2021. As rates began to rise in 2022, Risk became increasingly concerned about the combination of unrealized securities losses and the liquidity pressures that could arise if banks were required to convert those securities into cash. 

[Board Staff 3] explained that, after the first quarter of 2022 until SVB failed in March 2023, Risk and Surveillance staff warned the Director of the LFBO portfolio and the head of the FRBSF supervisory team that there was a high-risk possibility that SVB could suffer losses because of its extraordinary exposure to interest-rate risk and liquidity risk. The Risk and Surveillance staff also warned the Deputy Director for Supervision. The SVB examination team at the FRBSF was principally responsible for supervising SVB during the period leading up to its failure, subject to the oversight of the LFBO Director, the Deputy Director for Supervision, the Board S&R Director and the VCS.41

“Looking back at examination reports, examiners just didn’t see the risk early enough,” [Board Staff 3] said. “I think it had been too long. I think people had just fell [sic] asleep on interest-rate risk, and people just didn’t remember the ’80s and ’90s.” 

[Board Staff 3]’s account resists the proposition that supervisors were unaware of the danger. “I think some people saw it,” [Board Staff 3] said. Reserve Bank personnel had published analyses concerning long-duration securities holdings, and Risk and Surveillance had begun calling attention to unrealized losses and the liquidity consequences that could follow. But, [Board Staff 3] observed, that analysis “didn’t seem to translate into a supervisory response.”

“I think people saw the risk,” [Board Staff 3] explained. “They just assigned a different probability of something bad happening.” In [Board Staff 3]’s characterization, Risk believed that the magnitude of the potential losses meant that “the probability didn’t matter.” Even a scenario regarded as unlikely warranted supervisory attention where its consequences could be severe.

[Board Staff 3] remembered this disagreement carrying into discussions with the leadership of the large-bank supervisory portfolio. SVB was not the sole bank that faced extraordinary interest-rate risk and liquidity risk, but SVB nevertheless became a specific focus of Risk’s concern. [Board Staff 3] recalled that the Risk staff had urged the Director of the LFBO portfolio and the head of SVB’s supervisory team at the FRBSF to downgrade SVB’s liquidity rating months before the bank failed.

Risk and Surveillance staff also warned the Board of Governors about SVB’s vulnerability to losses in a presentation to the Board in mid-February 2023, yet that warning did not produce an intervention by the Board S&R staff.

When news of SVB’s intended capital-raising transaction in March 2023 reached the Director of the Risk and Surveillance staff, according to [Board Staff 3]’s recollection, the Director exclaimed, “That bank’s going to fail. Oh my God, that bank’s going to fail.”

 

3.Board S&R and FRBSF supervisory staff failed to require SVB to reduce the duration of its bond portfolio, to restructure its deposits to be less runnable, or to improve its readiness to obtain emergency liquidity if needed, even though they knew or should have known about SVB’s high vulnerability to massive losses, a run, and potential failure.

 

This finding is based in large part on a written statement provided to Starling by the former Independent Directors of SVB (“Directors’ Statement”),42 an interview conducted with former SVB independent director Mary Miller,43 the analysis by CRA commissioned by Starling,44 OIG report findings,45 and interview comment offered to Starling by [Board Staff 15].46

In the Directors’ Statement, the Independent Directors of SVB indicate that Board S&R and FRBSF supervisory staff never directed that SVB reduce the duration of its bond portfolio or its exposure to uninsured depositors. 

Indeed, following on their 2021 examination of SVB, Fed supervisory staff downgraded SVB’s composite CAMELS rating as well as its Management component rating from 2 (Satisfactory) to 3 (Less than Satisfactory). Notably, they did not downgrade the Liquidity or Sensitivity to market (i.e., interest-rate) risk components, suggesting that they were satisfied with SVB’s exposure to interest-rate risk and liquidity risk and the manner in which SVB sought to manage those risks. Nor did the supervisory staff require SVB to reduce the duration of its bond portfolio or its exposure to uninsured depositors. Instead, it issued three MRAs related to the process by which SVB managed its liquidity risk. 

In an interview with Starling, Mary Miller, an independent director at SVB and chair of the board’s audit committee, explained that Board S&R and FRBSF supervisory staff imposed 54 MRAs or MRIAs on SVB between 2020 and 2022. None of these MRAs or MRIAs directed SVB to reduce the duration of its bond portfolio or its exposure to uninsured depositors. Instead, most of them related to technology and operational matters, including BSA/AML, cyber risk, and identity-access management. Remediation consumed substantial board attention and SVB devoted considerable resources to this effort.

We note that the OIG Report is largely consistent with this assertion:

The [Federal Reserve’s] March 2019 examination report [of SVB] highlighted liquidity risk management concerns and noted that the bank’s funding structure was primarily concentrated in private equity and venture capital firms whose deposits experience larger-volume, irregular cash flows. The report notes that management had not formally addressed the risks associated with these funding concentrations and that SVB’s liquidity risk management should be consistent with the bank’s risk appetite, complexity, risk profile, and scope of operations. As a result, examiners issued SVB a Matter Requiring Attention (MRA) related to liquidity risk management. Despite noting these concerns, FRB San Francisco examiners assigned SVB a CAMELS liquidity component rating of 1 because SVB had ample sources of funds on favorable terms to meet liquidity needs. In our opinion, examiners did not balance the lack of appropriate liquidity risk management with the bank’s financial results. We believe examiners missed an opportunity to downgrade the liquidity component.47

. . . 

In August 2022, FRB San Francisco examiners . . . issued CAMELS ratings for SVB. Examiners downgraded SVB’s composite rating and management component rating to 3 and the liquidity component [from 1] to 2 . . . Despite the surveillance team’s reports, examiners commented in the supervisory letter that SVB’s balance sheet structure mitigated some of the risks associated with the recent rapid growth by having 64 percent of the bank’s total assets in cash and fixed income securities, of which 93 percent were invested in U.S. Treasuries and government agency-issued securities and did not consider the changing market environment and its potential effect on the bank’s financial condition.48

. . .

After identifying SVB as one of the institutions most at risk, the Board surveillance team continued to warn of the effect of rising rates on banking conditions through surveillance reports available to System examiners and presentations and communications to supervision officials and staff. However, the SVB DST did not shift its attention to the rising interest rates and SVB’s unrealized losses on investment securities and did not sufficiently assess the extent of the potential effect to the bank’s capital and liquidity. In our opinion, examiners should have heeded the warning signs and taken immediate actions to closely scrutinize the risks of rising interest rates on SVB’s investment securities portfolio and the sufficiency of the bank’s liquidity.49

According to both the Barr Report and the analysis by CRA commissioned by Starling, SVB had repeatedly failed its liquidity stress tests and was not ready to use the Federal Reserve’s standing repo facility or to borrow substantial amounts of emergency liquidity from the FHLBs or the Discount Window. And yet, the Board S&R and FRBSF supervisory staff did not take decisive action to require SVB to improve its contingent-liquidity readiness.50

This is consistent with the views expressed during our interview with [Board Staff 15], who noted that even the exam findings at SVB that had focused on interest-rate and liquidity risk were nevertheless not aimed at addressing those risks in a manner calculated to reduce them. Supervisors were not, he said, directing SVB to hedge substantially more of the duration risk in its securities portfolio or to reduce the portfolio’s duration. Instead, “It was a lot of process and procedure and governance and controls criticism that may or may not have helped.”

4.The failure of Board S&R and FRBSF supervisors to require SVB to reduce its interest-rate risk or liquidity risk did not result from EGRRCPA or a less assertive supervisory culture promoted by related supervisory policy priorities.

This finding is based primarily on Starling’s interviews with [Board Staff 2] and [Board Staff 14], summaries of which are contained in the Appendix to this Preliminary Report.51

The Barr Report included four principal findings,52 the fourth being that the Economic Growth, Regulatory Relief, and Consumer Protection Act (“EGRRCPA”), and related policy priorities implemented during the tenure of former VCS Quarles, promoted a less assertive supervisory culture.53 According to [Board Staff 14], the four conclusions that had been presented in the Barr Report were to have been read in a descending order of significance. The fourth finding, concerning EGRRCPA, tailoring, and related shifts in supervisory policy, was the least important of the findings.

Moreover, [Board Staff 14] described the fourth conclusion as having been “poorly understood.” The Barr Report, he explained, “did not say there was a bad law” [EGRRCPA]. Its criticism, rather, was implementation: “the way it was implemented,” he believes, created “step functions in how supervision was done” and resulting confusion “as firms move between portfolios.”

Further, [Board Staff 2] insisted, former VCS Quarles never instructed examiners to reduce the intensity of supervision. Rather, Quarles’s emphasis was on ensuring that supervisory actions were grounded in applicable law, regulation, and legitimate supervisory authority. Those policies may have changed things at the margins, [Board Staff 2] said, “but the core really stay[ed] the same.” It should be noted that [Board Staff 2] was the Barr Report writer tasked most specifically with an examination of supervisory lapses that may have contributed to SVB’s failure and/or S&R’s lack of alacrity in addressing the risk thereof.

[Board Staff 2] was emphatic about former VCS Quarles’s likely response had the immediate danger to SVB been presented to him clearly: “I’ve heard Quarles speak about the [Barr] Report and sort of the key takeaway for the tone of supervision stuff, that he never said, ‘Turn down supervision.’ And I agree, he did not,” [Board Staff 2] said. “And had he seen that risk and we recognized it and we told him, ‘Hey, there’s this big risk at SVB,’ he would have said, ‘Go do something about it.’ He may have picked up the phone and called them himself,” [Board Staff 2] argued. “That was my impression of the man.”

[Board Staff 2] allowed that there may have been a distinction between what Quarles had intended and what supervisory personnel may have believed his expectations to be. [Board Staff 2]’s decades of experience in bank examination led him to conclude that examiners sometimes “overreact” to or misinterpret the priorities announced by any Vice Chair for Supervision. “Honestly, I would’ve stopped at Key Takeaway 2” [of the Barr Report], [Board Staff 2] concluded. “The bank screwed up, we missed it, end of story,” he asserted.

[Board Staff 14] did not retreat from the conclusion that the stance of supervisory policy had shifted along with a change in VCS. “Was that part of the story? I think it was,” [Board Staff 14] said. “I think the evidence supports that it was.” But he immediately qualified the significance of that finding: “Was it the driving factor? No.” 

In [Board Staff 14]’s view, public and congressional attention was disproportionately focused on the Barr Report’s fourth conclusion. “It seemed misguided from the beginning that some thought the fourth conclusion was the most important thing,” he observed. Nor did [Board Staff 14] understand the fourth finding principally as an indictment of former VCS Quarles. His own explanation was more institutional. Different policymakers bring “different risk appetites” to the supervisory function, he said, and the resulting differences are “not a question of right or wrong”; they reflect choices about “how do I want to prioritize and make trade-offs?”

5.Board S&R has struggled with a culture of uncertainty regarding decision-making authority that dates back decades, and it is this culture that may explain most pointedly why Board S&R and the FRBSF supervisory staff failed to take more prompt and decisive action to require SVB to remediate the vulnerabilities that contributed to its failure.

This finding is based largely on the testimony of former Governor Daniel Tarullo, former VCS Quarles and [Board Staff 2], [Board Staff 3], [Board Staff 14], and [Board Staff 15]. 

If Board S&R and FRBSF supervisory staff knew of the vulnerabilities that put SVB at risk and yet failed to take decisive action to require SVB to reduce the size and duration of its bond portfolio, restructure its deposits to be less runnable, or require SVB to improve its readiness to obtain emergency liquidity, then we must ask — why not?

The Barr Report asserts that one principal reason was the enactment of EGRRCPA in 2018, which raised the asset thresholds for the applicability of Enhanced Prudential Standards for the large banks, and a less assertive supervisory culture that stemmed from a shift in consequent policy priorities advanced by former VCS Quarles.

An alternate view emerges through Starling’s interviews with both former VCS Quarles and former Governor Tarullo: excessive risk aversion, indecisiveness, and finger-pointing are common features of the supervisory culture in Board S&R that dates back decades, both contend. They further insist that no past oversight governor or VCS role-holder is responsible for creating this culture — and, notably, none has been successful in trying to change it. 

This culture is characterized by a fear of taking action that may later be viewed as mistaken and being held accountable for such. This is exacerbated by a cultural norm to which examiners “know” they are expected to conform: an examiner is not supposed to cross the line from supervising to “managing” a firm. Decisive supervisory actions that would have mandated the risk remedies that appear to have been necessary at SVB were considered by many of the Board S&R staff whom Starling interviewed as likely to have been viewed as crossing that line.

In the absence of 100% certainty that such an action would later be viewed as having been “right,” staff may have concluded that inaction was the better option. In this connection, former Governor Tarullo remarked that direction from the top is essential. When senior leadership’s policy position is unclear, he said, “the safest thing [for supervisory staff to do] is probably not to push it very much.”

Tarullo believes that this culture of indecision can have disastrous results. He pointed to the Board’s Surveillance function as having identified the mounting risks at SVB in June 2022. “What the heck happened?” he asked. “Nothing happened apparently, but why?” 

When SVB’s internal models produced unfavorable results concerning interest-rate risk and liquidity stress, management appears to have changed the assumptions used in the models. “A redder flag is hard to imagine,” former Governor Tarullo commented. And yet no decisive action was taken.

6.Emphasis on collegial decision-making through committees and a pervasive divorcing of responsibility, authority, and accountability across the Federal Reserve supervisory system impeded agreement on whether or how to take decisive remedial action at SVB.

Where finding number 5, immediately above, focuses on cultural predilections that promote hesitance to act, this finding highlights a lack of clarity around the structural ability to act at all. It is based largely on the testimony of various Board staff, especially [Board Staff 4], [Board Staff 13], and [Board Staff 10], as well as former Governor Daniel Tarullo, and former VCS Quarles.

Supervision is formally delegated to the Reserve Banks, while supervisory expectations, policy priorities, escalation pathways, legal review, and reputational accountability run up through the Board. This structure seems to have created a jurisdictional no-man’s land; Reserve Bank personnel might possess delegated authority but nevertheless seek alignment with the Board before exercising it, while Board personnel might exert substantial influence over supervisory judgments while maintaining that direct responsibility for such remains with the Reserve Bank.

A consequent lack of clarity around responsibility, authority, and accountability — within Board S&R and between the Board and the Reserve Banks — appears to have contributed to the culture of indecision that is directly salient in connection with the failure of SVB.

Former VCS Quarles offered the related view that the decision-making process requisite to drive action at Board S&R has long been mired in diffused authority, exercised through a “miasma of committees,” amidst an institutional reluctance to prioritize supervisory attention. This view was echoed by many S&R staff interviewed by Starling.

  • [Board Staff 13] described supervisors as emphasizing stronger risk-management frameworks rather than insisting upon concrete risk-reduction measures, in keeping with the account offered by the Independent Directors of SVB in the Statement they provided Starling.
  • [Board Staff 4] recalled that [Board Staff 14] observed, “No one knows who can make a decision to permit something [within Board S&R], but every single person here holds a ‘pocket veto’ to prevent something.”
  • [Board Staff 10] described the consequent recurring tension: Risk teams raise alarms and highlight concerns, while portfolio teams may respond that such matters are “not a priority.”

This history of making critical decisions by committee is consistent with a cultural predilection to avoid accountability for perceived mistakes in actions taken. While an individual may be deemed negligent, reckless, or willful, a committee diffuses accountability.

In connection with SVB, the decision whether to require the bank to reduce the size or duration of its bond portfolio, restructure its deposits to be less runnable, or to improve its readiness to borrow large amounts of emergency liquidity when needed, would have been made by an informal committee that involved the Director of the LFBO portfolio, the Deputy Director for Supervision, the Director of S&R, the head of the FRBSF examination team, and various other officers and managers within Board S&R and members of the FRBSF examination team. As they considered whether to take action and what action to take, they appear to have been unable to reach consensus on whether to do anything and, if so, what to do. 

To cite one specific example, FRBSF had recommended in late 2022 bringing an informal (confidential) enforcement action against SVB to require it to remediate its liquidity risk management. But instead of finalizing that enforcement action before SVB failed in March 2023, Board S&R and the FRBSF supervisory staff decided to delay issuing such an action until they could complete a targeted examination on liquidity risk. Unfortunately, SVB failed before the informal committee had completed its deliberation.

Commenting on this, former Governor Tarullo observed that, if a formal or informal committee does not reach a decision on an issue, “the VC or the Director of the Supervision and Regulation Division may have to make the decision.” This distinction is consequential. The supervisory lapses described in our initial interviews, reported upon herein, did not arise because too many people were permitted to contribute views. Rather, they appear to have followed when peer consultation had no terminus, when the absence of agreement prevented action by default, and when no identified official could determine whether or when the institution should take action.

7.Social media did not fuel the run at SVB.

 

This finding is based on the Social Media Analysis dated July 24th 2026 performed by Charles River Associates (“CRA”), attached as Addendum Item 6 to this Preliminary Report (“CRA Social Media Analysis”).

The Barr Report asserts that the run on SVB was “fueled” by social media.54 That assertion was not substantiated by any data or analysis. At Starling’s request, CRA conducted “an independent, data-driven analysis to examine the effect that social media may have had on SVB’s failure, if any.”55 CRA first defines “social media” as “publicly reviewable commentary” to distinguish it from “private communications channels such as group texts, emails, and chats.”56 Based on a forensic analysis of Twitter commentary about SVB in the months leading up to and the days immediately following March 10th 2023, CRA made the following key findings:

  • “The topic of SVB did indeed go viral on social media, but only after the bank run was well underway.”
  • “Several prescient posts were identified in the weeks and months preceding SVB’s failure, discussing fundamental problems at SVB — including exposure to the tech sector downturn and a high concentration of uninsured deposits.”
  • “CRA also examined SVB-related posts during the run period itself, to understand whether social media ‘fueled’ the bank run. CRA defined the SVB run period as 12 AM EST on March 8, 2023 through 12pm EST on March 10, 2023.”
  • “The vast majority of posts during the run period came from third-party accounts commenting on peripheral themes.”
  • “CRA also identified a smaller subsection of relevant first-person Twitter posts about the VC community advising founders to move money out of SVB.”
  • “In general, these tweets confirm a wave of coordinated withdrawal advice coming from the VC firms to tech founders occurring on March 9. However, these social media posts also indicate that most of that withdrawal advice happened outside the publicly viewable social media channels, with people subsequently posting on Twitter about the flurry of communication that had occurred earlier.”
  • “By the evening of March 9, 2023, mainstream media outlets (including Bloomberg, Axios, TechCrunch and The Information) had all published articles indicating that VC firms were advising their founders to withdraw money from SVB.”

“In sum, the analysis reveals that early posts in 2022 and early 2023 highlighting issues at SVB received limited attention and had no measurable effect on SVB itself or the markets,” CRA concludes. While CRA did identify Twitter posts from individuals discussing a withdrawal of funds from SVB during the bank run period, CRA’s analysis indicates that these posts account for “less than 0.6% of all SVB-related posts during the bank run period, and less than 0.02% of the total volume of Twitter activity referencing SVB from October 2022 through March 31, 2023…”

Moreover, the majority of first-person posts reviewed by CRA indicate that the run was “initiated via private communications (DMs, emails, texts, internal corporate messaging platforms, and phone calls) from VCs to startup founders on or before March 9, 2023…” As such, CRA’s analysis “does not support the characterization of the SVB run as ‘social media fueled.’”57

Nevertheless, the assertion that social media fueled the run at SVB was repeated, almost verbatim, in 2023 reports about SVB’s failure published by the Basel Committee on Banking Supervision (“BCBS”)58 and the Financial Stability Board (“FSB”).59

The Basel Committee’s The 2023 banking turmoil and liquidity risk: a progress report (October 2024) further notes “the role of social media and the digitalisation of financing in hastening the speed and impact of a bank’s distress” in connection with “liquidity risk dynamics.”60 Withdrawals at SVB, the BCBS argues, “were amplified by widespread access to digitalisation and social media,” adding that this, in turn, prompts “considerations for supervisors.” 

“Technological developments and social media could accelerate future bank runs, with implications for liquidity risk management practices and supervision,” the FSB asserts in a press release announcing its 2024 paper.61 “Furthermore, the implications of technological developments and social media for deposit stickiness suggest that there could be more such runs in the future.”it offers in the underlying report.62

Perhaps the first effort to provide an empirical foundation for the “social-media-fueled” account of SVB’s failure was an academic working paper, “Social Media as a Bank Run Catalyst,” by J. Anthony Cookson, Corbin Fox, Javier Gil-Bazo, Juan Felipe Imbet, and Christoph Schiller, dated April 18th 2023 — less than six weeks after SVB failed.63

The paper opens with the categorical assertion: “Social media fueled a bank run on Silicon Valley Bank (SVB)...” Press attention followed immediately. On April 24th — four days prior to the release of the Barr Report — Axios reported that “Twitter fueled” the run at SVB,64 while Fortune described the paper as the first major study to examine the issue and announced that its authors had confirmed that social media contributed to SVB’s collapse.65 Such coverage gave the paper’s causal formulation wide public visibility. 

Thereafter, the paper was presented widely within policy circles, lending the paper’s conclusions greater authoritative weight. For instance, it was posted to the FDIC website in connection with its 22nd Annual Bank Research Conference, at which the authors presented the study, on September 28th 2023.66 The paper received on-going attention from policymakers throughout the balance of 2023, and its authors were invited to present their findings at several events hosted by the Fed and other regulatory agencies.

Following further revision and – critically – peer review, the Cookson paper was published in the February 2026 volume of the Journal of Financial Economics.67 It is notable that the categorical assertion that social media “fueled” the run at SVB disappears entirely. Rather, the authors conclude, “We find that negative Twitter sentiment does not amplify bank run risks.”

References
  1. See, e.g., SVB, Call Reports, Schedule RC-O (from March 31st 2019 until December 31st 2021) and Barr Report, 18-20.
  2. SVB, Call Reports, Schedule RC-B (from March 31st 2022 until December 31st 2022). The year-end 2022 call report showed total bonds of approximately $117bn (book value), with approximately $91bn (book value) being classified as HTM (78%).
  3. SVB, Call Reports, Schedule RC, lines 2a, 2b, 12 (from June 30th 2021 until December 31st 2022).
  4. Addendum Item 5 — Analytical Account of the Failure of Silicon Valley Bank as Viewed in Public Data, 7.
  5. Ibid., 4; Board of Governors of the Federal Reserve System, Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank (Washington: Board of Governors, April 2023), https://www.federalreserve.gov/publications/review-of-the-federal-reserves-supervision-and-regulation-of-silicon-valley-bank.htm, 4, 59. 
  6. SVB, Call Reports, Schedule RC-B (from June 30th 2021 until December 31st 2022). 
  7. SVBFG, Quarterly Report on Form 10-Q, 91 (June 30th 2021); SVBFG, Annual Report on Form 10-K, 66 (December 31st 2022).
  8. Frank J. Fabozzi, ed., The Handbook of Fixed Income Securities, 197-98 (McGraw Hill 7th ed. 2005); Andrew Metrick, “The Failure of Silicon Valley Bank and the Panic of 2023”, 38 J. Econ. Persp. 133, 137 (2024).
  9. SVBFG, Current Report on Form 8-K, Exhibit 99.1, 3, note 1 to Investment Securities table (October 20th 2022) (relevant fixed to floating interest rate swaps fell from $11.3bn at the end of Q3 2021 to zero by end of Q3 2022).
  10. Ibid., lines 3 and 4 of the table under the heading “Investment Securities,” which shows the duration of the AFS securities in June 2022 and September 2021 with and without the hedge.
  11. Accounting Standards Codification (“ASC”) 815-20-25-43; ASC 320.
  12. ASC 320.
  13. Ibid.
  14. See, e.g., SVB, Call Report, Schedule RC-B, line 8, difference between Column A and Column B (September 30th 2022).
  15. SVB, Call Report, Schedule RC-O (December 31st 2022).
  16. 12 U.S.C. 1823(c)(4)(A).
  17. 12 U.S.C. 1823(c)(4)(G).
  18. SVB, Call Report, Schedule RC-B, line 8, sum of the difference between Column A and Column B and between Column C and Column D (March 31st 2022). Its unrealized losses on its bond portfolio were about $1bn at year-end 2021. SVB, Call Report Schedule RC-B, line 8, sum of the difference between Column A and Column B and between Column C and Column D (December 31st 2021).
  19. SVB, Call Report, Schedule RC-R Part I, line 19 (March 31st 2022).
  20. SVB, Call Report, Schedule RC-B, line 8, sum of the difference between Column A and Column B and between Column C and Column D (September 30th 2022).
  21. SVB, Call Report, Schedule RC-R Part I, line 19 (September 30th 2022).
  22. Board of Governors, Open Market Operations (showing federal funds rate ranging from 0-0.25% at March 16th 2020 to 4.50-4.75% at February 2nd 2023), https://www.federalreserve.gov/monetarypolicy/openmarket.htm.
  23. Like most Category III and Category IV banks, SVB opted out of reflecting AOCI in its capital the way Category I and Category II banks are required to do. See 12 C.F.R. Part 217. As a result, not only were the unrealized losses on its AFS portfolio not reflected in its reported OCI, they were not reflected in its reported total capital.
  24. FDIC, Dissecting Depositor Flight: An Analysis of the Spring 2023 Bank Failures, 6 (May 2026). The study reports that SVB received $54bn of withdrawal requests on March 9th and honored $41bn of them. See also OIG Report, 9–10 (approximately $42bn of withdrawals on March 9th).
  25. Board of Governors of the Federal Reserve System, Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank, i.; 
  26. Board of Governors of the Federal Reserve System, Office of Inspector General, Material Loss Review of Silicon Valley Bank, Evaluation Report 2023-SR-B-013 (Washington: Board of Governors, September 25th 2023), https://oig.federalreserve.gov/reports/board-material-loss-review-silicon-valley-bank-sep2023.htm, 10.
  27. Ibid.
  28. FDIC Press Release (March 10th 2023), https://www.fdic.gov/resources/resolutions/bank-failures/failed-bank-list/silicon-valley.html.
  29. SVB, Call Reports, Schedule RC and RC-B (from March 31st 2022 until December 31st 2022).
  30. SVB, Call Reports, Schedule RC-B, line 8, difference between Column A and Column B and between Column C and Column D (March 31st 2022 and June 30th 2022).
  31. SVB, Call Reports, Schedule RC-B, line 8, difference between Column A and Column B and between Column C and Column D (September 30th 2022 and December 30th 2022).
  32. Board of Governors of the Federal Reserve System, Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank, 4, 59.
  33. Addendum Item 5 — Analytical Account of the Failure of Silicon Valley Bank as Viewed in Public Data, 4, citing the Board of Governors of the Federal Reserve System, Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank, 4, 59.
  34. Board of Governors of the Federal Reserve System, Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank, 22-23 & note 40 (After noting that “[i]n 2022, as interest rates began to rise, SVBFG saw a rapid increase in unrealized losses in both its HTM and [AFS] portfolios (figure 9),” footnote 40 dismissed the risk of unrealized losses on its HTM securities, stating that “[w]hile unrealized losses must be disclosed in financial statements, they do not change the assets’ value on the balance sheet itself.”). This statement seems to confuse accounting standards with real economic value.
  35. Board of Governors of the Federal Reserve System, Office of Inspector General, Material Loss Review of Silicon Valley Bank, Evaluation Report 2023-SR-B-013 (Washington: Board of Governors, September 25 2023), https://oig.federalreserve.gov/reports/board-material-loss-review-silicon-valley-bank-sep2023.htm.
  36. Because SVB accounted for roughly 99% of the consolidated assets of SVBFG including all of its AFS and HTM bonds, they are economically indistinguishable for purposes of this statement. Compare SVB, Call Report, Schedules RC and RC-B, 19, 21 (December 31st 2022) (showing total assets, AFS securities, and HTM securities (at book value) of $209bn, $28.5bn, and $91.3bn, respectively) to SVBFG, Annual Report on Form 10-K, 95, 123, 125 (December 31st 2022) showing total consolidated assets, AFS securities, and HTM securities (at book value) of $211.8bn, $28.6bn, and $91.3bn, respectively). 
  37. Board of Governors of the Federal Reserve System, Office of Inspector General, Material Loss Review of Silicon Valley Bank, 26-27.
  38. Board of Governors of the Federal Reserve System, Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank, 67.
  39. Ibid., 14, note 26, 70.
  40. Testimony of VCS Barr before the Senate Banking Committee (March 28th 2023), Transcript 44 (in response to a question from Senator Sinema as to whether he first heard about SVB’s extraordinary exposure to interest-rate risk in October or November 2022), https://www.govinfo.gov/content/pkg/CHRG-118shrg54561/pdf/CHRG-118shrg54561.pdf.
  41. Appendix A-13, [Board Staff 15], 4 & 12 June 2026.
  42. Addendum Item 2 — Board S&R and FRBSF Supervision.
  43. Addendum Item 4 — SVB Directors’ Statement.
  44. Appendix A-8, Mary Miller, 17 March 2026.
  45. Addendum Item 5 — Analytical Account of the Failure of Silicon Valley Bank as Viewed in Public Data.
  46. Board of Governors of the Federal Reserve System, Office of Inspector General, Material Loss Review of Silicon Valley Bank, 24-28.
  47. Appendix A-13, [Board Staff 15], 4 & 12 June 2026.
  48. Board of Governors of the Federal Reserve System, Office of Inspector General, Material Loss Review of Silicon Valley Bank, 24.
  49. Ibid., 27.
  50. Ibid., 27-28.
  51. Board of Governors of the Federal Reserve System, Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank, 4, 59; Addendum Item 5 — Analytical Account of the Failure of Silicon Valley Bank as Viewed in Public Data, 4.
  52. Appendix A-2, [Board Staff 2], 5 May 2026; Appendix A-11 [Board Staff 14], 30 April 2026.
  53. See text accompanying note 1 of this Preliminary Report.
  54. Board of Governors of the Federal Reserve System, Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank, i, iii, 1, 10, 11, 35-36.
  55. Board of Governors of the Federal Reserve System, Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank, 4, 24.
  56. Addendum Item 6 — Social Media Analysis Re The Failure of Silicon Valley Bank, 2.
  57. Ibid., 5.
  58. Addendum Item 6 — Social Media Analysis Re The Failure of Silicon Valley Bank, 12-15.
  59. Basel Committee on Banking Supervision, Report on the 2023 Banking Turmoil, (BIS October 2023), https://www.bis.org/bcbs/publ/d555.pdf, 10.
  60. Financial Stability Board, 2023 Bank Failures: Preliminary Lessons Learnt for Resolution, (Basel: Financial Stability Board October 10th 2023), https://www.fsb.org/uploads/P101023.pdf, 19.
  61. Basel Committee on Banking Supervision, The 2023 Banking Turmoil and Liquidity Risk: A Progress Report, (BIS October 2024), https://elischolar.library.yale.edu/cgi/viewcontent.cgi?article=6292&context=ypfs-documents2, 1.
  62. Financial Stability Board. (2024, October 23). FSB analyses interest rate and liquidity risks and the role of technology and social media on depositor behaviour [Press release]. https://www.fsb.org/2024/10/fsb-analyses-interest-rate-and-liquidity-risks-and-the-role-of-technology-and-social-media-on-depositor-behaviour/
  63. Financial Stability Board, Depositor Behaviour and Interest Rate and Liquidity Risks in the Financial System: Lessons from the March 2023 Banking Turmoil, (Basel: Financial Stability Board October 23rd 2024) 10th https://www.fsb.org/uploads/P231024.pdf, 17.
  64. J. Anthony Cookson et al., “Social Media as a Bank Run Catalyst” (working paper, April 18th 2023), 1, https://www.fdic.gov/system/files/2024-07/cookson-paper.pdf.
  65. Emily Peck, “Twitter Fueled Run on Silicon Valley Bank, New Paper Finds,” Axios, April 24th 2023, https://www.axios.com/2023/04/24/twitter-svb-bank-run.
  66. Will Daniel, “Twitter Poses a Risk to the Financial System and Helped Fuel the SVB Run, Says First Major Study to Investigate the Topic,” Fortune, April 24th 2023, https://fortune.com/2023/04/24/twitter-financial-system-risk-silicon-valley-bank-run-study/.
  67. Federal Deposit Insurance Corporation, “22nd Bank Research Conference Agenda,” September 28–29th 2023, https://www.fdic.gov/center-financial-research/22nd-bank-research-conference-agenda.
  68. J. Anthony Cookson et al., “Social Media as a Bank Run Catalyst,” Journal of Financial Economics 176 (2026): 104218, https://doi.org/10.1016/j.jfineco.2025.104218.

Join The Discussion

Sign in and be the first to comment.

See something that doesn't look quite right?

We strive to provide high quality and accurate content at all times. With that said, we realize that sometimes links break, new information becomes available, or there is something that you feel we may have missed.

If you see something that you think we should be aware of, we would love to hear from you. Feel free to drop us a note below and leave your name and contact info if you'd like to hear back from us.

Thank you for being a key part of the Starling Insights community!