Complexity and Unknown Unknowns

March 2023 Mid-Month Commentary

The past couple of weeks have witnessed an incredible volatility in both digital assets and TradFi assets. The VIX, an index that measures short term US equity volatility expectations published by the Chicago Board of Options Exchange, jumped to over 25% from an average of around 19% experienced over most of 2023 YTD. BTC, representing about 45% of the total digital assets market cap, traded over a range from the lows of $19,800 to over $28,000, a trough-to-peak deviation over 41% in just under two weeks. To characterize these as “normal” activities of risk assets observed from the tails of their respective return distributions is to miss the nuance of the increasing complexity and the unknown unknowns that are present.

The phrase “unknown unknowns” was coined by the US Secretary of Defense Rumsfeld back in 2002 in reference to Iraq’s alleged supply of weapons of mass destruction. It falls under a broader category of Knightian uncertainty, a concept attributed to economist Frank Knight who in 1921 made the distinction between risk and uncertainty. Risk is quantifiable, with stochastic parameters, while uncertainty refers to occurrences that may not be quantified or even expected. Both the US banking liquidity crisis and the hack of the Euler DeFi lending protocol seem to be prime examples of this unquantified uncertainty.

US Banking Liquidity Woes

Silvergate Bank had been a close partner to the digital assets industry by providing critical fiat currency on- and off-ramps. Late on March 1, Silvergate Capital Corporation, the parent of Silvergate Bank, announced that it would not be able to file its annual 10-K by its allotted mid-March extension date. A week later on March 8, Silvergate announced its intention to wind down its bank. US Senator Elizabeth Warren, a vocal critic of digital assets, quickly tweeted that “Silvergate Bank’s failure is disappointing, but predictable” due to its exposure to cryptoassets.

Despite Senator Warren’s assertion, it soon become evident that Silvergate’s failure was only partly due to crypto contagion, with the other part being the result of a traditional run on the bank, i.e., a liquidity crisis. By that weekend, the Federal Deposit Insurance Corporation (FDIC) had shuttered the doors of and placed into its receivership both Silicon Valley Bank and Signature Bank. Fears of a banking contagion spilled over to First Republic Bank, motivating a consortium of large banks on Mar 16 to pledge a total of $30B in cash infusion for First Republic Bank. Europe has not been spared the contagion either and as Credit Suisse convulsed, the Swiss National Bank urged the remaining Swiss banking giant, UBS, to take over its smaller rival this past weekend on the 19th (which it did).

Many of the details of these bank failures are quickly coming to light as these are publicly traded entities who regularly post their financial statements. A quick peek at the most recent balance sheets of the three failed and one teetering bank indicated that the amount of cash and equivalents each had was far less than the amount of cash depositors demanded within a very short time window during early March. The path to profits for a typical bank practicing the fractional reserve business model is to borrow short (from depositors) and lend long in the form of issuing and holding loans or long-dated securities to capture the interest rate spread. For example, SVB Financial Group’s, the parent company of Silicon Valley Bank (SVB), annual 10-K filing showed that the company had $13.8B in cash and cash equivalents compared to a total deposit size of $173.1B, the majority of which were non-FDIC insured. Much of its assets were tied up in long-dated assets including $26B in available-for-sale securities, $91B in held-to-maturity securities, and $74B in loans. When depositors tried to pull $42B from SVB in one day on March 9, the bank simply did not have the cash readily on hand to meet the demand.

Euler Hacked

Meanwhile over in DeFi, Euler Finance, a decentralized lending platform, was hacked on March 13 whereby over $190M was stolen from the protocol. The hacker used flash loans (a DeFi construct that allows a person to borrow money without collateral and simultaneously execute multiple legs of a transaction including the opening and closing of the loan) to build up large loan positions on Euler. They then took advantage of a vulnerability in a smart contract function, tricking Euler to flag the loans as under-collateralized. In the last step, the hacker played the role of the liquidator to close the loans and profit by taking possession of the collateral at favorable rates. By repeating this process, the hacker walked away with various digital assets including DAI, WBTC, stETH, and USDC.

Complexities, Incentive Misalignments, Unintended Consequences

While the TradFi banking crisis and the DeFi hack are two distinct events, they both highlight how our financial systems have become incredibly complex and interconnected, with information flowing ever more quickly, thereby leading to consequences that are harder to predict with the potential to be more impactful.

To be fair, financial systems were never simple. Fractional reserve banking underpins economic growth. Banks owe it to their shareholders to seek profits. In normal economic environments with an upward sloping yield curve and a stable deposit base, borrowing short to invest long is a sensible way to mint profits. A graph from the Economist publication shows how commonplace this activity is among US banks, although the same can be said about banks globally.

Source: “How deep is the rot in America’s banking industry?”, The Economist, 2023-Mar-18

What is less clear are the pitfalls that may be present when we stray outside of “normal” economic environments. This is where we find ourselves today – on the other side of normal. Post-pandemic, general expectations were indicating for decent economic growth, accompanied by moderate levels of inflation, allowing central banks to gradually wind down the assets on their respective balance sheets from years of quantitative easing. But it has been anything but normal. We have a war raging in Ukraine that disrupted the energy and commodities supply chains, causing inflation to rise abruptly. With unemployment in major economies near its lows, central banks globally aggressively raised interest rates, causing long duration, risk asset prices to buckle. It is these very assets that have quickly damaged the left side of the bank balance sheets, leaving the banks with a far weaker liquidity profile to meet the demands of the depositors. Coupled with the ability for news or rumors to travel instantaneously across various media platforms and it is not surprising to see that depositors were trying to withdraw $42B from SVB in one day.

Banking is a heavily regulated industry and in fairness, its top regulator, the Federal Reserve, strenuously tracks risk in this sector. As outlined in its November 2022 Financial Stability Report, promoting the “stability of the financial system” is one of its five key functions. The date of the publication suggests that the data is only through September 2022 at best, but it nonetheless gives us insights into what metrics the Fed is monitoring and how those metrics were trending. The Fed uses a four-part monitoring framework that focuses on asset valuations, business and household borrowings, financial leverage, and funding risks, and considers both domestic and international trends as well. Post-GFC (global financial crisis), the solutions that the Fed has employed have included requiring higher quality capital, stress testing, new liquidity requirements, and the implementation of countercyclical capital buffers.

From the report, it is clear that the Fed is well aware of the risks present in the financial system and the impact on the banking system. While equity capital seems to be adequate, the Fed report did point out falling asset prices and worsening liquidity conditions (as measured by the market depth in the chart below), even among high quality assets such as the two- and ten-year Treasuries over the past year.

In the Funding Risk section, the Fed highlighted six types of liabilities that would lead to a liquidity crisis. Notably, it called out that uninsured deposits at banks totaled $7.9T by Q2 2022, equivalent to over 30% of the US GDP.

While Silicon Valley Bank was a “smaller” and wasn’t even considered a Category IV bank, thereby possibly receiving less scrutiny than the larger banks, it nevertheless received its fair share of warnings from the Fed over the past year.

Over in DeFi, there is no central authority monitoring financial or smart contract risks but there are grassroot effort. DeFiSafety is one such organization that evaluates DeFi protocols based on each protocol’s “transparency and adherence to best practices”. According to DeFiSafety’s Process Quality Reviews (PQRs) documentation, evaluations are conducted using its five review categories of smart contracts and team, code documentation, testing, admin controls, and oracles. Pre-hack, Euler Finance received a total PQR score of 95%. It stood out notably on having its protocol sufficiently audited, has in place a $1M bug bounty, and $10M of smart contract insurance coverage from Sherlock. After being hacked for over half of its TVL (total value locked), Euler now has an updated PQR score of 65%.

So, what’s an investor supposed to do?

We need to acknowledge how complex our financial systems have become, interconnected they are, and susceptible to seemingly exogenous factors they’ve become. The concept of risk needs to be multi-dimensional, encompassing more than price volatility to include counterparty, liquidity, and with digital assets, smart contract risks. While price volatility and liquidity risks are more easily quantifiable and counterparty or smart contract risks are more qualitative, properly prioritizing and rank ordering the various dimensions of risks remains a useful exercise.

Diversification remains a key tool for investors, but one needs to understand the effective breadth of one’s investments. Seemingly diversified investments may be aligned along just a few dimensions. For example, investments in automobile and semiconductor companies are seemingly separate but the two industries have become more vertically integrated. Investing in Orca, a DEX, and Star Atlas, a metaverse game that is still undergoing development, is seemingly diversified until one sees that both are dapps that run on the Solana blockchain and subjected to the chain’s ability to gain adoption.

Investors who rely on stochastic models of return, risk, and higher statistical moments may want to pay some attention to the works of Richard Bookstaber, an author and finance professional who’s worked in chief risk officer roles and policy making. His book, The End of Theory, and other papers call out the limits of our traditional heuristics due to the “four horsemen”. These four include emergent phenomena (unpredictable dynamics arising out of agents pursuing simple incentives), computational irreducibility (inability to reduce real-world dynamics into a closed-form mathematical model), non-ergodicity (varying variations of behavior due to interactions of individuals in and with the system), and radical uncertainty (the unquantifiable and unknown unknowns).

Risk is on the other side of the return coin. A fuller measurement, or at least assessment, of risk is much needed when considering return opportunities. These past couple of weeks, nay, all of last year, highlight how important this is. This is especially true for digital assets where return opportunities seem plentiful, but risks take on newer forms and may be underappreciated.