Perspective on Risk - Jan. 11, 2023

Crypto Snuck Into the System; Extend & Pretend (Belt & Roads Edition); LGD Estimates; Infrastructure Lending; Lessons From Southwest Airlines; Quantifying Operational Risk; Ironic Internet System

Jan 11, 2023

I’m kinda proud of this one; a bit different.

Crypto Did Sneak Into The Financial System

During the whole crypto debacle, the one thing that I’ve been most heartened by was that the industry had not been allowed into the regulated financial system.

Now we do know that Silvergate Bank snuck through the door and was given a Fed master account. Per Silvergate’s, 10-K, 58% of deposits are held by 96 crypto exchanges.

But now the Kate Barry of the American Banker caught that Silvergate actually borrowed from the FHLB amidst the crypto run:

Silvergate received $4.3 billion from the Federal Home Loan Bank of San Francisco late last year, company filings show. The billions in liquidity provided by the FHLB in the fourth quarter alone helped La Jolla, Calif.,-based Silvergate stave off a further run on deposits. The crypto-friendly bank now holds roughly $4.6 billion in cash — the bulk of which came from Home Loan Bank advances, according to select financial metrics that Silvergate released last week.

The FHLB system is often called “the lender on 2nd to last resort.” In practice, banks will much more readily go to the FHLB for funding that come to the Fed’s Discount Window.

Extend & Pretend (Belt & Roads Edition)

China Props Up Belt-and-Road Borrowers Via Unusual Channel (WSJ)

Alternative title: Extend & Pretend on a Global Scale

The PBOC’s swap network is the largest of its kind, according to the World Bank. The PBOC said in a 2021 report that it has swap facilities with 40 countries with a combined capacity of almost 4 trillion yuan, or about $570 billion. … The PBOC’s swap lines are broader in scope, and can be used by foreign central banks to address balance of payment needs.
The transactions, researchers say, are part of a broad but opaque effort by Chinese authorities to prop up governments that borrowed heavily from Chinese banks as part of Beijing’s $1 trillion Belt and Road Initiative to finance infrastructure projects and win influence across the world. … The swaps are often rolled over, sometimes for years
“The big difference in how China uses these swap lines is it effectively uses it as a bailout mechanism,” said Christoph Trebesch

Some Information To Calibrate LGD Estimates

In economic capital modeling, one tries to estimate the losses that occur in very rare cases *such as 1-in-1000). To do this, one needs, at a minimum, to estimate the probability of default of each borrower, the likelihood that multiple borrowers will default during the same time period (correlation), and the amount lost when collecting on the defaulted loan.

The Quarterly Journal of Economics has published Asset Specificity of Nonfinancial Firms by Amir Kermani and Yueran Ma that should have some useful information for those of you that model:

We develop a new data set to study asset specificity among nonfinancial firms. Our data cover the liquidation values of each category of assets on firms’ balance sheets and provides information across major industries. First, we find that nonfinancial firms have high asset specificity. For example, the liquidation value of fixed assets is 35% of the net book value in the average industry. Second, we analyze the determinants of asset specificity and document that assets’ physical attributes (e.g., mobility, durability, and customization) play a crucial role. Third, we investigate several implications. Consistent with theories of investment irreversibility, high asset specificity is associated with less disinvestment and stronger effects of uncertainty on investment activities. We also find that the increasing prevalence of intangible assets has not significantly reduced firms’ liquidation values.

That last sentence alone is worth investigating. Michael Mauboussin helpfully tweeted this set of graphs.

Infrastructure Lending

Hot area, Building Portfolios with Infrastructure: Performance, Cash Flows & Portfolio Allocation by Shen and Blanc-Brude have a new paper for those interested.

We explore the performance and cash ﬂow properties of infrastructure investments, both at the asset and the fund level. We use PGIM IAS’ asset-allocation framework (OASIS) to determine the appropriate allocation to infrastructure and other illiquid investments (e.g., private equity) given a CIO’s liquidity risk tolerance.

The paper has some useful intra- and inter-asset correlation statistics on this non-public class.

Lessons From Southwest Airlines

The airline industry has been having problems getting the planes into the air. First, Southwest had its holiday outage, and recently the entire system was affected by the NOTAM1 system resulting in ‘ground stop.’

It’s too early to learn lessons from the NOTAM outage, but perhaps not for the Southwest issues last December. Let’s dig in.

The issues appear to have at least two components: a breakdown of antiquated back-end scheduling system, and the complexity of Southwest’s business model.

Skysolver Scheduling System

During a major winter storm over the Christmas travel period, the system Southwest uses to schedule flight crews, Skysolver, broke down.

"The process of matching up those crew members with the aircraft could not be handled by our technology," Watterson said. …
Southwest ended up with planes that were ready to take off with available crew, but the company's scheduling software wasn't able to match them quickly and accurately, Watterson added.2

Numerous reports, though likely from the same source, describe SkySolver as ‘outdated’ or ‘end-of-life’

On a call with employees Monday, Southwest Chief Operating Officer Andrew Watterson explained that the company's outdated scheduling software quickly became the main culprit of the cancellations once the storm cleared, according to a transcript of the call that was obtained by CNN from an aviation source.3

It appears that customization by Southwest may have led to a failure to stay on the upgrade path. Southwest also reportedly ran the system on its own mainframes; the current SkySolver brochure shows that it is now a cloud-native application.

Southwest relies on crew-assignment software called SkySolver, an off-the-shelf application that it has customized and updated, but that is nearing the end of its life, according to the airline. The program was developed decades ago and is now owned by General Electric Co.4

GE may not have invested in keeping SkySolver at the leading edge, as ‘it was originally a Navitaire product’ that was acquired by GE.

Multiple systems are involved in crew scheduling, according to a spokesman for GE Aerospace. The company said its software isn’t an end-to-end solution. Rather, it’s a so-called backend algorithm that airlines can supplement with other software. The algorithm gathers input from other systems to provide recommendations to resolve crew-related disruptions, according to GE Aerospace.5

One critical issue is that Southwest’s use of SkySolver relied heavily on telephone communication. This resulted in a ‘meltdown’ of the telephone-based system.

systems failed to keep up because flight attendants use a system largely reliant on phone lines, Montgomery told KHOU. That system can get clogged up, forcing crews to wait on hold for hours waiting for a new assignment.6

“You simply can’t make enough phone calls, you can’t make thousands of phone calls at once,” Montgomery said. Southwest, for its part, issued a new statement on Tuesday saying “operational conditions” and unprecedented volume challenged its flight scheduling tools.7

Clearly, SkySolver couldn’t scale to the size of the problem.

Southwest’s software wasn’t designed to solve problems of that scale, Chief Operating Officer Andrew Watterson said Thursday. …
During the winter storm, amid a huge volume of changes to crew schedules to work through, SkySolver couldn’t handle the task of matching crew members and which flights they should work, executives of the Dallas-based carrier said. 8

As with many companies, priority for IT investment was devoted towards front-end customer-facing activity:

In 2017, Southwest completed a migration to Amadeus Altéa passenger service system, one of two industry standard platforms widely used throughout the aviation industry.9

Southwest recently completed an upgrade of its new reservation system and had been working through multiyear upgrades to systems used in its operations. But it had focused on maintenance and ground operations ahead of crew scheduling, said Southwest’s Mr. Watterson. “At the time that seemed like a proper sequence,” he said. 10

There also appears to have been issues coordinating different systems:

One of the major issues, according to current Southwest CEO Bob Jordan, is that the airplane and route optimization program known as "the Baker," which is designed to automate disruption recovery, is not integrated with their crew scheduling system.
These stovepiped and non-automated systems may have worked fine when Southwest was a small regional carriers, but not today when it’s become one of the world’s largest airlines.11

So, we have the usual prioritization of client-facing IT over core back-end systems, a reliance on ‘antiquated’ technology, perhaps due to excessive customization that inhibited upgrades, and inadequate capacity planning and an inability to scale up when needed.

Southwest clearly faced what is euphemistically referred to as ‘technical debt.’ Technical debt is the sum total of maintenance to IT components (hardware, software, network) that should have been performed but wasn’t. That just means they had old, outdated systems.

Southwest’s Business Model Added Complexity

Southwest’s business model is different from other airlines that predominantly rely on a ‘hub-and'-spoke’ network.

Southwest Airlines, instead, uses a point-to-point model in which "aircraft may make multiple stops along a route as they ply their way across the country," he added. "So, when an extensive bad weather system hits, Southwest is affected in ways that other airlines are not."

This makes logistics more difficult.

A confounding factor is that Southwest's operations give them significant advantages, as Peter Greenberg noted above. However, their point-to-point model is far more likely to collapse under severe disruption. There is a formula in route optimization, (n * ( n − 1)) /2), to calculate the number of routes where the variable of n stands for the number of point-to-point cities (destinations). If an airline offers flights to 100 destinations, 4,950 routes will be needed to cover all destinations in the network.
Southwest has roughly 6,000 pilots and approximately 10,000 cabin crews, composed of two pilots and three or four flight attendants per flight. Unless this personnel (equipment, ground crew and gate attendants) are in scheduled places at the scheduled time, Southwest's tight operations can quickly fall apart because if any of those resources are misplaced or off-shift, FAA regulations restrict the plane from flying passengers.
Scheduling airlines is a complex system with constraints concerning union rules, federal regulations and airline policies when assigning crews and pilots to flights. However, Southwest's system couldn't track where its crew members and pilots were after so many flights were canceled. 12

Now, Southwest ironically had taken a number of steps that reduce their overall risk. Unlike competitors, they only fly one type of aircraft and are known for extensively cross-training their staff. Seems like they just under-estimated this risk.

Forbes Has An Interesting Summation

In 20 Ways To Avert Your Company’s Own Southwest Meltdown Doug Laney lays out a number of lessons. Much of this is speculation at this point, but I thought I’d highlight a four of his points:

Executive Management Experience Versus Brand
Herb Kelleher, known for his commitment to operational excellence … His successors, on the other hand, come from accounting and financial backgrounds and have spent the last decade focusing on what they know: revenue management enhancements.
Workforce Woes
Cultural inconsistencies can result in critical labor shortages, with employees quitting, striking, calling in sick, or the new trend: “quiet quitting” in which employees show up for work and put in minimum effort. Even worse, operational mandates that put employees in peril will exacerbate labor issues—especially when there are mistrust issues with management. This played out as Southwest forced ground personnel to work longer shifts in frigid weather.
Becoming a Technology Company
No mention of technology priorities has been made in any of Southwest's annual report letters from the chairman and CEO over the last ten years. This sends a clear message to the rest of the organization (and a prodrome for investors) that the overall business strategy is business as usual.
Scenario Planning, Stress Testing, and Simulation
When it comes to highly integrated systems and procedures involving human interaction, many advanced organizations are developing “digital twins” that provide visibility into how the business (or a product) performs under a wide variety of “what-if” circumstances.

Quantifying Operational Risk

A new paper by two Board economists, Filippo Curti and Marco Migueis; The Information Value of Past Losses in Operational Risk. This stuff will probably never become credible to CFOs and CEOs. Unless you can tangibly ties internal capital levels to the quality of controls, it is just academics.

Operational risk is a substantial source of risk for US banks. Improving the performance of operational risk models allows banks’ management to make more informed risk decisions by better matching economic capital and risk appetite, and allows regulators to enhance their understanding of banks’ operational risk. We show that past operational losses are informative of future losses, even after controlling for a wide range of financial characteristics. We propose that the information provided by past losses results from them capturing hard to quantify factors such as the quality of operational risk controls, the risk culture, and the risk appetite of the bank.