The Cartographers of Finance

Plus: Root S-1, Morgan Stanley, Lending Club

Welcome to another issue of Net Interest, my newsletter on financial sector themes. Every Friday I distil 25 years experience of looking at financial institutions into an email that explores key themes trending in the industry. I’m grateful to all of you who have signed up, but don’t keep it to yourselves! Tell your co-workers, your cousins, your friends. Thank you.

The Cartographers of Finance

In nineteenth century England, cholera was a killer. For the first half of the century, no-one knew what caused it. The prevailing hypothesis was miasma—noxious bad air. John Snow, a prominent doctor working in London, had a different theory. He suspected the disease was spread via water. In 1854 an outbreak took hold in Soho and Snow went to investigate.

Snow pulled out a map of Soho and began to mark on it where the fatalities were. This map – reproduced below – would go on to become famous as one of the first examples of how geographic analysis can give us a better understanding of the world. It shows a cluster of black rectangles, one for each cholera case, around Broad Street, near to the local water pump. Snow drew in a line to indicate the area within walking distance from the pump and, as per his theory, that’s where most of the cases cropped up. 

Since Snow’s time, maps have been used to visualise data across many areas—even financial services.

The practice really took off following the financial crisis. Coming out of the crisis, policymakers were determined not to let something like it happen again. They set about creating a more resilient system, less prone to shock. The headline focus was capital, but beneath that were two additional elements: stress testing and mapping. 

Most authorities now stress test their banks on an annual cycle, sometimes more regularly in the event of, say, a pandemic. The Federal Reserve will release the results of its second stress test of US banks within the next few months. The process policymakers adopted has various drawbacks, but the principle is clear. There are few better ways at gauging resilience than chucking stuff at a system, even if it is on a desktop, and seeing what sticks. 

The second response – mapping – is less widely appreciated. The ability to see what’s going on across the system as a whole is key and the way to do that is through maps.

Since the crisis, policymakers have attempted to plot every contour of the financial system:

To a finance nerd, these maps are a thing of beauty.

Take this one:

Through its kaleidoscope of colour, it shows the linkages between banks in Europe and banks elsewhere in the world. The black dots in the middle each represent a European bank; the lines represent their exposures to other banks—blue for the US, green for the UK, red for China. The thicker the line, the greater the exposure. It shows that European banks have few (yet sizable) exposures to Chinese banks and lots of exposures to US and UK banks. The bond between European banks and US banks is clear. 

Or the next one, which shows banks’ dollar-denominated cross-border claims. It differs from the one above because it looks at banks’ exposures to all entities in a different country, not just to other banks, and it drills down only on their dollar exposures. It’s an important map because historically a shortage of dollars outside the US has caused all manner of crises. This was one vector for the spread of the financial crisis into Europe in 2008 (as wonderfully recounted by Adam Tooze in his book, Crashed). 

Unsurprisingly US banks sit in the middle of this map as the largest lender of US dollars to non-US counterparts. The bigger the node, the higher the value of the dollar flows going through it. However, a lot of dollars also flow between non-US entities—shown by the yellow arrows. The good news here is that the flows are a lot smaller than they used to be. 

This week, the US Securities and Exchange Commission published a whole series of maps that break down flows within the US credit markets. The level of detail in these maps is breathtaking. They trace the movement of $54 trillion of credit assets from issuance through intermediaries to their eventual home as at the end of 2019. 

The overall picture looks like this:

But we can zoom in on a particular market for more granularity. Let’s look at the residential mortgage market. It’s laid out below. 

The mortgage map shows that non-banks originate around 70% of all mortgages in the US (companies like Rocket and United Wholesale). However, banks are still very much involved not just as originators of the other 30% but as owners of a big chunk of the mortgage-backed securities that non-bank product is processed into. The federal government also has a big role to play as guarantor of those mortgages through the GSEs (Fannie Mae, Freddie Mac and Ginnie Mae) and as the ultimate owner of many of the securities through the Federal Reserve (QE). However much you know it, looking at the map drives home just how invested the US government is in its mortgage market.

The SEC’s mapping is an exercise that may not have been possible in the past. The computing power available to mine all this data has grown at the same time as the complexity of the system has reduced. Since the crisis, less financial transactions are done bilaterally and more are done via central clearing houses, allowing flows to be more readily tracked. 

These maps are useful because by improving the legibility of the system, we get a clearer perspective of what’s going on within it.

Sometimes this can lead to profound realisations. It wasn’t until the financial crisis was underway that we realised that risk distribution is healthy only up to a point. If everyone distributes risk in the same way, the diversity of the system as a whole goes down and systemic risk is – paradoxically – exacerbated. “We’re in the distribution business, not the storage business,” they all said. Yet what works for a single entity doesn’t work in aggregate. A map of the system may have revealed that.

A map of the system also reveals other patterns. Resilience is affected not just by the component parts in a system or by its aggregate risk exposure, but also by the pattern of connections between entities. Mapping reveals the shape of the network underpinning the financial system. Some of the charts above show clear ‘super spreaders’ prominent in the network structure—large universal banks with lots of connections. These banks can propagate risk and are consequently put under special surveillance by regulators who regularly measure their connectivity. [1]

The basis for mapping is gaining traction more widely. It is being harnessed as a tool to enhance resilience in the face of technological risk as well as financial risk.

Operational Resilience

Over the weekend of 21/22 April 2018, UK bank TSB embarked on migrating its 5.2 million customers onto a new IT platform. It didn’t go well. Mobile banking collapsed, branch technology stopped working, some customers gained access to others’ accounts, cards stopped being accepted in ATMs abroad, phishing attacks spiked, and customer service waiting times lengthened.

In response, UK policymakers issued requirements for financial services companies to strengthen their operational resilience. They highlighted an increasingly interconnected operating environment and expressed concern that the failure of a shared piece of connectivity or the loss of access to a major cloud provider, say, could lead to major disruption.

The risk was further exposed when UK banking-app customers were locked out of their accounts in June, after the failure of Wirecard. As the FT wrote, “Wirecard was not considered particularly risky from a regulatory perspective. Its risk was hidden in its interconnectedness across the market.”

So seriously are policymakers taking this issue of operational risk and resilience that they now rate it equivalent in importance to financial stability.

Their proposals call for comprehensive mapping of systems. Implementation of the rules has been delayed because of Covid, but more mapping is inevitable. Again, shape will tell a story. Some network structures are more resilient to random failures, such as those caused by natural disasters; others are more resilient to targeted incidents, such as hacks.

Clearly mapping doesn’t provide complete transparency—as anyone who has taken the tube from Embankment to Charing Cross can attest. There are two issues here.

The first is that the markets these maps represent do not operate in isolation. Just as the London Underground connects to the overground, the US credit market connects to short-term funding markets. Richard Bookstaber, a risk manager who has worked for Morgan Stanley, Bridgewater and others, worked with colleagues to draw up a grand map of the markets for the US Office of Financial Research. He proposed a multilayer map that shows how risks can emerge and spread across the US financial system. The three layers of his map represent short-term funding, assets, and collateral flows. It looks like one of those maps you see at a shopping mall, showing the stores on each of the floors. 

The second issue is that these maps may be a lot more descriptive than they are predictive. That’s because the terrain shifts in a way it doesn’t on the tube network (thankfully). Bookstaber’s multilayer map is employed to explain the Bear Stearns collapse after it happened; the SEC analysis is used to better understand what happened to credit markets in March, in the aftermath of the Covid shock.

In short, their maps came too late—just as John Snow’s did for the residents of Soho all those years ago. Soho deaths peaked on 1 September 1854 but Snow wasn’t able to get the Broad Street pump taken out of service until 8 September. By then, most of the people at risk had either caught the infection or fled the area. 

There is an argument that simply having the map provides a security blanket of sorts. It’s an argument made by Dr. Kimmo Soramäki, the founder of a company which sells mapping software to regulatory supervisors. He says that after the financial crisis he and his team did a lot of work mapping bank exposures and building models of contagion. The problem is that they couldn’t replicate the cascading defaults that would have happened without central bank and government intervention. He concludes that the root cause of the crisis was therefore not the actual exposures or risks, but the fear caused by the lack of measurement and valid information on them. Mapping provides that measurement and information and so serves to dispel the fear. 

As Captain Barbossa from Pirates of the Caribbean says, “You’re off the edge of the map, mate. Here there be monsters!”

There’s a version of the future in which the entire financial system is mapped out and updates in real time, just as traffic maps update to accommodate roadworks. If this map enhances the legibility of the system it may mitigate risk. It’s a map that Andy Haldane, Chief Economist of the Bank of England, dreams about:

“I have a dream. It is futuristic, but realistic. It involves a Star Trek chair and a bank of monitors. It would involve tracking the global flow of funds in close to real time (from a Star Trek chair using a bank of monitors), in much the same way as happens with global weather systems and global internet traffic. Its centrepiece would be a global map of financial flows, charting spill-overs and correlations.”

[1] Most banking networks appear to have disassortative network structures whereby big banks are disproportionately connected to small banks and vice versa. In 1989 epidemiologist Sunetra Gupta led a study showing that such network structures are more liable to spread infection. For more on this idea, read Adam Kucharski’s book, The Rules of Contagion.

Forwarded this? Sign up here

Share Net Interest

More Net Interest

Root S-1

Root filed an S-1 prospectus this week as it prepares to follow fellow insurtech, Lemonade, to the market. While Lemonade’s S-1 was full of jocularity (“Lemonade is not everyone's cup of tea”), Root’s is more staid. That reflects its founder, an insurance man: 

“The Root story began in my childhood. I grew up in an insurance household, and unlike other kids who might have had a paper route, I began working at my father’s insurance company at the age of 14.”

An insurance veteran he may be, but that doesn’t stop him positioning his company as a technology company: “Root is a technology company revolutionising personal insurance.” At the heart of the company is data. The company boasts of a massive data set sourced from over 10 billion miles of driving history. It uses that data to price auto insurance risk, using predictive models it reckons are 10x better than others’.

Right now that pricing advantage is not translating into profit. The company reported a loss ratio of ~100% in 2019. However, it reckons that as renewals take a greater share of premiums, the loss ratio will come down. Renewals were 47% of last twelve month premiums in June 2020 against an industry average of 80%. The early cohorts are not necessarily indicative of how future customers will behave, but the company has seen loss ratios decline by 27% between the first and fourth term.

Meanwhile, Root is looking to grow. The trouble with insurance is that the more you grow the more regulatory capital you need to put aside—growth comes with a capital drag. That’s not a challenge other technology companies have to contend with. To alleviate this, Root is giving away some of the upside to growth by ceding 70% of premium to reinsurers. 

Meanwhile, it grapples with the challenge other technology companies do have to deal with—customer acquisition. Root paid $929 in sales and marketing per new policy in 1H20. But with average premium per policy of $909 and one-year retention of 84%, the numbers may be OK.

Morgan Stanley / Eaton Vance

Not one week after completing its E*Trade acquisition, Morgan Stanley announced the purchase of asset manager Eaton Vance. The pair take the firm closer towards wealth and asset management and further from its investment banking roots. It’s an evolution that began in 2010 with the acquisition of Smith Barney and now close to two thirds of group revenues will be derived from managing other people’s money. 

The deals cement Morgan Stanley’s vertically integrated approach to its market. Since the birth of asset management, firms have cycled between vertical integration and openness. The advantages of an open structure are that it eliminates potential conflicts of interest and broadens customer choice. The advantage of vertical integration is that it can lower cost.

The irony here is that establishing a modularised, open structure has never been easier with the technology that is available. The UK is exploring an ‘open finance’ initiative to take advantage of that technology. Yet banks like Morgan Stanley have excess capital that needs a home and so the financial incentive leans the other way. 

Lending Club

Lending Club, one of the pioneers of peer-to-peer lending has announced that it will stop offering retail peer-to-peer at the end of the year.

The problem with peer-to-peer lending is that although it looks like a two-sided marketplace, it doesn’t derive the positive feedback that other two-sided marketplaces enjoy. Just because there are lots of lenders on the platform, it doesn’t mean that there should be lots of borrowers. Contrast that with other platforms in domains like ride sharing, e-commerce, credit cards.

This means that peer-to-peer lenders constantly have to shimmy one side of the platform or the other. They bolstered the lender side by attracting institutional money from hedge funds; they poked the borrower side by reducing loan standards. Banks emerged many centuries ago as a solution not to have to do this. Lending Club has realised as much and on 25 September it filed an application with the Federal Reserve to become a bank holding company. The peer-to-peer experiment was failing anyway; now it’s over.