The devil we don’t know is lurking in the financial system.
At the start of last week, Knight Capital Group was a battle-tested trading firm whose electronic trading systems had performed steadfastly through several crises since the late 1990s. Unlike Lehman Brothers, it was not the sort of firm that took on too much debt to buy assets that could later collapse in value.
Knight’s business model was to match buyers and sellers of stocks more quickly and skillfully than its rivals could, hardly a business that is associated with firm-destroying risk. Yet seemingly out of nowhere, Knight made a huge amount of trades last Wednesday that ended up saddling the firm with $440 million of losses and forcing it into a financially punitive rescue that was sealed on Monday.
The episode tests the limits of what a market can possibly know. And that’s why it’s uniquely unnerving.
Nearly all trades boil down to a disagreement between buyers and sellers over how much a company or security is worth. In a healthy market, investors feel they have ways to deepen their understanding of what constitutes the value of a share or bond. They dive into balance sheets. They quiz management. Investors spend a lot of time debating their theses. That back and forth is what markets thrive on.
Even in the biggest manias, like the mortgage bubble of the last decade, there were skeptics, armed with data. Hedge funds had enough information to bet against subprime securities. Investors went public with criticisms of Lehman and other shaky banks, using the firms’ own financial statements against them. In the aftermath of the crisis, banks quickly became targets if they had troublesome positions. MF Global’s European trades helped deplete confidence in the firm.
But Knight had almost no red flags. Its balance sheet doesn’t show big build-ups of assets of potentially dubious value. Its income statement is plain compared with those put out by much bigger Wall Street firms.
The market even appeared to give Knight some credit for that. At the end of June, the company’s stock market value was three-fourths of its net worth, as measured by its balance sheet. That’s hardly outstanding, given that investors usually give a company a value above its net worth if they like its prospects. But it was much better than Morgan Stanley, a much larger, more complex firm, which was then trading at half its net worth.
Of course, algorithmic trading, Knight’s core business, had plenty of critics. Recent incidents showed that it could be precarious for practitioners. But could outsiders ever have known that such trading had the potential to land a firm with crushing losses in a matter minutes? Knight’s filings hardly quantified that risk. In its last annual report, the firm said merely that a major system failure might lead to “potentially costly incidents” or “substantial financial losses.”
The likely truth of the matter is that Knight had no idea how much it could lose if its trading went awry. How could it? Predicting the losses from a rogue algorithm is like predicting the losses from a tornado that turns into a hurricane. In fact, it’s not even clear if it was the technology in isolation. In a filing with the Securities and Exchange Commission on Monday, Knight said there might have been a human element. The filing stated that “the company experienced a human error and/or a technology malfunction related to its installation of trading software.”
Knight’s failures were perhaps worse than those of banks that allow human traders to get out of hand. Though JPMorgan Chase announced big trading losses on derivatives in May, several media reports had previously noted the souring positions, which had taken months to amass. Knight’s losses stemmed from positions that came about in minutes.
No doubt, bank executives will say that the way to prevent the similar debacles is to improve “risk management,” the industry term used for firms’ efforts to assess how much they could lose under different situations. But in a piece of research that enraged Wall Street this year, Moody’s Investors Service said that “there are no financial ratios to measure the effectiveness of risk management.” In other words, outsiders have no sure way of knowing which banks are good at policing their own risks.
Looking at Knight, regulators are likely to argue that banks have built up hefty loss buffers since that crisis that will be able to absorb surprise losses. But that cushion, called capital, is calculated with a relatively static approach using a known amount of assets. Capital gets overwhelmed if, like Knight, a bank acquires a huge amount of additional assets in a matter of minutes.
The hope is that the Knight debacle was a freak accident. But in today’s market, how could we ever know?