Thursday, January 31, 2008

The Rise of Systemic Financial Risk

MIT's Andrew Lo describes how one rogue trader can impact global markets.
Market watcher: Andrew Lo, director of MIT’s Laboratory for Financial Engineering, says that the growing complexity of world markets makes it more likely that aberrations like the Societe Generale fraud will rock world markets.
Credit: MIT

Yesterday, a week after Societe Generale disclosed a $7.2 billion loss by a single rogue trader, Bank of France chairman Christian Noyer declared to a French senate finance committee, "None of the controls within Societe Generale seem to have worked as they should have."

But beyond the evident failure of internal control technologies lie wider vulnerabilities in the global financial system. It is possible that the deeds of 31-year-old Jerome Kerviel at Societe Generale triggered global stock sell-offs, says Andrew Lo, director of MIT's Laboratory for Financial Engineering. And that points to widening systemic risk in ever more complex financial markets.

Technology Review: First, what do we know about the failures of those Societe Generale controls?

Andrew Lo: They are still trying to piece together the different methods he used, but apparently, it was his intimate knowledge of Societe Generale's systems infrastructure that allowed him to circumvent various controls. From news reports, it appears he was able to access internal financial databases and not only alter the stated holdings of the accounts he was trading, but was also able to circumvent the checks and reconciliation processes that were put into place to make sure these were accurate. Apparently, the standard reconciliation processes did run, but he was able to alter the records both before and after these processes ran so as to avoid detection and maintain his portfolio.

TR: Can't we just build better software and other technologies to prevent a recurrence?

AL: Yes, but anytime there is an interface between technology and human behavior, you open yourself up to the potential for fraud. Systems don't build themselves: humans program them. A big event like this happens every so often, and then people say, "Gee, we have to spend more time and money to improve our systems," and the systems become safer. Once the systems become safer, we get lulled into a false sense of security and complacency. And eventually, we experience a rude awakening when the next disaster strikes. I would argue that it is impossible to prevent these disasters with 100 percent certainty.

TR: Okay, so bad things will happen. I take it you are mainly concerned about the ripple effect when they do?

AL: Exactly. The financial system as a whole is getting more complex. Financial institutions rely on ever more elaborate systems architecture and electronic communications across different counterparties and sectors. The number of parties involved, the nature of transactions, the volume of transactions as the market grows--taken together, the dynamics among these aspects of financial markets imply that the complexity is growing exponentially. No single human can comprehend that complexity. And as the system grows more complex, it is a well-known phenomenon that the probability of some kind of shock spreading through the system increases as well. Systemic shocks become more likely. Today, we are looking at some significant exposure to relatively rare events.

TR: In what way was the Societe Generale matter such a shock?

AL: One natural hypothesis is that the global sell-off that happened early last week was a direct outcome of Societe Generale's unwinding of these rogue trades. We don't have any conclusive evidence yet, but it's not an outlandish conjecture given the circumstances surrounding the massive fraud that was allegedly committed. According to Societe Generale, the problem was discovered on Saturday [January 19], and the firm began unwinding their portfolio at the first possible opportunity. If it turns out that this "unwind" was on the scale of a billion dollars or more, it is plausible that the unwind itself triggered the global sell-off--first in Asia, then in Europe, and then in the U.S.

TR: So one person, in this case Mr. Kerviel, can move the entire global financial system.

AL: It's a larger-scale version of what happened in August of 2007--in particular, August 7, 8, and 9. A large number of quantitative equity hedge funds lost money on those dates simultaneously, yet there is no market event that you can point to that can explain why these funds lost money at the same time. But looking at circumstantial evidence, we [at MIT] pieced together a story that one large quantitative equity fund decided to unwind its portfolio, for reasons we don't know for sure, but which we conjecture to be related to credit problems from the subprime mortgage market. Because the conjectured liquidation involved a big fund that needed to be liquidated quickly, this implies that the impact of the liquidation on other similarly positioned quantitative equity funds would be negative--and large. You get a snowball effect. Everybody is heading for the exit door at the same time, and you get a crash. But in August 2007, it was not a crash of the market as a whole, but of portfolios that are similarly structured to the fund that started the snowball.

TR: So how can we mitigate these kinds of wider risks?

AL: Probably the best way to reduce the impact of systemic shocks is to provide investors with some transparency as to their likelihood and severity, and let the investors decide how much risk to bear. This is probably best accomplished by creating a government organization like the National Transportation Safety Board, charged with the mandate of analyzing every financial blowup or crisis and producing publicly available reports that describe the nature of the crisis, the circumstances leading up to it, and proposed methods for avoiding such incidents in the future. In the same way that the NTSB has improved the safety of air travel by sifting through the wreckage of every airplane crash and publishing a detailed study of its findings and recommendations, a Capital Markets Safety Board would give investors more insight into the risks of any given investment. Over time, the aggregate information produced by the CMSB would shed additional light on the nature of systemic risks for the entire global financial system.

Cheap Hydrogen

A new process uses sunlight and a nanostructured catalyst to inexpensively and efficiently generate hydrogen for fuel.

Solar gases: A parabolic trough can focus sunlight on nanostructured titania, improving the efficiency of a new system for generating hydrogen by splitting water.
Credit: John Guerra, Nanoptek

, a startup based in Maynard, MA, has developed a new way to make hydrogen from water using solar energy. The company says that its process is cheap enough to compete with the cheapest approaches used now, which strip hydrogen from natural gas, and it has the further advantage of releasing no carbon dioxide.

Nanoptek, which has been developing the new technology in part with grants from NASA and the Department of Energy (DOE), recently completed its first venture-capital round, raising $4.7 million that it will use to install its first pilot plant. The technology uses titania, a cheap and abundant material, to capture energy from sunlight. The absorbed energy releases electrons, which split water to make hydrogen. Other researchers have used titania to split water in the past, but Nanoptek researchers found a way to modify titania to absorb more sunlight, which makes the process much cheaper and more efficient, says John Guerra, the company's founder and CEO.

Researchers have known since the 1970s that titania can catalyze reactions that split water. But while titania is a good material because it's cheap and doesn't degrade in water, it only absorbs ultraviolet light, which represents a small fraction of the energy in sunlight. Other researchers have tried to increase the amount of sunlight absorbed by pairing titania with dyes or dopants, but dyes aren't nearly as durable as titania, and dopants haven't produced efficient systems, says John Turner, who develops hydrogen generation technologies at the National Renewable Energy Laboratory (NREL), in Golden, CO.

Nanoptek's approach uses insights from the semiconductor industry to make titania absorb more sunlight. Guerra says that chip makers have long known that straining a material so that its atoms are slightly pressed together or pulled apart alters the material's electronic properties. He found that depositing a coating of titania on dome-like nanostructures caused the atoms to be pulled apart. "When you pull the atoms apart, less energy is required to knock the electrons out of orbit," he says. "That means you can use light with lower energy--which means visible light" rather than just ultraviolet light.

The strain on the atoms also affects the way that electrons move through the material. Too much strain, and the electrons tend to be reabsorbed by the material before they split water. Guerra says that the company has had to find a balance between absorbing more sunlight and allowing the electrons to move freely out of the material. Nanoptek has also developed cheaper ways to manufacture the nanostructured materials. Initially, the company used DVD manufacturing processes, but it has since moved on to a still-cheaper proprietary process.

NREL's John Turner says that Nanoptek's process is "very, very promising." And Harriet Kung, the acting director of the DOE's office of basic energy sciences, which has funded Nanoptek's work, says that the strained-titania approach is "one of the major exciting advances" since titania was first discovered to be a photocatalyst in the 1970s.

If it works as expected, the technology could help address one of the fundamental problems with using hydrogen as fuel. Hydrogen is attractive because it is light, and burning it only produces water. But today most hydrogen is made from natural gas, a process that releases considerable amounts of carbon dioxide. The other main option is electrolysis. But even if it's powered by clean energy, such as electricity from photovoltaics, electrolysis is inefficient and expensive. Guerra says using strained titania, and Nanoptek's inexpensive manufacturing process, makes the process cheap and efficient enough to compete with processes that create hydrogen from natural gas. What's more, Guerra says, the Nanoptek technology can be located closer to customers than large-scale natural-gas processes, which could significantly reduce transportation costs, thereby helping make the technology attractive. And if in the future carbon emissions are taxed or regulated, Nanoptek's carbon-free approach is another advantage.

Turner says that in addition to making hydrogen for fuel-cell vehicles, Nanoptek's process--if it is indeed efficient and inexpensive, as the company claims--could also be important for large-scale solar electricity. If solar is ever to be a dominant source of power, finding ways of storing the energy for night use will be essential. And hydrogen, he says, could be a good way to store it.

Programming Advanced Materials

Ordered nano order: Sequences of DNA attached to gold nanoparticles (upper image) program the particles’ self-assembly into novel crystals (lower image). X-ray diffraction confirms the crystals--partly squashed by the electron microscopy that produced these images--to be perfect lattices of tens of thousands of particles.
Credit: Oleg Gang

In 1996, scientists at IBM and Northwestern University used single-stranded DNA as if it were molecular Velcro to program the self-assembly of nanoparticles into simple structures. The work helped launch the then-nascent nanotechnology field by suggesting the possibility of building novel materials from the bottom up. Twelve years later, researchers from Northwestern and Brookhaven National Laboratory report separately in the journal Nature that they have finally delivered on that promise, using DNA linkers to transform nanoparticles into perfect crystals containing up to one million particles.

"The crystal structures are deliberately designed," says Northwestern's Chad Mirkin, one of the materials scientists who pioneered DNA linking in the 1990s and a coauthor of one of today's reports. "This is a new way of making things."

Ohio State University physicist David Stroud calls the work "quite valuable." He predicts that the breakthrough will enable the assembly of new materials with novel optical, electronic, or magnetic properties that have, until now, existed only in the minds and models of materials scientists. "Even now I'm surprised they could do it," says Stroud.

To date, efforts at programmed nanoparticle self-assembly in three dimensions have produced mostly disordered clumps. These clumps can have value; indeed, Mirkin's startup company NanoSphere has used the technology to develop medical diagnostics that have gained approval from the Food and Drug Administration.

But more complex and exotic materials imagined by Stroud and others require ordered structures. The hang-up, says Stroud, is that nanoparticles are immense relative to the atoms that form most crystals. As a result, the nanoparticles move relatively slowly, especially with DNA strands attached. When cooled to allow the complementary strands of DNA to link up, the nanoparticles tend to get frozen into a disordered arrangement before they can find their way to the orderly lattice of a crystal.

The authors of the new reports--a team at Northwestern led by Mirkin and chemist George Schatz, and physicist Oleg Gang's team in Brookhaven National Laboratory's functional materials center, in Upton, NY--overcame the particles' sluggishness by using longer DNA strands that give the particles more flexibility during crystal formation. "Typically, we think that crystallinity requires very rigid structures, so one could imagine it's necessary to have a very rigid DNA shell on the particles to have good crystals," says Gang. "In reality, it's the opposite."

While the details of the Northwestern and Brookhaven systems differ, both pad out their DNA strands with sequences that act as spacers and flexors, in addition to complementary sequences on the DNA ends that bind particles together. The groups start by binding one of two types of DNA to gold nanoparticles. The DNA types are complementary to each other. These two pools of modified particles are then mixed and cooled. DNA strands with complementary DNA form a double helix, tying together their respective nanoparticles, while identical DNA strands act like springs to repel their respective particles. The spacers on each DNA strand, meanwhile, allow bound particles to twist and bend so each particle in the mix can bind the largest number of complementary particles.

The result is exactly what theory predicts: a crystal lattice in which each particle of one type is surrounded by eight of the others marking the corners of a cube. Mirkin's group further demonstrated that tweaking the temperature and DNA sequences could nudge the same mix of particles to form a distinct crystal structure in which each particle has 12 neighbors.

Mirkin says that he and his team are just getting started. "To me, it's really only the start rather than the ending," he says. Over the past three years, Mirkin's group has been demonstrating methods to place different DNA linkers on different faces of nonspherical particles, such as triangle-faced prisms and virus particles. That, he says, should enable programming of more complex materials with repeating patterns of three or more components. "The really intriguing possibility here is the ability to program the formation of any structure you want," says Mirkin.

Stroud says that the structures already produced will be useful as the DNA-programmed assembly is extended to particles other than gold. Applications could include photonic crystals, in which the precise periodicity of particles can tune the overall materials to manipulate specific wavelengths of light, and photovoltaics that capture a broader range of the solar spectrum.

The structures are highly porous--10 percent particles and DNA and 90 percent water. That could hinder applications in which water is undesirable. Drain out the water, and the crystals collapse. Gang says that one could stabilize the crystals by filling the lattice with a polymer, but he is also exploring alternate stabilization schemes that would preserve the lattice's open space.