HOME Recent Events RMI in the News ALUMNI
  Issue 4 | Archive  

August 2010

Risk Management: Lessons from the Crisis

By Myron S Scholes, Nobel Laureate, Frank E. Buck Professor of Finance, Emeritus, Stanford University

In the aftermath of the 2008 financial crisis, regulators of the financial system are reemphasizing risk management to minimize risk and prevent future crises. However, risk management is not about risk minimization but rather optimization of risk and returns. Since making returns for risk are two sides of the same coin, financial institutions cannot achieve returns without taking risks. Hence, the level of risks to take is a business decision. There are seven components to risk management; the first four components 1) capital allocation models, 2) considerations for capital structure, 3) optimization tools, and 4) stress testing are a system of what financial institutions necessarily must do to optimize risk, while the last three, 5) feedback systems, 6) reporting systems and 7) firm reporting structure and compensation policies are feedback systems to monitor how well firm managers are doing it. All the components are interrelated and their developmental process is evolutionary as each component will be augmented over time. There is an eighth component - regulatory oversight - although it is an external monitoring system not within the control of the firm. The topics discussed in this article are capital allocation model and capital structure considerations.

The capital allocation model determines how much capital to allocate to a division or strategy. Only after capital is assigned to provide certain services can financial institutions know if they have made sufficient returns on their capital to continue with the allocation. In the past, financial institutions used to assign capital based on asset category, e.g. unrated paper, bonds and equities, and their net present value. This method satisfies investment needs but describes only the best case scenario as such models generally ignore uncertainty and risk over future cash flows and capital needs. With the advent of portfolio theory, capital began to be allocated with respect to risk using value-at-risk (VaR) models. While the models are dynamic, being under constant evolution from Bank of International Settlement (BIS), it assumes capital sharing through low correlations, which is inappropriate during shocks when capital is scarce as correlations among business entities approach one. Furthermore, with all financial institutions using the same VaR models, interdependences among bank actions would be created. VaR models are also one-period models being, where capital (putty) is committed to a position (clay) at the start of a period and the position is exited for capital (putty) only at the end of the period. The output of the model for one period is the input for the next period and the models assume no costs in switching from clay to putty. Such models assume no costs in switching from clay to putty when in reality, costs of both capital and liquidity increase during times of shocks. The cost increases coupled with losses from certain strategies due to the shocks in Period 1 may cause a lack of capital in Period 2 to invest in lucrative opportunities realized in Period 1. The one-period models will not capture such changes and when the models are used sequentially, the input for the next period would be based on the inaccurate output of the previous period. The changes in the opportunity set of the financial institutions, due to changes in correlation structure, will also not be captured by the models.

Since financial institutions are in the business of providing liquidity, the focus of capital allocation should be on survival so that the institutions have sufficient capital to take advantage of the high price of liquidity. To provide staying power during a crisis period, a financial institution would need to consider the possibility of needing to liquidate assets in the future to reduce risk as asset values fell and its risks increased given relatively unchanging liability values. One method would be to buy a put option. When uncertainty and the price of liquidity increases, and asset values fall, the value of the put option would increase, and assets would automatically be liquidated through exercise of the put option. The value (cost) of the put option is then the price of providing liquidity and should be added as a dynamic cushion to the static cushion of capital assigned to each risk strategy in addition to VaR capital or other capital allocations. Theoretically, its addition to capital protects the firm against liquidity shocks and is incorporated as part of the investment cost. However, there is difficulty in using this model, as the intermediary would need to articulate the states of the world when liquidity (protection) is needed before being able to price the put option. This is both diffuse and is computationally difficult. Furthermore, the protection may not be readily available for sale in the market and, if self generated, option dynamics do not handle jumps in underlying prices because of liquidity shocks.

The second method of allocating capital is to look at tail distributions and allocate capital to withstand "nearly all" chaotic events. However, this method has measurement issues as well, since using simple statistical inferences may give too much weight to events such as the recent financial crisis and analytical models which use normal or evolutionary distributions are inaccurate. Hence, it would be prudent to rely on more computationally intensive methods such as the bootstrapping, which weighs events in the empirical dataset according to their magnitude and likely occurrence, for example, the crisis of 2008 might occur once every thirty years while the Asia crisis might occur once every ten years. The empirical distribution is generated by picking outcomes from the total pool of possible events in the dataset. The level of stress capital is selected based on empirical distribution at a set probability level. Not only does the stress capital protect against shocks and provides staying power at times of shocks, but also the capital allocation for each business activity is static, and therefore, does not give excessive weights to recent events. The model is superior to arbitrarily allocating capital or to VaR, which ignores stress when correlations increase dramatically, and it also precludes financial institutions from increasing risk during low volatility years to "target volatility" to enhance returns.

In providing liquidity, financial institutions act as intermediaries by buying securities or providing services to make a market for their clients by buying and holding inventory until they are able to liquidate it (at expectation of profit) when buyers enter the market. Buyers and sellers are asynchronous in their sale and purchase timings. Hence, the business of financial institutions is about compressing time to reduce the impact of mismatches between buyers and ultimate sellers (or vice versa) and in most cases is closely related to proprietary trading. Like speculators, financial institutions are paid by the hedgers (producers) as they bear omega (I call this omega in contrast with alpha) risk - the opportunities forgone for holding inventory before selling it to the end-users - for the hedgers. Hedgers pay speculators for their services. And, therefore, in providing omega services, intermediaries react to opportunities to profit. (Alpha generates predicted returns). Financial institutions would only provide such services if they have a fixed point or valuation of underlying securities, adding to inventory as prices deviate further from this valuation given capital, timing of return to equilibrium and uncertainty. When shocks occur, however, both liquidity prices and valuations change simultaneously. Intermediaries no longer have a fixed point for, with shocks it takes time to reassess equilibrium valuations and determine what part of the price changes are changes in liquidity and what part are changes in valuations. They stop intermediating - time stops. The volatility is too great to make assessments in calendar time. And, when uncertainty is resolved, the new valuations will differ from the old valuations. Shocks (Heisenberg's uncertainty principle of finance) cause a stop in the provision of omega services since financial institutions will not intermediate or will intermediate at wide spreads until they gain confidence as to the new equilibrium values.

To optimize omega risk, financial institutions must make a capital structure decision on how much flexibility to incorporate into their operating finance policies. Prior to the crisis, financial institutions reduced their operating and financing flexibility relying on their own and government assertions such as the belief of macro economists that the "Great Moderation", and other evidence that previous shocks had muted effects and that risks were transferred spreading risks to reduce the import of a crisis. Risks had been tamed or controlled. With recent experience, and evidence mounting of a relatively steady reduction of global risk and the muted effects of shocks over the last fifty years, individuals, and financial institutions took on more risks and built in less and less flexibility into their operating and financing policies. This makes sense if the world was really tamer than previously. Option values fall when volatility falls. This encourages more risk taking. The Bayesian approach encourages investors to adjust priors based on information. Their views might encourage them to assume greater risk and not incur the cost of providing for flexibility. However, shocks have informational value, despite their rare occurrence, and this value increases if beliefs are such that market participants believe that we are in a harmonious low risk environment. Information theorists, however, garner most information from shocks. Both are incomplete descriptions of reality. A stress risk environment building in flexibility to weather shocks prevents underweighting past shocks and overweighting current shocks.

In a chaotic world, with infrequent shocks, our Bayesian assumptions, which cause us to overweight more recent experience to affirm our beliefs, lead us down the wrong path by allowing us to assume a more orderly world where risks are well understood and tamed than they might be. While regulators have promised a V-shaped recovery to old values and paths, that can't be correct. Risks have changed, cash flows have changed, and uncertainty has increased because of the advent of new financial regulations. Post-crisis valuations are different from pre-crisis valuations. Hence, the regulators' assumptions that the economy will mean-revert, given sufficient liquidity, is not economically correct. Without deep understanding or study, the speed with which regulators impose new regulations on the financial system also leads regulators to enact rules and regulations that may be far from the mark and impose needless costs and risks on the system that reduce omega services to the detriment of the economy. For instance, if regulators arbitrarily require twice the capital reserves compared to pre-crisis levels, returns from intermediation businesses will fall and system assynchronicity will increase ĘC markets will become more volatile. Prices will need to deviate further from equilibrium values for intermediaries to step in. I was hoping that the crisis might provide the impetus to find new understanding and more efficient ways to manage capital and to provide intermediation services. Most entities are searching for new methods using the components of the framework that we introduced above. New innovations and improvements in methods will occur as a result of the chaos. We will see a new order from the chaos.

Back To Newsletter

Published quarterly by Risk Management Institute, NUS
Editor: Ivy Wang (rmiwy@nus.edu.sg)