By Julia Schaumburg, Vrije Universiteit Amsterdam
Let us assess the breakdown risk of a building. Most of us would consider specific individual factors such as engineering statics and the age of the building, and others would also include environmental parameters such as the weather and the territory in the area where it stands. It is equally important, however, to identify the load-bearing walls of the building, and to investigate and monitor their condition within the system of the object’s other components. The breakdown of one load-bearing wall may bring down the others and therefore the entire building. Being blind to possible domino effects in the maintenance of buildings might lead to catastrophic consequences.
The bankruptcy of Lehman Brothers in September 2008 is like the breakdown of a load-bearing wall in a building that already had some cracks in its walls. The story began unfolding when the US subprime crisis forced the government and Federal Reserve to bail out government-sponsored mortgage firms Fannie Mae and Freddie Mac. But it was Lehman’s breakdown that marked the beginning of the global financial crisis featuring massive bailout costs across both the US and Europe, downturns in stock markets all over the world and economic recession in many countries. All risk models had underestimated Lehman’s interconnectedness, which made it a part of the financial system’s foundation.
A big lesson for us to learn is that we must quantify an entity’s potential to bring down the system. But the task of determining the degree of interconnectedness — a major determinant of systemic risk — is not a trivial one. The current approach involves official annual bank stress tests, which are based on exhaustive amounts of data from large banks. A problem arises because a large part of this data is confidential. Moreover, these stress test models are complex and running them more frequently is costly for both banks and regulators.
An alternative is to build a model that considers share prices, which are in the domain of public information and contain the market’s perception of a company’s state. There have been previous attempts to use share price information to quantify systemic risk, but our approach is suited to complement the official stress tests because we directly incorporate the estimated network of risk spillovers into the measure of systemic risk contributions through stress scenario analyses. For instance, in a given market environment if Deutsche Bank increases its risk exposure, how would the rest of the sector be affected? How large is the expected impact?
Market fundamentals and balance sheet characteristics are often used as conditioning variables in systemic risk models that use publicly available data. They comprise the environmental and the individual-specific variables, respectively. But their effectiveness in clarifying the contribution of financial companies to systemic risk is reduced when considering multiple interactions and stakeholders. In contrast, the innovative feature of our model is that Deutsche Bank’s risk may not only increase because of a change in its own balance sheet structure or the macroeconomic factors, but may also increase because of higher fragility of its competitors, with which it is interconnected via financial contracts and the interbank market. We estimate a time-varying downside risk measure, i.e., the potential loss on its share price that would only be exceeded with a small probability within the next week (or day, month, etc.). The available conditioning variables in our model include, among others, the downside risks of possibly all other firms. Unlike in the case of a building, where the location and neighbourhood of each component can be assessed unambiguously, estimating a network of risks requires a suitable statistical procedure. Fortunately, statistical regularization techniques exist, that are designed for many variables and allow a data-driven selection of the relevant risk drivers of each firm. By doing this for all the firms, we can estimate the complete network of risk spillovers. The resulting network-augmented downside risk measures are then used, in a second step, to estimate the impact and significance of each firm’s systemic risk contribution.
This model can be updated frequently, to allow for timely assessment of the financial system’s risk induced by the fragility of the main players. The evolution of the estimated risk contributions can be followed by regulation authorities on a weekly or monthly basis with minimal effort. Studying the network of influences through a thorough analysis of all major players provides relevant qualitative information on loss spillovers within the financial sector. It is an effective way to monitor the load-bearing walls, when the aim is to prevent the breakdown of a complex building.
Julia Schaumburg
Vrije Universiteit Amsterdam
www.atomiumculture.eu
Hay 0 Comentarios