The Great Recession made it very clear that financial institutions are linked through many parts of their balance sheets. This web of linkages between institutions played a key role in exposing banks and broker-dealers to fundamental risks, as well as propagating shocks that may have arisen only at only one institution.
Along these lines, work
by Rama Cont and Eric Schaaning focuses on understanding the amplification channel that affects bank balance sheets, when assets are liquidated in a fire sale by another institution. The focus of the paper is novel, as much research have examined the financial network amplification when either an individual firm has an asset de valuation shock or the same firm responds to a apparent (or potential) bank run. In this sense, Cont and Schaaning focus on the feedback effects that result from other institutions’ activities.
The work is particularly useful when we consider macroprudential policy. For example when thinking about designing central bank stress tests, policymakers should not only account for how banks respond to their own stress but further measure how institutions react to one another.
To measure these effects, the authors design a network model where institutions hold more liquid and less liquid assets. When an institution in the network suffers an asset devaluation shock, it forces the original bank to de-lever by selling either type of asset, to obey leverage limits. Due to this fire-sale, prices drop further, which then affects neighboring banks, causing them to also de-lever. This spiraling process, which is bounded in the model, generates an amplification mechanism that is both realistic and new to the literature.
While the research is groundbreaking, open questions remain. The authors mainly consider shocks to asset valuations (eg. market liquidity). What if there are shocks to the funding side of the balance sheet (eg. funding liquidity)? Furthermore, a key building block of the model is the sensitivity of asset valuations to fire-sales. Empirically, however, how can we properly measure this elasticity without more granular data?
The above paper uses a model-based approach to understand the balance sheet-based linkages of firms. Other research utilizes an empirically driven, returns-based approach. Gerard Hoberg presented work
with Kathleen Hanley that focuses on the key drivers of stock return covariances across financial institutions.
In particular, they examine systemic risks that are found in 10K filings. Using advanced textual analysis tools to parse these documents, the authors first identify 18 categories of risks and provide a category-related score for each financial institution. When a pair of institutions has elevated scores with respect to a category, this suggests that both firms are jointly exposed to a common underlying risk.
Hanley and Hoberg then examine whether the commonality of these risks, across firms, manifests itself into the covariance of stock returns, while accounting for other characteristics of these firms.
The research is particularly novel on two fronts: first, it is one of the first to parse and interpret the risks found in financial disclosure forms and second, it makes advancements in textual analysis to classify firm risks into multiple semantic categories. A key finding of the paper is that the percentage of return covariance, driven by common underlying risks, is both a significant and forward-looking risk measure for financial returns.
Through the exercise, other questions arise. For example, how can we interpret the output of the themes? Many of the themes are connected and are not orthogonal to one another. Also do these themes change in their interpretation over time by managers running these institutions? Interest rate risk, for example, might be interpreted differently today relative to the early 2000’s as the sensitivity of
fixed income markets to central bank policy has increased dramatically in recent years.
— By Ram Yamarthy