Abstract
In November 2011, the Financial Stability Board (FSB) and the International Monetary Fund (IMF) designated 29 "systemically important financial institutions" (SIFIs) whose potential failure posed a significant threat to global economic stability. This designation is based on three primary criteria: size, interconnectedness, and complexity. While size and interconnectedness are well-quantified, "complexity" has remained largely qualitative. This paper introduces a rigorous mathematical framework to measure the internal complexity of a financial institution by analyzing its control hierarchy as a network. Using a dataset of SIFI control structures from 2011 and 2013, we demonstrate that structural complexity is distinct from asset size and that firms have significantly streamlined their hierarchies in response to post-crisis regulatory pressures.
Introduction: The "Too Complex to Manage" Problem
The 2008 financial crisis highlighted a critical vulnerability in global finance: many institutions had grown so complex that neither their own management nor external regulators fully understood their internal operations. This "opacity" creates a coordination burden; if a firm’s subsidiaries are scattered across disparate jurisdictions and industries without a clear organizational logic, a localized liquidity crisis or operational failure can escalate into a global catastrophe before intervention is possible.
While regulators have long focused on "too big to fail," this research addresses the "too complex to manage" problem. By treating corporate hierarchies as directed trees, we can move beyond vague descriptions of complexity and provide a consistent metric for comparing disparate business models.
Methodology: Modeling Control Hierarchies
We represent each financial institution as a rooted directed tree. In this model:
- Nodes represent the parent company and its various subsidiaries.
- Edges represent a control relationship, defined as $\ge 51\%$ ownership and a board seat.
- The Root is the ultimate parent entity (the SIFI itself).
Measuring Complexity via "Perfect Tree Similarity"
To quantify complexity, we introduce the concept of a Perfect Tree. In an ideal, low-complexity structure, every subsidiary would belong to the same country or industry (Standard Industrial Classification, or SIC code) as its immediate parent. Such a structure minimizes the need for cross-border or cross-sectoral regulatory coordination.
To determine if a firm is truly "complex" or merely large, we use a bootstrapping approach. We take a firm's actual tree topology and randomly shuffle the country and industry labels 1,000 times. By comparing the actual firm’s alignment to this random distribution, we can distinguish between firms that are "organized" (closer to a perfect tree) and those that are "jumbled" (more complex than random chance would suggest).
Data and Analysis
The study utilizes an anonymized dataset provided by Kingland Systems, capturing two snapshots of the global financial landscape: May 26, 2011, and February 25, 2013. The sample includes 29 large institutions: 19 SIFIs, 5 non-SIFI banks, and 5 insurance companies. For each entity, the data includes the country of incorporation and SIC codes, which were paired with consolidated asset data from Bloomberg.
Key Findings
1. Complexity is Not Size
A central finding of this research is that a firm’s asset size is a poor predictor of its structural complexity. The rank correlation between asset size and "perfect tree" similarity was low (ranging from -0.32 to -0.36). Some of the world's largest SIFIs maintain relatively clean, organized hierarchies, while smaller insurance companies often exhibit high levels of "jumbled" complexity. This suggests that size-based regulations alone may fail to address the risks inherent in structurally opaque firms.
2. Bureaucratic Deepening
Between 2011 and 2013, the hierarchies of these institutions became "deeper." The number of levels in some corporate trees increased to as many as 20. This suggests that post-crisis acquisitions and reorganizations were often "grafted" onto existing structures, creating additional layers of bureaucracy even as firms attempted to stabilize.
3. Regulatory Impact and Improvement
Despite the increase in depth, firms became significantly more organized geographically. Between 2011 and 2013, 25 out of 29 firms moved closer to the "perfect tree" ideal for geography. In 2011, nine firms had structures that were statistically "jumbled"; by 2013, only one remained in that category. SIFIs, in particular, made the most progress in reducing geographical and industry complexity, likely as a strategic response to the increased capital requirements and supervisory scrutiny associated with their designation.
4. Sectoral Differences
Insurance companies were found to be consistently more complex than banks. Despite having fewer subsidiaries and smaller balance sheets on average, their structures were more likely to be "jumbled" across different industries and jurisdictions, presenting a unique challenge for systemic risk oversight.
Conclusion and Policy Implications
By quantifying the level of complexity within a firm, this approach provides a foundation for more nuanced SIFI designations. It allows regulators to identify firms that pose a risk not just because of their size, but because of their structural opacity.
However, a move toward "perfect trees" involves a trade-off. While simpler, country-aligned structures are easier to supervise, they may reduce the benefits of geographical diversification, potentially making firms more sensitive to local business cycles. Future research must determine the optimal balance between supervisory clarity and the risk-mitigation benefits of a diversified corporate structure.
Summary updated