Bank Income Diversification

This thesis examines the impact of bank income diversification on bank performance and systemic stability. The research is driven by the post-2007 Global Financial Crisis debate on the implications of universal banking for financial stability. Many policymakers and economists called for a separation of commercial and investment banking activities, similar to the former Glass-Steagall Act in the United States (1934-1999). The argument is that risky investment banking activities of systemically important banks may impose negative externalities. While some jurisdictions adopted more restrained structural regulations inspired by the Glass-Steagall model, the debate about reinstating such separation or implementing similar specialized banking systems persists. The thesis, therefore, seeks to determine whether the historical American experience and similar specialized banking systems in other countries offer a valid model for improving financial stability in the wake of the global financial crisis and whether diversification strategies in banking positively or negatively affect this stability. The study's main hypothesis is that specialized banking regimes enhance financial stability by limiting the prevalence and severity of banking crises. This is tested using both qualitative and quantitative methods.

This section reviews the Glass-Steagall Act and similar specialized banking systems from a historical and macroeconomic perspective. The aim is to evaluate whether the introduction of these systems was primarily driven by financial stability concerns and whether they produced positive effects on financial stability. A historical analysis of ten countries that previously had specialized banking systems in place indicates that financial stability was not the main motivation for their adoption. The study further assesses the effects of such regimes on financial stability using a panel of 30 high-income jurisdictions from 1970 to 2011. The findings do not support the common belief that specialized banking systems reduce crisis probability; the evidence for a positive relationship is limited. However, the results do suggest that specialized banking systems experienced less severe banking crises compared to universal banking regimes. Specific examples are provided for the United States, Belgium, and the United Kingdom, revealing the diverse political and economic factors that shaped the adoption and eventual dismantling of specialized banking structures. The analysis highlights the complexities of attributing causality between banking regulations and financial stability outcomes.

This section explores theoretical models and existing empirical studies that address the relationship between bank income diversification and financial outcomes. It discusses the arguments put forth by Boot and Thakor (1997) suggesting that universal banking systems can be less efficient due to lower financial innovation compared to specialized systems. The section also acknowledges studies (Detragiache, 1998; Iftikhar, 2015) linking financial liberalization with increased financial fragility. A critical review of the literature reveals inconsistencies in empirical findings regarding the effects of bank income diversification. These inconsistencies stem from the use of varied methodological approaches and varying results from studies using similar methodologies. The research then prepares the groundwork for a more extensive and robust meta-analysis of the existing empirical literature on bank diversification, aiming to resolve the observed inconsistencies and offer a more definitive view on the subject.

This section investigates the historical motivations behind the establishment of specialized banking systems in various countries. The primary goal is to assess whether financial stability concerns were the primary driver for implementing these regulations. The study examines the historical context surrounding the adoption of such systems in ten specific countries: Australia, Belgium, Britain, Canada, France, Italy, Japan, New Zealand, Taiwan, and the United States. A qualitative historical analysis is performed to determine the relative importance of financial stability considerations compared to other political and economic factors. The findings challenge the prevailing notion that these regulations were primarily designed to bolster financial stability. The research delves into specific policy decisions, analyzing the interplay of political influence, economic conditions, and the interests of various banking stakeholders. The analysis considers the period between the Great Depression (post-1929) and the beginning of the era of financial liberalization (1970).

This subsection focuses specifically on the United States and the historical context of the Glass-Steagall Act of 1933. This legislation served as a crucial reference point for post-2007 discussions on separating investment and commercial banking. The study details the provisions of the Act, including its impact on commercial banks' involvement in corporate securities and the separation between deposit-taking and securities activities. Key features such as the establishment of the Federal Deposit Insurance Corporation (FDIC) and restrictions on interest payments on checking accounts are mentioned. While the Act is often cited as bolstering financial stability, this section critically evaluates that claim, referencing conflicting interpretations found in the literature. Authors such as Edwards (1942) and Benston (1990) provide contrasting perspectives on the Act's motivations, highlighting the influence of political factors and the self-interest of specific banking institutions. This section also addresses the role of the Rockefeller family and their National City Bank (later Citibank) in shaping the Act’s passage.

This part extends the analysis beyond the United States, exploring the experiences of other countries with specialized banking systems. Belgium is presented as a notable case, initially characterized by universal banking in the 19th century, but later adopting stricter separations in the wake of the 1929 crash, influenced by the US's actions. A similar pattern is shown to exist in other countries, and the analysis investigates whether the motivations behind adopting these systems in these countries were primarily for financial stability purposes. The research further examines the historical development of specialized banking in the United Kingdom, detailing the division of labor between clearing banks and merchant banks, and contrasting their roles. Additionally, the evolution of specialization in the French banking system is examined, acknowledging the complex interplay of tradition, the role of Paris as an international financial center, and the impact of political factors on regulatory choices. These comparative case studies contribute to a broader understanding of the factors influencing the adoption and effectiveness of specialized banking regulations across different contexts, showing that financial stability motives were not always the main drivers.

This section presents an empirical investigation into the relationship between specialized banking systems and financial crises. Using a comprehensive dataset of high-income countries (and non-sovereign jurisdictions) after 1929, the study examines whether these regimes had any empirical effects on financial stability. The main hypothesis tests whether specialized banking systems are associated with reduced crisis prevalence and severity. A macro-level analysis is conducted on a panel of 30 high-income countries between 1970 and 2011, focusing on the probability and severity of banking crises. The methodology carefully defines and measures crisis depth, duration, and severity using GDP per capita data from the Penn World Table. The analysis also incorporates control variables to account for factors such as inflation, economic openness (trade and financial openness), and the level of credit to the private sector. The research considers the influence of deposit insurance on the occurrence and severity of crises, comparing the study's findings with those of previous research by Demirgüç-Kunt and Detragiache (2002).

This section begins by outlining the existing empirical literature on bank income diversification, noting its heterogeneity and inconsistent findings. Following the 2007 Global Financial Crisis, the debate regarding the merits of restricting banks' securities activities intensified. While both proponents and opponents of such regulations claimed that the existing literature supported their positions, a closer examination reveals a lack of clear consensus. The diverse methodologies employed in existing studies contribute significantly to this heterogeneity, with varying results even among studies using similar approaches. This section sets the stage for the meta-analysis by highlighting the limitations and inconsistencies present in the existing body of research, thereby justifying the need for a more systematic and comprehensive review. The diverse range of dependent and independent variables used in previous studies is noted, adding to the complexity of drawing general conclusions.

The core methodology of this section is a meta-regression analysis, following the approach developed by Stanley and Jarrell (1989), Stanley (2008), and Stanley and Doucouliagos (2012). The analysis incorporates 932 regression results from 34 primary studies examining the effects of bank income diversification, specifically focusing on non-interest activities like fee and trading income. The study aims to provide more robust evidence on the effects of these income types and explain the heterogeneity observed in the literature. The data collection process involved searching databases such as Ideas, ScienceDirect, SSRN, and Google Scholar for relevant studies using keywords like “bank income diversification” and “universal banking.” The authors also reviewed references from initially identified studies and utilized the RePEc database to expand their sample. A critical aspect of this methodology was the construction of comparable effect variables across different studies, which involved calculating partial correlation coefficients to measure the relationship between diversification and various bank performance measures. The section highlights the challenges associated with constructing these comparable effect variables due to the diverse dependent variables used in the primary studies.

This section presents the results of the meta-regression analysis, focusing on the effects of bank income diversification on risk, profitability, and risk-adjusted profitability. The analysis examines the effects of fee income and trading income separately, recognizing the heterogeneity of existing research. The findings include an assessment of publication bias, using methods like funnel graphs and statistical tests to determine if the reported results are skewed due to publication bias. The study uses moderator variables to account for study quality, sample characteristics (such as focusing on large banks or specific regions), and methodological approaches used in the primary studies. The researchers used cluster-robust precision-weighted least squares to account for within-study data dependency. A general-to-specific model selection approach is adopted to determine which moderator variables are most influential in explaining the variations in observed effects. The findings are examined for consistency across different diversification and performance measures. The analysis of publication bias reveals the extent to which existing research might be biased and highlights the importance of correcting for this bias in drawing conclusions about the effects of bank diversification.

This part focuses on the sensitivity of the findings to different assumptions about the values of the moderator variables. A sensitivity analysis explores the robustness of the results by examining various combinations of moderator values. The authors compare best-practice research designs, where values of moderator variables are set based on what represents ideal research methodology, with other possible research designs. The analysis assesses the consistency of the results across different models and investigates the impact of individual moderator variables on the observed effects. The purpose of this section is to gauge the overall stability of the findings and to determine the extent to which the conclusions are dependent on the specific choices made in the model specification. The overall combined effects are calculated, considering the potential impact of different research designs and addressing the challenges of inconsistent findings in the literature. The aim is to determine if certain research designs lead to consistently positive or negative predictions of the effects of diversification.

This section details the selection criteria and data collection methods used in the study's analysis of large international banks. The study focuses on large banks with over USD 100 billion in total assets, chosen as a proxy for systemic importance given the variations in official definitions across countries. This criterion balances the need to focus on institutions most relevant to post-crisis regulatory debates with maintaining an adequate sample size. The resulting dataset comprises an unbalanced panel of ninety listed banks with 979 annual data points spanning from 2003 to 2015. The length of individual time series varies, ranging from four to thirteen years, which is further reduced due to using dynamic panel data analysis with incorporated lags to a total of 799 usable observations. The data used include income statement data on fee and trading income, complemented with investment banking deal volumes and assets under management data from annual reports and secondary sources. This dataset constitutes one of the most comprehensive panels of assets under management data in existing literature. The Bloomberg Default Risk indicator is used as a dependent variable, offering an advantage of not being solely based on return volatility and thus allowing to capture longitudinal changes in risk. This choice of sample and data collection methods distinguishes this study from much of the existing literature, enabling more robust analysis on the issues at hand.

This section discusses the measurement of bank income diversification, highlighting the limitations of existing approaches and proposing an alternative approach. The study critiques the common methods using banks' dependence on non-interest income as opposed to interest income, or those separating fee and trading income. It also addresses those methods that construct diversification indices by assuming equal shares across all types of income. The author argues that these approaches have disadvantages. Assuming an equal split of income types to define maximum diversification is considered arbitrary, especially when different income types are measured in different units. By using business volumes as proxies for related fee income, the author assumes fees from each business line are the product of a constant but unknown multiplier and the respective business volume, with a different multiplier for each income type. Herfindahl indices, commonly used, are therefore deemed to be unsuitable for measuring diversification. As a result of these considerations, an improved methodology for measuring diversification is proposed in the study, to enable more accurate and reliable analysis of its impact on other variables of interest.

The methodological choices in this section are driven by the need to address potential endogeneity in banks' diversification decisions. The study acknowledges that diversification can influence bank performance, but bank performance can also influence management decisions regarding diversification. The use of ordinary least squares estimators in the presence of endogeneity can lead to biased results due to dynamic panel bias (Nickell, 1981) and autocorrelation. To mitigate this, the study employs the system generalized method of moments (system GMM) by Arellano and Bond (1991), which allows using predetermined instruments and controls for endogeneity. Furthermore, the study employs principal component analysis (PCA) of the instrument matrix to reduce the number of instruments, improving efficiency and robustness. This PCA approach, as proposed by Bontempi and Mammi (2015), is chosen for its ability to utilize information from all available instruments while minimizing information loss. This refined methodology allows for a more accurate and robust assessment of the relationship between diversification and bank performance, going beyond limitations of previous research on this topic.

This section presents the empirical results of the analysis, focusing on the effects of diversification on profitability and default risk. The findings include results for both non-interest diversification and detailed diversification measures, using both standard system GMM and the variant with PCA scores. The results show significant positive diversification effects on profitability in most models, which are considered economically significant despite the small magnitude of the coefficients. Regarding default risk, measured by the Bloomberg Default Risk indicator, the findings suggest a significant negative effect of leverage, particularly in models including individual banking activities. This negative association between leverage and default risk, along with a positive relationship between leverage and profitability, is analyzed as potentially counterintuitive, proposing an explanation based on post-crisis regulatory constraints and risk-taking behavior of banks. The results are compared and contrasted with the findings from the earlier chapters, offering a nuanced understanding of the complex relationship between diversification and financial stability. This comprehensive approach allows for a thorough assessment of the impact of income diversification on large international banks.

This section synthesizes the findings from the three main parts of the thesis: the historical analysis of specialized banking systems, the meta-regression analysis of bank diversification literature, and the analysis of large international banks. The overarching question guiding the synthesis is whether the evidence supports the post-crisis policy debate's implicit assumption that specialized banking is superior to universal banking from a financial stability perspective. The main conclusion is that the overall evidence does not strongly support this hypothesis. However, the results are not entirely consistent across all three studies. Some findings do lend support to the idea that specialized banking systems might offer some advantages, such as evidence of more severe banking crises in universal banking systems and indications from the meta-regression analysis that trading income is associated with higher bank risk. Nevertheless, the evidence favoring specialized banking does not outweigh the other findings.

The study highlights key findings relevant for policy implications. The meta-regression analysis indicates that larger banks are more likely to increase risk-adjusted profitability when diversifying into fee and trading income. This finding contrasts with the results suggesting a potential risk increase from trading income in smaller banks or within specialized banking systems. The analysis of large international banks reveals risk-reducing effects from diversification, contrasting with findings about higher risk associated with trading income from the meta-analysis. On a macro level, the study suggests that universal banking systems may have a lower likelihood of banking crises. This finding further challenges the initial hypothesis that specialized banking inherently leads to greater financial stability. These seemingly conflicting findings call for a cautious interpretation and a nuanced approach to policy-making, considering factors like bank size and the nature of diversification activities.

This concluding section acknowledges that not all the results are perfectly consistent across the three studies which may result from the differences in methodologies and sample choices. While some results do favor specialized banking systems, the evidence is far from conclusive, and alternative interpretations are warranted. This underscores the need for further investigation into the complex relationship between bank structure, income diversification, and financial stability. The study's findings suggest that the simplistic notion of preferring specialized banking over universal banking solely for financial stability purposes may be oversimplified. The impact of specific diversification strategies and the role of bank size need more careful consideration. Future research should explore these issues in greater depth and refine the measures for diversification and bank performance. The study’s findings, though not uniformly conclusive, offer valuable insights into the complexities of banking regulation and financial stability. Further exploration of the potential moral hazard arising from deposit insurance in emerging economies is also flagged as an avenue for future research.