For all the author calculations within chapter 3, the top 20 US (and European in the case of CISS) banks was derived from the exchanged listed banks based on market capitalisation as of the 2nd January 2007. This is prior to the financial crisis which caused significant changes to the market capitalisation of the banks in the sample. The observation period is between 1988 and 2015.
In order to estimate CISS for both the US and Europe, a portfolio of indexes was created within Bloomberg Professional Service PORT function (and their historic simulation capability) applying a similar data and method to Hollo et al. (2012). The following data was all obtained from Bloomberg Professional Service:
• Bond Market data
– Realised volatility of the US and German 10-year benchmark government bond index. Germany was selected as this countries bond was used by Hollo et al. (2012)
– Yield spread between the above government bonds and the A-rated non- financial corporations within that country (10-year maturity)
– 10-year interest rate swap spreads bracket
• Equity Market data
– Realised volatility of the S&P500 and Euronext 100, non-financial sector stock market index
• Financial Intermediaries data
– Yield spread between A-rated financial and non-financial corporations (10-year maturity)
– Realised volatility of the idiosyncratic equity return of the US and European bank sector index over the respective market indexes (S&P500 and Euronext 100)
• Foreign Exchange Market data
– Realised volatility of the Euro exchange rate vis-`a-vis the US dollar
• Money Market data
– Realised volatility of the 3-month Euribor rate and 3-month Fed Funds rate
– Interest rate spread between 3-month Euribor and 3-month French T- bills rate (Europe CISS). France was selected as this countries T-bill was used by Hollo et al. (2012)
– Interest rate spread between 3-month Fed Funds rate and 3-month US Government T-bill rate (US CISS)
In order to estimate DIP (Figure 3.2), a similar methodology to X. Huang et al. (2009) was applied using MatLab15a code obtained from Bisias, Flood, Lo, and Valavanis (2012). The method to estimate DIP is two fold, firstly a probability of default is required and secondly a forward looking correlation metrics. For the probability of default data rather than using the method explained in X. Huang et al. (2009), this was obtained using Bloomberg Professional Services DRSK function. See Leeney (2015) for the methodology, this data has been used by a number of authors such as Cetina and Loudis (2016); Cetina, Paddrik, and Rajan (2017); Laurent, Sestier, and Thomas (2016); Nirei, Sushko, and Caballero (2016) inter alia. This default likelihood model is based on the Merton distance-to-default (DD) measure (Merton, 1974), along with additional economically and statistically relevant factors. To produce the forward looking correlation metrics X. Huang et al. (2009)’s method was applied using the geometric return for the top 20 US banks in Stata12. Then both were combined in MatLab15a to obtain DIP.
MES (in Figure 3.3) is defined as the average return of its equity (Ri) during the
worst 5% of days of an overall market return (Rm), where the market is proxied by
the S&P500 index. The following equation is applied:
M ESb = 1 N umberof days X T Rit
Where T is the system is in its 5% tail. Weekly geometric return of the banks as a portfolio and the S&P500 were used. Equation 3.8 was calculated using Stata12.
In estimations CoVaR (see Figure 3.3) weekly bank (as a portfolio) equity returns were used as well as the following US Country level data:
• 3 Month Repurchase Agreement Rate
• 3 Month Treasury Bill Rate
• Weekly return in the 10 year and 3 month Treasury Bill spread
• Weekly return in the Chicago Board Options Exchange SPX Volatility Index (VIX)
• Weekly return in the S&P500
Note all returns were geometric, all data was collected from Bloomberg Professional Service. CoVaR was calculated following the methodology discussed earlier (see equation 3.4.3) using the Econometrics Toolbox within MatLab15a.
Chapter 4
Banking Efficiency Determinants
Abstract
The aim of this paper is threefold, firstly to conduct an empirical literature review on the banking sector efficiency over the last two decades, thereby identifying banking level risk and regulatory variables used to assess cost efficiency. Secondly, apply Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis (SFA) to measure efficiency within the Basel jurisdictions banks. Thirdly, to investigate the cost efficiency of United States banking sector by employing System Generalised Methods of Moments (GMM) regression analysis on a panel data of 233 commercial banks over the period of 2000 to 2015. This paper found that: (i) within the GMM analysis econometric measures of efficiency provided more statistically significant regression models than when using accounting based measures of efficiency; (ii) credit and liquidity risk is negatively associated with cost efficiency; and that (iii) regulations designed to mitigate these risks also negatively affect efficiency.
JEL Classification: G21, D24, N20
4.1
Introduction
As financial institutions have changed over the decades from the traditional transformation business model to a more contemporary and diverse model the
comparison of productive performance has become more difficult. Further the regulation landscape across the banking sector has transformed at both transnational and domestic levels, changing the market structure via the likes of consolidation and opening new markets to foreign banks. This provided new challenges to academics and regulators, to answer what factors are key to growth and productivity of banks. It may not always be possible for a financial institution to ever become fully efficient, because several of the inputs may not be under full control of management. With special reference to the US banking sector1, this
study examines the determinants of cost and productivity efficiency2 among US
bank holding companies (BHCs). Applying the two most commonly used measures, the non-parametric, Data Envelopment Analysis (DEA) and the parametric, Stochastic Frontier Analysis (SFA).
The rest of this paper is organised as follows: Section 4.2 provides on overview of the theoretical concept of banking efficiency and covers a broad range of empirical findings. Section 4.3 outlines this paper’s research hypotheses derived from the gaps or inconclusive evidence highlighted in the empirical literature review. Section 4.4 contains two steps, first of all, it provides a discussion of the main parametric and non-parametric methods of calculating efficiency, as well as applying these approaches to the data. Then, step two discusses the generalized method of moments (GMM) regression methodology and the variables used to identify the determinants of bank efficiency. Section 4.5 discusses the main findings in the context of the US Banks. Finally, Section 4.6 summarises this paper’s findings.
1This is due to data availability amongst the full Basel jurisdictions sample, efficiency scores
were also calculated for Japanese, Indonesian and French Banks.
2These types of efficiency are chosen over profit efficiency due to the assumption that banks
need to enhance cost efficiency to survive. Further not all financial institutions types are motivated by profitability.