4. Data description
4.1 Data sources
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
10
scale prevails, meaning that the current production scale of the bank is optimal.
3.4 Cost efficiency
The ratio of the cost inefficiency for the i-th bank, relative to the potential cost, defined by the frontier function, given the output vector, xi, is used to define the cost efficiency of i-th bank:
CE𝑖 = 𝐶𝑖∗
𝐶𝑖 = exp (X′ijt𝛽)
exp (X′ijt𝛽+𝑢𝑖)= exp (−𝑢𝑖) (2) Where Ci* is minimum cost and Ci is actual cost. The equation 2 is a measurement of cost efficiency, of which the value is equal or less than one after the ratio is reciprocal. Finally, we will apply cost efficiency method to analyze the difference of cost efficiency in distinct regions.
4. Data description
4.1 Data sources
We first introduce how to construct the pivotal variable of the CSR index obtained from EIRIS Sustainability survey, spanning 2010-2014. The survey covers wide range of corporate social responsibility collected by questionnaire aiming at investigating the involvement of CSR for different firms. In EIRIS, the questions are customized to measure the risks and performances of companies in 41 sectors on 38 ESG ( Environment, Social and Governance) criteria based on international standards, which are further divided into 6 domains of corporate social responsibility, including environment, human rights, human resources, community involvement, business behavior, and corporate governance. The response rate of the questionnaire differs across banks and years. According to Brammer and Pavelin (2004) and Brammer et al. (2006), this survey by EIRIS offers the largest and most sound multidimensional social performance coverage. In the questionnaire, there are lots of questions covering from the field of subsidiaries to the proportion of stocks owned by the employees etc. We use the criteria in Wu and Shen (2013) to categorize those questions. For qualitative questions, some are two-scale “yes” or
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
11
Table 4.1 Bank distribution across countries and over time (Group 1)
“no”, while some are three scale (e.g, “many, some, or none”) and four to six scale choices. The more complicated thing is that some positive replies favor CSR,7 whereas others imply oppositely.8 For convenience, we refer to these two kinds of questions as positive and negative attitude questions, respectively. The transformation of text answer in the survey into a single aggregate value is simple by adding all the transformed number. Appendix 1 shows the detailed conversion of CSR scores.
After deleting incomplete data, we obtain 121 sample banks from 22 countries.
We next collect accounting statements for those banks from Orbis Bank database, originally Bankscope. All dollar-valued variables are measured by millions of US dollars and deflated by the consumer price indices, extracted from the World Bank, of individual countries with base year 2010.Because some banks fail to disclose the item of non-performing loan (NPL), we separate our environmental factors into two groups. One of them includes NPL along with other exogenous variables, called group1. The other group, called group 2, replaces NPL by loan to asset ratio, together with other exogenous variables. Group 1 has 114 sample banks from 19 countries with 423 bank-year observations, where most of the sample banks come from Asia (Japan), North America (the United States), and Europe. Group 2 has 121 banks from 22 countries with 457 bank-year observations and the additional observations are from Austria, Belgium, and Singapore. The distribution of banks across countries and over time is summarized in Tables 4.1-4.4
7 For example, the question ”How clear is the company’s commitment to community and charitable work?” has the following reply choices:
“Advanced”, ”Good”, ”Intermediate”, ”Basic”, ”Limited ”and “Little or no.”
8 For another example, the question “What is the level of potential exposure to bribery issues?” has the reply choices: “High”, “Medium”, and “Low.”
Country 2010 2011 2012 2013 2014 Total
Australia 4 6 6 6 6 28
Canada 0 6 6 6 5 23
Cyprus 0 1 1 0 0 2
Denmark 2 2 1 2 2 9
France 0 3 3 3 3 12
Greece 1 2 2 3 2 10
Hong kong 2 6 6 6 6 26
Israel 1 3 3 3 3 13
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
12
Table 4.2 Bank distribution across regions and over time (Group 1)
Table 4.3 Bank distribution across countries and over time (Group 2)
Japan 31 23 23 23 23 123
Netherlands 0 1 1 1 1 4
Norway 1 1 1 1 0 4
Portugal 1 0 0 1 1 3
South Korea 1 3 4 4 4 16
Spain 5 3 0 3 3 14
Sweden 4 4 4 4 4 20
Switzerland 2 2 2 2 2 10
Ukraine 0 1 1 0 0 2
United Kingdom 1 2 2 2 0 7
United States 21 18 20 19 19 97
Total 77 87 86 89 84 423
Region 2010 2011 2012 2013 2014 Total
Asia 34 32 33 33 33 165
Australia 4 6 6 6 6 28
British 1 2 2 2 0 7
Middle East Asia 1 3 3 3 3 13
North America 21 24 26 25 24 120
North Europe 7 7 6 7 6 33
South Europe 7 5 2 7 6 27
West Europe 2 7 7 6 6 28
Other 0 1 1 0 0 2
Total 77 87 86 89 84 423
Country 2010 2011 2012 2013 2014 Total
Australia 4 6 6 6 6 28
Austria 2 2 2 2 2 10
Belgium 0 0 0 1 0 1
Canada 0 6 6 6 5 23
Cyprus 0 1 1 0 0 2
Denmark 3 3 2 3 3 14
France 0 3 3 3 3 12
Greece 1 2 2 3 2 10
Hong kong 2 6 6 6 6 26
Israel 1 3 3 3 3 13
Japan 32 24 24 23 23 126
Netherlands 0 1 1 1 1 4
Norway 1 1 1 1 0 4
Portugal 1 0 0 1 1 3
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
13
Table 4.4 Bank distribution across regions and over time (Group 2)
As reported in Table 1 set, there are 114 different banks across 19 countries, with the majority coming from Asia (Japan), North America (United States) and Europe. While group two covers 121 distinct banks across 22 countries with extra sample banks from Austria, Belgium and Singapore. To sum up, there are 423 bank-year observations in group one and 457 bank-bank-year observations in group two.
Later, we will show the empirical result of these two groups and then to select a better outcome as our determined model.
Based on the intermediation approach, which regards a bank as a financial intermediary between depositors and borrowers and mirrors the familiar ”T”
account model of banking used to explain the process of money creation.
Accordingly, we identify three inputs and three outputs. The inputs categories are comprised of labor (X1), physical capital (X2), and borrowed funds (X3). Since the item of number of employees is missing for many banks, we instead use total assets net of fixed assets as the proxy. 9 Total loans (Y1) and investments (Y2) are two
9 As data on the number of employees are either missing or unavailable for many sample banks, the price of labor is defined as the ratio of personnel expenses to total assets. In other words, the item of
Singapore 1 1 1 1 1 5
South Korea 1 3 4 4 4 16
Spain 5 3 0 3 3 14
Sweden 4 4 4 4 4 20
Switzerland 4 4 4 4 4 20
Ukraine 0 1 1 0 0 2
United Kingdom 1 2 2 2 0 7
United States 21 18 20 19 19 97
Total 84 94 93 96 90 457
Region 2010 2011 2012 2013 2014 Total
Asia 36 34 35 34 34 173
Australia 4 6 6 6 6 28
British 1 2 2 2 0 7
Middle East Asia 1 3 3 3 3 13
North America 21 24 26 25 24 120
North Europe 8 8 7 8 7 38
South Europe 7 5 2 7 6 27
West Europe 6 11 11 11 10 49
Other 0 1 1 0 0 2
Total 84 94 93 96 90 457
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
14
traditional outputs. It is noteworthy that non-interest income (Y3) is gradually growing essentially nowadays as it reveals a bank’s degree of product diversification and also constitutes critical sources of revenue it generates from non-traditional activities. Spreading out risk and decreasing production cost through resource sharing are two advantages brought by non-interest income. Hence, we define non-interest income as a type of output. Table 4. 5 shows all variable definitions.
Table 4.5 Variable definitions
As for the environmental variables, we consider those factors that affect the efficiency of a bank exogenously and are not traditional inputs as stated above.
They are included to articulate the particular characteristics of each country’s financial industry, macroeconomic and regulatory conditions, as well as to reflect each bank’s traits. Unfortunately, there is no consensus on the selection of those variables and there are few theories for us to refer to when selecting environmental variables. Following Berger et al.(1993), Mester (1993), Allen and Rai (1996), Lozano-Vivas et al. (2001) and Lozano-Vivas et al.(2002), Huang et al. (2011), we
total assets is used as the proxy to the number of employees. Altunbas et al. (2000, 2001), Weil (2004), Fries and Taci (2005), and others utilize the same approach.
Variable name Description
Total loan (Y1) Short-term and long-term loans
Investments (Y2)
Other earning asset, loans and advances to banks, reverse repos and cash collateral, all securities, investment in property and insurance asset
Non-interest income (Y3)
Other operating income,
total non-interest operating income equity-accounted profit/ loss - operating Labor (X1) Total assets net of total fixed assets
Physical capital (X2) Total fixed assets (including property, plant and equipment)
Funds (X3) Deposits and short-term funding
Price of labor (W1) Total personnel expenses/ total assets Price of physical capital (W2) Other operating expenses/ total fixed assets Price of funds (W3) Total interest expenses/ total funds
Total cost (TC) W1X1+W2X2+W3X3
Cost share of labor (S1) W1X1/TC Cost share of capital (S2) W2X2/TC Cost share of funds (S3) W3X3/TC
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
15
relate the inefficiency term u to eight environmental variables to account for the impact of the exogenous discrepancy among countries and banks on technical inefficiency. Besides, as this study focuses the effect of CSR on cost efficiency, CSR and its squared term are two of the environmental variables. Following Barnett and Salomon (2006), the squared term of CSR is considered to capture the possible non-linear relationship between CSR and efficiency.
Environmental variables are summarized as follows.
1. Equity to asset ratio (ETA): The variable is referred to as an indicator of regulatory condition of a country’s banking industry. According to Hughes and Mester(1993), Mester (1996), Berger and Mester (1997), and Huang (2000), the variable is considered as fixed netput in either a cost or a profit function. Equity capital is also known as financial capital which offers a buffer against potential portfolio losses and acts as a substitute for deposits and borrowed money to fund loans. There are always two types of bank managers’ attitude toward risk, risk-averse and risk-neutral respectively. For risk-averse managers, they will try to mitigate the insolvency risk by enhancing the level of ETA. Meanwhile, financial soundness gets improved at the expense of compelling the chosen level of equity to deviate from the one required by cost minimization. On the other hand, a risk-neutral bank manager, running banks with lower ETA ratio, implying higher loan to equity leverage, is possibly more willing to implement policies aiming at the promotion of production efficiency and financial performance than a risk-averse manager. To sum up, since ETA can unveil the risk-preferences of bank managers, which may be correlated with production efficiency, we include it as one of the environmental variables. The expected sign of ETA is uncertain.
2. ROA (return on assets): ROA is defined as average return over assets and is used as an indicator of profitability, which may relate to the competitiveness in each banking industry. Previous works show the predicted relationship between ROE (return on equity) and efficiency is positive in a competitive scenario, i.e., the higher the profits, the higher the efficiency. However, since return on equity can be manipulated by managers by distorting the financial leverage. Hence, we decide to use ROA to measure the profitability. We expect the effect would be positive as ROE. This variable is also examined with the same orientation by Berger et al. (1993), Mester (1993) and Allen and Rai (1996).
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
16
3. Net interest margin: The net interest margin measures the difference between interest paid and interest received, divided by the amount of interest-generating assets. Although net interest margin somehow gauges the profitability of a bank, its association with efficiency remains ambiguous because of the broad product line that banks launch and provide nowadays. According to Lin et. Al (2012), banks that are diversified can lower down the fluctuation on net interest margin.
Yet, the lower volatility of net interest margin does not guarantee the higher income or higher efficiency. Lepetit et al. (2005), show that higher income from other activities is associated with lower lending rates, which suggests that banks may actually use loans as a sacrifice to replace with loanable funds.Thus; the effect of net interest margin on efficiency is in-determinant.
4. Loan/Asset: This variable enables to assess the attitude of bank’s strategy. The higher the degree of the ratio, the more loans are lent out financial risk, and consequently higher financial risk. However, higher level of loan to asset ratio may contribute to more profits for banks if non-performing loans remain stable.
So the correlation between loan to asset and cost efficiency remains blurring.
5. Real GDP growth rate: Real GDP growth rate is used to reflect the overall economic conditions. The expected correlation of this variable with bank costs is ambiguous since GDP interferes with the demand and supply factors of the banking production function at the same time (Perera, Skully, and Wickramanayake, 2007) and there is no guarantee that higher GDP growth rate result in higher demands for banking services. Hence, the sign of real GDP growth rate is not yet sure.
6. (log)Real GDP per capita: It’s defined as the ratio of a nation’s real GDP to its population, transformed by taking the natural logarithm. This macroeconomic indicator represents the proxy to the overall economic condition, affecting both demand and supply sides of banking activities, including deposits and loans, which may impact the efficiency of banks. When real GDP per capita increases, the demand for banking services boost and so does the supply of loanable funds fueled by savings. Interest rates will increase, leading to higher profits and cost efficiency for banks. Thus, this variable is expected to be positively related to efficiency.
Table 4. 6 shows the overall sample statistics for the cost function, while Table 4.7 represents the CSR score clusters of overall sample across regions in the sample
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
17
period. Table 4.9 shows that the mode of CSR scores is equal to 20 and the mean of CSR is 43.4. This implies that the CRS index distribution is skewed to the right.
The separate CSR indices for individual years are present in Appendix 2.1-2.5 Table 4.6 Sample statistics for the entire sample
Number of observations: 457
a Measured in millions of real US dollars with base year 2010 Table 4.7 CSR index cluster of overall samples
Table 4. 8 CSR index statistic description of overall samples
Variable name Mean Standard deviation
Total loansa 89249.2 1524642
Investmentsa 57773.5 372457
Non-interest incomea 643.453 6295.93
Price of labor 0.02126 0.16584
Price of physical capital 2.11774 13.3169
Price of funds 0.06209 0.59206
Total costsa 7186.17 132080
CSR Score 1-11 11-21 21-31 31-41 41-51 51-61 61-71 71-81 81-91 91-101
101-111 Total
Asia 2 60 41 14 16 11 21 8 0 0 0 173
Australia 0 0 1 2 8 1 1 9 3 3 0 28
British 0 0 0 0 0 0 0 0 2 4 1 7
Middle East
Asia 0 3 1 1 2 2 2 2 0 0 0 13
North
America 4 23 27 16 13 19 6 6 5 1 0 120
North
Europe 9 0 0 0 6 13 8 2 0 0 0 38
South
Europe 0 1 1 0 7 2 6 4 5 1 0 27
West
Europe 0 5 4 6 7 4 1 1 16 5 0 49
Other 0 0 0 0 0 0 0 2 0 0 0 2
Total 15 92 75 39 59 52 45 34 31 14 1 457
Mean Standard
Error Median Mode Standard
deviation Variance Skewness Min Max Number of
observations
43.40919 1.164332 43 20 24.89057 619.5405 -0.91045 0.398423 1 102 457
kurtosis
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
18