Análisis sensorial

Bank regulations are expected by policy makers to be an effective tool in guaranteeing the soundness of the financial system (e.g. Dietsch and Lozano-Vivas 2000; Demirgüç- Kunt et al. 2008; Houston et al., 2011; Delis, 2012). By restricting bank activities from securities underwriting, real estate investment and insurance underwriting, bank regulations mitigate the potential conflicts of interest between these activities and their fundamental banking business (Barth et al., 2004; Demirgüç-Kunt et al., 2008; Barth et al., 2001; Pasiouras et al., 2009; Demirguc-Kunt et al., 2012). These can secure bank stability (Barth et al., 2004; Haw et al., 2010) and avoid market crises (Barth et al., 2004; Pasiouras et al., 2009). But their impact on banking performance is mixed (Beck et al.,

2006b; Pasiouras et al., 2009; Diamond and Rajan, 2001; Barth et al., 2002; VanHoose, 2007; Demirgüç-Kunt et al., 2008; Delis et al., 2011).

The objective of policy makers to minimize the costs associated with existing financial- markets safety nets (Rojas-Suarez, 2002) may come at a cost of bank performance (Diamond & Rajan, 2001; Barth et al., 2002; VanHoose, 2007; Demirgüç-Kunt et al., 2008). Particularly, strict restrictive capital requirements may improve bank soundness (Kim et al., 2005; Pasiouras et al., 2009), i.e., mitigating AP3 and AP2, but damage bank lending capabilities (Diamond and Rajan, 2001; VanHoose, 2007), i.e., stimulating AP1. Previous research focusing on the nonbanking sector has highlighted the costs of over- regulation in damaging managerial incentives to exert effort to search for profitable investment projects, and deliver performance (e.g. Chhaochharia and Grinstein, 2007; Bruno and Claessens, 2010; Zhang, 2007). Especially, regulations in the banking industry are complex, and the costs of understanding and then complying with these rules are extremely high (Marsh & Norman, 2015). Such high costs of compliance can damage bank competition advantages and thus lead to lower credit ratings (Pasiouras et al., 2006). Inflexible bank regulations can even stimulate more AP1 agency problems which lead to managerial opportunistic incentives to inappropriately assess operational risk and mislead outsiders via poor and distorted disclosure (Barakat and Hussainey, 2013). This can therefore cause adverse selection problems for CRAs to assess and evaluate bank performance for investors and ultimately enhance costs of borrowing and damage bank performance.

Overall, the objective of CRAs to provide investors with an adequate measurement of the risks involved in banks, may not be in line with policy makers’ objectives which are to minimize the costs associated with existing financial-markets safety nets (Rojas-Suarez,

an inflexible and strict bank regulation system, CRAs’ benefits in enhancing bank performance for shareholders can be offset, resulting in the mitigation ofAP3 and AP2 at the costs of AP1 and concerns of over-regulation (Marsh & Norman, 2015). Therefore, our third hypothesis is as follows:

Hypothesis 3: Credit rating agencies enhance bank performance less when bank

regulations are stricter in emerging markets.

2. 4 Data and Methodology

We focus our empirical analysis on all commercial and savings banks listed on stock exchanges from 11 South-East Asian countries including Australia, China, Hong Kong, India, Indonesia, Japan, Sri Lanka, Malaysia, Philippines, Singapore and Thailand, for the period from 2000 to 2012. We collect our bank data from BankScope. Credit Rating Scores are collected from Thomsen Reuters Eikon, and investor protection quality data from the World Bank database. Macroeconomic data are obtained from World Development Indicators. After removing observations with missing variables, we have an unbalanced panel dataset with 2,398 observations including 389 commercial and savings banks.

2. 4.1 Dependent Variable

The dependent variable in the empirical analyses is bank cost inefficiency. A key advantage of the cost efficiency index is that it considers worth, costs, or benefits of a bank at the same time (Shaban and James, 2017), and make the estimates less exposed to the influence of random events and measurement errors (Kumbhakar and Lovell, 2003). Previous studies estimating the efficiency of banks (Brissimis et al., 2008; Delis et al.,

2011) use the technical efficiency measurement. We use the cost efficiency measurement because it is a wider concept than technical efficiency, referring to both technical and allocative efficiency (Pasiouras et al. 2009). Empirically, when analyzing bank cost inefficiency, we opt for stochastic frontier analysis (SFA) rather than data envelop analysis (DEA), following Barth et al. (2002), Pasiouras et al. (2006) and Demirgüç-Kunt et al. (2008). The SFA method is better than the DEA approach because it simultaneously accounts for relevant inputs and outputs of a bank, as well as for differences in the input prices, which allows us to distinguish between inefficiency and other stochastic shocks in the estimation of efficiency scores (Pasiouras et al., 2009). The SFA approach, by incorporating both error and inefficiency in a composite error term, allows us to estimate a global frontier while accounting for cross- country differences (Aigner et al., 1977).

More specifically, the model for examining the cost efficiency frontier is as follows:

ln 𝐶_𝑖,𝑡 = 𝑓(𝑃_𝑖,𝑡𝑄_𝑖,𝑡𝑁_𝑖,𝑡𝑍_𝑖,𝑡) + 𝑣_𝑖,𝑡+ 𝑢_𝑖,𝑡 𝑖 = 1,2, … , 𝑁; 𝑡 = 1,2, … , 𝑇 (1)

Where 𝐶𝑖,𝑡 the total cost for bank 𝑖 at year 𝑡; 𝑃𝑖,𝑡 is a vector of inputs; 𝑄𝑖,𝑡 denotes a vector

of values of outputs, 𝑁_𝑖,𝑡 is a vector of fixed netputs while 𝑍_𝑖,𝑡 is a vector of control variable. The term 𝑉_𝑖,𝑡 is symmetric error and represents that management of a bank cannot deal with this random fluctuation. 𝑈𝑖,𝑡 captures the effects of inefficiency relative

to the stochastic cost frontier; it is assumed to be independently distributed on one-side, meaning that this effect has the potential to enhance the cost of banks over the best- practice level.

ln 𝐶𝑖,𝑡 = 𝛼0+ ∑ 𝛼𝑖 𝑖 ln 𝑃𝑖,𝑡+ ∑ 𝛽𝑖 𝑖 ln 𝑄𝑖,𝑡 + 1 2∑ ∑ 𝛼𝑖𝑗 𝑗 𝑖 ln 𝑃𝑖,𝑡ln 𝑃𝑗,𝑡 +1 2∑ ∑ 𝛽𝑖𝑗 𝑗 𝑖 ln 𝑄𝑖,𝑡ln 𝑄𝑗,𝑡+ ∑ ∑ 𝛿𝑖,𝑡 𝑗 𝑖 ln 𝑃𝑖,𝑡𝑄𝑗,𝑡+ ∑ 𝜑𝑖 𝑖 ln 𝑁𝑖,𝑡 +1 2∑ ∑ 𝜉𝑖,𝑗ln 𝑃𝑖,𝑡 𝑗 𝑖 ln 𝑁𝑗,𝑡+ ∑ ∑ 𝛿𝑖,𝑗 𝑗 𝑖 ln 𝑄𝑖,𝑡ln 𝑁𝑗,𝑡+ ∑ 𝜀𝑖 𝑖 ln 𝑍𝑖,𝑡 + 𝑣_𝑖,𝑡+ 𝑢_𝑖,𝑡 (2)

In terms of the cost frontier model, we not only impose the restrictions of standard linear homogeneity and symmetry, we also consider the time and country effects. As mentioned above, concerning the specification of the efficiency frontier, we decide the bank’s total cost (𝐶), which is calculated as a total expense (non-interest expenses plus interest expenses), as the dependent variable. Following Sealey and Lindley (1997), we choose two outputs, which include loans (net of provisions, 𝑄1) and other earning assets

(government securities, bonds, investment, CDs and T-bills, 𝑄2). Furthermore, consistent

with previous studies of bank efficiency, we select the following two inputs: price of labour (𝑃1), calculated as the ratio of personnel expense to total assets; price of financial

capital (𝑃2), calculated as total interest expense to total interest bearing borrowed funds.

It can be seen that equity is an alternative funding for a bank and has the potential of affecting the bank’s cost. Following Berger and Mester (1997), we use the equity of each bank in the model as a fixed netput (𝑁) to control for differences in risk preference. Analysing the efficiency frontier in a cross-country sample, it is crucial to control variables that can capture country-level heterogeneity so GDP per capita is chosen as an indicator of the dynamism of each economy.

In order to avoid heterogeneous error, we estimate the bank cost inefficiency (BCIE) separately in the advanced countries group and the developing countries group. Advanced

countries include Australia, Hong Kong China, Japan and Singapore, and others are in developing countries group. The inefficiency scores using the cost efficiency frontier model are summarised in Table 1 by country in Panel A and by year in Panel B accordingly. The full sample overall mean BCIE score equals 0.252, which means on average a bank in our sample needs to improve by 25.2% to achieve its full cost-efficiency. Banks in China are the best performers with inefficiency scores at about 0.201, in line with the results from Berger et al. (2009a). Banks in the Australia are the second best performers with scores around 0.212. Banks in Japan and Singapore have the largest cost inefficiency levels, with scores of 0.268 and 0.259 respectively, in line with the results from Drake et al. (2009). Japan and Singapore, which are well-developed, are prone to establishing investment banks to stimulate their economic evolution and thus their savings and commercial banks may not receive enough attention in terms of efficiency maximization. Our results show that Sri Lanka and India have inefficiency scores 0.231 and 0.251 respectively, in line with Sathye (2003). Compared with the results from Perera et al.’s (2007) study focussing on Sri Lankan and Indian banks before 2004, our results show there is significant bank efficiency improvement in these two countries after 2004.

Table 1, Panel B shows that banks on average have the worst performance in 2003 in South-East Asian countries during the period of 2000-2012, which is similar to the result of Thoraneenitiyan and Avkiran (2009). This reflects that Asian bank industries struggled to improve their low efficiency before the entry of foreign banks (Park and Weber 2006). Table 1 Panel B also shows that on average, banks in South-East Asian countries have better performance in the post- 2007-2008 financial crisis period than the pre financial

crisis period such as year 2006. This shows the impacts of enhanced attention on improving bank efficiency after the financial crisis in Asia (Campello et al. 2010)7.

Table 1A shows the Chi-square test for Inefficiency. The result indicates Inefficiency scores are homogeneous, as the P-values are higher than 5 percentage. This is because that when we estimate the bank inefficiency (IE) separately in the advanced countries group and the developing countries group.

Table 1 Bank cost inefficiency (BCIE) estimates Panel A By Country

N Mean Min Max

Advanced countries

Australia 121 0.212 0.092 0.391

Hong Kong China 148 0.220 0.054 0.552

Japan 111 0.268 0.029 0.817 Singapore 59 0.259 0.047 0.501 Developing countries India 531 0.251 0.055 0.707 Indonesia 437 0.249 0.055 0.667 China 464 0.201 0.062 0.613 Sri Lanka 31 0.231 0.092 0.325 Malaysia 63 0.257 0.060 0.455 Philippines 214 0.231 0.064 0.543 Thailand 222 0.255 0.052 0.682 Panel B By year 2000 50 0.200 0.060 0.407 2001 74 0.303 0.155 0.447 2002 80 0.306 0.135 0.409 2003 95 0.250 0.117 0.817 2004 130 0.213 0.092 0.469 2005 165 0.220 0.029 0.347 2006 196 0.251 0.103 0.467 2007 207 0.244 0.092 0.635 2008 223 0.239 0.070 0.411 2009 246 0.207 0.066 0.490 2010 273 0.198 0.061 0.278 2011 331 0.225 0.067 0.257 2012 328 0.230 0.073 0.291 Average 2398 0.252 0.029 0.817

7 _{Campello et al. (2010) indicated that firms planned deeper cuts in tech spending, employment, and capital}

Table 1A

Chi-square test for Inefficiency

Variables IE T-test 1.387 P-value 0.172 Advanced 439 Developing 1959 Number 2398

Notes:95% Conf. Interval