Determinants of profitability for US and EU commercial banks: a dynamic panel data analysis

(1)

Determinants of profitability for US and EU commercial banks: a dynamic

panel data analysis

MSc International Financial Management Thesis

**Mitchell Angenent *, supervised by Dr. M. Hernandez Tinoco**

Faculty of Economics and Business, University of Groningen, The Netherlands ARTICLE INFO ABSTRACT JEL classification: C23 E44 G21 G32 O57 Keywords: Bank profitability Internal determinants External determinants European Union United States Panel data GMM Word count: 10915 This thesis defines the determinants that influence the profitability of commercial banks in the European and American banking sector. Such a cross-country analysis also enables one to determine possible differences between both regions. The determinants are divided into three groups, which are: bank-specific, industry-specific, and macroeconomic. Using panel data of 210 commercial banks (103 from the 28 member states of the European and 107 from the United States) between 2011-2015, this thesis uses the GMM estimator in order to empirically test the dynamic model which often arises in such a study. The findings suggest that for the overall sample a commercial bank’s profitability is affected by the parameters of its: size, efficiency and the credit risks. By comparing the sample of the European Union and the United States it is presented that in the United States the size of the commercial bank and the inflation are important while in the European Union the probability is affected by the capital of the commercial bank. Both samples present evidence for the determinants of efficiency, credit risk, liquidity and concentration.

(2)

1. Introduction

Within the European Union, hereafter called EU, there have been several processes of de-regulation and innovation in the two decades prior to the recent financial crisis. This period led to an increasingly integrated and more competitive European banking market (Fiordelisi et al., 2010). On the other hand, the banking sector of the United States, hereafter called US, is often described as a sector with a lack of competion where a few commercial banks have high market power (Akins et al., 2016). Banks that have a certain degree of market power can increase its performance (profits) by raising prices without lowering their costs (Bikker and Bos, 2004). Competition is one of the various factors that can influence the overall profitability of commercial banks. Such factors are defined by academia as determinants. Most research conducted in this field define two main groups in which the influences can be placed: internal (bank-specific) and external (industry-specific or macroeconomic).

(4)

concentration. Evidence for concentration is also presented in the separate US sample. This sample also presents evidence for the macroeconomic variable inflation and the bank-specific variables of size, efficiency, credit risk, and liquidity.

The remainder of this thesis is structured as follows. The next chapter discusses the thorough literature review of notable empirical research in the field of determinants of bank profitability. Subsequently, chapter 3 presents the research objective of this study in the form of research questions and hypotheses. The data set, methodologies, and variables that are used to present empirical results are presented in chapter 4. Consequently, chapter 5 presents these results of the analyses of the full sample, as well one to compare the US and the EU sample. In addition, this chapter also reports on fixed effects. Finally, section 6 concludes the paper by summarizing the findings of this thesis. Moreover, it discusses the implication of management and presents limitations of the findings.

2. Literature Review

(5)

sectors and the role they play in the economy. The first notable difference between both sectors is that, on average, US banks have a higher profitability since the sub-prime crisis period (which is the time period of this study) (ECB, 2013). Amongst other factors, this is mainly achieved by lower loan losses. Another significant difference presented by the ECB (2013) is the difference in size between both sectors where the EU sector is larger. This is also presented by WSBI-ESBG (2015). Both argue that the EU sector is larger due to the differences of accounting standards in both regions (US GAAP vs. IFRS). Under US GAAP, the assets of banks are presented at a lower net value than IFRS. For example, the total assets of the eight biggest US commercial banks are worth over $10 trillion under US GAAP while this would have been around $16 under IFRS (ECB, 2013). Another reason concerning size is the increase of the shadow banking system. In such a system, financial firms other than commercial banks have been rapidly growing and have taken over the traditional banking system (Schildbach et al., 2013). The final notable aspect where the US and EU differ is competition within the sector. More information on this topic is provided in the industry-specific section of this literature review. 2.2 Determinants of Bank Profitability The determinants of the bank profitability are often divided into two groups. They are either: (1) internal (bank-specific) or (2) external (industry-specific and/or macroeconomic). The theoretical framework presents empirical research that is helpful in assessing the relationship between the groups of determinants and commercial bank profits. The research papers in this review vary in the datasets, environments and time periods used. Furthermore, there is a division between either a single-country analysis (Bourke, 1989) (Molyneux and Thornton, 1992) (Anthanasoglou et al., 2008) (Dietrich and Wanzenried, 2011) or a cross-country analysis (Anthanasoglou et al., 2006) (Dietrich and Wanzenried, 2014) (Petria et al., 2015) (Islam and Nishiyama, 2016). However, even though significant differences exist across these papers, there are common types of determinants identified. These are introduced and explained in the following sub-sections.

2.2.1. Internal (Bank-specific)

(6)

applicable, ratios. Mutual bank-specific variables that are used in previous literature are: (1) size, (2)

capital, (3) efficiency, (4) credit risk, and (5) liquidity and associated risks.

Anthanasoglou et al. (2006) find statistically significant evidence for a positive linear relationship

between size and profit for commercial banks from Southern-European countries. Such a linear relationship can be explained through the economies of scale theory. According to this theory there are differences in costs, product diversification, and risk diversification by the size of a commercial bank (Anthanasoglou et al., 2006). Thus, the larger the bank, the better it can deal with such factors and gain a higher profit. Petria et al. (2015) also provide statistical evidence for this positive relationship within 27 member states of the EU. Dietrich and Wanzenried (2011) study 372 Swiss commercial banks in the pre-crises period (1999-2006) and crisis years (2007-2009) and find evidence in line with the economies of scale theory. However, this is up until a certain point. Using dummy variables for small-, medium-, and large-sized banks the study presents empirical evidence that during the pre-crisis period larger banks were more profitable than medium-sized banks. This is because economies of scale enable banks to benefit from higher product and loan diversification possibilities in order to enhance performance. During the crisis years, the larger banks were less profitable than both small- and medium-sized firms. Dietrich and Wanzenried (2011) justify this inverse relationship by the reputational problems faced by mainly larger banks in Switzerland during the crisis. On the other hand, Anthanasoglou et al. (2008) find that the relationship between size and profit for Greek banks is not significant and not important as banks often try to grow faster at the expense of their profits or are focused on different goals (e.g. improving market share). Dietrich and Wanzenried (2014) and Islam and Nishiyama (2016) also find no statistical evidence for a linear or inverse-linear relationship between size and commercial bank profitability.

(7)

Nishiyama (2016). Dietrich and Wanzenried (2014) study 118 countries divided into low-, middle-, and high-income in the period of 1998-2012 and find partially significant results. The parameter of the overall sample of all countries and the sample of high-income countries is positive and significant. Dietrich and Wanzenried (2011), amongst others, find no significant impact of capital on profits before the crisis. This is negative and significant for the period during the crisis (e.g. less attractive investment opportunities). The previous studies that test the efficiency of a commercial bank use the ratio of its cost to income. Such a ratio looks at the operating costs of a bank (e.g. salaries of employees). Anthanasoglou et al. (2008) find evidence for a significant, negative relationship between efficiency and profitability as a higher cost to income ratio presents a less overall efficiency. A low ratio shows the improvement management of the bank’s operating costs which can lead to a less negative association as the efficiency is increased. Empirical evidence by Dietrich and Wanzenried (2011) (2014) and Petria et al. (2015) is in line with this relationship as all papers present significant and negative parameters for efficiency. Credit risk is defined as one of the most important risks associated with a bank and is defined as the uncertainty of future credit losses around their expected levels (Brown and Wang, 2002). Such future credit losses are often incurred by the inability of the debtors to meet their obligations to the bank. Consequently, Anthanasoglou et al. (2008) (2008) and Petria et al. (2015) find statistical significant evidence for a negative impact of bank profits. Islam and Nishiyama (2016) also find a negative relationship. However, this is statistically insignificant.

(8)

relationship between liquidity and profitability. However, this relationship is statistically insignificant. Their justification for such a result is that higher liquidity ratios could also lead to lower rates of returns. 2.2.2. External Industry-specific The industry-specific variable that is used by almost all notable research papers in the field of bank profitability determinants is market concentration within the banking sector. Most of the previous work that study market concentration as a variable base their expectations on the structure-conduct-performance hypothesis, hereafter called SCPH. The SCPH argues that the market concentration can lower the collusion costs between firms which, in turn, leads to higher profits for all market participants (Evanoff and Fortier, 1988). In other words, the more concentrated the market, the less competition. Such an environment enables the small number of banks in the industry to achieve a joint price-output configuration that approaches a monopoly and increases the profits (Anthanasoglou et al., 2006). Bourke (1989) and Molyneux and Thornton (1992) are one of the first studies that included the market concentration as a possible determinant of profitability. Both present evidence for a significant, positive association in line with the SCPH. The somewhat more recent empirical papers by Anthanasoglou et al. (2006) and Dietrich and Wanzenried (2011) also find a significant positive coefficient.

Contrary to the SCPH, both Dietrich and Wanzenried (2014) and Petria et al. (2015) find evidence for a negative relationship between market concentration and profitability. In this case, a higher concentration of the market could be the result of tough competitions within the industry and, therefore, create a negative association (Dietrich and Wanzenried, 2014). The coefficient of market concentration of Anthanasoglou et al. (2008) and Islam and Nishiyama (2016) is also negative. However, both present insignificant results. Studying this industry-specific variable could potentially contribute to explain possible differences between the US and EU banking sector. As explained, Europe’s banking sector is defined as highly competitive while the US is more concentrated.

Macroeconomic

(9)

most comprehensive measure of macroeconomic developments is the economic growth of a country or region (Rumler and Waschiczek, 2010). Dietrich and Wanzenried (2011) find evidence for a positive relationship between economic growth and profitability. During a period of economic growth, the demand for lending often increases which could lead to higher profits for commercial banks. Dietrich and Wanzenried (2014) confirm this relationship in middle- and high-income countries where the economic growth is also positive and significant. Petria et al. (2015) argue that the significant positive coefficient presented is, besides the increased number of loans granted, also caused by an increase in customer deposits. On the other hand, Islam and Nishiyama (2016) find evidence for an inverse relationship. This relationship is based on the reasoning that economic growth guarantees a country’s or region’s economic stability. In such stable environments, the overall business risks of commercial banks reduce and thus bank could charge less. This eventually leads to lower profits. Another notable macroeconomic variable is inflation. The effect of inflation on bank profits depends on whether expected inflation changes are properly anticipated. If so, banks could, for example, adjust their interest rates in order to increase revenues at a faster rate than their costs (Perry, 1992). Bourke (1989), Molyneux and Thornton (1992), Anthanasoglou et al. (2006) (2008), and Islam and Nishiyama (2016) find evidence that inflation has a positive, significant effect on bank profit. A positive, significant relationship is also presented by Dietrich and Wanzenried (2014) who argue such a relationship can also be caused by customers who fail to fully anticipate on the inflation changes. Petria et al. (2015) provide no significant evidence on the relationship between inflation and commercial bank profitability.

3. Research Questions and Hypotheses Development

3.1 Research Questions

(10)

(11)

those funds. It is not expected that the parameter of capital presents a negative association with profits in this thesis. As this thesis focuses on the more recent years in which the market started to recover, it is expected that such above-mentioned events did not occur. Therefore, it is hypothesized that: H2: The capital of a commercial bank is positively associated with its profits. Anthanasoglou et al. (2008), Dietrich and Wanzenried (2011) (2014), and Petria et al. (2015) test the efficiency of a commercial bank by studying its cost to income ratio. In general, the higher the costs the lower the profits. Therefore, it is straightforward that the expected sign presented is negative: H3: The efficiency of a commercial bank is negatively associated with its profits. Continuing to credit risk, a negative influence is also expected for this parameter. Anthanasoglou et al. (2006) argue that increased exposure to credit risk often results in a decreased profitability for commercial banks. This is caused by the increased probability that customers are not paying loans or other products of the commercial bank. Such a negative association is especially expected after the financial crisis as people are becoming more aware and careful about risks than before and during the crisis (Dongheon and Baeho, 2015): H4: The credit risk of a commercial bank is negatively associated with its profits.

(12)

For the industry-specific determinants market concentration, the results of previous work are varying. Bourke (1989), Molyneux and Thornton (1992), Anthanasoglou et al. (2006), and Dietrich and Wanzenried (2011) present evidence in line with the SCPH that the market concentration has a positive sign. As opposed to SCPH, Dietrich and Wanzenried (2014) and Petria et al. (2015) find evidence for a negative relationship between concentration and profitability. This negative sign is explained as, not concentration, but competition increases profitability within the industry. Therefore, based on previous work, the expected sign is unpredictable. Besides determining the overall determinants of bank profitability this article also aims to find differences between the EU and US. While it is difficult to determine expected differences between both regions for most variables, this is not the case for market concentration. As mentioned, the EU is seen as a more competitive environment while the US is more concentrated. This leads to the following hypotheses:

H6A: The market concentration of the banking sector is positively associated with the commercial bank’s profit.

H6B: The market concentration of the US banking sector is positively associated with the commercial bank’s profit.

H6C: The market concentration of the banking sector is negatively associated with the commercial bank’s profit.

H6D: The market concentration of the EU banking sector is negatively associated with the commercial bank’s profit.

(13)

Finally, all research papers in the literature review that use inflation as parameter present a positive relationship with the profits of commercial banks. Therefore, this thesis also expects a positive sign on the parameter of inflation:

H8: The inflation of the country or region is positively associated with the commercial bank’s profit.

4. Methodology, Data, and Sample Description

This section explains the applied methods that are used in order to answer the research questions and hypotheses. Furthermore, all variables that are used in the models are introduced and justified. Finally, the data and sample with regards to the studied banks are introduced as well as the reasoning on the choices made. 4.1 The Model In order to answer the research questions and to test the hypotheses, this study follows the approach by Anthanasoglou et al. (2008) and Dietrich and Wanzenried (2011) by applying the following linear equation (1): 𝛱_"# = 𝑐 + 𝛽₎𝑋_"#) + ),-+ 𝛽_.𝑋_"#. / . ,-+ 𝛽₀𝑋_0#0 1 0 ,-+ 𝜀_"#, 𝜀_"# = 𝜈_" + 𝜇_"# (1) 𝛱_"# is defined as the probability of bank 𝑖 at time 𝑡, with 𝑖 = 1, … , 𝑁, 𝑡 = 1, … , 𝑇, 𝑐 is defined as the constant term, 𝑋"#’s are all the explanatory variables of this study, and 𝜀"# is the disturbance which is calculated as 𝜈" (unobserved bank-specific effect) + 𝜇"# (idiosyncratic error). All the explanatory variables are grouped into: (1) bank-specific (𝑋_"#)), (2) industry-specific (𝑋_"#>_{), and (3) macroeconomic} (𝑋_"#0_).

(14)

𝛱_"# = 𝑐 + 𝛿𝛱_"#,#@- 𝛽₎𝑋_"#) + ),-+ 𝛽_.𝑋_"#. / . ,-+ 𝛽₀𝑋_0#0 1 0 ,-+ 𝜀_"#, (2) 𝛱_"#,#@- is defined as the one-period lagged profitability and 𝛿 is defined as the speed of adjustment to equilibrium. A value of 0 < 𝛿 < 1 implies that there is an indication that profits persist. However, they eventually return to their normal, average level. Furthermore, a value of 𝛿 » 0 indicates that the industry is significantly competitive as there is a high speed of adjustment. On the other hand, a value of 𝛿 » 1 indicates very slow adjustments and a less competitive structure. 4.1.1 Random Effects (RE) vs. Fixed Effects (FE) This study uses panel data which is a pooling of observations on a cross-section of commercial banks over several time periods (Baltagi, 2005). Most literature that study panel data and static relationships follow the common estimation models applicable. The two that are often used are the fixed effect and the random effect models. The fixed effects model focuses on a specific set of N firms (or commercial banks) where the inference is restricted to only their behaviour. It assumes that there is correlation to the independent variables. On the other hand, the random effects model draw N randomly from a large population (Baltagi, 2005). Here the assumption is that there is no correlation with the independent variables. In order to test if your data requires an analysis by using random effects of fixed effects one can run the Hausman test. The null hypothesis of this test is that model required is random. The alternative hypothesis is that the model requires the fixed effects (Torres-Reyna, 2007). 4.2.2. Generalized Method of Moments (GMM)

(15)

(16)

(17)

(18)

4.3.2. Sample This study is a panel data study which covers the period between 2011-2015. In order for a bank to be included in the sample it has to meet three criteria, which are: (1) the bank is located in either the US or the EU, (2) the bank is a commercial bank, and (3) the bank has to be ranked in the top 100 concerning total assets for at least one of the five studied years. This leads to an initial sample of 293 commercial banks of which 171 are from the EU and 122 from the US. However, a major disadvantage of Orbis Bank Scope is that is does not offer an extensive database on all the studied bank-specific variables. This means that for some banks the specific data on variables was not available. Therefore, these banks are excluded from the sample. The final dynamic panel data sample consists of 210 commercial banks of which 103 are from the EU and 107 are from the US. 4.3.3. Descriptive Statistics Table 1. Descriptive statics of the combined commercial banks sample. The table reports all the descriptive statistics of the dependent and independent variables that are used in the analysis. The variables that are denoted in percentages are divided by 100 to provide decimal results. Table 1 presents the descriptive statistics of the combined sample of both EU and US commercial banks. An initial look at the minimums and maximums show that there are significant differences in each of the bank-specific variables. This is caused by the differences between the EU and US sample which are presented in table 1 and 2. Some interesting facts of this table show that the commercial banks show a positive ROAA of 0.68%. Thus, on average the banks had a slightly positive return on the average total assets between 2011-2015. Furthermore, the total loans of the banks are a significant part of the total assets with around 60%.

Dependent variable: bank profitability Mean Std. Dev. Min Max

(19)

Table 2. Descriptive statistics of the US commercial banks. The table reports all the descriptive statistics of the dependent and independent variables that are used in the analysis. The variables that are denoted in percentages are divided by 100 to provide decimal results. Table 2 and 3 present the descriptive statistics of respectively the US and EU samples. Thoroughly analysing both tables present some notable differences for the variables. First of all, the US commercial banks generate more profit from their assets than their European counterparts with respectively around 1% vs. 0.21%. The size of banks is presented by looking at the natural logarithm of their total assets. Comparing both show that, on average, EU commercial banks have more assets than US banks. The equity to total assets which indicates the capitalization of commercial banks ranges significantly in both samples. Especially in Europe where is ranges from -12% to 70%. This also applies for the cost to income ratio. While the commercial banks in the US have more loans as a part of total assets they have less loan-loss provisions. Thus, in the US more customers of banks pay more on time as the loan-loss provisions are used as an allowance for uncollected payments. As explained in the literature review and methodology sections, the higher the HHI the higher the market concentration (or the lower the competition). The descriptive statistics show that the US banking sector is more concentrated with lower competition as the parameter is with 0.12 notable higher than in the EU. This supports the argumentation of Akins et al. (2016). As for the macroeconomic variables used in this study one could state that on average the GDP growth and inflation have a higher mean in the US.

(20)

Table 3. Descriptive statistics of the EU commercial banks. The table reports all the descriptive statistics of the dependent and independent variables that are used in the analysis. The variables that are denoted in percentages are divided by 100 to provide decimal results.

5. Results

This section of the paper presents the analysis of the study of commercial bank profitability and the key drivers in both the EU, US and a combined sample in the period of 2011-2015. The panel data analysis is conducted by using the xtabond2 program in STATA14. This program is especially useful when one has to implement the GMM to a panel data set. Before presenting the results of the combined sample and the separate US and EU sample, a short explanation is provided on fixed and random effects.

5.1. Hausman Test, Fixed Effects, and Random Effects

Even though this study uses the GMM estimator for the analysis of determinants, it is useful to include the random or fixed effects in order to show their shortcomings. As explained, in order to find if the panel data samples require fixed or random effects one should use the Hausman test (results are Appendix II). For each of the three studied samples one can reject the null hypothesis in favour of the alternative. Thus, the fixed effects model should be applied. Some statistically significant results are presented in each sample. However, as mentioned such results are not very robust. The fixed effects model can only correctly estimate the dynamic variable of lagged profits when there is no heterogeneity.

(21)

5.2 Combined Sample Results Table 4. Regression results from the Combined sample of the EU and the US The table reports the results from the GMM estimations of the effect of bank-, industry-specific, and macroeconomic determinants of bank profitability. The dependent variable is the return on average assets (ROAA). The full sample includes 210 banks. ***, **, and * indicate significance at the 1%, 5% and 10% respectively. The AB test AR(1) and AR(2) refer to the Arellano-Bond test (1991) that the average autocovariance in the residuals of order 1 and 2 is 0 (H0: no autocorrelation). The Sargan-Hansen test is the test used for identifying over-identifying restrictions in GMM dynamic model

estimation (H0: over-identification is valid).

Table 4 presents the results for the relationship between profitability and the parameters of the combined sample of EU and US by using the GMM estimator. This estimator is a very robust alternative for other analysis such as the fixed and random effects. Before the results are discussed, the results of the goodness of fit of the regression are examined. The Arellano-Bond test deals with autocorrelation. The equation of AR(1) indicates that there is a negative first-order autocorrelation as one can reject the null hypotheses of no autocorrelation. However, this does not imply inconsistent estimates within the model. According to Arellano and Bond (1991) the estimates are inconsistent when there is a second-order autocorrelation. As the p-value of AR(2) is 0.3880, one fails to reject the null hypothesis of no autocorrelation. The Sargan-Hansen test provides evidence that their might be over-identifying restrictions that are not valid as the p-value is 0.000. Normally, over-identification is caused by too many instruments. However, the test results in STATA explain that the test is not robust but that it is not weakened by many instruments. Roodman (2009) states that the Sargan-Hansen test should often not be taken for granted as it is prone to weakness. This test grows weaker as the number of conditions and instruments applied in the regression increase. This makes it harder to satisfy all. Even though this weakness is identified, one has to be cautious with the results that are provided by the Sargan-Hansen test. Dependent variable: ROAA

Variable: Coefficient Std. error p -value

(22)

The coefficient of the lagged dependent variable (L.ROAA) has to most explanatory power in the model of the combined sample with a coefficient that equals 0.406. The high significance at the 1% level confirms that profit persistence should be taken into account when one is explaining the profits of commercial banks in the EU and the US. It should be mentioned that the variable of lagged profits does not predict and affect the ROAA. The coefficient is used in order to present that commercial banks are able to generate a positive profit this year as it also did so the year before. The results are in line with Anthanasoglou et al. (2008) who present a profit persistence coefficient of 0.26 which is also significant at the 1% level.

(23)

(24)

Table 5. Regression results from EU and US separately. The table reports the results from the GMM estimations of the effect of bank-, industry-specific, and macroeconomic determinants of bank profitability. The dependent variable is the return on average assets (ROAA). The results of the US are presented in column (1) and are based on a sample of 107 commercial banks. The results of the EU are presented in column (2) and are based on a sample of 103 commercial banks. ***, **, and * indicate significance at the 1%, 5% and 10% respectively. The AB test AR(1) and AR(2) refer to the Arellano-Bond test (1991) that the average autocovariance in the residuals of order 1 and 2 is 0 (H0: no autocorrelation). The Sargan-Hansen test is the test used for identifying over-identifying restrictions in GMM dynamic model estimation (H0: over-identification is valid). The parameter of size is negative but insignificant for the EU sample. Hence, one cannot present any evidence as the insignificant coefficient indicates that size does not affect commercial bank profitability in the EU. However, it is worth mentioning that the sign of the result is negative. The US also present a negative association of -0.002. However, this result is significant at the 1% level. As opposed to the hypothesis, the association is thus negative instead of positive. Even though Islam and Nishiyama (2016) presented no significant results on bank size they took a negative association into consideration beforehand. They argue that, once a commercial bank becomes extremely large, the operation efficiency might become inefficient due to bureaucratic complexity (Islam and Nishiyama, 2016). Dietrich and Wanzenried (2011) found significant evidence for the linear relationship between commercial bank profit and size but also argued a potential negative association. They argue that such an association could be caused due to various costs related to managing a large firm (e.g. agency costs) (Dietrich and Wanzenried, 2011). Thus, in the US the size of a commercial bank has a negative influence on its profits and there is no evidence for a linear relationship between both coefficients.

As hypothesized, the capital of EU commercial banks has a positive influence on its profits. The parameter of equity over total assets equals 0.045 and is significant at the 10% level. A positive association is also presented for US commercial banks; however, it is insignificant. Therefore, one cannot identify the capital as a determinant of profits in the US. This result is in line with Bourke

Dependent variable: ROAA US EU

(1) (2)

Variable: Coefficient Std. error p -value Coefficient Std. error p -value

L.ROAA -0.289 0.222 0.195 0.340*** 0.094 0.000 LN(total assets) -0.002*** 0.001 0.004 -0.000 0.000 0.706 Equity over total assets 0.015 0.069 0.823 0.045* 0.024 0.061 Costs over total income -0.044*** 0.010 0.000 -0.023*** 0.005 0.000 Loan loss provisions over total loans -0.669*** 0.217 0.002 -0.225*** 0.066 0.001 Total loans over total assets -0.041** 0.017 0.015 -0.047*** 0.014 0.001 HHI 0.799*** 0.210 0.000 0.065** 0.271 0.018 GDP growth 0.169 0.115 0.143 0.094 0.102 0.353 Inflation -0.059** 0.230 0.011 0.270 0.338 0.425 AB test AR(1) z = -2.45 - 0.014 z = -4.09 - 0.000 AB test AR(2) z = -1.15 - 0.250 z = -0.87 - 0.385

(25)

(1989), Molyneux and Thornton (1992), Anthanasoglou et al. (2006) (2008), Dietrich and Wanzenried (2014), Petria et al. (2015), and Islam and Nishiyama (2016). The effect of efficiency is negative and significant at the 1% level for both US and EU commercial banks. Hence, the parameter can be defined as a determinant of profitability in both regions. The coefficient of the cost income ratio in the US is with -0.044 slightly lower than in the EU where it is -0.023. Consequently, the negative impact on profitability is higher in the US. This result is somewhat surprising and unexpected as the descriptive statistics of both samples presented a, on average, higher percentage of costs to income in the EU (68.62%) than in the US (58.34%). One would expect that the higher the costs, the lower the profits. So, even though the ratio is higher in the EU than in the US the overall efficiency has a more significant impact on the profits of commercial banks in the US than in the EU.

Continuing to credit risk, the parameter of loan loss provisions to total loans is negative and significant for both the EU and US. Hence, one can reject the null hypothesis is favour of the alternative and state that credit risk is a determinant of commercial bank profit. Such empirical evidence is also provided by Anthanasoglou et al. (2006) (2008) and Petria et al. (2015). Furthermore, the result show that the negative impact of credit risk in the US is significantly larger than in the EU (-0.669 vs. -0.225).

Due to contrasting results of previous paper the impact of liquidity were predicted to be either positive of negative. The regression outcomes show that this is impact is both negative and statistically significant for the US and the EU. Consequently, contrasting to Islam and Nishiyama (2016) this study does present significance on the negative association which they could not. The negative association is explained as, even though high liquidity ratios reduce the risk, it also reduces the loanable funds of commercial banks and the overall earnings potential (Islam and Nishiyama, 2016).

(26)

(27)

this research paper was to determine the parameters that influence the profitability of commercial banks after the crisis. So far, most of the previous work has focused on the decades before the crisis and the crisis period of 2007-2009. To the knowledge of the author this study is also one of the first to, besides finding determinants of the overall sector, show differences in the determinants between the EU and the US. The literature review showed that the determinants of a commercial bank after divided into three groups, which are: bank-specific (internal), industry-specific (external), and macroeconomic (external). Using panel data between 2011-2015 this research paper the independent variables that are associated with the three groups for 210 commercial banks. Out of these 210 banks, 103 are from the 28 member states of the EU and 107 are from the US. This research papers follows the empirical model that is used by both Anthanasoglou et al. (2008) and Dietrich and Wanzenried (2011). As explained by Berger et al. (2000), there is a tendency of profit persistence within the commercial banking sector. Therefore, the overall equation tested is adjusted in order to fit the dynamic model that is created by that persistence. In a dynamic model, it is easy to produce various biases and inconsistent estimates when one deals with a small-time dimension and a large sample. Furthermore, there is possibility of have endogeneity and heterogeneity. In order to deal with such issues, the GMM estimator by Arellano and Bond (1991) is added to the empirical model

(28)

evidence for such a relationship. The inflation parameter within the combined sample is negative, which is a different result as the hypothesis expected. This negative result is in line with Dietrich and Wanzenried (2014) that commercial banks within the sample did not properly anticipate on inflation rate changes. Within the EU sample, the paper found significant associations for the parameter capital. As hypothesized, a positive association is presented. There are also parameters that present significant results in the EU and US sample. First of all, commercial bank efficiency. As explained, this is analysed by the cost to income ratio. Both sample present a negative association which was hypothesized. Results in line with the hypotheses are also presented for the parameters of credit, liquidity and market concentration. Management Implications The implications for the management of commercial banking is that this study presents determinants with a significant magnitude on profitability within the EU and the US. Regarding the bank-specific determinants, the management of EU commercial should be especially focusing on the variables of capital, efficiency, liquidity, and credit risk. On the other hand, US commercial banks’ management should carefully look at size, efficiency, liquidity and credit risk. Where the bank-specific determinants can be properly managed by management of commercial banks, this is not the case for industry and macroeconomic variables. Therefore, the implication for management is to focus on proper hedging tools against those types of risk. US banks should particularly pay attention to competition and inflation, while EU banks should only focus on competition.

Limitations

(29)

7. Reference List

Akins, B., Li, L., Ng, J., Rusticus, T. 2016. Bank competition and financial stability: evidence from the financial crisis. Journal of Financial and Quantitative Analysis, 51 (1): 1-28. Anthanasoglou, P., Delis, M., Staikouras, C. 2006. Determinants of bank profitability in the Southern Eastern European region. Bank of Greece Working Paper, 47: 1-35. Anthanasoglou, P., Brissimis, S., Delis, M. 2008. Bank-specific, industry-specific, and macroeconomic determinants of bank profitability. Journal of International Financial Markets, Institutions and

Money, 18 (2): 121-136.

Arellano, M. Bond, S. 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies, 58: 277-297.

Arellano, M. Bover, O. 1995. Another look at the instrumental variable estimation of error-component models. Journal of Econometrics, 68: 29-51. Baltagi, B. 2005. Econometrics analysis of panel data third edition. John Wiley and Sons Ltd. Berger, A., Bonime, S., Covitz, D., Hancock, D., 2000. Why are bank profits so persistent? The roles of product market competition, informational opacity, and regional/macroeconomic shocks. Journal of Banking and Finance, 24: 1203-1235.

Bikker, J., Bos, J. 2004. Trends in competition and profitability in the banking industry: a basic framework. DNB Working Paper, 18: 1-67.

Bourke, P. 1989. Concentration and other determinants of bank profitability in Europe, North America and Australia. Journal of Banking and Finance, 13 (1): 65-79. Brown, C., Wang, S. 2002. Credit risk: the case of First Interstate Bankcorp. International Review of Financial Analysis, 11: 229-248. Dietrich, A., Wanzenried, G. 2011. Determinants of bank profitability before and during the crisis: evidence from Switzerland. Journal of International Financial Markets, Institutions and Money, 21: 207-327. Dietrich, A., Wanzenried, G. The determinants of commercial banking profitability in low-, middle-, and high-income countries. The Quarterly Review of Economics and Finance, 54: 337-354. Dongheon, S., Baeho, K. 2015. Liquidity and credit risk before and after the global financial crisis: evidence from the Korean corporate bond market. Pacific-Basin Finance Journal, 33: 38-61.

ECB. 2004. Development in banks’ loan-loss provisions over recent years. ECB Economic and

Monetary Development Bulletin, 3: 37-38.

(30)

ECB. 2013. Structural characteristics of the Euro area and US banking sectors: key distinguishing features. ECB Banking Structures Report, 11: 1-39.

Evanoff, D., Fortier, D. 1988. Re-evaluation of the structure-conduct-performance paradigm in banking. Journal of Financial Services Research, 1: 277-294.

Fiordelisi, F., Marques-Ibanez, D., Molyneux, P. 2010. Efficiency and risk in European banking.

Working Paper Series ECB, 1211: 1-39. Hillier, D., Clacher, I., Jordan, B., Westerfield, R., Ross, S. 2014. Fundamental of corporate finance second revised edition. McGraw-Hill Education Europe. Islam, S., Nishiyama, S. 2016. The determinants of bank profitability: dynamic panel evidence from South Asian Countries. Journal of Applied Finance and Banking, 6 (3): 77-97. Mileva, E. 2007. Using Arellano-Bond dynamic panel GMM estimators in STATA: a tutorial. Fordham University Economics Department: 1-9.

Molyneux, P., Thornton, J. 1992. Determinants of European bank profitability: a note. Journal of

Banking and Finance, 16: 1173-1178. Perry, P. 1992. Do banks gain or lose from Inflation? Journal of Retail Banking, 14: 25-30 Petria, N., Capraru, B., Ihnatov, I. 2015. Determinants of banks’ profitability: evidence from EU 27 banking systems. Procedia Economics and Finance, 20: 518-524. Roodman, D. 2009. How to do xtabond2: an introduction to difference and system GMM in Stata. The Stata Journal, 9 (1): 86-136 Rumler, F., Waschiczek, W. 2010. The impact of economic factors on bank profits. Monetary Policy and The Economy, 4: 49-67. Schildbach, J., Wenzel, C., Speyer, B. 2013. Bank performance in the U.S. and Europe. Deutsche Bank Research, 9: 1-20.

Torres-Reyna, O. 2007. Panel data analysis: fixed and random effects using STATA. Princeton

(31)

Appendix I.

This appendix presents an overview of all the dependent and independent variables that are used in this thesis. First of all, in presents a description of all variables. Next, it shows the expected effect as well as the data source from which the applicable information is gathered. Table 6. Overview of the variables used and their description, expected sign, and data source.

Variables Description Expected sign Data source

Dependent variable: commercial bank profitability

Return on average total assets (ROAA) Net income over average total assets Orbis Bank Focus Independent variables

Internal (bank-specific)

Size Natural logaritm of total assets + Orbis Bank Focus

Capital Equity to total assets + Orbis Bank Focus

Efficiency Cost to income ratio - Orbis Bank Focus

Credit risk Loan-loss provisions over total loans - Orbis Bank Focus

Liquidity Loans to assets +/- Orbis Bank Focus

External (industry-specific)

Market concentration Herfindahl-Hirschman index +/- Andrew Chin calculator

External (macroeconomic)

(32)

Appendix II.

This appendix shows the fixed effect model applied to the three studied samples. Before, the fixed effects were analyzed one first had to use the Hausman tests. For each sample one can reject the null hypothesis in favour of the alternative. Thus, the fixed effects model should be used. Table 6. Overview of the fixed effects model results for the combined, US, and EU sample. The table reports the results from the fixed effects model estimations of the effect of bank-, industry-specific, and macroeconomic determinants of bank profitability. The dependent variable is the return on average assets (ROAA). The results of the combined are presented in column (1), the results of the US sample in column (2), and the results of the EU sample in column (3). ***, **, and * indicate significance at the 1%, 5% and 10% respectively. The Hausman test shows whether to apply fixed or random effects (H0: random effects). Dependent variable: ROAA Combined US EU (1) (2) (3)

Variable: Coefficient Std. error p -value Coefficient Std. error p -value Coefficient Std. error p -value

L.ROAA -0.054** 0.029 0.065 -0.300 0.065 0.000 0.000 0.033 0.992 LN(total assets) -0.000 0.002 0.756 0.000 0.002 0.948 -0.002 0.003 0.599 Equity over total assets 0.152*** 0.016 0.000 0.085 0.030 0.005 0.161*** 0.020 0.000 Costs over total income -0.020*** 0.002 0.000 -0.043 0.008 0.000 -0.019*** 0.002 0.000 Loan loss provisions over total loans -0.320*** 0.031 0.000 -0.172 0.084 0.042 -0.378*** 0.035 0.000 Total loans over total assets -0.000 0.005 0.966 0.007 0.007 0.312 -0.006 0.009 0.495 HHI 0.101 0.205 0.621 -0.127 0.716 0.859 0.607 0.416 0.145 GDP growth 0.022 0.051 0.667 -0.096 0.179 0.593 0.059 0.099 0.552 Inflation 0.564 0.165 0.733 0.140 0.501 0.780 0.250 0.304 0.411

Determinants of profitability for US and EU commercial banks: a dynamic panel data analysis