7/6/2019
Course: MSc Thesis IFM (EBM022A20)
Assignment: MSc International Financial Management Thesis
Title: Differences in bank risk taking levels between developing
countries and developed countries and its determinants
Name: Michael Toxopeus
Student number: S3268543
Supervisor: Dr. L. Dam
Co-assessor: Dr. E. Karmaziene
Date of submission: June 7, 2019
Field key words: Bank risk taking, developing countries, emerging
countries, developed countries, Europe
Word count: 15,140 (including tables, footnotes and references,
excluding abstract and appendices)
Abstract
2
1.
Introduction
Academic research concerning the banking sector has always had a prominent role among scholars. However, with the recent global financial crisis, considered to be the largest since the 1930s Great Depression, this topic might have gained even more momentum. To give an indication of the magnitude of the European banking crisis, Black et al. (2016) find that the systematic risk of European banks reached an all-time high of €500 billion in 2011. Many academics suggest that the excessive risk taking levels of financial institutions, and banks specifically, drive such crises (e.g. Vazquez and Federico, 2015; Vallascas et al. 2017 and Barry et al. 2011). Additionally, many differences exist between developing countries and developed countries. As being part of a country’s economy, banks in developing countries and banks in developed countries have specific characteristics depending on their own two unique environments. These banks’ unique environments, along with their specific characteristics, could induce differences in the risk taking levels between these two types of banks. However, only a few academics have empirically investigated this research question. Thus, the goal of this study is to investigate whether there is a difference in the risk taking levels between banks in developing countries and banks in developed countries. Additionally, various country-level variables are employed to determine whether they are responsible for driving this difference.
3 higher, it is interesting to solve the ambiguity concerning this topic by filling this gap in the literature. Additionally, the first research question provides the opportunity to introduce a second research question. Given that a difference between the risk taking levels of banks in developing countries and banks in developed countries exists, it is interesting to examine which country-level variables drive this difference. Therefore, proxy variables for a country’s level of economic development and proxy variables for a country’s level of financial systems’ development are employed. Specifically, the inflation rate and the growth rate of GDP per capita proxy for a country’s level of economic development. Furthermore, GDP per capita and financial depth proxy for a country’s level of financial systems’ development.
The classification of the International Monetary Fund (IMF) is adopted for identifying the developing countries and developed countries. Furthermore, the Z-score is identified as the main proxy variable for capturing a bank’s risk taking level. Additionally, four proxies for a bank’s risk taking level are employed as robustness tests. Also, based on the literature review, various control variables are identified and consequently employed. Bank Focus is the data source for the bank-level variables, while World Bank is the data source for the country-level control variables. To answer both research questions, linear regressions are employed. Finally, several selection criteria are employed which ultimately results in unbalanced panel data of 939 banks in mainland Europe during 2011-2017. Specifically, 285 banks are based in developing countries, whereas 654 banks are based in developed countries.
4 Z-score, since only the leverage risk indicator and the asset portfolio risk indicator yield similar results. In contrast, the profitability indicator and the riskiness of a bank’s loans indicator do not support hypothesis 1. In fact, they show that banks in developed countries have higher risk taking levels than banks in developing countries. Furthermore, with the Z-score as the dependent variable, inflation, growth of GDP per capita and GDP per capita have a statistically significant effect. These country-level variables drive the difference between the risk taking level of banks in developing countries and banks in developed countries. Various robustness tests broadly corroborate these results. Thus, for a country’s level of economic development, inflation and growth of GDP per capita have a statistically significant effect. This supports hypothesis 2, since these proxies drive the difference between the risk taking level of banks in developing countries and banks in developed countries. Also, for a country’s level of financial systems’ development, only GDP per capita has a statistically significant effect. This partially supports hypothesis 3, since this proxy drives the difference between the risk taking level of banks in developing countries and banks in developed countries.
5 for scholars with respect to the difference in risk taking levels between banks in the two types of countries. Third, besides showing the existence of a difference in the risk taking levels between banks in developing countries and banks in developed countries, this study identifies several country-level variables which drive this difference. Specifically, this applies to a country’s level of economic development and a country’s level of financial systems’ development. Fourth, the time period of this study, 2011-2017, is fairly recent. Thus, it focuses primarily on the period after the end of the financial crisis. Consequently, it provides novel and up-to-date insights into different features concerning a bank’s risk taking level.
The rest of this paper proceeds as follows. Chapter 2 contains a literature review and serves as a background for establishing the hypotheses. Chapter 3 presents the hypotheses, along with an explanation of the research methodology. Chapter 4 gives an explanation of the collected data, supplemented with a description of the summary statistics and the correlation matrix. Chapter 5 presents the results, accompanied with a discussion in the context of existing literature. Chapter 6 concludes, accompanied with this studies’ (economic) implications and policy relevance. Finally, several limitations and directions for future research are discussed.
2.
Literature review and background
6 due to less effective market mechanisms, banks in developing countries are more inclined to moral hazard than banks in developed countries. Hrychiewicz (2014) also shows that public guaranteed bailouts are followed by increased levels of moral hazard. Consequently, although moral hazard has a different definition than risk taking, it can be argued that banks in developing countries have higher risk taking levels than banks in developed countries. In a similar vein, Kaminsky and Reinhart (1999) show that, historically, emerging countries have a larger probability of experiencing a crisis than developed countries. This could suggest that banks in emerging countries have higher risk taking levels than banks in developed countries. In summary, banks in developing countries are more inclined to moral hazard than banks in developed countries. Additionally, the probability of a crisis occurring in an emerging country is higher than in a developed country. Thus, both findings could be reflected in higher risk taking levels of banks in developing countries, compared to banks in developed countries.
7 owners than private banks. As put forth by Jensen and Meckling (1976), this higher effectiveness in separating managers from shareholders can create increased information asymmetry. In turn, this can cause differences in the risk taking levels between these two types of banks. Finally, privately-owned banks are commonly characterized by a less thorough separation between management and shareholders. Consequently, this allows shareholders to gain access to (private) information of a bank’s management more easily. Barry et al. (2011) show that when private owners (banking institutions, families or individuals) have a larger equity stake, a bank’s asset- and default risk are lowered. Although their sample consists of banks in Western-European (developed) countries, their reasoning - and results - about the effect of ownership structure on a bank’s risk taking level are relevant for this study. Especially since Garciá-Kuhnert et al. (2015) show that Eastern-European banks recently went through a significant process of much privatization and restructuring. Therefore, it is important to specify that many of this studies’ developing countries are based in Eastern-Europe. Consequently, many of these banks may have recently been privatized. In summary, combining the findings of Barry et al. (2011) and Garciá-Kuhnert et al. (2015) could result in lower levels of risk taking for banks in developing countries, compared to banks in developed countries.
8 exposure to credit risk of privatized banks in developing countries. Specifically, newly privatized banks are exposed to greater credit risk in the short-run. However, over a longer period of time, these privatized banks are exposed to lower credit risk. In summary, there has been a recent significant privatization process among banks in Eastern-Europe, accompanied with a substantial increase in credit provided to the private sector (Garciá-Kuhnert et al. 2015). Additionally, Boubraki et al. (2005) show that the extent of exposure to credit risk depends on different time-periods. Depending on the period of time (short-run or long-run) after the privatization process, these findings suggest either an increase- or decrease in the risk taking levels of banks in developing countries, compared to banks in developed countries.
9 a significant increase in the risk taking level of banks that were privatized in developing countries, compared to banks in developed countries. In summary, Mohsni and Otchere (2014) and Otchere (2009) both show that privatized banks in developing countries have higher risk taking levels than privatized banks in developed countries. Also, Garciá-Kuhnert
et al. (2015) show that substantial levels of privatized banks in Eastern-Europe exist. When
investigating all banks, combining both findings could result in the observation that banks in developing countries have higher risk taking levels than banks in developed countries.
Scholars also dedicated much research on the effect that a foreign-owned bank has on a bank’s risk taking levels, see e.g. Barry et al. (2011). They find research which shows that foreign-owned banks score better on performance benchmarks, especially in developing countries. Furthermore, Garciá-Kuhnert et al. (2015) find that, in their sample, approximately 56% of the shareholders of banks in Eastern-Europe have headquarters in a different country than where the bank itself is based. Additionally, they find these results to be similar to previous research. Chen et al. (2017) corroborate their findings and shows that during the preceding few decades a considerable increase in the number of foreign-owned banks in emerging countries took place. Specifically, the amount of foreign-owned banks increased by almost 75%, and their market share increased twofold.
10 reflect this. Second, increased moral hazard could arise, which is related to agency problems arising from the separation between management and shareholders. Specifically, this can arise when managers are allowed to keep profits in the subsidiary itself, but are also authorized to share losses with the consolidated company as a whole. In summary, there are contradictory arguments as to the effect of a foreign-owned bank on a bank’s risk taking level. Some argue that it has an increasing effect, whereas others argue that it has a decreasing effect.
Besides the theoretical ambiguity, the empirical research on the foreign-owned type of bank also provides inconclusive evidence. Demirgüc-Kunt et al. (1998) show that, in developing countries and developed countries, when a foreign bank enters a country, the probability that the host country experiences a banking crisis is decreased. This decreased probability of a banking crisis, when a foreign bank enters a host country, could be driven by this foreign-owned type of bank’s lower risk taking levels. In contrast, de Nicolò and Loukoianova (2007) show that foreign-owned banks actually have risk taking levels which are substantially higher than domestic, privately-owned banks. However, Chen et al. (2017) synthesize various empirical studies that demonstrate the opposite, thus concluding that the effect that a foreign bank entry has on a host country’s financial stability remains inconclusive. Ultimately, Chen
et al. (2017) show that, in emerging countries, foreign-owned banks negatively impact bank
11 Boubraki et al. (2005) further extend the topic of ownership and note that the financial sector is imperative to a country’s economic development. Given that the banking sector is a crucial part of the financial sector, the government has a vital role in its regulation- and monitoring process. Additionally, Boubraki et al. (2005) present research which shows that in developing countries government ownership of banks is particularly high. While investigating the same topic, Iannotta et al. (2013) show that government-owned banks have a lower default risk than privately-owned banks. They attribute this to the fact that these banks simply are government-owned, thus allowing them to access capital at lower costs while also providing them with protection from market discipline. However, Hrychiewicz (2014) opposes this point of view by synthesizing research which shows that government-owned banks take more risk than private banks. In fact, Hrychiewicz (2014) finds various empirical studies which show that a government-owned bank is less profitable and less efficient, and tries to compensate this by taking more risk. In summary, Boubraki et al. (2005) show that government ownership is particularly high in developing countries. However, because of contradictory evidence, it is difficult to conclude which effect a government-owned bank has on a bank’s risk taking level: some studies show an increase in risk taking, while others show a decrease in risk taking.
12 ambiguous. On the one hand, Iannotta’s et al. (2013) sample of European government-owned banks (primarily in developed countries) shows a lower probability of default risk. On the other hand however, Hrychiewicz (2014) presents several empirical studies which show that government-owned banks have higher risk taking levels than private banks. Furthermore, the recent financial (and European banking) crisis affected banks both in developing countries and developed countries. In summary, Hrychiewicz (2014) shows that in developing countries, governmental involvement in the event of a banking crisis is more common than in developed countries. Additionally, she shows that banks that dealt with governmental interventions increased their risk taking levels after the financial crisis. Combining these findings could result in the observation that banks in developing countries have higher risk taking levels than banks in developed countries. However, taking into account the contradictory results of Iannotta et al. (2013) and Hrychiewicz (2014) of the effect of a government-owned bank on a bank’s risk taking level, causes this prediction to be less certain. Ultimately, because of these contradictions, it is challenging to predict whether banks in developing countries have higher risk taking levels than banks in developed countries.
Williams (2014) investigates the relationship between bank risk and (national) governance in Asia. He argues that differences in the banking systems’ structure between developing countries and developed countries can be inherent to differences in the bank’s risk taking levels. Also, he explains that it is not clear whether his differences in bank’s risk taking levels between developing countries and developed countries are driven by the developed status of these countries or by their English legal origin system. He also shows that in countries that were affected by the Asian Financial Crisis, improvements in national governance do not reduce bank-risk in the short-run. However, in fact they do reduce bank-risk in the long-run.
13 Mohsni and Otchere (2014) and Hryckiewicz (2014) are one of the few exceptions. Mohsni and Othcere (2014) show that banks in developing countries have higher risk taking levels than banks in developed countries, as exemplified by their lower Z-scores. When separately including country-level and bank-level control variables, as well as combining all variables, they conclude the same. They attribute this difference in the degree of risk taking levels of banks in the two types of countries to the higher variability of ROA and the higher ROA of banks in developing countries, compared to banks in developed countries. However, in her global sample of banks that were bailed out by the government, Hryckiewicz (2014) provides contradictory results. With the majority of these banks based in developing countries, she shows that banks in developing countries have lower risk taking levels than banks in developed countries. Extensive monitoring by international organizations, which helped out these developing countries during financial crises, could drive these lower risk taking levels.
In summary, the empirical studies show contradictory results as to whether the risk taking level of banks in developing countries are higher or whether the risk taking level of banks in developed countries are higher. Thus, given the combination of the scarce empirical literature and their contradictory results, it remains challenging to predict which bank has higher risk taking levels: banks in developing countries, or banks in developed countries. However, it should be noted that Hryckiewicz’ (2014) sample only takes into account banks that were bailed out by the government.
3.
Hypotheses and research methodology
14 countries have higher risk taking levels than banks in developed countries. Additionally, two empirical studies examined the difference in the risk taking level between banks in developing countries and banks in developed countries in isolation. Whereas Mohsni and Othcere (2014) show that banks in developing countries have higher risk taking levels than banks in developed countries, Hryckiewicz (2014) finds the opposite. However, Hryckiewicz (2014) sample only takes into account banks that were bailed out by the government, whereas the sample of Mohsni and Otchere (2014) does not. In summary, the empirical literature on this topic is scarce, accompanied with the fact that the literature review presents various conflicting theories. Therefore, the goal of this study is to solve the ambiguity of this topic by empirically investigating the following research question: are there differences in the risk taking levels between banks in developing countries and banks in developed countries? Based on the literature review, hypothesis 1 is developed as follows: “The risk taking level of banks in developing countries is higher than the risk taking level of banks in developed countries.” Hypothesis 1 is formally expressed in regression equation (1) below:
Bank risk𝑖,𝑡−1 = 𝛽0+ 𝛽1 Developing country𝑖,𝑡+ 𝛽2 Size𝑖,𝑡−1+ 𝛽3 Liquidity𝑖,𝑡−1+ 𝛽4 Income diversification𝑖,𝑡−1+ 𝛽5 Managerial efficiency𝑖,𝑡−1+
𝛽6 Operating efficiency𝑖,𝑡−1+ 𝛽7 Inflation𝑖,𝑡 + 𝛽8 GDP growth per capita𝑖,𝑡+ 𝛽9 GDP per capita𝑖,𝑡+ 𝛽10 Financial depth𝑖,𝑡 + 𝜀𝑖,𝑡
15
are lagged. The following bank-level control variables are employed: Size (β2) is the natural
logarithm of total assets (in USD millions), Liquidity (β3) is the ratio of liquid assets to total assets and Income diversification (β4) is the ratio of non-interest income to operating revenue. Additionally, Managerial efficiency (β5)is the mean of the ratio of total operating expenses to total operating revenues, while Operating efficiency (β6) is the cost to income ratio. The following country-level control variables are employed: Inflation (β7) represents the change in inflation rate as reflected by the Consumer Price Index (CPI) and GDP growth per capita (β8) is the logarithm of GDP growth per capita. Additionally, GDP per capita (β9) is the logarithm of GDP per capita in current US dollars and Financial depth (β10) represents the share of domestic credit to the private sector as a percentage of GDP. Finally εi,t represents the error term. Table A1, in the Appendix, provides a detailed break-down of the variable definitions, supplemented with a description of the variable calculations and their data sources.
16
3.1. Variable description
The main dependent variable of interest for this paper is the risk taking level of banks, although it cannot be observed directly. However, the literature documents different proxies which adequately capture the risk taking level of banks. A thorough literature review shows that Z-scores are commonly employed as a proxy variable to capture risk taking in the banking sector (e.g. Barry et al. 2011; Chen et al. 2017; Garciá-Kuhnert et al. 2015 and Williams, 2014). The Z-score is calculated as follows: (return on assets + ratio of equity over total assets) divided by the standard deviation of the return on assets. The Z-score measures a bank’s default risk: it is the inverse probability that a bank defaults. A higher Z-score demonstrates that a bank’s stability is higher. Conversely, a lower Z-score demonstrates that a bank’s probability of default is higher. Specifically, the Z-score is the number of standard deviations below the average a bank’s return on assets must fall, before that bank’s capital reserves are depleted (Chen et al. 2017). To reduce the effect of outliers, and the skewness of the values, the natural logarithm of the Z-score is employed. This approach is also adopted by Chen et al. (2017) and Mohsni and Otchere (2014).
17 over a five-year rolling window period. They find no substantial differences: their main findings continued to remain statistically significant.
Although the literature primarily employs Z-scores as a proxy variable for a bank’s risk taking level, a synthesis of research shows that other proxies are employed for checking their results. Chen et al. (2017) decompose the components of the Z-score individually. As robustness tests, they employ (1) the return on assets (a profitability indicator), (2) the ratio of equity over total assets (a leverage risk indicator) and (3) the standard deviation of return on assets (an asset portfolio risk indicator). Barry et al. (2011) and Mohsni and Otchere (2014) both follow a relatively comparable approach. Besides the Z-score and its individual components, the literature provides various other indicators related to a bank’s risk taking level as robustness tests. Indicators related to the quality of a bank’s loans are frequently employed. In fact, Barry et al. (2011) employ loan loss provisions, while Hryckiewicz (2014) employs loan loss reserves and Chen et al. (2017) and Mohsni and Otchere (2014) employ non-performing loans. Based on a review of these studies, and the availability of data in Bank Focus, the loan loss reserves to gross customer loans and advances is employed as a final robustness test.
However, other decisive factors could potentially drive the difference in risk taking levels between banks in developing countries and banks in developed countries. To this end, the literature review provides several bank-level and country-level control variables. These control variables are employed to guarantee that the results of this study are not corrupted by these factors.
18 government in case of crises. Garciá-Kuhnert et al. (2015) employ the same variable, because Boyd and Runkle (1993) show that a bank’s size proxies for its capability to diversify risk. Second, Chen et al. (2017) employ the ratio of liquid assets to total assets to control for liquidity. Research shows that liquidity can have a two-sided effect on a bank’s risk taking level. Consequently, ambiguity among scholars remains as to the effect of liquidity on a bank’s risk taking level. On the one hand, a bank may employ its liquid assets to protect their loans from being affected in times of crises (Cornett et al. 2011). On the other hand, a bank may use its liquidity when expecting a higher volatility on its returns in the future (Alger and Alger, 1999). Third, although it is legitimate to assume that more diversification results in a reduction of risk and generates more stable revenues, several empirical studies contradict this assumption. Therefore, to control for a bank’s diversification of income, Chen et al. (2017) employ the ratio of non-interest income to operating revenues. Fourth, Barry et al. (2011) control for managerial differences in efficiency between banks, by employing the mean of the ratio of total operating expenses to total operating revenues. Fifth, Hryckiewicz (2014) notes that banks that are less efficient are prone to higher risk taking levels to compensate for their lower financial performance. In fact, she finds several studies which show that operating inefficiency increases a bank’s risk taking levels. Therefore, the cost to income ratio is employed to control for operating efficiency. Chen et al. (2017) also control for operating efficiency, since studies show that a higher operating inefficiency results in increased risk taking levels of banks.
19 is because Athanasoglou et al. (2008) show that the inflation rate is a factor in determining a bank’s profit. Thus, it could potentially affect a bank’s risk taking level. Furthermore, following several studies, Williams (2014) employs the growth rate of GDP per capita to control for a country’s economic cycle effects. Second, two country-level control variables which proxy for a country’s level of financial systems’ development are discussed. Vithessonthi (2014) argues that a country’s level of financial development may have an effect on the risk taking level of banks. This could create a difference between the risk taking levels of banks in developing countries and banks in developed countries. Thus, Williams (2014) employs GDP per capita as a control variable, because a synthesis of research shows that countries with higher GDP per capita have more developed financial systems. He notes that this control variable is essential to include, since a more developed financial system allows national banks to better manage different dimensions of a bank’s risk. Also, Kanagaretnam et
al. (2019) and Boubraki et al. (2005) control for GDP per capita. Finally, Chen et al. (2017)
country-20 level variables (except for the developing country dummy variable) are winsorized at their top and bottom 1% distribution levels. This is common in academic research concerning financial data and adheres to Garciá-Kuhnert et al. (2015) and Chen et al. (2017). Additionally, given that financial data often suffers from endogeneity problems, the endogeneity concerns are addressed by lagging the winsorized bank-level variables by one year. Not only the bank-level control variables are lagged (Chen et al. 2017), but also the dependent variables which proxy for a bank’s risk taking level are lagged. However, the regressions are also rerun, in which case the dependent variables are not lagged.
4.
Data and summary statistics
21 Since Bank Focus provides no data at all prior to 2011, the sample period is 2011-2017. This results in an initial sample size of 7,615 banks in both developing countries and developed countries. Furthermore, since the goal of this study is to investigate the difference in risk taking levels for all banks between two types of countries, no specific type of bank is excluded from the sample. However, various other sample selection criteria were in fact employed. First, to avoid sample selection (survivor) bias, both active banks and inactive banks are selected. Second, to avoid having the same bank in the dataset twice, only consolidated banks are taken into account. Therefore, two consolidation codes provided by Bank Focus are employed: C1 and C2. These codes integrate a bank’s controlled subsidiaries- and branches statements, thus resulting in their consolidated form. This selection criteria results in a substantial drop in the amount of banks, consequently leaving 1,408 banks in the sample. Third, to avoid that the results of this study are corrupted by differences in accounting standards, only banks that adhere to International Financial Reporting Standards (IFRS) are selected. Again, this results in a substantial drop in the amount of banks, consequently leaving 1,047 banks in the sample. Fourth, the selection of the loan loss reserves to gross customer loans and advances as a proxy variable for a bank’s risk taking level also results in a substantial drop in the amount of banks. Specifically, 108 banks are excluded from the sample, since they do not have this information available. After applying the discussed selection criteria, the final sample yields unbalanced panel data of 939 banks. Specifically, 285 banks are based in developing countries, whereas 654 banks are based in developed countries. Due to the application of these selection criteria, the sample does not consist of any banks based in Kosovo, Montenegro and Macedonia.
4.1. Summary statistics
22 summary statistics of both the entire sample and of the two types of countries, before winsorizing and before lagging, are shown in Tables A2 and A3 in the Appendix, respectively. Based on Table 2, the summary statistics of the two types of countries are discussed. As previously discussed, the Z-score proxies for a bank’s risk taking level. A higher Z-score demonstrates that a bank’s stability is higher. Conversely, a lower Z-score demonstrates that a bank’s probability of default is higher. The most striking observation is the difference between the two types of countries with respect to their Z-scores. The Z-score for banks in developed countries (3.928) is substantially higher than the Z-score for banks in developing countries (3.181). This statistic demonstrates a substantial difference in the risk taking level between banks in developing countries and banks in developed countries. The other proxies for a bank’s risk taking level provide a similar picture: banks in developing countries have higher risk taking levels than banks in developed countries. Specifically, banks in developed countries (0.438) score higher on the profitability indicator than banks in developing countries (0.315). Consequently, this demonstrates their ability to achieve a higher profitability. Furthermore, banks in developing countries have a higher leverage risk (13.719), a higher asset portfolio risk (1.586) and a higher riskiness of a bank’s loans (10.877), compared to banks in developed countries (10.453, 0.603 and 4.738, respectively).
23 keeping inflation low (1.274), having a higher GDP per capita (42,885) and demonstrating a substantially higher financial depth (100.753), compared to developing countries (6.067, 10,417 and 50.730, respectively). However, developing countries (2.126) are in fact superior in realising a higher GDP per capita growth compared to developed countries (0.814).
Variable Mean Std. Dev. Median Min Max N
Developing country 0.304 0.460 0 0 1 6573
Z-score 3.716 1.405 3.815 -0.132 6.911 3527 Profitability 0.402 2.258 0.515 -13.410 6.394 4994 Leverage Risk 11.412 9.132 9.382 -0.099 64.166 4994 Asset Portfolio Risk 0.891 1.964 0.241 0.008 14.077 3751 Riskiness Bank Loans 6.554 8.765 3.617 0.021 53.716 4904 Size (in USD millions) 9.137 2.061 9.106 4.512 14.189 4995 Liquidity 20.958 16.831 16.027 1.095 79.467 4994 Income Diversification 42.516 25.794 40.547 -32.457 113.726 4983 Managerial Efficiency 0.650 0.285 0.619 -0.256 2.063 4990 Operating Efficiency 65.297 27.112 61.898 9.569 203.978 4979 Inflation (CPI) 2.727 3.486 1.738 -1.125 15.534 6580 GDP Per Capita Growth 1.214 2.214 1.055 -3.638 7.924 6580 GDP Per Capita 33002 22383 31953 2640 104103 6580 Financial Depth 85.516 33.897 87.490 27.078 177.015 6570 Table 1: Summary statistics of all banks in developing countries and developed countries during 2011-2017, after winsorizing
24
Table 2: Summary statistics split for banks in developed country and banks in developing country observations during 2011-2017, after winsorizing
Developed Developing
Variable Mean
Std.
Dev. Median Min Max N Mean Std.
Dev. Median Min Max N
Developing 0 0 0 0 0 4578 1 0 1 1 1 1995 Z-score 3.928 1.338 3.986 -0.132 6.911 2525 3.181 1.426 3.177 -0.132 6.911 1002 Profitability 0.438 1.786 0.461 -13.410 6.394 3527 0.315 3.113 0.868 -13.410 6.394 1467 Leverage Risk 10.453 9.035 8.248 -0.099 64.166 3527 13.719 8.953 11.964 -0.099 64.166 1467 Asset Portfolio Risk 0.603 1.476 0.167 0.008 14.077 2651 1.586 2.687 0.600 0.008 14.077 1100 Riskiness Bank Loans 4.738 7.127 2.651 0.021 53.716 3456 10.887 10.598 7.621 0.021 53.716 1448
Size (in USD
millions) 9.608 2.016 9.605 4.512 14.189 3528 8.004 1.697 8.030 4.512 13.229 1467 Liquidity 20.142 17.277 14.424 1.095 79.467 3527 22.921 15.539 19.167 1.095 79.467 1467 Income Diversification 45.804 26.663 43.875 -32.457 113.726 3517 34.629 21.637 33.100 -32.457 113.726 1466 Managerial Efficiency 0.660 0.283 0.630 -0.256 2.063 3523 0.626 0.290 0.585 -0.256 2.063 1467 Operating Efficiency 66.271 26.668 63.029 9.569 203.978 3514 62.960 28.019 58.460 9.569 203.978 1465 Inflation (CPI) 1.274 1.132 1.113 -1.125 4.982 4578 6.067 4.597 6.472 -1.125 15.534 1995 GDP Per Capita Growth 0.814 1.818 0.841 -3.638 7.924 4578 2.126 2.715 2.378 -3.638 7.924 1995 GDP Per Capita 42885 19799 40875 13640 104103 4578 10417 3998 10743 2640 16007 1995 Financial Depth 100.753 28.981 95.251 40.830 177.015 4569 50.730 10.682 52.680 27.078 75.229 1995
25
4.2. Correlation matrix
Besides the summary statistics, reviewing the correlations between the individual independent variables is crucial, especially since linear regressions are performed in this study. One of the most important assumptions of this method is that there is no correlation between independent (control) variables. However, inevitably there is always some correlation between the variables in reality. This should not present a problem though, and this should not result in a loss of precision from the model. However, these assumptions only hold when the correlations between the variables are relatively low and do not exceed a pre-specified threshold. When these variables in fact exceed this pre-specified threshold, they are highly correlated. Consequently, this results in multicollinearity (Brooks, 2014). Ultimately, this could induce wrong conclusions based on the regression results. In academic research, a threshold which is commonly employed to classify high correlation between the variables, is above 0.8.
26 Variable Z-score Profit ability Levera ge Risk Asset Portfolio Risk Riskin ess Bank Loans Size (in USD millio ns) Liqui dity Income Diversi fication Mana gerial Effici ency Opera ting Effici ency Inflation (CPI) GDP Per Capit a Growt h GDP Per Capit a Finan cial Depth Z-score 1.000 Profitability 0.255 1.000 Leverage Risk 0.013 0.153 1.000 Asset Portfolio Risk -0.603 -0.124 0.330 1.000 Riskiness Bank Loans -0.410 -0.226 0.289 0.555 1.000
Size (in USD
millions) 0.198 -0.059 -0.438 -0.264 -0.298 1.000 Liquidity -0.143 0.063 0.060 0.130 0.181 -0.143 1.000 Income Diversification -0.067 0.083 0.147 0.106 0.022 -0.061 0.291 1.000 Managerial Efficiency -0.215 -0.410 0.039 0.147 0.098 -0.143 0.111 0.191 1.000 Operating Efficiency -0.225 -0.429 0.052 0.157 0.105 -0.144 0.115 0.203 0.995 1.000 Inflation (CPI) -0.228 -0.022 0.106 0.204 0.247 -0.246 0.121 -0.156 -0.103 -0.106 1.000 GDP Per Capita Growth 0.081 0.150 0.043 -0.042 -0.054 -0.087 -0.067 -0.036 0.022 0.020 -0.293 1.000 GDP Per Capita 0.209 0.028 -0.142 -0.195 -0.357 0.288 0.016 0.125 0.016 0.018 -0.370 -0.223 1.000 Financial Depth 0.151 -0.016 -0.100 -0.139 -0.205 0.271 -0.006 0.118 0.026 0.028 -0.383 -0.192 0.711 1.000
Table 3: Correlation matrix for all employed variables during 2011-2017, after winsorizing
27
5.
Results and analyses
29
Table 4: Regression results: differences in risk taking levels between banks in developing countries and banks in developed countries during 2011-2017
Variable (1) Z-score (2) Z-score (3) Z-score (4) Z-score
Constant 3.871*** (0.050) 4.511*** (0.243) 1.542* (1.074) 1.922** (1.028) Developing country -0.722*** (0.093) -0.744*** (0.097) -0.488*** (0.179) -0.459*** (0.172) Size (in USD millions) 0.035**
(0.021) 0.036* (0.022) Liquidity -0.006*** (0.002) -0.005** (0.002) Income Diversification -0.004*** (0.001) -0.003*** (0.001) Managerial Efficiency 3.453*** (0.873) 3.943*** (0.931) Operating Efficiency -0.045*** (0.009) -0.050*** (0.010) Inflation (CPI) 0.007 (0.016) -0.005 (0.015)
Growth of GDP per capita (log) -0.015
(0.078)
0.003 (0.077)
GDP per capita (log) 0.519**
(0.248) 0.590*** (0.242) Financial depth -0.000 (0.002) -0.001 (0.002) R-square (within) 0.0000 0.0195 0.0001 0.0160 R-square (between) 0.0665 0.1614 0.0868 0.1914 R-square (overall) 0.0463 0.1199 0.0367 0.1236 No. of observations 2,793 2,783 2,381 2,371 No. of groups 856 855 831 829
30 Four different proxy variables for a bank’s risk taking level are shown in Table 5. Models (1) – (6) are individual components of the Z-score and represent the profitability indicator, the leverage risk indicator and the asset portfolio risk indicator, respectively. Models (7) and (8) represent the riskiness of a bank’s loans. Identical models, but without lagging the dependent variables, are shown in Table A7 in the Appendix. Without the control variables, leverage risk (3.056), asset portfolio risk (0.941) and the riskiness of bank’s loans (5.514) are positive and statistically significant at the 1% level. The negative profitability coefficient (-0.156) is the only exception, although it is not statistically significant. By and large, these regression models provide similar evidence as the regression models based on the Z-score: banks in developing countries have higher risk taking levels than banks in developed countries.
31
Table 5: Robustness tests: differences in risk taking levels between banks in developing countries and banks in developed countries during 2011-2017
Variable (1) Profitability (2) Profitability (3) Leverage risk (4) Leverage Risk (5) Asset portfolio risk (6) Asset portfolio risk (7) Riskiness bank loans (8) Riskiness bank loans Constant 0.370*** (0.073) -3.743*** (1.212) 10.702*** (0.351) 12.574*** (5.011) 0.653*** (0.073) 5.027*** (1.351) 5.230*** (0.308) 63.014*** (5.185) Developing country -0.156 (0.134) 0.828*** (0.208) 3.056*** (0.643) 3.952*** (0.989) 0.941*** (0.134) 0.271 (0.233) 5.514*** (0.564) -2.536*** (0.973)
Size (in USD millions) -0.066*** (0.028) -3.371*** (0.135) -0.178*** (0.030) -0.895*** (0.133) Liquidity 0.010*** (0.003) -0.025*** (0.009) 0.002 (0.003) 0.037*** (0.010) Income diversification 0.007*** (0.002) 0.031*** (0.005) 0.002* (0.002) 0.015*** (0.006) Managerial Efficiency 15.754*** (1.007) -9.653*** (2.497) -6.787*** (0.919) -29.519*** (2.916) Operating Efficiency -0.194*** (0.011) 0.094*** (0.026) 0.080*** (0.010) 0.324*** (0.031) Inflation (CPI) -0.031** (0.016) -0.133*** (0.042) 0.025* (0.018) 0.109** (0.049) Growth of GDP per capita (log)
0.200*** (0.077) 0.103 (0.184) -0.005 (0.087) -0.009 (0.217) GDP per capita (log) 1.482*** (0.282) 5.902*** (1.116) -0.795*** (0.316) -11.417*** (1.170) Financial depth -0.000 (0.002) 0.022*** (0.008) -0.000 (0.002) -0.004 (0.009) R-square (within) 0.0000 0.1747 0.0000 0.2763 0.0000 0.0162 0.0000 0.0888 R-square (between) 0.0015 0.3668 0.0237 0.2015 0.0535 0.2521 0.0929 0.2295 R-square (overall) 0.0004 0.2830 0.0313 0.2014 0.0452 0.1799 0.0918 0.2025 No. of observations 4,222 3,209 4,222 3,209 2,948 2,405 4,150 3,156 No. of groups 917 889 917 889 869 844 916 886
32 Table 6 presents different regression models to determine which country-level variables drive the difference in a bank’s risk taking levels between the two types of countries. The inflation rate and growth of GDP per capita proxy for a country’s level of economic development. Furthermore, GDP per capita and financial depth proxy for a country’s level of financial systems’ development. Model (1) represents the main proxy variable for a bank’s risk taking level: the Z-score. Additionally, models (2) – (5) represent robustness tests with different proxy variables for a bank’s risk taking levels. Identical models, but without lagging the dependent variables, are shown in Table A8 in the Appendix. In model (1), with the Z-score as the dependent variable, inflation (-0.019), growth of GDP per capita (-0.170) and GDP per capita (0.637) have a statistically significant effect. Thus, these country-level variables drive the difference in the risk taking levels between banks in developing countries and banks in developed countries. However, financial depth does not have such an effect (0.000).
33
Table 6: Regression results: country-level drivers of differences in risk taking levels of banks in developing countries and banks in developed countries during 2011-2017
Variable (1) Z-score (2) Profitability (3) Leverage risk (4) Asset portfolio risk (5) Riskiness bank loans Constant 3.889*** (0.031) 0.457*** (0.042) 10.278*** (0.167) 0.582*** (0.041) 4.803*** (0.156) Developing country -0.575*** (0.073) 0.277*** (0.096) 4.107*** (0.378) 0.703*** (0.096) 5.539*** (0.353) Inflation (CPI) -0.019** (0.009) -0.076*** (0.012) -0.128*** (0.048) 0.035*** (0.011) 0.032 (0.044) Constant 3.910*** (0.033) 0.470*** (0.041) 10.567*** (0.179) 0.592*** (0.041) 4.984*** (0.168) Developing country -0.526*** (0.077) 0.097 (0.091) 2.742*** (0.393) 0.754*** (0.095) 4.324*** (0.367) Growth of GDP per capita (log) -0.170* (0.088) 0.126 (0.092) 1.069*** (0.399) 0.194* (0.112) 1.572*** (0.372) Constant 0.957 (0.645) -5.785*** (0.822) 19.765*** (3.254) 7.111*** (0.835) 55.149*** (2.945) Developing country -0.265** (0.107) 0.754*** (0.136) 2.162*** (0.537) -0.028 (0.139) -1.246** (0.487) GDP per capita (log) 0.637***
(0.141) 1.344*** (0.179) -2.092*** (0.708) -1.420*** (0.182) -10.956*** (0.640) Constant 3.855*** (0.111) 0.220 (0.147) 9.876*** (0.572) 1.012*** (0.147) 6.085*** (0.540) Developing country -0.664*** (0.077) -0.016 (0.102) 3.632*** (0.397) 0.699*** (0.101) 5.088*** (0.374) Financial depth 0.000 (0.001) 0.002 (0.001) 0.003 (0.006) -0.004*** (0.001) -0.013** (0.005)
34 5.1. Results in the context of existing literature
Although the regressions based on the Z-score provide full support for hypothesis 1, the robustness tests only partially support hypothesis 1. Nevertheless, these results are generally in line with the expectation based on the literature review and the subsequent hypotheses development section. Specifically, because the majority of the discussed theories in the literature review provide support for hypothesis 1: banks in developing countries have higher risk taking levels than banks in developed countries. However, the literature review also presents several theories which provide reasonable evidence of contradicting hypothesis 1. Specifically, the following theories may suggest that banks in developed countries have higher risk taking levels than banks in developing countries.
35 Third, Garciá-Kuhnert et al. (2015) show that a substantial part of the Eastern-European banks are foreign-owned (56%). Chen et al. (2011) corroborates this by showing that for the preceding few decades there has been a substantial increase in the presence, and market share, of owned banks in emerging countries. The following characteristics of these foreign-owned banks may result in their lower risk taking levels: superior access to international capital, more adequate screening of borrowers, and management with more expertise and sophistication. In fact, Demirgünc-Kunt et al. (1998) show that when a foreign-owned bank enters a host country, the probability of the occurrence of a banking crisis is lowered. The substantial presence of foreign-owned banks in Eastern-Europe, combined with the finding by Demirgünc-Kunt et al. (1998), could possibly cause that, in this study, banks in developed countries have higher risk taking levels than banks in developing countries.
36 banks in developing countries. A potential driver of this observation is the extensive monitoring of international organizations which provided assistance to these developing countries during financial crises. Thus, this explanation could also apply to this specific study.
Similar to the discussion of the results of hypothesis 1 in the context of existing literature, the same procedure is repeated for hypotheses 2 and 3. This study shows that the inflation rate has a statistically significant effect in explaining the difference in the risk taking level of banks in developing countries and banks in developed countries. However, previous studies provide inconclusive evidence about the effect of this proxy variable of a country’s level of economic development on a bank’s risk taking level. Specifically, Chen et al. (2017) do not find that the inflation rate has a statistically significant effect on a bank’s risk taking level. However, in several regression models, Williams (2014) does find such a statistically significant effect, whereas in several other regression models he does not.
37 variable of a country’s level of financial systems’ development in explaining the difference in the risk taking level of banks in developing countries and banks in developed countries.
6.
Conclusion
The goal of this study is to answer the following research question: is there a difference in the risk taking levels between banks in developing countries and banks in developed countries? Additionally, various country-level variables are employed to determine whether they are responsible for driving this difference. The Z-score is employed as a proxy variable to adequately capture a bank’s risk taking level. Additionally, various proxies for a bank’s risk taking level are employed as robustness tests. Besides a bank’s risk taking level, other factors could drive the difference in risk taking levels between banks in developing countries and banks in developed countries. Therefore, several bank-level and country-level variables are employed as control variables. Finally, several selection criteria are employed which ultimately results in unbalanced panel data of 939 banks in mainland Europe during 2011-2017. Specifically, 285 banks are based in developing countries, whereas 654 banks are based in developed countries.
38 tests only partially support hypothesis 1. Specifically, only two of the four proxy variables support hypothesis 1. In contrast, the other two proxy variables show that banks in developed countries have higher risk taking levels than banks in developing countries. Nevertheless, the literature review also provides various reasonable theories which suggest that banks in developed countries have higher risk taking levels than banks in developing countries. In summary, whereas the results based on the Z-score provide full support for hypothesis 1, the robustness tests only provide partial support for hypothesis 1.
39 study is in line with previous studies by showing that GDP per capita has a statistically significant effect. Finally, although this study finds that financial depth only exhibits a statistically significant effect in two situations, these results are partially inconsistent with previous studies.
6.1. Implications and policy relevance
Besides the contributions of this study to the literature, it also has several (economic) implications and relevance for policy makers and (financial) regulators. First, several scholars (e.g. Vazquez and Federico, 2015; Vallascas et al. 2017 and Barry et al. 2011) argue that excessive risk taking of financial institutions, and banks specifically, drove the recent financial crises. However, this study demonstrates that a difference exists in the risk taking levels between banks in developing countries and banks in developed countries. This is an important observation for policy makers and financial regulators: e.g. for individual countries’ Central Banks and supranational organizations such as the European Central Bank. Given the findings of this study, a distinction has to made with respect to supervisory measures for banks in the two types of countries. Attention has to be focused more closely on banks in developing countries compared to banks in developed countries, given their higher risk taking levels. Specifically, supervision has to be more tight and strict for banks in developing countries compared to banks in developed countries.
40 higher risk taking levels than banks in developed countries. When conducting business with these banks, these groups need to be aware of their higher risk taking levels and implement cautious strategies accordingly. Fourth, several country-level variables are identified which drive the difference in the risk taking levels between banks in developing countries and banks in developed countries. Specifically, this applies to the inflation rate, growth of GDP per capita and GDP per capita. Governmental policy makers and financial regulators can implement strategies, procedures and policies to strengthen these factors. This results in lowering the risk taking levels of banks, which contributes to a higher stability of a country’s economy and thus lowers the probability of the occurrence of a banking- or financial crisis.
6.2. Limitations and directions for future research
41 potentially decreased the robustness and consistency of the results. Fourth, although a great amount of attention has been paid to include various relevant control variables, the possibility of omitted variables in the regression still exists. As can be observed from the regression tables, there is in fact room to include more relevant bank-level and country-level control variables. This could provide more robust results, specifically by including additional control variables which could possibly drive the difference in the risk taking levels between banks in developing countries and banks in developed countries.
Several suggestions are made with respect to future research on this topic. First, while this study only takes into account countries in mainland Europe, it is interesting to examine the same research question by extending the sample to different regions or even to a global level. This is especially interesting to investigate since this study has a vast amount of banks based in Eastern-European developing countries. Thus, with the inclusion of more banks from developing countries from different regions, it interesting to test whether the results change or remain similar. Second, the inclusion of more relevant control variables in future research is important, since this provides more robust results. This holds especially for the country-level control variables, to examine whether they are responsible for driving the difference in risk taking levels between banks in developing countries and banks in developed countries. This is an essential recommendation, because it provides the opportunity to policy makers and financial regulators to develop more detailed and more specific policies to decrease the risk taking levels of banks. Consequently, this strengthens a country’s economic stability and thus lowers the probability of a crisis occurring.
43
Bibliography
Afonso, G., Santos, J., Traina, J., 2014. Do too-big-to-fail banks take on more risk? Federal reserve bank New York economic policy review 20, 1 – 18.
Alger, G., Alger, I., 1999. Liquid assets in banks: theory and practice. Boston college working papers in economics, no. 446.
Athanasoglou, 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, 121 – 136.
Barry, T., Lepetit, L., Tarazi, A., 2011. Ownership structure and risk in publicly held and privately owned banks. Journal of banking and finance 35, 1327 – 1340.
Black, L., Correa, R., Huang, X., Zhou, G., 2016. The systemic risk of European banks during the financial and sovereign debt crises. Journal of banking and finance 63, 107 – 125.
Boubakri, N., Cosset, J-C., Fischer, K., Guedhami, O., 2005. Privatization and bank performance in developing countries. Journal of banking and finance 29, 2015 – 2041.
Boyd, J.H., Runkle, D.E., 1993. Size and performance: testing the predictions of banking firms of theory. Journal of monetary economics 311, 47 – 67.
Boyd, J., de Nicolo, G., 2005. The theory of bank risk taking and competition revisited. Journal of finance 60, 1329 – 1343.
Brooks, C., 2014. Introductory econometrics for finance. Cambridge, United Kingdom: Cambridge university press.
44 Cornett, M.M., McNutt, J.J., Strahan, P.E. Tehranian, H., 2011. Liquidity risk management and credit supply in the financial crisis. Journal of financial economics 101, 297 – 312.
Dam, L., Koetter, M., 2012. Bank bailouts and moral hazard: evidence from Germany. The review of financial studies, 25, 2343 – 2380.
Demirgünc-Kunt, A., Levine, R., Min, H., 1998. Opening to foreign banks: issues of efficiency, stability, and growth. The implications of globalization of world financial markets. The bank of Korea, Korea, 83 – 105.
García-Kuhnert, Y., Marchica, M-T., Mura, R., 2015. Shareholder diversification and bank-risk taking. Journal of financial intermediation 24, 602 - 635.
Hoshi, T., Kashyap, A., 2010. Will the U.S. bank recapitalization succeed? Eight lessons from Japan. Journal of financial economics, 97, 398 – 417.
Hrychiewicz, A., 2014. What do we know about the impact of government interventions in the banking sector? An assessment of various bailout programs on bank behaviour. Journal of banking and finance 46, 240 – 265.
Iannotta, G., Nocera, G., Sironi, A., 2013. The impact of government ownership on bank risk. Journal of financial intermediation 22, 152 – 176.
International Monetary Fund, 2019. Retrieved from:
https://www.imf.org/external/datamapper/NGDP_RPCH@WEO/OEMDC/ADVEC/WEOWO RLD
45 Kaminsky, G.L., Reinhart, C.M., 1999. The twin crises: the causes of banking and balance-of-payments problems. American economic review 3, 473 – 500.
Kanagaretnam, K., Lobo, G., Wang, C., Whalen, D., 2019. Cross-country evidence on the relationship between societal trust and risk taking by banks. Journal of financial and quantitative analysis 54, 275 – 301.
La Porta, R., Lopez-de-Silanes, F., F., Shliefer, A., 2002. Government ownership of banks. Journal of finance 57, 265 – 301.
Laeven, L., Levine, R., 2009. Bank governance, regulation and risk taking. Journal of financial economics 93, 259 – 275.
Mohsni, S., Otchere, I., 2014. Risk taking behaviour of privatized banks. Journal of corporate Finance 29, 122 – 142.
de Nicolò, G., Loukoianova, E., 2007. Bank ownership, market structure and risk. IMF working paper series WP/07/215.
Nier, E., Baumann, U., 2006. Market discipline, disclosure and moral hazard in banking. Journal of financial intermediation 15, 333 – 362.
Otchere, I., 2009. Competitive and value effects of bank privatization in developed countries. Journal of banking and finance 33, 2373 – 2385.
Vallascas, F., Mollah, S., Keasey, K., 2017. Does the impact of board independence on large bank risks change after the global financial crisis? Journal of corporate finance 44, 149 – 166.
46 Vithessonthi, C., 2014. The effect of financial market development on bank risk: evidence from Southeast Asian countries. International review of financial analysis 35, 249 – 260.
47
Appendix
Table A1: Variable definitions, a description of the variable calculations and their data sources
Variable Description Data source
Dependent variable: Bank risk
Z-score
Based on a 3-year rolling window of the standard deviation of Return on Average Assets
Natural logarithm of the Z-score: (return on average assets + ratio of equity over total assets) / standard deviation of return on average assets
Bank Focus and own calculations
Robustness tests
Profitability indicator Return on average assets Bank Focus
and own calculations
Leverage risk indicator Ratio of equity over total assets Bank Focus
and own calculations Asset portfolio risk indicator Standard deviation of return on average assets Bank Focus
and own calculations Riskiness bank loans indicator Loan loss reserves / gross customer loans and
advances
Bank Focus
Independent variable
Developing country Dummy variable which is equal to 1 for a
developing/emerging country and 0 for a developed country
International Monetary Fund (IMF)
Bank-level control variables
Size Natural logarithm of total assets, in USD millions Bank Focus
Liquidity Ratio of liquid assets to total assets (%) Bank Focus
Income diversification Ratio of non-interest income to operating revenue
(%)
Bank Focus
Managerial efficiency Ratio of total operating expenses to total operating revenues (%)
Bank Focus
Operating efficiency Cost to income ratio (%) Bank Focus
Country-level control variables
Inflation (CPI) Inflation, consumer prices (annual %) World Bank
Growth of GDP per capita GDP per capita growth (annual %) World Bank
GDP per capita GDP per capita (current US dollars) World Bank
48
Table A2: Summary statistics of all banks in developing countries and developed countries during 2011-2017, before winsorizing
Variable Mean Std. Dev. Median Min Max N
Developing country 0.304 0.460 0 0 1 6573
49
Table A3: Summary statistics split for banks in developed country and banks in developing country observations during 2011-2017, before winsorizing
Developed Developing
Variable Mean
Std.
Dev. Median Min Max N Mean Std.
Dev. Median Min Max N
Developing 0 0 0 0 0 4578 1 0 1 1 1 1995 Z-score 3.931 1.381 3.986 -5.120 9.346 2525 3.164 1.501 3.177 -4.106 8.300 1002 Profitability 0.397 3.214 0.461 -72.672 67.203 3527 -0.093 7.676 0.868 -181.637 39.818 1467 Leverage Risk 10.209 16.245 8.248 -690.524 95.604 3527 13.065 18.800 11.964 -496.185 97.822 1467 Asset Portfolio Risk 0.666 2.343 0.167 0.000 54.060 2651 1.993 6.274 0.600 0.003 99.069 1100 Riskiness Bank Loans 4.886 8.347 2.651 -1.756 95.177 3456 11.081 11.510 7.621 0.001 89.639 1448
Size (in USD
millions) 9.606 2.040 9.605 2.465 14.845 3528 7.997 1.712 8.030 3.148 13.229 1467 Liquidity 20.206 17.546 14.424 0.002 97.032 3527 22.988 15.818 19.167 0.243 99.980 1467 Income Diversification 45.767 37.435 43.875 -704.059 497.465 3517 34.091 38.179 33.100 -532.332 631.362 1466 Managerial Efficiency 0.799 7.727 0.630 -24.875 449.533 3523 0.582 1.381 0.585 -45.729 9.487 1467 Operating Efficiency 66.732 46.341 63.029 -602.797 980.992 3514 62.257 56.148 58.460 -870.465 948.667 1465 Inflation (CPI) 1.272 1.136 1.113 -1.736 4.982 4578 6.828 7.936 6.472 -1.545 59.220 1995 GDP Per Capita Growth 0.801 1.881 0.841 -8.998 8.465 4578 2.091 2.954 2.378 -9.444 9.475 1995 GDP Per Capita 43021 20255 40875 13640 119225 4578 10401 4030 10743 1832 16007 1995 Financial Depth 100.827 29.185 95.251 40.830 187.241 4569 50.651 10.870 52.680 21.778 75.229 1995
50
Table A4: Correlation matrix for all employed variables during 2011-2017, after winsorizing, for banks in developed countries Variable Z-score Profita bility Levera ge Risk Asset Portfoli o Risk Riskine ss Bank Loans Size (in USD million s) Liquidi ty Income Diversi fication Manag erial Efficien cy Operati ng Efficien cy Inflatio n (CPI) GDP Per Capita Growth GDP Per Capita Fin anc ial De pth Z-score 1 Profitability 0.2064 1 Leverage Risk 0.001 0.1882 1 Asset Portfolio Risk -0.5532 -0.1365 0.4141 1 Riskiness Bank Loans -0.3103 -0.2059 0.3119 0.5036 1
Size (in USD
millions) 0.143 -0.1024 -0.425 -0.2487 -0.2017 1 Liquidity -0.1032 0.0813 0.057 0.1072 0.0699 -0.1023 1 Income Diversification -0.135 0.112 0.2744 0.1928 0.0772 -0.1987 0.3592 1 Managerial Efficiency -0.2262 -0.376 0.0685 0.1821 0.1236 -0.1526 0.1589 0.2779 1 Operating Efficiency -0.2394 -0.3959 0.0838 0.1958 0.1346 -0.1527 0.1649 0.2897 0.9943 1 Inflation (CPI) 0.0014 0.0802 -0.0232 -0.0771 -0.1829 -0.0288 0.019 -0.1003 -0.0517 -0.0549 1 GDP Per Capita Growth 0.0345 0.1152 0.0769 -0.0063 0.0446 -0.1763 0.0849 -0.0002 0.0105 0.0088 -0.057 1 GDP Per Capita 0.0639 0.0958 -0.0248 -0.0853 -0.2472 0.0516 0.1135 -0.0237 -0.038 -0.0402 0.3562 -0.3023 1 Financial Depth -0.0339 0.0169 0.0402 0.0172 0.0312 0.037 0.0739 -0.0105 -0.0138 -0.0154 0.0465 -0.2025 0.4906 1
51 Variable Z-score Profita bility Levera ge Risk Asset Portfoli o Risk Riskine ss Bank Loans Size (in USD million s) Liquidi ty Income Diversi fication Manag erial Efficien cy Operati ng Efficien cy Inflatio n (CPI) GDP Per Capita Growth GDP Per Capit a Fin anc ial De pth Z-score 1 Profitability 0.3756 1 Leverage Risk 0.1942 0.1041 1 Asset Portfolio Risk -0.6517 -0.1282 0.1592 1 Riskiness Bank Loans -0.4525 -0.2948 0.1428 0.5548 1
Size (in USD
millions) 0.0633 0.0244 -0.3572 -0.1555 -0.2268 1 Liquidity -0.1862 0.0331 0.0175 0.139 0.3518 -0.1824 1 Income Diversification -0.0997 0.063 -0.0423 0.1145 0.1517 0.0203 0.178 1 Managerial Efficiency -0.26 -0.488 0.0073 0.1443 0.123 -0.2394 0.0035 -0.1006 1 Operating Efficiency -0.2664 -0.5056 0.0156 0.1522 0.1273 -0.2449 0.0057 -0.0848 0.9971 1 Inflation (CPI) -0.1838 -0.0895 -0.0068 0.1584 0.156 -0.0578 0.1728 -0.028 -0.1467 -0.1475 1 GDP Per Capita Growth 0.2262 0.1745 -0.0445 -0.1218 -0.2365 0.1362 -0.3116 -0.0261 0.0562 0.0536 -0.6167 1 GDP Per Capita 0.0993 -0.065 -0.0943 -0.1716 -0.1892 0.365 -0.0044 -0.0222 0.0555 0.0599 -0.1276 -0.012 1 Financial Depth 0.0703 -0.0628 -0.032 -0.0615 -0.0626 0.1307 0.0237 -0.1816 -0.0189 -0.0188 0.1362 -0.1293 0.149 1
Table A5: Correlation matrix for all employed variables during 2011-2017, after winsorizing, for banks in developing countries
