Government ownership of banks : the European analysis

(1)

Government ownership of banks: the European analysis

FEB

Name Suzanne van Muijden

Number 10274073

BSc in Economics and Business Specialization Finance

Supervisor Mr. Timotej Homar Completion 1st_{July, 2014}

(2)

1 Table of Contents 1. Introduction ... 2 2. Literature review ... 3 3. Empirical methodology ... 6 4. Empirical data ... 7 5. Empirical result ... 11 5.1 Empirical Results ... 11 5.2 Robustness check ... 20

(3)

1. Introduction

Because of the recent Eurozone debt crisis, several governments were forced to take over control of some of its banks. In most EMU countries national bank bailout plans have been implanted during the crisis, which means that there was a shift to state-owned banks. For example, the EU/IMF bailout plans have rescued Greece, Portugal and Ireland from insolvency (Eichler & Hielscher, 2012). There was a shift to state-owned banks, which raises the question how state-state-owned banks perform in general.

Ownership structure is widely accepted as an important determinant of firm performance in finance literature. Recent studies like Cornett et al. (2010) conclude that state-owned banks in Asian countries performed less favorable than privately owned banks during the Asian crisis, and Barth et al. (2002) also find a negative link between government ownership and favorable banking outcomes, such as that state-owned banks have higher non-performing loans ratios. Jia, (2009) states that

government owned banks are less prudent, measured by a higher loans to assets ratio and a lower deposits to loans ratio. Although most results are based on Asia or developing and emerging countries, most studies on government ownership show a significant difference in bank performance such as bank efficiency and profitability between state-owned and privately owned banks, with the difference being in favor of the privately owned banks. Questions arise if these results are time- and country-specific and since banks are important financial providers, there is a need to find out if government ownership will cause banks to perform less than privately owned banks in the European countries and if this difference is only significant in times of financial turmoil.

This paper is organized as follows. Section 2 summarizes the relevant literature used for this study. Section 3 describes the empirical methodology and hypothesis. Section 4 presents the empirical data and some summary statistics. Section 5 evaluates the empirical results and some additional robustness checks. Finally, Section 6

(4)

2. Literature review

The recent study of Cornett et al. (2010) examines how government ownership of banks affect bank performance. Their sample is based on banks of Asian countries from 1989 through 2004, a period including the Asian financial crisis which started in 1997. Cornett et al. (2010) use several performance measures to indicate performance differences. Overall, compared to state-owned banks, they conclude that privately owned banks were more profitable, held more core capital and had less credit risk. More specific, Cornett et al. (2010) state that the performance deterioration of state-owned banks was significantly greater than that of privately state-owned banks during the crisis from 1997 through 2000, and that these differences between government owned and privately owned banks were more significant in those countries with greater political corruption and those that were hardest hit by the Asian crisis. Nonetheless, the gap closed in the post-crisis period of 2001-2004 (Cornett et al., 2010). Since this study on government ownership in European countries does not include a post-crisis period, it is only possible to examine how state ownership affects the banks

performance in general or in times of turmoil. The study of Cornett et al. (2010) suggests that the performance differences are more significant in countries with greater political corruption. Although measuring corruption is highly difficult, I can state that political corruption will not be so much of an issue in European countries, which in turn ascertains that this study will probably generate less significant

performance differences between state-owned and privately owned banks. However, because the crisis period from 2008 through 2012 is included, I suspect to find a greater gap in performance between the different types of banks in these particular years.

Barth et al. (2004), who examined the relationships between a broad array of bank regulations and supervisory practices and bank performance, also conclude that government ownership is positively linked with political corruption. Barth et al. (2004) locate a negative link of state-owned banks with favorable banking outcomes such as that the level of non-performing loans is positively related with government ownership. While they find these links, it does not retain an independent, robust association with banking outcomes as efficiency and stability when controlling for other features of the regulatory and supervisory environment. They do not find

(5)

evidence that state-owned banks are associated with positive outcomes, not even in weak institutional settings.

Cull & Pería (2013) examined the impact of bank ownership in developing countries on credit growth before and during the 2008-2009 crisis. They made the distinction between foreign-, domestic- and government-owned banks. Cull & Pería (2013) find that in Latin America the state-owned bank’s lending growth was much higher than that of domestic and foreign banks. In Eastern Europe bank lending significantly declined due to the 2008-2009 crisis, but the bank lending did not significantly differ for government owned banks in comparison with foreign owned and domestic owned banks. In contrast to their study, I focus on European countries.

Dinç (2005) studied if the actions of state-owned banks are motivated by political concerns. He examined if in times of elections politicians were tempted to use state-owned banks for political purposes, in particular how the effect is on bank lending. It shows that banks, which are state-owned, increase their lending in election years relative to privately owned banks. Most of the countries in the sample are European Union members, which have elections for the European Parliament.

Because the study is based on national election years, and these European Parliament elections are not as important, they blur differences between the years in which national elections take place and the other years (Dinç, 2005). Though these results are based on the years of national election, given the potential blur of differences, it should indicate that in general the loan to total assets ratio will be higher for state-owned banks relative to privately state-owned banks. Besides that, also the ratio of government securities to total assets is higher for state-owned banks, as it complements the fact that banks which are government owned finance the

government to a greater degree than privately owned banks (Cornett et al., 2010). Some studies questioned the effect of a change in bank ownership on bank performances. Megginson (2005) did a study on the privatization of banks and concluded that bank privatization has resulted in clear performance improvements. However, he states a few essential minimum conditions that must be addressed in order for bank privatization to be successful.

Besides Megginson (2005), Taboada (2011) also evaluated the change in bank ownership structure. He examined if this has an impact on the allocation of capital. Taboada (2011) finds that the change to foreign ownership from government

(6)

and Megginson (2005) studied the movement away from state ownership. Because they examine banks that are privatized for these studies, they have a sample of banks, which did not change their ownership structure for no reason. The performance differences between several banks with different ownership structures, allows the examination of general differences between state-owned and privately owned banks. This study could lead to actual dissimilar results.

Iannotta et al. (2013) use a sample of European banks to evaluate the difference in bank risk between state-owned and privately owned banks. They distinguish default risk - the probability that a bank’s creditors suffer losses - from operating risk, which is the probability that a bank’s asset value will be below the value of its liabilities. Iannotta et al. (2013) find that, on average, state-owned banks have a lower default risk, which contradicts the findings of Cornett et al. (2010), stating that government owned banks had greater credit risk. Moreover, Iannotta et al. (2013) conclude that state-owned banks have a higher operating risk, which they think does not necessarily makes the case against the government ownership of banks stronger. Because state-owned banks should be affected by market failures – like any other government owned firm – and therefore are contributing to economic

development, this behavior would lead to inevitably asset quality deterioration and an increase of their risk profile (Iannotta et al., 2013). Besides this, Laeven & Levine (2009) conduct an empirical assessment of theories concerning bank ownership structures, their risk taking and national bank regulations. They find that banks tend to take greater risks when having more powerful owners.

Jia (2009) studied the relationship between bank ownership and bank

prudence in China. He uses loan to asset ratio, deposit to loan ratio and excess reserve as proxies of bank prudence. Jia, (2009) finds that state-owned banks tend to have less excess reserves, higher loan to asset ratios and lower deposit to loan ratios, which means that government owned banks are significantly less prudent than joint-equity banks. Joint-equity banks are banks that have several different owners.

Shehzad et al. (2010) studied the effect of ownership concentration on impaired loans and capital adequacy ratios for a sample of around 800 banks. They find that ownership concentration significantly affects loan quality, and conclude that the effect on the non-performing loans ratio is negative, at least if ownership is above 50% of the shares.

(7)

3. Empirical methodology

To study the performance differences between state-owned and privately owned banks, financial statement data is needed, next to the shareholder information on each bank to distinguish state ownership from privately owned banks. Several performance measures are used to answer if bank performance differs between state-owned and privately state-owned banks.

The following model should be used on testing the statistical relationship:

Yi,t= αi,t + β1Dgov,i,t + β2,tlnsizei,t + β3Dgov,i,t x Dcrisis + β4Dcrisis + β5Dcountry + β6Dtype + εi,t

Yi,t stands for the particular bank performance measure that is used as a

dependent variable of bank i in year t. The lnsize variable is bank size measured as a logarithm of total assets. Dgov,i,t presents a dummy for state ownership, which is equal to 1 if a bank is state-owned and 0 otherwise. Dgov,i,t x Dcrisis, presents a dummy for state ownership in the crisis period, which is equal to 1 if a bank is state-owned in the crisis period. The rest of the explanatory variables are used as control variables and are described in the data section. Εi,t is the error term and the βi are vectors of coefficient estimates.

If the government owned banks differ from the banks that are privately owned, the coefficient β1 should be significantly different from zero. For example, a positive

significance of β1 if Yi,t is a measure of bank profitability, means that state-owned

banks are more profitable than privately owned banks. If the coefficient β2 is

significantly different from zero, state-owned banks in the crisis period perform differently than privately owned banks.

As in most of the other studies on government ownership, I assume there to be significant differences in performance between state-owned and privately owned banks and these differences to be in favor of privately owned banks. However, I expect these differences to be less significant than, for example, those that were found by Cornett et al. (2010) and Barth et al. (2004). This should indicate that the

coefficient β1, in the model as described above, would be significantly less than zero

for bank profitability measures and the opposite (significantly greater than zero) for the bank efficiency ratio.

(8)

4. Empirical data

This study uses year-end financial statement data from 2005 through 2012 for the 28 European Union countries (Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, The Netherlands, Poland, Portugal, Romania, Slovenia, Slovakia, Spain, Sweden and the United Kingdom) obtained from the BankScope database. Moreover, the shareholder information for each bank is gathered from the BankScope database. Because of the lack of available information from the Owner database, the BankScope dataset on Ultimate, Immediate and Global owners is also used. From the last two databases mentioned, banks were divided into state-owned and privately owned.

A bank is classified as being state-owned if government ownership is at least 20%, following La Porta et al. (1999), who state that a corporation has a controlling shareholder (ultimate owner) if direct and indirect ownership is 20 percent or more. When a foreign government holds indirectly this 20 percent or more in a particular bank, it is considered as privately owned1. All other banks are considered as privately owned. In addition to the distinction between state-owned and privately owned banks, I will examine whether the extent of state ownership has an impact on the

performance differences between state-owned and privately owned banks. This will be done by including an explanatory variable of the percentage of state ownership in the regressions. The 8-year sample gives the opportunity to not only look at the performance differences between state-owned and privately owned banks in the years of the financial crisis, but it should also examine the greater picture of performance differences between state and private ownership.

Because of the lack of information in the BankScope Shareholder dataset, the banks of the European Union countries that did not have the required shareholder information were dropped. Unfortunately, I only have shareholder information for each bank and not for each year. This is why it is not possible to look at the impact on the bank performance of a change in ownership structure for each bank over the years. However, I expect that the ownership structure will not be affected by major changes.

1_{If
a
foreign
government
holds
directly
20
per
cent
or
more
in
a
particular
bank,
it
is
considered} as a state-‐owned bank. However, I did not find any banks, which had direct foreign ownership of 20 per cent or more.

(9)

Besides this, it was reasonable to keep all different types of banks and financial institutions to compensate for the infrequent information.

Table 1 – banks per country

Country Number of total

banks Number of state-owned banks Austria 15 1 Belgium 28 0 Bulgaria 6 0 Croatia 4 1 Cyprus 4 0 Czech Republic 10 0 Denmark 31 1 Estonia 2 0 Finland 3 1 France 209 10 Germany 130 9 Greece 1 0 Hungary 3 0 Ireland 3 0 Italy 22 0

Table 2 – type of banks

Banks Total number of banks State-owned banks

Bank Holding & Holding Companies 8 1

Clearing Institutions & Custody 5 0

Commercial Banks 275 10

Cooperative Bank 55 0

Finance Companies 92 3

Group Finance Companies 3 0

Investment Banks 32 1

Islamic Banks 2 0

Private Banking Companies 31 1

Real Estate & Mortgage Bank 33 2

Savings Bank 42 3

Securities Firm 13 1

Specialized Governmental Credit Institution 23 10

Total 614 32 Latvia 1 0 Lithuania 1 0 Luxembourg 13 0 Malta 3 0 The Netherlands 11 2 Poland 15 0 Portugal 8 1 Romania 9 0 Slovakia 4 0 Slovenia 5 2 Spain 19 1 Sweden 7 0 United Kingdom 47 3 Total 614 32

(10)

Table 1 presents the total number of banks, the total number of state-owned banks and the number of total banks and state-owned banks for each country separately. There is a total bank sample of European Union countries of 614, from which 32 are defined as state-owned banks. France and Germany have the most recorded information on banks available. Together they stand for a number of 339 total banks, which is more than half of the entire sample. Moreover, the number of state-owned banks of these two countries together sum up to 19, which is around 60% of the total number of state-owned banks. Table 2 presents all different types of banks, for each type of bank the total number of banks and the number of state-owned banks.

Several performance measures are used as dependent variables in different regressions to test for the possible differences in performance between state-owned and privately owned banks. For bank profitability, I will use return on assets which is net income divided by total assets. A measure of operating pre-tax cash flow divided by total assets is also used to measure profitability differences between state-owned and privately owned banks. Besides that profitability is measured, several measures of loan quality are used since credit quality is a dominant source of risk in the banking industry. The measures of loan quality used are:

- Loan growth;

- Impaired loans to gross loans; - Loan loss reserves to gross loans; - Loan loss reserves to impaired loans; - Loan loss provisions to gross loans

I will look at all of these loan quality ratios because the data on non-performing loans is often missing. The ratio of loan loss provisions to gross loans is used to look at the impact of credit quality management against the portfolio size of loans. To examine if there is a difference in bank efficiency, a measure of non-interest expenses divided by gross revenues will be used which shows to which extent operating revenues are absorbed by operating expenses.

A bank’s liquidity is measured by examining several measures. First, a ratio of liquid assets to total assets is used. Liquid assets stand for loans and advances and trading assets with a maturity of less than three months. Several other measures of bank liquidity that are used are:

(11)

- Loans to deposits and short-term funding; - Liquid assets to deposits and short-term funding

The loans to total assets ratio is used to look at differences in the bank’s investment in loans. A measure of core capital to total assets stands for the ability of a bank to meet regulated capital standards. The asset growth rate will also be a dependent variable and a measure of government securities to total assets is used to examine whether state ownership makes a bank finance the government to a greater extent than privately owned banks do.

One of the explanatory variables is bank size, which are total assets corrected for currency differences, by measuring total assets in dollars for each bank. Other explanatory variables will be a state ownership dummy, a crisis period dummy, a state ownership in the crisis period dummy, country dummies and bank type dummies. The dummy on government ownership in the crisis period indicates if state ownership lets a bank behave differently in the years of the financial crisis.

Table 3

Variable Obs Mean SD Min Median Max

Size (total assets in $ millions) 4094 45,035.97 208,662.2 1.39 2,666.11 2,992,885

Operating pre-tax income/Assets 3003 0.93 2.19 -6.72 0.71 11.69

Return on Assets 4093 0.58 2.02 -8.82 0.44 9.89

Core Capital/Assets 1509 10.46 13.18 -16.67 7.28 98.45

Asset growth rate 3685 12.27 42.93 -70.62 5.12 600.48

Loan growth 4094 94.45 671.09 -99.67 2.93 6203.92

Loans/Assets 4094 55.86 27.47 0.02 61.91 98.23

Impaired loans/Gross loans 1609 6.27 7.02 0.02 4.21 42.48

Loan loss reserves/Gross loans 2362 4.16 5.67 0.00 2.54 58.39

Loan loss reserves/Impaired loans 1581 80.94 79.11 4.29 61.90 585.71

Loan loss provisions/Gross loans 3715 1.42 12.23 -41.54 0.32 251.22

Loans/Customer deposits 3317 145.71 144.08 0.43 106.67 800.91

Loans/Deposits & Short-term funding

4053 125.17 295.32 0.10 81.40 2600

Liquid assets/Total assets 4081 25.04 23.37 0.02 16.64 95.72

Liquid assets/Deposits & Short-term funding 4048 48.16 77.66 0.32 23.85 481.57 Non-interest expenses/Gross revenues 4038 68.20 34.28 2.41 65.04 230.77 Government securities/Assets 1763 6.04 7.61 0.00 3.05 41.62

(12)

Table 3 presents the mean, standard deviation, min, max, median and number of observations for all variables used in this study for the period of 2005-2012. All variables except size are corrected for outliers, which is done by winsorizing the variables by 1%. As explained by the minimum and maximum values of each variable, especially for loan growth, dispersion for each variable is high.

5. Empirical result 5.1 Empirical Results

Because it is necessary to look at the differences between state-owned and privately owned banks, table 4 presents the mean and number of observation for each type of ownership. It also reports the difference in means between state-owned and privately owned banks for these variables. Because the crisis period of 2008-2012 can have a significant impact on the differences in bank performance, I also calculated the means and differences in means for the two separate periods. I will test for differences in means between state-owned and privately owned banks of the various performance measures with a t-test that does not assume equal variances.

Table 4 makes clear that there is a significant difference in size between government owned and privately owned banks, government owned banks are on average much larger. However, this difference seems not significant in the pre-crisis period. Overall, table 4 suggests that government owned banks have on average less loan quality but are more efficient and hold more reserves for non-performing loans, which is safer. The return on assets of state-owned banks in the crisis period of 2008-2012 is significantly lower than that of privately owned banks, this ratio averages 0.41% for privately owned banks and only 0.07% for state-owned banks. These results support those of Cornett et al. (2005) stating that government ownership of banks creates an opportunity for political bureaucrats to follow their objectives dictated by political interest, though these objectives are in conflict with firm value maximization. A point noticeable is the government securities to total assets ratio. From other studies like Dinç (2005) and Cornett et al. (2010) I conducted a theory that state-owned banks finance the government more heavily than privately owned banks. However, the results on the differences in means in table 4 suggest the opposite. The government securities to total assets ratio is on average significantly

(13)

higher for privately owned banks than for state-owned banks in the crisis period, and also for the entire period the ratio is (less but still) significantly higher for privately owned banks. The rest of the performance measures suggest on average no significant difference between state-owned and privately owned banks.

Table 4 – differences in means, t-tests

Variable 2005-2007 2008-2012 Total period

Privately owned banks State-owned banks Difference in mean Privately owned banks State-owned banks Difference in mean Privately owned banks State-owned banks Difference in mean

Size (total assets in $ millions) 39,146.58 (1278) 57,676.53 (81) -18,529.94 45,123.98 (2593) 89,222.80 (142) -44,098.82** 43,150.56 (3871) 77,764.29 (223) -34,613.73** Operating pre-tax/Assets 1.40 (964) 1.08 (67) 0.32 0.71 (1864) 0.50 (108) 0.21 0.94 (2828) 0.72 (175) 0.22 ROA 0.95 (1278) 1.05 (80) -0.11 0.41 (2593) 0.07 (142) 0.33* 0.59 (3871) 0.43 (222) 0.16 Core capital/Assets 9.05 (43) 3.78 (3) 5.27 10.45 (1377) 11.48 (86) -1.03 10.41 (1420) 11.22 (89) -0.81 Asset growth rate 18.28 (1127) 8.47 (69) 9.82** 9.65 (2363) 9.77 (126) -0.11 12.44 (3490) 9.31 (195) 3.13 Loan growth 40.41 (1278) 97.06 (81) -56.65 122.37 (2593) 69.40 (142) 52.97 95.31 (3871) 79.45 (223) 15.86 Loans/Assets 55.88 (1278) 57.82 (81) -1.95 55.59 (2593) 59.53 (142) -3.94* 55.68 (3871) 58.91 (223) -3.23* Impaired loans/Gross loans 4.58 (321) 3.95 (22) 0.63 6.67 (1206) 8.08 (60) -1.41 6.23 (1527) 6.97 (82) -0.74 Loan loss reserves/Gross loans 3.16 (602) 4.76 (39) -1.60** 4.46 (1629) 5.21 (92) -0.75 4.11 (2231) 5.07 (131) -0.97*

*,**,*** Significantly different from zero at the 10%, 5% and 1% levels, respectively, using a two-tailed test. The means of the ratios in the table are all computed percentages. . The number of observations is included between parentheses.

(14)

Table 4 (continued)

*,**,*** Significantly different from zero at the 10%, 5% and 1% levels, respectively, using a two-tailed test. The means of the ratios in the table are all computed percentages. The number of observations is included between parentheses.

Because bank profitability is important for measuring differences in

performance between state-owned banks and privately owned banks, table 5 reports the regression results on the bank profitability measures as dependent variables. The regressions run on the 2005-2012 period. The explanatory variables are a state ownership dummy (Dgov), the natural logarithm of bank size (Size), crisis period

dummy (Dcrisis), a cross product of the crisis period and state ownership, country dummies and bank type dummies. The crisis period dummy, country dummies and bank type dummies are used to control for bank type-specific, country-specific and macroeconomic factors. The country and bank type dummies are all included in the regressions but are both not reported in the table to conserve space. The state

ownership dummy, Dgov, is set equal to 1 if government ownership of a bank is at least

Variable 2005-2007 2008-2012 Total period

Privately owned banks State-owned banks Difference in mean Privately owned banks State-owned banks Difference in mean Privately owned banks State-owned banks Difference in mean Loan loss reserves/Impaired loans 97.19 (315) 142.45 (22) -45.26** 75.87 (1184) 73.18 (60) 2.70 80.35 (1499) 91.76 (82) -11.41 Loan loss provisions/Gross loans 0.31 (1173) 0.51 (74) -0.21 2.02 (2351) 1.00 (134) 1.02 1.46 (3507) 0.83 (208) 0.63 Loans/Customer deposits 134.07 (1059) 262.53 (48) -128.47*** 141.05 (2112) 314.97 (98) -173.93*** 138.72 (3171) 297.73 (146) -159.02*** Loans/Deposits & Short-term funding 128.42 (1269) 322.05 (81) -193.63*** 111.56 (2561) 229.44 (142) -117.88*** 117.15 (3830) 263.08 (223) -145.93*** Liquid assets/Total assets 25.05 (1276) 26.96 (81) -1.91 25.09 (2582) 22.85 (142) 2.24 25.08 (3858) 24.34 (223) 0.74 Liquid assets/Deposits & Short-term funding 46.22 (1267) 80.87 (81) -34.66*** 46.74 (2558) 72.36 (142) -25.62*** 46.56 (3825) 75.45 (223) -28.89*** Non-interest expenses/Gross revenues 65.38 (1266) 50.83 (78) 14.55*** 70.51 (2555) 61.23 (139) 9.28*** 68.81 (3821) 57.50 (217) 11.31*** Government securities/Assets 5.57 (549) 6.12 (38) -0.55 6.37 (1093) 4.63 (83) 1.74* 6.10 (1642) 5.10 (121) 1.00

(15)

20% and 0 otherwise. The crisis dummy included in the regression is set equal to 1 if the statement year of the particular computed performance measure is between 2008-2012 and 0 otherwise. This dummy is included to see what the impact is of the crisis on bank performance and controls at the same time for time-specific factors in the regression of government ownership on bank performance. Table 5 presents two regressions on bank profitability. The first regression, presents the return on assets as the dependent variable. The state ownership dummy is not significantly different from zero, which should indicate that there is no difference in return on assets between state-owned banks and privately owned banks. However, the cross product dummy is significantly smaller than zero at the 5% level, indicating that banks which were state-owned in the crisis period had a return on assets that was 0.43 percentage points lower than privately owned banks in this period. This should suggest that in times of

financial turmoil, privately owned banks are more profitable than state-owned banks. Table 5 also reports that size is significantly negative at the 5% level. This indicates that larger banks are less profitable, namely a 1% change in the bank size is

associated with a 0.00030 percentage points decrease in the return on assets. The R-squared is 0.1142, which means that 11.42% of the dependent variable, return on assets, is explained by the explanatory variables in the regression.

Table 5 – Bank profitability

Explanatory variables Return on Assets Operating pre-tax income/Assets

State ownership 0.0850

[0.1816]

-0.1572 [0.1821]

State ownership in crisis period -0.4331**

[0.2135] 0.0001 [0.2198] Size (ln) -0.0298* [0.0161] -0.0586*** [0.0203] Crisis -0.4997*** [0.0621] -0.6301*** [0.0807] Intercept 1.6341** [0.8207] 7.5901** [3.1639]

Country dummies Included Included

Bank type dummies Included Included

Obs 4093 3003

R2 _0.1142 _0.1821

*,**,*** Significantly different from zero at the 10%, 5% and 1% levels, respectively. The robust standard errors are presented between brackets under each coefficient.

(16)

One other measure of bank profitability, the pre-tax operating income to assets ratio, is reported in the second regression of table 5. Here, both the coefficient of the state ownership dummy and the of state ownership in crisis period are not

significantly different from zero, which means that the ownership structure does not have an impact on the operating pre-tax to assets ratio and therefore suggests that state-owned and privately owned banks do not differ in bank profitability. However, the coefficient of bank size and the crisis period are both significantly negative at the 1% level, so that a 1% larger bank is associated with a 0.000598 percentage points operating pre-tax income to assets ratio. In the second regression in table 5, bank profitability is suggested to be 0.64 percentage points lower for all banks in the crisis period.

Table 6 presents regressions on loan quality measures. I will only evaluate some of the important ones. One measure of loan quality is: impaired loans to gross loans. Impaired loans are non-performing loans, which indicates that a higher ratio implicates a lower loan quality. Table 6 presents a negative but not significant

coefficient for the state ownership, which suggests that the non-performing loans ratio does not differ between state-owned and privately owned banks in general.

Table 6 – Loan quality

Explanatory variables Impaired loans/Gross loans Loan loss reserves/Gross loans Loan loss reserves/Impaired loans Loan loss provisions/Gross loans State ownership -1.1283 [0.9859] 1.4979*** [0.5824] 53.0768** [25.6974] 0.0801 [0.3654] State ownership in crisis period 1.8522* [1.1209] -0.9158 [0.7351] -51.5299** [25.9345] -1.1193** [0.5003] Size (ln) -0.5709*** [0.0862] -0.3591*** [0.0653] -2.8643*** [0.9350] -0.0789 [0.1050] Crisis 2.1435*** [0.3783] 1.1900*** [0.2126] -22.1356*** [5.3348] 1.5108*** [0.3038] Intercept 4.0765*** [1.5503] 22.7724*** [4.8265] 111.8432*** [14.8471] 9.9599 [8.2012]

Country dummies Included Included Included Included

Bank type dummies Included Included Included Included

Obs 1609 2362 1581 3715

R2 _0.1803 _0.1582 _0.1195 _0.0434

(17)

Therefore, the theory of Barth et al. (2002), who find that non-performing loans have a positive link with government ownership, will not be followed by these results. However, the coefficient of state ownership in the crisis period is significantly positive at the 10% level, which indicates that banks that were state-owned in the crisis period, had an 1.85 percentage points higher non-performing ratio than privately owned banks in the crisis and thus had less loan quality. The regression on the

impaired loans to gross loans ratio also shows that the coefficient of bank size is significantly negative at the 1% level and the crisis period coefficient is significantly positive at the 1% level. A 1% increase in bank size in the crisis period indicates a 0.0057 percentage points decrease in the non-performing loans ratio. From this can be concluded that the bigger the bank the better the loan quality is. The positive

significance of the crisis period coefficient indicates that in the period of financial crisis banks had less loan quality, an increase of the impaired loans ratio of 2.14 percentage points.

The loan loss reserves to gross loans ratio evaluates the degree to which loan losses are allowed. The higher the ratio the lower the loan quality is considered to be, however, I could not conclude this. Banks that have higher loan loss reserves could expect more loan losses but banks with lower reserves could actually experience these higher loan losses. This means that a higher loan loss reserves to gross loans is not necessarily a bad thing. Table 6 shows that the loan quality ratio significantly differs between state-owned and privately owned banks, and both the bank size and crisis period coefficients are significant at the 1% level. State-owned banks have a 1.50 percentage points higher loan loss reserves to gross loans ratio than privately owned banks. Hence, it can only suggest lower loan quality. Bank size has a positive effect, indicating an 1% increase in size is associated with a 0.0036 percentage points increase in loan quality and the crisis period has a negative effect, stating that banks have an 1.19 percentage points higher loan loss reserves to gross loans ratio in the crisis period, which could be explained by the fact that loans are expected to default to a greater extent in times of financial turmoil.

Both the loans to customer deposits ratio and the loans to deposits and short-term funding ratio are commonly used to measure a bank’s liquidity. If the ratios are too high, it is possible that banks do not have enough liquidity to cover unforeseen fund requirements. If the ratios are too low, banks might not be earning as much as

(18)

ownership coefficients of both regressions are significantly positive at the 5% level, suggesting that state-owned banks have an 145.29 or 123.72 percentage points, respectively, higher ratio than privately owned banks and these could indicate that state-owned banks might fund loans with non-deposit liabilities. The difference is not significant in the crisis period, which could indicate that both state-owned and

privately owned banks have trouble to meet fund requirements.

Next, some other bank liquidity measures are considered, the liquid assets to total assets ratio and the liquid assets to deposits and short-term funding ratio. In table 7, both show no significant difference in bank liquidity between state-owned and privately owned banks. Besides that, both ratios on liquid assets show no significant coefficient for both the crisis period and bank size.

Table 7 – Bank liquidity

Explanatory variables Loans/Customer deposits Loans/Deposits & Short-term funding Liquid assets/Total assets Liquid assets/Deposits & Short-term funding State ownership 130.59*** [32.7347] 119.7101** [59.3615] -1.6951 [2.9986] 4.6788 [11.4864] State ownership in crisis period 49.9998 [38.7466] -60.0799 [76.7119] -2.0140 [3.4530] 2.9753 [15.1318] Size (ln) 4.1446*** [1.1337] -4.5520** [2.2316] -0.2003 [0.1704] -0.6683 [0.6260] Crisis 2.2889 [4.7626] -20.3528** [9.6695] -0.3156 [0.7233] -2.1232 [2.4111] Intercept 266.6387*** [27.1342] -122.9206* [72.9045] 21.9438*** [4.0864] 121.8615*** [38.9473]

Obs 3317 4053 4081 4048

R2 _0.1896 _0.1069 _0.1737 _0.1271

The first regression in table 8 presents a regression for bank efficiency, the non-interest expenses to gross revenues ratio, which indicates that the lower the ratio the higher bank efficiency is. The state ownership dummy has a significantly negative coefficient, which means that state-owned banks are in general more efficient than privately owned banks. Precisely, government owned banks have an around 12.00

(19)

percentage points lower ratio than state-owned banks. The coefficient of state ownership in crisis period is not significantly different from zero, meaning that state-owned banks in the crisis do not differ in efficiency from privately state-owned banks. The bank size coefficient is significantly negative and this indicates that an 1% increase in bank size is associated with a 0.0164 percentage points decrease in the non-interest expenses to gross revenues ratio which in turn leads to greater bank efficiency. The first regression in table 8 also shows a significantly positive coefficient of the crisis period, which suggests that all banks in general have a higher non-interest expenses to gross revenues ratio and thus are less efficient. The explanation behind this is that in times of turmoil, banks suffer a loss in revenues but the non-interest expenses, such as personnel expenses, do not change. An increase of 5.34 percentage points of the ratio indicates a decrease in bank efficiency.

Table 8 – Bank efficiency, core capital and some other measures

Explanatory variables Non-interest expenses/Gross revenues Core capital/Assets

Asset growth rate Government

securities/Assets State ownership -11.9977*** [2.6273] 1.5153 [1.9173] -10.7052** [4.3471] 0.1274 [1.3811] State ownership in crisis period 5.9841 [3.8840] 0.8578 [2.0774] 13.8893*** [5.4183] -1.3924 [1.4714] Size (ln) -1.6398*** [0.27373] -1.4963*** [0.1605] 0.6362 [0.4346] 0.2195*** [0.0830] Crisis 5.3369*** [1.0348] -0.2713 [1.4182] -10.4405*** [1.4593] 0.5807* [0.3402] Intercept 137.3529*** [46.6279] 19.9741*** [7.7920] 170.4049* [89.1354] 10.6348*** [2.8558]

Obs 4038 1509 3685 1763

R2 _0.1148 _0.1796 _0.0724 _0.2799

The second regression in table 8 presents a regression on the core capital to assets ratio. As stated in the methodology section, the greater the core capital to total assets ratio, the greater the ability of a bank to meet regulated capital standards. Table

(20)

period are not significantly different from zero. This means that the ratio of core capital to assets does not differ between state-owned and privately owned banks. The only significant explanatory variable is bank size, which suggests that an increase of 1% in bank size is associated with a 0.0150 percentage points decrease in the core capital ratio. Hence, larger banks are suggested to have a lower ability to meet the regulated capital standards. A possible explanation may be that the conjectural government guarantee that is enjoyed by larger institutions is higher (Cornett et al., 2010).

Asset growth is evaluated in the third regression of table 8. I observe all coefficients but one, the bank size coefficient, to be significantly different from zero. Government owned banks are considered to have an assets growth rate that is 10.71 percentage points lower than that of privately owned banks (at the 5% level).

However, in the crisis period this effect seems to rotate such that, at this point in time, state-owned banks have a 13.89 percentage points greater growth rate (at the 1% level). Banks suffer, in general, a 10.44 percentage points decrease in the asset growth rate in the crisis period. The R-squared of the asset growth regression is low,

indicating that the asset growth rate is explained by only 7% of the explanatory variables.

The last regression in table 8 presents the government securities to total assets ratio, which examines if government owned banks finance the government to a greater extent than privately owned banks do. Where Cornett et al. (2010) find a significant difference between state-owned and privately owned banks, I find that the significant difference between state-owned and privately owned banks does not exist. However, the coefficient of bank size and the crisis period are both significantly positive, which indicates that a 1% larger bank has a higher government securities to total assets ratio of 0.0022 percentage points and that all banks have a 0.58 percentage points higher government securities to assets ratio in the crisis period.

Table 9 presents regressions on the loans to assets ratio and loan growth. The coefficient of state ownership in loan to assets ratio regression is significantly different at the 5% level, which indicates that state-owned banks have a 7.35

percentage points higher loans to assets ratio than privately owned banks. However, the loan growth ratio does not significantly differ between government owned banks and privately owned banks. These findings do not follow those of Dinç (2005), who concludes that state-owned banks increase their lending in election years, and Cull &

(21)

Pería (2013) who realize that lending growth is significantly greater for state-owned banks than for privately owned banks in Latin America, but they do not find this difference to be significant for Eastern Europe banks. The opposite findings of Cull & Pería (2013) could be because their study is focused on developing countries and due to stronger ties between the government and state-owned banks in Latin America. Besides that, the state-owned banks in Latin America were larger than those in Eastern Europe and had greater profitability. This enabled the government owned banks in Latin America to increase their lending, which is also shown in the second regression in table 9 by the positive significance of the bank size coefficient. A 1% increase in bank size is associated with a 0.59 percentage points increase in loan growth. The loan growth also significantly differs in the crisis period; banks had a 34.35 percentage points higher loan growth ratio.

Table 9 – Loan measures

Explanatory variables Loans/Assets Loan growth

State ownership 7.3541**

[3.2886]

26.0641 [76.9578] State ownership in crisis

period -1.5851 [3.7461] -86.3022 [93.7447] Size (ln) -0.2087 [0.1960] 59.2330*** [8.9615] Crisis 0.5516 [0.8337] 34.3530** [14.4587] Intercept 29.9043*** [6.4728] -585.6338*** [110.324]

Obs 4094 4094

R2 _0.2169 _0.0605

5.2 Robustness check

I performed some additional regressions to check if the degree of government

(22)

et al. (2010), who find that the effect of ownership concentration of at least 50% is negative on the non-performing loans ratio.

In the sample of banks in this study, all state-owned banks have government ownership of 50% or more. Therefore, a wholly state ownership dummy is included that is equal to 1 if government ownership is 98% or more and 0 otherwise. Hence, this dummy defines banks with total government ownership.

Table 10 presents some additional regressions for some of the dependent variables used before, in which it includes the wholly state ownership dummy. The first regression in table 10 reports a regression on the return on assets. Both the state ownership and total state ownership are not significantly different from zero,

indicating that the degree of state ownership does not matter for the bank profitability and that state-owned banks in general do not significantly differ from privately owned banks in return to assets. However, the state ownership in crisis period is still

significantly negative at the 5% level as showed in table 5.

Table 10 – Additional regressions

Explanatory variables Return on Assets Non-interest expenses/Gross loans

Loan growth Impaired

loans/Gross loans Government securities/Assets State ownership 0.0641 [0.1958] -6.5927** [2.7556] 63.0385 [84.4025] -1.6410 [1.2729] -2.6490** [1.2972] Wholly state ownership 0.1081 [0.1925] -27.0183*** [4.7668] -182.6949** [80.2639] 1.6833 [1.8295] 9.8486*** [2.0394] State ownership in crisis period -0.4329** [0.2139] 5.8943 [3.7608] -88.2405 [93.5385] 2.0603* [1.2208] -0.7011 [1.3365] Size (ln) -0.0299* [0.0161] -1.6179*** [0.2730] 59.3591*** [8.9714] -0.5707*** [0.0862] 0.2198*** [0.0836] Crisis -0.4995*** [0.0621] 5.2799*** [1.0351] 33.9983** [14.4456] 2.1432*** [0.3786] 0.5969* [0.3391] Intercept 6.6724** [2.6343] 137.1905*** [46.6287] -586.5251*** [110.3468] 1.3485 [1.3166] 10.6467*** [2.8576]

Country dummies Included Included Included Included Included

Bank type dummies Included Included Included Included Included

Obs 4093 4038 4094 1609 1763

R2 _0.1142 _0.1189 _0.0610 _0.1807 _0.2931

(23)

The second regression in table 10 presents the bank efficiency ratio. Here, the wholly state ownership coefficient is significantly negative at the 1% level, which means that totally state-owned banks have a 27.02 percentage points lower ratio and thus have greater bank efficiency. Besides that, the state ownership coefficient is still significantly negative, indicating that state-owned banks in general have greater bank efficiency than privately owned banks. The R-squared in the additional regression on bank efficiency is higher than the r-squared of the bank efficiency regression in table 8, suggesting that the regression in table 10 is more explained by the explanatory variables.

Although the loan growth is not significantly different between state-owned and privately owned banks in general, the wholly state ownership coefficient is significantly negative. This is shown by the third regression in table 10. Total government ownership is associated with a 182.69 percentage points lower loan growth. This result contradicts the findings of Dinç (2005) who stated that the government increased lending because of political interests. Greater government ownership suggests a greater influence of political interests and thus greater loan growth.

The fourth regression in table 10 reports a regression with a measure of loan quality as dependent variable. Both the state ownership coefficient and the coefficient of wholly state ownership are not significantly different from zero, such that loan quality does not differ between state-owned and privately owned banks in general and that loan quality is not significantly different for banks who are totally state-owned. This contradicts the findings of Laeven & Levine (2009) who stated that banks with more powerful owners tend to take greater risk. However, in the crisis period, state-owned banks have significantly more non-performing loans than privately state-owned banks, which was already shown by the regression in table 6.

Although I did not find a significant difference in government securities to assets in the regression of table 8, the last regression in table 10 shows that both the state ownership coefficient and the wholly state ownership coefficient are

significantly different from zero. State-owned banks have in general a lower government securities to assets ratio, which would mean that state-owned banks finance the government less than privately owned banks. However, the total state ownership coefficient is significantly positive at the 1% level, which means that

(24)

supports the findings of Cornett et al. (2010) and Dinç (2005) who stated that government owned banks focus more on political interests and the finance the

government to a greater extent. The R-squared is larger than that of the last regression in table 8, indicating that the ratio in table 10 is more explained by the explanatory variables.

Table 11 presents bank liquidity regressions. For both measures of bank liquidity, the state ownership are not significant, suggesting that in general, state-owned banks do not differ from privately state-owned banks in liquidity. Nonetheless, the wholly state ownership coefficient of both regressions in table 11 are significantly negative and seem to have an impact. Both show a 8.66 and 33.31 percentage points, respectively, decrease in bank liquidity when banks are wholly owned by the

government. The R2 of both regressions are larger than the regressions in table 7 on the same bank liquidity measures, such that the dependent variables in table 11 are more explained by the explanatory variables.

Table 11 – Additional liquidity regressions

Explanatory variables Liquid

assets/Total assets Liquid assets/Deposits & short-term funding State ownership 0.0570 [3.0885] 11,4249 [12.3448]

Wholly state ownership -8.6573**

[4.0326]

-33.3141*** [11.5351] State ownership in crisis

period -2.1058 [3.4578] 2.6260 [15.1302] Size (ln) -0.1943 [0.1707] -0.6445 [0.6264] Crisis -0.3326 [0.7234] -2.1945 [2.410] Intercept 21.3065*** [4.0639] 40.0587 [26.2985]

Obs 4081 4048

R2 _0.1746 _0.1284

(25)

6. Conclusion and discussion

In this paper, I examine the effect of government ownership on bank performance. Moreover, I use several performance measures to indicate performance differences between state-owned and privately owned banks of 28 European Union countries from 2005 through 2012, including the crisis period from 2008 through 2012. The study covers a total number of 614 banks, and 32 of these banks are defined as state-owned. The performance measures are used as dependent variables and the

explanatory variables are state ownership, state ownership in crisis period, bank size and I control for country-, time- and bank type-specific factor by including some country- and bank type dummies. State-owned banks have, in general, less asset growth and more loans to assets. I find that bank profitability does not differ between state-owned and privately owned banks in general, though the return on assets is significantly smaller for owned banks in the crisis period. In general, state-owned banks are more efficient, but there seems no significant difference in the crisis period. An explanation could be that, in times of turmoil, banks suffer a loss in revenues but the non-interest expenses, such as personnel expenses, do not change. I find that if a bank is totally government owned, the efficiency is even better. It seems that state-owned banks have lower loan quality than privately owned banks in the crisis but this difference is not suggested to be significant in general. The loan loss reserves to gross loans ratio is higher for state-owned banks, but this should not indicate a negative impact on loan quality per se. State-owned banks could have higher loan loss reserves because they expect more non-performing loans, but privately owned banks could have less loss reserves but actually experience more impaired loans. Moreover, state-owned banks have significantly higher loans to deposits and short-term funding ratios, which could be because state-owned banks might fund loans with non-deposit liabilities and not because government owned banks are less liquid. The liquid assets ratios show no significant difference, which would mean that state-owned banks and privately owned banks do not differ in liquidity. However, I find a negative impact of wholly state ownership, suggesting that wholly government owned banks have less liquid assets and thus are less liquid.

I do not find a significant difference in loan growth between state-owned and privately owned banks, though, when including wholly state ownership into the regression, totally state-owned banks are suggested to have a much lower loan

(26)

growth. This result contradicts the findings of Dinç (2005) who states that the government increased lending because of political interests. Greater government ownership suggests a greater influence of political interests and thus greater loan growth. The government securities to total assets ratio examines if government owned banks finance the government to a greater extent than privately owned banks do. Where Cornett et al. (2010) find that state-owned banks finance the government more than privately owned banks in general, I find no such significant difference. However, this study suggests that wholly state-owned banks do have a significantly higher ratio, which indicates that they finance the government to a greater extent. This result supports the findings of Dinç (2005) and Cornett et al. (2010) who stated that government owned banks focus more on political interests and finance the government to a higher degree.

Because shareholder information was often missing, I was left with a small sample of total banks and with few state-owned banks to analyse, which might be the cause of the non-significant differences between state-owned and privately owned banks. Besides that, shareholder information was not available for each year; there was only information on a particular bank in general which could have led to different outcomes. It could be, for instants, that a bank was government owned in some years of the total examined period and privately owned for the remaining years. The coefficients of the explanatory variables could therefore be biased, since large

changes in the performance measured in a subsequent year could be due to the change in ownership structure. I suggest future research to include the differences in

ownership per year, such that the results will be more explained. I think it is important to stress why differences in bank performance occur.

(27)

References

Barth, J.R., Caprio, G. & Levine, R., 2004. Bank regulation and supervision: What works best? Journal of Financial Intermediation, 13(2), pp.205–248.

Cornett, M.M. Guo, L., Khaksari, S. & Tehranian, H., 2010. The impact of state ownership on performance differences in privately-owned versus state-owned banks: An international comparison. Journal of Financial Intermediation, 19(1), pp.74–94.

Cull, R. & Martínez Pería, M.S., 2013. Bank ownership and lending patterns during the 2008–2009 financial crisis: Evidence from Latin America and Eastern Europe. Journal of Banking & Finance, 37(12), pp.4861–4878.

Dinç, I.S., 2005. Politicians and banks: Political influences on government-owned banks in emerging markets. Journal of Financial Economics, 77(2), pp.453–479. Iannotta, G., Nocera, G. & Sironi, A., 2013. The impact of government ownership on

bank risk. Journal of Financial Intermediation, 22(2), pp.152–176.

Jia, C., 2009. The effect of ownership on the prudential behavior of banks – The case of China. Journal of Banking & Finance, 33(1), pp.77–87.

Laeven, L. & Levine, R., 2009. Bank governance, regulation and risk taking. Journal

of Financial Economics, 93(2), pp.259–275.

Megginson, W.L., 2005. The economics of bank privatization. Journal of Banking

and Finance, 29(8-9 SPEC. ISS.), pp.1931–1980.

La Porta, R. Lopez-De-Silanes, F. & Shleifer, A., 1999. Corporate ownership around the world. , pp.471–517.

Shehzad, C.T., de Haan, J. & Scholtens, B., 2010. The impact of bank ownership concentration on impaired loans and capital adequacy. Journal of Banking &

Finance, 34(2), pp.399–408.

Taboada, A.G., 2011. The impact of changes in bank ownership structure on the allocation of capital: International evidence. Journal of Banking & Finance, 35(10), pp.2528–2543.