THE EFFECTS OF INCOME SOURCE DIVERSIFICATION ON THE VOLATILITY OF BANKS’ PROFITABILITY

Uprkos iscrpnim istraživanjima koja proučavaju uticaj diversifikacije poslovnih aktivnosti banke na njene performanse, još uvijek nije postignut konsenzus po pitanju toga da li je ovaj uticaj pozitivan ili negativan. U radu se identifikuju prednosti i nedostaci diversifikacije poslovnih aktivnosti i izvora prihoda u bankarstvu, a iskazani stavovi se potkrijepljuju ekonometrijskim dokazima. Istraživanje se sprovodi na uzorku od 20 sistemski važnih banaka čije se bilansne performanse prate za period od 2005-2014 godine. Uz primjenu linearnog regresionog modela ispituje se da li nivo učešća nekamatnog prihoda u ukupnim prihodima banke direktno utiče na volatilnost njene profitabilnosti. Istraživanje pruža praktične dokaze i doprinosi debati o prednostima i nedostacima primjene modela univerzalnog bankarstva.


Strategic Choice of the Bank: Diversification vs. Specialization
Constant changes at the financial market have made it difficult for the banks to make strategic decisions, including the decision related to the choice between diversification and specialisaton in banking.While some banks choose narrow specialization, trying to offer a specific, limited range of services to their customers, a large number of banks opt for universal banking and begin to offer new services as a result of their growing presence at the capital market.
Reference literature points to the numerous advantages that universal banking brings.The implementation of this business model allows banks to diversify risks and increase revenues.The fact that they can satisfy almost all financial needs of their clients increases the negotiating power of the bank and contributes to the greater customer loyalty.Universal banking expands the range of business opportunities of banks.However, a wider range of business opportunities also means a wider spectrum of risks that these banks are exposed to.Universal banks are often defined as the fundamental problem of the global financial system, which allowed the incorporation of the investment banks' risk into commercial banks, all at the expense of taxpayers who paid for the risky moves of these banks.
While commercial banks mainly rely on interest income, universal banks rely on noninterest income as well, which entails new risks.With the deregulation of the banking sector and the expansion of banking activities into various types of businesses, the share of non-interest income in total income of banks increases.After the crisis, the research has been increasingly focusing on the advantages and disadvantages arising from the involvement of universal banks in proprietary trading and other activities at the securities market.
The impact that non-interest income has on the stability of banks' performance, primarily on the stability of their total revenues and profitability, has become an increasingly common topic.Accordingly, this paper examines whether there is a positive relationship between the level of share of non-interest income in the bank's total income and the volatility of its profitability.The objective is to determine whether the level of non-interest income share in the bank's total income directly affects the volatility of its profitability.This way it can be indirectly determined whether the involvement of commercial banks in the investment banking business leads to diversification benefits, or whether banks thereby increase their risk.

Literature Review
Numerous authors have studied universal banking and tried to determine whether such a business model of the bank increases the risks that the institution is exposed to.The impact of income diversification on the banks' performance has been mainly researched from three different aspects.The first considers whether the diversification of revenue leads to economies of scale, the second one examines the link between income diversification and operational performance of the bank, while the third explores the relationship between diversification of income and banks' risk (Zhou, 2014).
Despite the exhaustive research investigating the impact of the diversification of a bank's business activities on its performance, there is still no consensus on whether this impact is positive or negative, because there are arguments that confirm the opinions of both sides (Gambacorta and Rixtel, 2013).While one argues that this reduces the banks' risk, because banks increasingly diversify their sources of income and become less dependent on the interest income, others argue that this increases the risk, because the banks' reliance on a volatile income source increases (in comparison with a relatively more stable interest income).Noninterest income can bring new operational, credit, legal and market risks, as well as liquidity risk (Zhou, 2014).The proponents of the opinion that diversification positively affects the bank's performance also argue that cross-selling of various products increases revenues and, therefore, the bank's profit.Portfolio  and interest income of a bank were negative or at least poorly correlated, non-interest income could diversify the bank's income and could have positive impact on the trade-off between risk and profitability (Stiroh, 2002).Saunders and Walter (1994) are among the authors who studied this issue.Considering 18 studies that investigate whether non-banking activities reduce the risk of bank holding companies, they conclude that 9 studies prove that this reduces risk, 6 studies prove that this does not reduce risk, while 3 studies provide mixed results.Templeton and Severiens (1992) researched the market data related to 54 banking holding companies for the period 1979-1986 and concluded that diversification (measured as a share of market value which is related to nonbank assets) is associated with a lower variance of shareholder returns.Similarly, De Young and Roland (2001) considered the relationship between banking profitability, volatility and participation of different types of income for 472 large commercial banks for the period 1988-1995.They concluded that all activities based on fees (income from all sources except interest, investments, deposits and trading activities) increase the volatility of banks' income and earnings.
Fiordelisi and Ibanez (2011) analyzed a sample of 709 European banks for the period 1997-2007 and concluded that with the growth of diversification (measured as a share of noninterest income in total income), the risk grows as well, including not only the risk which refers to that particular bank, but systemic risk or the risk of the entire financial system as well.Van Ewijk and Arnold (2013) analyzed a sample of 16,000 FDIC-insured banks in the US for the sample period 1992-2010.This way, they concluded that the banks involved in traditional activities are better prepared for a financial crisis than the banks that diversify their business activities.Stiroh (2002) analyzed the data for 14,503 banks for the period 1978-2000, with the aim of exploring the link between the growing reliance on non-interest income and the volatility of banking income and profit.The results of the work provide very little evidence that diversification leads to more stable profits or income.At the aggregate level, non-interest income has higher volatility than interest income.At the level of individual banks, noninterest income shows a rising correlation with interest income.Therefore, as the interest income correlation with non-interest income increases, diversification benefits declines.For example, administrative fees and commissions are highly correlated with interest income.From the aspect of risk and return, there is a clear negative link between the share of non-interest income and profit per unit of risk.Based on the reference literature so far, the same author concluded that there is very little evidence of diversification benefits.
Martel et al. ( 2013) examined the business models applied by international banks during the 2007-2009 crisis.They concluded that commercial banks were more resilient than universal banks, due to the fact that they had modest exposures to trading activities and derivatives, and therefore dominantly relied on a stable source of financing.In contrast, banks that were more involved in investment banking, especially investment banking-oriented universal banks, experienced significant changes in the balance sheet structure over time.Following the return on equity on the sample of 10 systemically important financial institutions for the period 2006-2010, the authors noted that majority of banks experienced a strong decline in profitability in 2008.However, commercial banking-oriented universal banks had a more stable return on equity and managed to maintain it at a high level in the period 2006-2010, which cannot be said for investmentoriented universal banks.Lepetit et al. (2008) analyzed a sample of 734 European banks for the period 1996-2002 in order to examine the relationship between product diversification and the banks' risk.They concluded that with a stronger involvement of banks in the activities resulting in non-interest income, the banks' operational risk is growing.They also concluded that this risk is higher for the banks with assets above GBP 1 billion.While the majority of empirical evidence is based on linear models, Gambacorta and Rixtel (2013) used a non-linear model and proved that there is a strong link between the bank's diversification rate (measured as the share of non-interest income in total revenues) and coefficient of Pajović M. The Effects of Income Source Diversification on the Volatility of Banks' Profitability od oko 50%, dok veći nivoi diversifikacije smanjuju profitabilnost.Pošto prinos na sopstveni kapital nije mjera profitabilnosti banke koja je usklađena sa rizikom, autori mjere i kako diversifikacija prihoda utiče na volatilnost prinosa na sopstveni kapital.U tu svrhu je korišćena regresija uporednih podataka za koeficijent varijacije prinosa na sopstveni kapital.Zaključak do kojeg dolaze je da diversifikacija prihoda ima pozitivan efekat na volatilnost prinosa na sopstveni kapital do nivoa od oko 70%.
Empirijsko istraživanje sprovedeno je na uzorku od 20 banaka: Citigroup, Wells Fargo, JP Morgan Chase, Bank of America, Credit variation of ROE.Namely, these authors used a sample of 108 international banks from 14 developed countries for the period 2000-2011.Their first conclusion is that diversification has a positive and growing effect on ROE up to the diversification level of about 50%, while higher levels of diversification reduce profitability.Since ROE is not a profitability measure compliant with risk, the authors also measure how the diversification of revenues affects volatility of ROE.For this purpose, they used the cross-sectional regression for the coefficient of variation of ROE.They concluded that diversification has a positive and increasing impact on ROE volatility up to a level of roughly 70%.
In order to study the impact of income diversification on the banks' risk, Zhou (2014) analyzed a sample of 62 Chinese banks for the period 1997-2002.The author used standard deviation of return on assets as a measure of risk, while the share of non-interest revenue in total revenues was the measure of income diversification.It was concluded that income diversification does not lead to a significant reduction in the banks' risk.The author explained that non-interest income brings additional risks and that the correlation between interest and non-interest income weakens the effects of diversification.

Data and Research Metodology
The established hypothesis claims that there is a positive relationship between the share of non-interest income in the total income of a bank and the volatility of its profitability.This way it can be indirectly determined whether the involvement of commercial banks in investment banking business leads to the manifestation of diversification benefits, or whether this way the banks increase their risk.In order to test the hypothesis, we used the linear regression model that measures the impact of non-interest income share in total income on the volatility of the banks' profitability.To test this link we used the econometric method -i.e. the least squares method.In order to evaluate the linear regression model, we resorted to the econometric and statistical software tool EViews version 7.
The empirical research was conducted on a sample of 20 banks: Citigroup, Wells Fargo, JP Morgan Chase, Bank of America, Credit Suisse, ING Bank, Morgan Steanly, Standard Chartered, Santander, Unicredit Group, Group Credit Agricole, HSBC, BNP Paribas, Barclays, Deutsche Bank, Societe Generale, State Street, Bank of NY Mellon and Goldman Sachs.Each of the mentioned banks was on the List of Global Systemically Important Banks published by the Financial Stability Board in 2014.The analysis uses the panel data collected from the annual financial reports of selected banks for the period 2005-2014.The planned sample period was 15 years, but the lack of data for some of banks is the reason why the monitoring period is shortened to 10 years.This time frame has been selected in order to include the years of the crisis, when the banks demonstrated their greatest weaknesses.The available data were collected for each bank separately and were taken over from the annual financial reports that the banks submit to their shareholders before the Annual Shareholders' Meeting.
The sample consists only of systemically important banks, due to the strong impact they can have on the stability of the whole system and economy.Several authors who studied this topic opted for a similar sample of banks, relying on the list of the Financial Stability Board.Masciantonio and Tiseno (2013) suggest that these banks experienced more radical changes than the small banks and were large enough to compete in a growing, global and integrated financial system.At the same time, these banks had the appropriate structure to react quickly to regulatory changes and financial innovations and to compete with the banks with different business models.
One of the most frequently used indicators of the banks' performance is return on equity (ROE).Since ROE is not a risk-adjusted measure of profitability, this paper analyses the impact of income diversification on volatility of ROE.The coefficient of variation of ROE was used as a measure of volatility of ROE.The survey uses panel data.Since the coefficient of variation of ROE has been chosen for a dependent variable, there is only one coefficient of variation for each bank in the period.Namely, for each of the twenty selected banks from the sample, Pajović M. The Effects of Income Source Diversification on the Volatility of Banks' Profitability kapital, postoji samo jedan koeficijent varijacije za svaku banku u periodu.Naime, za svaku od dvadeset odabranih banaka koje čine uzorak je izračunat odnos standardne devijacije njene profitabilnosti (mjerene prinosom na sopstveni kapital) i aritmetičke sredine, u periodu od 2005-2014.Dobijeni količnik, koeficijent varijacije ROE (CV_ROE), je zavisna varijabla u modelu.Nezavisna varijabla je racio diversifikacije (RD), mjeren kao prosječno učešće nekamatnog prihoda u ukupnim prihodima banke za dati period.Osim racija diversifikacije, uvedene su i tri kontrolne varijable: prosječan racio primarnog kapitala koji mjeri odnos kapitala i rizikom ponderisane aktive (TIER), koeficijent varijacije zarade po akciji za dati period (CV_EPS) i koeficijent varijacije profita prije poreza za dati period (CV_BD).
Kao što se vidi sa grafika 2, logaritamskom transformacijom postignuto je da varijabla prati normalan raspored kako bi se ispunio neophodan uslov za primjenu linearne regresije.Po istom principu je izvršena logaritamska transformacija kontrolnih varijabli, koeficijenta varijacije zarade po akciji i koeficijenta varijacije profita prije odbitka poreza.Pregled varijabli korišćenih u istraživanju prikazan je u Tabeli 1. we calculated the ratio of standard deviation of its profitability (measured by ROE) and the arithmetic mean, for the period 2005-2014.The resulting coefficient of variation of ROE (CV_ ROE) is the dependent variable in the model.The independent variable is the diversification ratio (RD), measured as the average share of non-interest income in total bank income for the chosen period.In addition to the diversification ratio, three control variables have been introduced: the average Tier 1 capital ratio which measures the banks' core equity capital to its total risk-weighted assets (TIER), the coefficient of variation of earnings per share for the chosen period (CV_EPS) and the coefficient of variation of profit before tax for the chosen period (CV_BD).
The initially defined model was confronted with a certain limitation at the very beginning.Namely, the coefficient of variation as a measure does not have a normal distribution, instead being skewed right.Consequently, the assumption of the normality of the dependent variable was not fulfilled, nor was it possible to apply the least squares method for the purpose of this model's evaluation.This can be seen from Chart 1, but also from Jarque-Bera test statistics.
In order to achieve the normality of the residuals, which is the assumption for the use of the classical linear regression model, we conducted a logarithmic transformation of coefficient of variation of ROE, followed by the distribution and Jacques-Bera statistics, as presented in Chart 2.
Izvor: Formulacija autora Logarithmic transformation of coefficient of variation of return on equity.Return on equity measures the ratio of net income to shareholders equity at the end of the reporting period.

Independent variable RD
The diversification rate, which measures the average share of noninterest income in the bank's total income.

Control variables
TIER Tier 1 capital ratio which measures the bank's core equity capital to its total risk-weighted assets.

log(CV_EPS)
Logarithmic transformation of the coefficient of variation of earnings per share.

log(CV_BD)
Logarithmic transformation of the coefficient of variation of profit before tax.
The final test was the multicollinearity test.The variance inflation factors were presented in Table 3.
After the multicollinearity test has been conducted, it was established that the value of the variance inflation factor (VIF) for all coefficients was below the limit value 5. Thereby, it was proven that the independent variables are not perfectly or significantly correlated with each other.

Analysis of Research Results: Conclusion
The regression results showed that the diversification ratio has a positive impact on the coefficient of variation of ROE.The estimated coefficient of the independent variable is 0.010899.This actually indicates that, if the diversification ratio increases by 1% (i.e. if the bank increases the share of non-interest income in total revenues by one percent), the logarithmic coefficient of variation of ROE will increase by 0.010899%.It has already been explained that the coefficient of variation as a measure does not have a normal distribution, instead being skewed right.Therefore, we conducted the logarithmic transformation of the coefficient of variation of ROE.Consequently, the interpretation of this model is more complicated (it can be said that there is a trade-off between the complexity of the model and the easiness of the coefficients' interpretation).Namely, for each individual value of the coefficient of variation of ROE it  The Effects of Income Source Diversification on the Volatility of Banks' Profitability logaritamska transformacija koeficijenta varijacije prinosa na sopstveni kapital.To za posljedicu ima otežanu interpretaciju modela (može se reći da postoji trade off između složenosti modela i lakoće interpretacije koeficijenata).Naime, za svaku pojedinačnu vrijednost koeficijenta varijacije prinosa na sopstveni kapital, moguće je izračunati procenat njenog povećanja u slučaju kada se logaritam te vrijednosti poveća za 0,010899% (to jest u slučaju rasta racija diversifikacije za 1%).
Pajović M. Uticaj diversifikacije izvora prihoda na volatilnost profitabilnosti banaka Bankarstvo, 2018, vol.47, br. 2 is possible to calculate the percentage of its increase in the case when the logarithm of that value increases by 0.010899% (i.e. in the scenario when diversification ratio increases by 1%).
The sign of the calculated coefficient is in line with the results obtained by some other, already mentioned authors.The sign of the coefficient indicates that there is a positive correlation between the share of non-interest income in the bank's total income and the volatility of its profitability.It is concluded that the engagement of commercial banks in investment banking does not lead to diversification benefits, but on the contrary, that the banks are thereby increasing their risks.Therefore, the hypothesis that claims that there is a positive connection between the share of non-interest income in the bank's total income and the volatility of its profitability is accepted.
It is important to underline certain limitations of this research.The first limitation is the sample size.Although the banks are monitored in the period from 2005 to 2014, the sample consists of 20 banks, and, therefore, the impact that one observation can have on the sample is high.The second limitation is the fact that the banks from the sample prepared their annual financial statements in different currencies (EUR, USD, GBP and CHF).In order to overcome this problem, only the relative measures were used in the paper.The third limitation is reflected in the fact that the coefficient of variation does not have normal distribution, instead being skewed right.Therefore, we conducted its logarithmic transformation.However, this may indicate that there is a non-linear correlation between an independent and a dependent variable.
The banks began to engage in a wide range of different activities by offering complex products and services.That weakened the market discipline.The strong and multiple connection between investment and commercial banks is one of the main reasons for the massive losses of banks that the taxpayers have had to cover.Many banks, using the situation of inadequate regulation, new products, new markets and opportunities for operating globally, enjoyed the support of taxpayers during the crisis only because they were too big to fail.When it comes to the large, complex, universal banks, which combine the services of commercial banking, investment banking and insurance within a single entity, the problem becomes even more prominent.
After the crisis there has been a large number of advocates for the introduction of a new law that would separate the activities of commercial and investment banks, such as the Glass-Steagall Act.Such initiatives have been launched not only in the United States, but also in Europe, where the universal banking model has always been dominant.This points out to the generally negative public opinion regarding universal banking.The new proposed and adopted measures aim to provide the stronger regulation of the financial sector.Bearing in mind that even before the crisis banking was the most regulated activity in the world, it can be concluded that the crisis is a regulatory failure which is not reflected in deregulation, but in inadequate regulation.The choice between specialized and universal banking is not as simple as it seems at the first glance: universal banking implies a complex network of relationships, and it is not always easy to determine whether it is the most suitable model for the organization of a single bank.

Table 3 :
Variance inflation factors Source: Author's calculation using software package EViews Pajović M.