LONG-RUN EFFECTIVE CORPORATE INCOME TAX RATES IN BANKS: A CASE OF THE REPUBLIC OF SERBIA

Efektivna poreska stopa (EPS) je jedna od najčešće korišćenih mera opterećenja porezom na dobitak. Iako se uobičajeno računa na godišnjem nivou, u poslednjoj deceniji koncept dugoročnih EPS postaje popularan. Ciljevi rada jesu poređenje godišnjih i dugoročnih EPS u bankama u Srbiji i poređenje uticaja determinanti EPS u kratkom i dugom roku. Rezultati istraživanja pokazuju da su godišnje i dugoročne EPS u bankama u Srbiji relativno niske. Iako propisana stopa poreza na dobitak iznosi 15%, najveći broj opservacija ima EPS niže od 5%. Značajan procenat opservacija ima godišnje EPS od 0%. Istraživanje je pokazalo da veće banke imaju statistički značajno više godišnje EPS. Međutim, u slučaju dugoročnih EPS, taj nalaz nije statistički značajan. Rezultati istraživanja mogu biti korisni menadžmentu banaka prilikom planiranja poreza na dobitak i poređenja poreskog opterećenja sa prosekom delatnosti, i nacionalnim poreskim vlastima prilikom reformisanja sistema oporezivanja banaka.


Introduction
The first two decades of the 21 st century have been highly challenging for the banking sector. These challenges primarily refer to the digitalization of banking operations and implementation of Internet and mobile banking concepts (Savić and Pešterac, 2019). However, after the global financial crisis, the taxation of banks has also become an important issue, since a number of banks have been bailed out with taxpayers' money. Shackelford and Shevlin (2001) argue that banks are very suitable for tax research due to high regulations imposed on them and the detailed disclosures they conduct. On the other hand, Goodspeed (2017) argues that the taxation of banks is an important issue that is not studied enough. Some previous research on banks in Serbia (for example, Knežević and Dobromirov, 2016;Marjanović et al., 2018) did not even include corporate income tax in the analysis, but measured bank profitability with income before taxation.
The motivation for running the research has been found in the paper of Dyreng et al. (2008) as they first implemented the concept of long-run effective tax rates (ETRs). They find considerable cross-industry differences in longrun ETRs. Furtherly, they find that financial institutions are among the industries with low long-run ETRs. In addition, banks in Serbia have experienced a turbulent period with major losses. In such conditions, the approximation of long-run corporate income tax burden can be useful.
The research subject of this paper is corporate income tax in banks in Serbia. After its restructuring at the beginning of the 21 st century, the banking system of Serbia is now modern, stable and relatively robust to crisis conditions (Domanović et al., 2018). The first objective of the paper is to analyze and compare short-run and long-run corporate income tax burdens of banks in Serbia. The second objective concerns the research on potentially different impacts of ETR determinants in the short and long run.
The paper contributes to previous research on the importance of long-run ETRs and, in particular, promotes their implementation in the banking sector. Prior research on the importance of long-run ETRs primarily regarded real-sector companies. To the author's knowledge, this is the first research on long-run ETRs in banks in Serbia.
These research results can be of interest to many interest groups. Bank management can benefit from knowing ETR dynamics in the long run, comparison of annual and long-run ETRs with the banking sector average, and knowing ETR determinants. National tax authorities can specially benefit when deciding on the modality of banking sector taxation. After the global financial crisis in 2008, tax authorities in many countries paid special attention to the analysis of the banking sector tax burden.
Aside from the introduction and conclusion, this paper consists of three parts. The first part features a detailed explanation of the research subject, divided in two subparts: measuring ETRs and ETRs in banks. The second presents the research methodology. Research results are given in third part of the paper.

Measuring the Effective Corporate Income Tax Rate
Due to its calculation simplicity and clear meaning, the ETR is probably the most widely used measure of corporate income tax burden. Unlike statutory corporate income tax rate as a rough tax burden measure, the ETR considers the statutory rate, the effects of tax exemptions, incentives and loans, as well as the effects of many tax avoidance mechanisms. Hanlon and Heitzman (2010) provide a detailed review of most widely used ETRs.
The ETR is usually calculated at the annual level, as an annual corporate income tax burden divided by the annual accounting result. The numerator is most commonly the total corporate income tax expense, current corporate income tax expense or corporate income tax expense paid. The denominator is usually the income before taxation. By combining the numerator and denominator, it is possible to define some of the most widely used ETRs: Dodatno, važan nedostatak EPS se odnosi na činjenicu da EPS (posebno kada je niža od propisane stope poreza na dobitak) ne razlikuje efekte državnih podsticaja i povoljnog poreskog tretmana određenih aktivnosti u kojima banka svakako učestvuje, s jedne strane, i tendencioznih aktivnosti izbegavanja poreza, s druge strane.

Efektivne stope poreza na dobitak u bankama
Važnost EPS u bankama se može posmatrati sa dva konkurentna gledišta. S jedne strane, porez na dobitak predstavlja rashod banke praćen stvarnim odlivom resursa u korist države. • cash ETR (corporate income tax paid/income before taxation). Dyreng et al. (2008) argue that it is more convenient to analyze long-run ETRs (for example, five-year or ten-year) instead of annual ETRs. One of the arguments regards high year-to-year variations of annual ETRs. They argue that long-run ETRs can mitigate such variations.
Long-run ETR is calculated as a sum of corporate income tax expense divided by the sum of results before taxation during the observed period. Dyreng et al. (2008) argue that this way of long-run ETR calculation is more convenient than using the average value of annual ETRs, since long-run ETR reduces the impact of extreme values of annual ETRs.
There is a good example in banking sector of Serbia that confirms the previous arguments. Table 1 shows the relevant data for one bank (that went bankrupt in 2018). These data confirm high variations of annual current ETRs during a five-year period and the important differences between long-run current ETR and average annual current ETR.
Annual ETRs do not have economic meaning in a situation when a company records loss before taxation. Such observations are usually eliminated in research. Henry and Sansing (2018) conclude that long-run ETRs can mitigate, though not fully eliminate, this problem. Long-run ETRs have greater usability than annual ETRs, if the portion of eliminated observations when calculating long-run ETRs is lower than the portion of eliminated observations when calculating annual ETRs.
Despite certain advantages, annual and long-run ETRs suffer from certain weaknesses (Hanlon and Heitzman, 2010). Probably the most important weakness of ETRs is the fact that they capture the effects of non-conforming tax avoidance (that reduces taxable income, while the accounting income before taxation remains constant), ignoring the effects of conforming tax avoidance (that reduces both the taxable income and the accounting income before taxation). This problem is mitigated using EBIT (income before interest and taxation), EBITDA (income before interest, taxation, depreciation and amortization) or net cash flow from operating activities, instead of income before taxation as the denominator.
Furthermore, an important weakness of ETRs concerns the fact that ETRs (particularly when they are lower than Vstatutory corporate income tax rate) do not distinguish between government incentives and the favorable tax treatment of certain activities that a bank engages in regardless, on the one hand, and tendentious tax avoidance activities, on the other hand.

Effective Corporate Income Tax Rates in Banks
The importance of ETRs in banks can be discussed from two competing views. On the one hand, corporate income tax represents a bank's expense with an actual outflow of funds in favor of government. Thus, high a corporate income tax burden may depreciate a bank's profitability (Dietrich and Wanzenried, 2014). Therefore, it is important for the banks to properly plan tax liabilities in order to minimize corporate income tax expenses and maximize profitability.

Metodologija istraživanja
The system of bank taxation has been criticized due to various points. Demirguc-Kunt and Huizinga (2001) argue that the problem lies in the fact that banking systems are dominated by multinational banking groups, while the taxation of income is conducted at the level of each separate country. This results in opportunities to avoid tax liabilities of multinational banks through arranging intragroup transactions that artificially shift income from countries with higher a tax burden to countries with a lower tax burden.
Roe and Troge (2018) conclude that a reform of the corporate income tax system of banks would contribute to the banking sector's stability. Such a reform should primarily mitigate the problem of asymmetric tax treatment between equity and borrowed sources of funding.
Prior findings on banks' ETRs considerably differ depending on the country in which the research was conducted. This is quite understandable, bearing in mind important cross-national differences in corporate income tax systems. In this regard, Ricotti et al. (2016) find that the ETRs of banks in five developed countries of the European Union are significantly different, despite the fact that the studied countries are members of the Eurozone and participants in the Single Supervisory Mechanism.
Gawehn and Muller (2019) find that banks in United States of America have significantly higher ETRs compared to other companies. On the other hand, Meeks and Meeks (2014) notice that the corporate income tax burden of banks in the United Kingdom considerably decreased during the second decade of the 21 st century.
In particular, the taxation of banks has been a current issue ever since the global financial crisis in 2008, seeing as how the recovery of many big financial institutions required state aid and the taxpayers' money (Stojković and Luković, 2019). In this regard, Weigand (2015) finds that ETRs of American and European banks decreased after the crisis, though American banks have higher ETRs than European banks.
Banks in Serbia have very low annual ETRsconsiderably lower than the statutory corporate income tax rate of 15%. Vržina (2018) uses four types of annual ETRs in the period between 2013 and 2017 and finds that the median of each ETR is lower than 2.5%. If these results are compared with the corporate income tax burden of companies from the real sector, quoted at the Belgrade Stock Exchange (Vržina, 2018a), it can be noted that the tax burden of banks is considerably lower.
Similarly to the companies from the real sector, banks in Serbia have many options to reduce ETRs. In addition, the global financial crisis and banking market deviations have resulted in losses for banks during the previous years. Such losses can be carried forward at the expense of future taxable incomes in order to reduce future corporate income tax liabilities. Vržina (2018) notes that the taxation of the banks' income in Serbia differs from the income taxation of real sector companies in terms of weaker capitalization, write-offs of loan receivables and the tax treatment of off-balance positions.
In order to accumulate funds for solving future banking crises and enabling the recovery of financial institutions, many countries imposed an additional tax burden on banks after the global financial crisis. Some of the countries imposed on their banks the payment of corporate income tax at higher statutory rate than companies from the real sector. The second option concerns the introduction of a special tax on banks, calculated after applying the statutory rate on a certain aggregate value (for example, value of loans, deposits, liabilities, owners' equity or income).

Research Methodology
Research in this paper spans a five-year period, from 2014 to 2018. A longer period (for example, ten years) was not used since it would have resulted in the overlay of the period when the statutory corporate income tax rate was 10% (until 1 January 2013) and the period when the statutory rate was 15% (as of 1 January 2013). The research includes commercial banks that were active in Serbia at the end of each observed year. There are 25 such banks identified and they are presented in Table 2.
The annual corporate income tax burden of banks was measured with the current ETR and cash ETR. The accounting ETR was not used since it contains the deferred corporate income tax expense as a non-cash income statement position. Furthermore, Vučković-Milutinović and Lukić (2013) find that a share of deferred corporate income tax in banks in Serbia is usually not materially significant. The longrun corporate income tax burden of banks is measured with long-run (five-year) current ETR and long-run cash ETR.
The paper examined whether the differences in the annual and long-run ETRs of banks can be explained with differences in certain bank features. In this regard, the link between bank size, leverage and profitability was examined, on the one hand, while the ETRs of banks were observed, on the other hand. Size, leverage and profitability are some of the traditional determinants of ETRs (Fernandez-Rodriguez and Martinez-Arias, 2014). For this purpose, t-tests of independent samples were conducted for this paper. Similar methodology was employed in some prior research (Dyreng et al., 2008;Vržina, 2018). Definitions of employed variables are given in Table 3.
Observing 25 banks during a five-year period, the sample initially consisted of 125 observations when analyzing annual ETRs and 25 observations when analyzing long-run ETRs. Due to unclear economic effect, observations with negative result before taxation were eliminated. Therefore, the final sample consists of 91 observations with annual ETRs and 16
Analiza distribucije EPS pokazuje da se godišnje i dugoročne EPS najčešće nalaze u intervalu između 0% i 5%. Grafik  observations with long-run ETRs. Using 91 of 125 observations with annual ETRs, the usability of the initial sample is at the level of 72.80%. On the other hand, this percentage for long-run ETRs is 64% (16 of 25 banks). This indicates that the implementation of long-run ETRs is of little importance for the mitigation of the negative observation problems when analyzing the corporate income tax burden of banks in Serbia.
The low usability of the initial sample in the long run is, primarily, a result of major losses of banks in 2014 and 2015. Namely, if long-run ETRs were calculated for a four-year period (2015-2018), 19 banks would have a positive long-run result before taxation, and their percentage of sample usability would be 76%. If long-run ETRs were calculated for a three-year period (2016-2018), the percentage of sample usability would be 80% (20 out of 25 banks). In other words, if a trend of increasing profitability remains in the banking sector of Serbia, the usability of long-run ETRs will be higher in the coming years.
This paper uses audited financial data retrieved from statutory financial reports of banks. These data are accessible at the Internet presentation of the Business Registers Agency of the Republic of Serbia (www.apr.gov.rs).

Research Results
Research results include descriptive statistics, distribution analysis of employed ETRs and results of independent sample t-tests. The results of descriptive statistics are presented in Table 4.
Data from the table show that annual ETRs in banks in Serbia are at a relatively low level, i.e. that they are considerably lower than the statutory corporate income tax rate of 15%. Such results are consistent with results of similar prior research (Vržina, 2018), though annual ETRs in this paper are even lower. Annual current ETR has a lower arithmetic mean, but higher median than annual cash ETR.
Long-run ETRs in banks in Serbia are, on the average, also considerably below the statutory corporate income tax rate. The arithmetic mean and median of long-run ETRs are higher than in the case of annual ETRs, which is understandable bearing in mind that long-run result before taxation includes certain losses before taxation. In addition, long-run ETRs have lower maximum values and lower standard deviations than long-run ETRs, confirming that using long-run ETRs can mitigate high year-toyear variations of annual ETRs.
The distribution analysis of ETRs shows that annual and long-run ETRs are most frequently in an interval between 0% and 5%. Figure 1 shows the distribution of observations according to the value of annual ETRs.
Naredni deo empirijskog istraživanja se odnosi na uticaj određenih karakteristika (veličina, zaduženost i profitabilnost) na EPS Annual cash ETRs have higher dispersion compared to annual current ETRs. High values of annual ETRs (higher than 20%) primarily refer to cash ETRs -there is only one observation with annual current ETR higher than 20% in the sample, compared to five observations with annual cash ETR higher than 20%.
The profitability of banks has considerably increased during the observed period. In 2018, only one sampled bank recorded loss before taxation, unlike 2014, when eleven banks recorded negative result before taxation. Such findings confirm the argument on the development of a modern and stable banking system in Serbia (Domanović et al., 2018).
Nine banks recorded income before taxation in each of the five observed years. Out of those, six banks had annual current ETR lower than 15% in each year, while four banks had annual cash ETR lower than 15% in each year. On the other hand, none of the banks had annual current ETR higher than 15% in each year, while only one bank had annual cash ETR higher than 15% in each year. Such results show that exceptionally low annual ETRs are more persistent compared to exceptionally high annual ETRs. In other words, high annual ETRs are usually a transitory occasion in banks. Chart 2 shows the distribution of observations according to the value of longrun ETRs. This distribution is relatively similar to the distribution of observations according to annual ETRs.
Eight banks have a longrun current ETR lower than 5%, while ten banks have longrun cash ETRs lower than 5%. It is interesting to note that only one bank had both long-run current ETR and long-run cash ETR higher than 15%. Two banks recorded long-run current ETRs of 0%, while four banks had long-run cash ETRs of 0%.
Presented results suggest that banks efficiently manage corporate income tax liabilities as they are in a position to keep relatively low ETRs in the long-run. In this regard, the minimization of corporate income tax liabilities leaves more resources for banks to reinvest or distribute to owners.
Results of independent samples t-tests for annual ETRs are presented in Table 5. Data from Table 5 show that larger banks have statistically significant higher annual current and cash ETRs. In taxation theory, a situation when larger companies pay more corporate income tax is known as political cost hypothesis. Vržina (2018) also finds the impact of political cost hypothesis in Serbian banking sector, but this finding is not statistically significant. However, it is important to note that the research methodology in that paper considerably differs from the methodology used here.
Banks with a higher share of liabilities in total assets have higher annual ETRs. However, this finding is not statistically significant. On the other hand, more profitable banks have statistically significantly higher annual current ETRs.
In order to determine whether the potential determinants of ETR have different impacts in the short and long run, additional t-tests were conducted for long-run ETRs. The results of these tests are presented in Table 6. For the purposes of analysing the determinants of long-run ETRs, bank size was calculated as an average of the natural logarithm of total assets (in 000 RSD), bank leverage was calculated as an average value of the ratio of total liabilities to total assets, while bank profitability was calculated as a ratio of average result before taxation and average total assets in the observed period.   Očigledno je da tradicionalne determinante EPS relativno loše objašnjavaju varijabilnost EPS banaka u Srbiji, posebno u dugom roku. Stoga je potrebno pronaći specifične karakteristike banaka koje mogu determinisati EPS. Na primer, stepen investiranja u državne dužničke hartije od vrednosti može umanjiti EPS, s obzirom na to da su prihodi od kamate na takve hartije (emitovane od strane Republike Srbije, autonomne pokrajine, jedinice lokalne samouprave i Narodne banke Srbije) oslobođeni poreza na dobitak. Prihodi od kamata na takve hartije obično predstavljaju najvažniju korekciju računovodstvenog dobitka pre oporezivanja u poreskom bilansu banaka.
Dugoročne EPS bi trebale biti izračunate i za ostale delatnosti u Srbiji u cilju poređenja dugoročnog opterećenja porezom na dobitak po delatnostima. Ukoliko bi naredni radovi proširili istraživanje na period pre 2013. godine, trebali bi koristiti razliku između propisane stope poreza na dobitak i EPS kao meru The impact of chosen bank features on ETRs is considerably different in the long run compared to the short run. In the long run, larger banks have higher long-run current ETRs, but lower long-run cash ETRs. Contrary to the analysis of determinants of annual ETRs, these differences between larger and smaller banks are not statistically significant.
Banks with a higher share of liabilities in total assets have lower long-run ETRs. This finding is contrary to the analysis of annual ETRs, but is also not statistically significant. In addition, more profitable banks have higher long-run ETRs. This finding is consistent with the analysis of annual ETRs, but is not statistically significant.
It is obvious that traditional determinants of ETRs are relatively insufficient in explaining the variability of ETRs of banks in Serbia, particularly in the long run. Therefore, it is necessary to find specific bank features that may determine ETRs. For example, the intensity of investment in government debt securities may reduce ETR since interest revenue from such securities (issued by the Republic of Serbia, an autonomous province, local government unit or the National bank of Serbia) is tax exempt. Interest revenue from such securities is usually the most important correction of accounting income before taxation in tax balance of banks.
Non-performing loans may also influence ETRs of banks. After the global financial crisis, the tax treatment of non-performing loans has become an important issue. Different countries imposed different provisions (Bholat et al., 2018), while Serbia has a relatively favorable tax treatment of non-performing loans. For example, banks in Serbia are allowed to deduct in the tax balance an amount of receivables that was not collected on the basis of real estate sale, as well as an amount of written-off receivables regarding non-performing loans (classified according to the criteria of the National bank of Serbia).

Conclusion
Research in this paper has been conducted in order to assess the annual and long-run (fiveyear) ETRs in banks in Serbia. For this purpose, the author used data on current corporate income tax expenses, corporate income tax paid and income before taxation of banks in the period between 2014 and 2018.
Research results show that, on the average, banks in Serbia have very low annual and longrun ETRs. Namely, the statutory corporate income tax rate in Serbia during the observed period was 15%, while most of the banks had ETRs lower than 5%. A significant portion of banks had ETR of 0%. Furthermore, long-run ETRs are, on the average, higher than annual ETRs, which is understandable, bearing in mind the fact that long-run result before taxation includes certain annual losses before taxation.
The analysis of determinants of annual and long-run ETRs showed that larger banks statistically have significantly higher annual ETRs. This finding is consistent with the political cost hypothesis. However, the statistical significance of this impact is lost in the long run. In addition, determinants of ETRs have different impact in the short and long run.
Research showed that banks in Serbia are highly successful in ETR minimization, both in the short and long run. High annual ETRs (higher than the statutory corporate income tax rate) are a rarity and not persistent in the long run. In other words, high annual ETRs are a transitory occasion in bank operations.
Long-run ETRs have limited usability in the banking sector of Serbia. The problem of observations with loss before taxation is not solved with the implementation of the long-run ETRs concept -on the contrary, the percentage of sample usability is higher in the case of annual ETRs than in the case of long-run ETRs. Furthermore, the number of banks in Serbia is relatively small, so the research sample was not large enough in the case of implementation of long-run ETRs.
Aside from the limited usability of long-run ETRs, research results should be interpreted in the light of other limitations. It is possible that the research results would have been different if the structure of the sample had been changed. In order to maintain the consistency of ETRs, this paper did not include samples of all active banks in Serbia, but only the banks that were constantly active during the observed period. In addition, due to important differences in national tax systems, research results should not be analogously transposed on the banking

Literatura / References
systems of other countries.
Long-run ETRs should also be calculated for other industries in Serbia in order to make a cross-industry comparison of the long-run corporate income tax burden. If a subsequent research were to expand the study to include the period before 2013, they should employ the difference between the statutory corporate income tax rate and the ETR as a measure of tax burden, due to the statutory corporate income tax rate being lower before 2013. Following the research of Fagbemi et al. (2018), a separate tax analysis of systemically important banks would also be useful.