WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

Use of Financial Ratios in selecting entities for Tax purposes - empirical study in Albania.

REZARTA SHKURTI (PERRI) Professor, University of Tirana, Albania Orcid ID: 0000-0002-2126-2339

ELENA MYFTARAJ (TOMORI) Professor, University of Tirana, Albania

ELSIA GJIKA Tax Directorate, Albania

Abstract: Information from financial statements and reported financial ratios have long been used to detect common phenomenon such as fraudulent financial statements, , and the relation between financial ratios and the level of tax risk of an entity. The focus of this study is to research the use of financial ratios that entities declare, in the detection of the magnitude of tax avoidance. In this paper we apply a binary logistic regression to detect which ratios differentiate between tax evading and non-tax evading entities. We analyse data from 183 tax audited Albanian entities for 2015 and 2016 years and calculate several financial ratios to determine the level of tax risk based on the tax evasion magnitude found by the tax audit of these entities. We apply univariate and multivariate analysis and find several important ratios that can indicate quite accurately the high risk of tax audit of an economic entity. We suggest including these ratios as risk indicators or “red flags” in the selection procedures employed by the tax auditors. As tax reporting and financial reporting have similarities across countries of the region, our findings may be useful for other Southern Eastern European Countries as well.

Keywords: tax audit, tax risk, financial ratios Received: March 10, 2021. Revised: April 8, 2021. Accepted: April 12, 2021. Published: April 14, 2021.

1. Introduction income and indirect taxes). To mitigate the overall level of the tax risk, the Tax The accurate assessment of the tax risk levels Administration employees, (the tax auditors), of economic entities is an important focus of engages in tax , whose primary focus is the tasks of any Tax Administration. It is the examination of an organization's or extremely important for the tax authorities to individual's tax return to verify that financial be able to manage the scarce tax audit information is being correctly reported. resources with effectiveness and efficiency, Usually, tax audits represent a considerable hence the special focus on determining the time demand for tax auditors as they imply tax risk which in turn would orient the scarce on-site controls and inspections for the tax audit resources towards the entities with a audited companies what is a major hassle for high tax risk. The tax risk from the them. Therefore, carefully selecting the tax perspective of the Tax Authorities is the risk audit subjects represents a major purpose in that companies may calculate, report, or pay effectively managing Tax Administration incorrect amount of taxes (including both resources and minimizing the probable false

E-ISSN: 2224-3496 297 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

identification of either tax evading or non-tax procedures of manually selecting the tax evading entities. audited companies. There have been attempts to link the To manually select the entities tax risk level of an entity with risk indicators which will be subjected to a tax audit, the tax (red flags) extracted from its financial auditors usually rely on continually statements and filed and reported monitoring the performance and activity of information. Therefore, this study aims to the companies such as the amount of the contributing towards a stream of research that reported , the amount of declared uses financial ratios and reported financial and paid Value Added Tax and other taxes information to detect high tax risk entities. payable. During this process they are also Usually, the selection procedures of oriented and guided by some Risk Indicators the companies that will be subjected to a tax which are traditional financial ratios easily audit go in two stages. In the first step the calculable from reported financial statements Risk Module of the Information System of – current ratio, acid test ratio, total debt ratio, the Tax Administration, based on indicators debt to ratio, times interest earned and red flags derived by its internal Business ratio, return on , and Intelligence algorithms, selects up to a certain operating margin. percentage of entities that will be audited. On a closer review of the Risk The remaining entities are manually selected Indicators, we observe two aspects which, by the tax auditors based on a close follow up given the high reliance that the tax auditors of the activities of the entities (in Albania this have on these indicators, can be considered represents a 70%1 to 30% ratio between both faulty: first, the Risk Indicators do not categories). In this paper we specifically represent a complete and comprehensive focus on the second step, the selection coverage of financial ratios from different procedures that tax auditors usually follow to perspectives of performance analysis; there select what they think should be the entities are liquidity ratios, debt ratios and that should be audited. To have a more profitability ratios, but no turnover ratios and objective framework to select the audited neither flow indicators. As a second companies, we propose to make use of observation, we find that the thresholds for indicators or red flags from the financial some of the ratios are far from the widely statements, hence the tax auditors will not be accepted optimal values for these ratios (for perceived as being subjective in their choices. instance, we see an optimal value of current According to the data published from the Tax ratio to be 0.42 times, which is quite far from Administration in Albania for 2017 the the optimal value of 1.50 times, widely amount of tax evasion discovered from accepted for this ratio). What we aim through entities manually selected from tax auditors this research is to “update” the list of is higher (23.274 Euro on average) than the financial ratios which could serve as red flags amount of tax evasion discovered in entities or risk indicators for the tax auditors, but not selected by the Risk Module of the only. The findings of this study also aim to Information System (20.558 euro on fill the gap that exists in the literature of average). This again supports the idea of how financial ratios use in earnings management, important it is having clear and unbiased fraudulent financial reporting and tax audit in

1 The ratio 70% to 30% in favor of selection from the Risk Finances of Albania and whether that rule is yielding the Module is a policy imposed from higher levels in decision- best results on the performance of the Albanian Tax making ranks of the Tax Administration and Ministry of Administration or not, it is a matter that goes beyond the scope of this study.

E-ISSN: 2224-3496 298 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

Albania and the South Eastern European incentives to engage in earnings management Region. Some relevant studies can be found in both dimensions, inflating earnings to in [1] and [2], but nevertheless we observe achieve the maximization of shareholders there is more to study in this regard. value and on the other hand deflating The rest of this paper is organized earnings to minimize the taxable income [3; as follows. In the following section we focus 4; 5; 6; 7; 8; 9; 10]. We find less studies, though on the literature review by adopting a having a specific focus on the connection relatively broad view. While studies focusing between financial ratios and tax risk. in finding linkages between audit risk (and Therefore, we adopt a broad approach in the especially tax audit risk) and financial ratios literature review section by considering all are very scarce, research in the broader studies about fraudulent financial reporting context of earnings management and and earnings management. Our special focus fraudulent financial reporting are more is on papers that reveal how we can use the common and offer rich insights as to the information in the reported financial probable indicators and variables that we statements to detect manipulated financial could use. Next, in the third section we statements. present the methodology by giving an For having a complete overview on overview of the sample selection procedures, previous results on this field we searched variables included, and the techniques used to through work from various authors, [11; 12; analyze the data. In the data analysis section, 13; 14; 15; 16; 17; 18; 19] whose research is we present the two analysis tools we have based on using financial ratios for detecting employed, the univariate and the multivariate manipulative financial statement. Their technique and the logistic regression models. findings suggest that some financial In the last section, we summarize the main statement figures are informative and that the findings of the study and come up with use of several financial ratios could be useful conclusions as to which are the most in detecting companies that might have been informative ratios regarding the tax risk level. involved in fraudulent activities and general Advantages of this study are the use of manipulation of financial statements. confidential information from Tax One study based on Finnish firms [20], which Authorities Database and Information are mostly known to manage earnings System (accessed via a Non-Disclosure downward (a similar situation as that of Agreement with this Institution) the use of Albania) suggested that fraudulent real cases and findings from recent tax companies in addition to managing earnings auditing processes. Having said that, as an downwards on a grander scale, empirical research, this study is limited to simultaneously manage earnings upwards in reflect the characteristics of Albanian a smaller scale. They show that firms tend to companies and is also limited to the time adjust the earnings and the tax audit risk, to frame constraints of the data set used to avoid “being caught” thus requiring a greater generate the model. attention from tax authorities to select more carefully the entities that will be subject to tax 2. Literature Review audits in the future. We find the literature to be complete in We see that the earnings management articles, papers and studies that focus on the and fraudulent reporting research usually fraudulent financial reporting and earnings employs either the discriminant analysis or management. These studies usually describe logistic/probit models. To achieve more techniques to detect manipulation and reliable results from empirical studies, and to

E-ISSN: 2224-3496 299 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

expand the scope of the research, authors inconsistencies, could lead to obtaining a have combined the use of big data sets with higher chance of preventing problems with advanced and innovative techniques. In one internal controls or detecting them when they study in 2002 [21], it was attempted to occur. The same study emphasized the identify the factors associated to Fraudulent importance of implementing continuous Financial Statements using the multi-criteria auditing embedded modules in the decision aid (MCDA) technique. Their Information System to help auditors have results reveal that the proposed MCDA higher assurance level of transaction integrity methodology outperforms traditional and reliability. Other authors also propose the statistical techniques which are based in only use of big data sets combined with the data one criterion. They also highlight the mining techniques to increase the importance of using the familiar financial effectiveness of selection of audited [25]. The ratios such as the total debt ratio, the techniques proposed by them were reported inventories to ratio, the net profit to be successful and could help the tax margin and the sales to total assets ratio as administration to better manage the process indicators to detect fraudulent financial of selecting the entities that will be audited statements. One year later, in 2003, in USA and minimizing error I and error II of false another study [22] indicated that the use of categorization of entities in risk of fraud. certain financial data from financial Advanced tools in detecting the statements could be successfully used to entities with the probable higher tax audit correctly assess the taxpayers by the tax risk, such as the neural networks were also auditors. used in a study in 2011 in Finland [26]. The In another study [23] the use of authors investigate how well an unsupervised computer-assisted audit techniques (CAATs) neural network method – the Self-Organizing to investigate for fraud and accounting Map (SOM) – can perform in detecting those irregularities detection based on big data, was companies that need to be tax audited. The employed. The study proposes an expert findings of the study report an system for identifying suspicious outperformance of SOM neural networks to irregularities in detailed financial data and ordinary approaches commonly used by the techniques for automatic analysis of tax inspectors, hence introducing a company on a large scale, identifying possibility of using such a technique to irregularities. They even name their system improve the overall performance of the tax Sherlock and demonstrate that if possessing auditing processes. huge data sets with detailed financial As another common theme in information it is possible to use advanced literature review on detecting fraudulent technological tools to build an expert system financial statements for tax purposes we that assists auditors in detecting irregularities encounter the use of the Benford’s Law as a and fraud. This idea could be useful not only rule that describes the distribution of first for Tax Administrations but also for the big (leading) digits in economic or accounting audit firms that have already collected data, but also in every kind of data. Two enough data to train the expert systems. studies in 2008 and 2018, [27 and 28] have In one study of 2011, [24] the use of studied whether Benford’s Law can be used sophisticated audit tests was analyzed by to improve selection of the entities prior to arriving at the conclusion that the tax audits, thus increasing effectiveness and combination of sophisticated audit tests with efficiency of this process. The findings from a graphical presentation of possible both studies report that manipulated financial

E-ISSN: 2224-3496 300 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

statements data deviate significantly from evidence on the phenomenon of earnings Benford’s Law rule of numbers and therefore management among Albanian companies, this law can be used as a tool for audit using the Jones model. The results confirm selection. that firms in the Albanian market are engaged Another stream of research in the in earnings management initiatives mainly literature review is the use of financial ratios for tax purposes, therefore demonstrating and other reported information derived from deflation of revenues and income. the financial statements as variables linked to the level of tax risk of entities. In a 2001 3. Methodology article, [29] the authors demonstrate by a The purpose of this research is to reveal real-life example the importance of using whether the information contained in the basic techniques like financial ratios and indicators could indicate vertical, horizontal, and ratio analyses that tax level risk for entities, thus helping tax can give an auditor hints in identifying auditors during the selection procedures and. probable fraudulent financial statements. To find if there is any link between financial Indexes such as sales in receivables index, information and tax risk, we will study the gross margin index, quality index, sales information represented in the financial ratios growth index, total to total assets that can be computed from reported financial index derived by the original Beneish model statements and the level of tax evasion that is were also discovered to be useful. Another subsequently discovered by tax auditors after study [30] examined the detection of they have conducting the tax audit. fraudulent financial reporting using both financial ratios and nonfinancial factors 3.1. Sample selection related to corporate governance and This study was performed by the end of 2018 employing a probit technique. This study and the first months of 2019. Therefore, the found that a combination of quantitative latest data available was from reports on tax financial ratio indexes and qualitative audits performed during 20172 (data for year corporate governance factors such as CEO 2018 was not yet reported by the time of power, management control systems, senior sample design). The data we have used in this management turnover insider stock trading, study is derived from tax audit reports for appears to work best for detecting fraudulent year 2017 for all regional tax directories of financial reporting. Studies, discovering Albania. This is confidential information that discrepancies between the accounting we could access only because of the current information and the real situation [31; 32], are engagement of one of the authors of the also useful in emphasizing the importance of paper. The data was made available to us reported financial information. upon a Non-Disclosure Agreement. Earnings management studies in In total, 1402 economic entities Albania have also been very limited. One were audited in Albania in 2017. We decided study in 2011 [33] shows the existence of to select a sample consisting of 15 percent of practices or attempts for earnings the total population, which is about 210 management on behalf of Albanian entities. The sample was randomly selected companies. Another Albanian author, [34] in using the random selection option of MS her study of 2013 also searches for empirical Professional Excel, as the list of audited

2 The tax audits performed in a certain year cover two focus on data from financial statements and tax return of previous financial years, and this means that the tax audit 2015 and 2016 for the audited entities. engagements conducted in 2017 and reported in 2018,

E-ISSN: 2224-3496 301 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

entities was received in an excel file format. whereas if the level of the magnitude of tax As the second step, after having the sample, evasion was found to be higher than 50.001 a list of selected entities, we accessed (again Albanian Lek (400 Euro), we have coded that under a confidential privileged access rights entity with 2. clause) the Information System of Tax Administration and downloaded the full 3.2.Variables included in the study. financial statements and tax returns for the Previous research on using financial ratios audited periods (2015 and 2016) for all the for detecting overall manipulative reporting companies in the sample. As a third step we revealed that the use of several ratios might carefully observed the data we had and be informative and useful in detecting decided to discard from the sample those companies that more than others might have entities which either were microenterprises been involved in fraudulent activities and (with less than 5 employees), or for which manipulation of financial statements. From there was not complete financial information reviewing these studies, [10; 16; 17; 18; 35; 36; in the database of Tax Administration (either 37; 38; 39; 40], we found that several common one statement or the other was missing). ratios were consistently reported to detect After this step of data selection, we were left fraud and manipulating activities. For with 183 entities at our final sample, for example, likely items to be subject to which we have complete financial statements manipulation are sales, accounts receivables and filed tax returns for year 2015 and 2016 altogether with the allowances, and inventory and also the Tax Audit Report which reports because of their subjective nature and the for the level of tax evasion, if any found after judgements involved with determining the the tax audit is performed on these entities. reported amount of these items. Usually, the Further, to facilitate the data fraudulent companies report inflated or analysis and to perform different tests, we deflated sales by inappropriately applying the have sub classified the sample first according revenues recognition standards; they do not to the industry activities and second match sales with the appropriate cost of according to their size. We have grouped the goods sold and this is reflected in ratios such entities by type of activity: construction & as profit margins (on a gross, operating, or net manufacturing, service & transport, and basis); or they tend to report the inventory at commerce, and coded them by category with a lower cost than cost or market value hence 1, 2 and 3 accordingly. A further impacting again the but categorization of entities was performed also the inventory to total assets ratio. Also, according to their size (size refers either to higher debt ratios are reported [35] to be total assets figure, total revenues or to the associated with fraud due to the probable number of employees). The size of the firm wealth transfer from debt holders to will later serve as a controlling variable managers which increases as during data analysis. We also divide the increases as shown in at least one study [41]. companies in the sample in two different Based on these findings from categories based on the level of magnitude of literature review we pooled a list of ratios that tax evasion which will be the dependent previous studies have found to be relevant in variable in this research. If the tax evasion the financial analysis for tax audit purposes. was from 0 – 50.000 Albanian Lek This list is further completed with other (approximately 400 Euro) we have ratios, mainly suggested from theoretical considered the entity to be risk-free or of literature (academic textbooks) and also minimal-risk and have coded it with 1; based on practical experience and insider

E-ISSN: 2224-3496 302 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

information of the authors. (With insider indicators in the list of ratios/variables of the information we refer to confidential study). information from Tax Administration The list of 10 financial ratios in database on the main areas identified as more table 1.1 will be the first independent prone to tax evasion for audited Albanian variables included in the study (later we will companies in 2017. For example, include 8 more variables to the list). We have deflating by non-reporting revenue was calculated these ratios for the firms included observed in 47% of all cases – 656 entities; in the sample and will try to use the dataset to errors in completing tax return file in 32 % of conduct analysis that can reveal which of the cases – 449 entities; other cases – 21 % are ratios contain information on identification of related to non-declaration of . Based probable tax evading entities. on these findings we inserted related

No. Category Financial Ratio Optimal Source of Codification Standard Value standardization of optimal vale

1 Current Ratio 150 % Theoretical 0 if less than 150% literature 1 if more than 150% 2 Acid test (quick) Ratio 100 % Theoretical 0 if less than 100% literature 1 if more than 100% 3 Liquid Ratio 50 % Theoretical 0 if less than 50%

Liquidity Ratios literature 1 if more than 50% 4 Non-Current Liabilities to Total 30 % Theoretical 0 if more than 30%

Assets Ratio literature 1 if less than 30% 5 Total Debt Ratio 50 % Theoretical 0 if more than 50% literature 1 if less than 50% 6 Times interest earned 5 times to 1 Theoretical 0 if less than 5

Debt Ratios literature 1 if more than 5 7 Net Profit Margin Varies according INSTAT data about 0 if less than average of to industries average profit industry margin according 1 if more than average to each industry of industry

8 Profit before Taxes (EBT) Margin Varies according INSTAT data about 0 if less than average of to industries average profit industry margin according 1 if more than average to each industry of industry

9 ROA 7 % adjusted with Theoretical 0 if less than 7% before inflation (1.6% in literature adjustments Profitability Ratios 2015 and 2.6% in 1 if more than 7% before 2016) adjustments 10 Return on Equity ROE Interest rate Theoretical 0 if less than interest rate percentage literature 1 if more than interest (2.25% in 2015 rate and 2% in 2016)

Table 1.1. List of financial ratios used as independent variables. Source: Primary data of the study. Formulas and optimal standard values of these ratios are based either on wide known theoretical literature of financial statement analysis, or on data derived from INSTAT, the Albanian Institute of Statistics

E-ISSN: 2224-3496 303 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

As mentioned, we also added eight other 2016. These other ratios are shown in table various financial ratios, for years 2015 and 1.2. No. Other financial ratios 1 NCA Ratios = Fixed Assets to Total Assets 2 Inventory Ratio = Inventory to Total Assets 3 (Inventory + Accounts Receivable) / Total Assets 4 Total Personnel Expenses to Total Revenues 5 Operating Profit Margin 6 Administrative Expenses to Total Revenues 7 Difference between taxable income and reported income to total Revenues 8 Cash Flow from Operating Activities to

Table 1.2: Financial ratios calculated for entities included in the sample for both years but not included in the univariate model due to lack of standardized information from INSTAT. This information is part of the database and is utilized only in the multivariate analysis. variable we will apply both a univariate and We included these financial ratios in the a multivariate technique analysis to our study, because based on our personal sample of 183 audited entities to find if there experience, and on findings from tax audits is any connection between the financial ratios we noticed that these ratios were important (independent variables) and the level of the and relevant during the audit process. tax risk (dependent variable). Nevertheless, we could only partially include these ratios as variables in the study due to 4. Data analysis the lack of optimal or standard values in literature and because we found no Based on the literature review where we find comparable data in the webpage of INSTAT, as the most widely used techniques either the Albanian Institute of Statistics, like we discriminant analysis or logistic regression, have for the variables included in table 1. we have decided to use the latter as the most The independent variable of the appropriate given the characteristics of the study is the magnitude of tax risk of tax sample. We have analyzed the data through audited entities. As this is an empirical study two different techniques. First, we have and is based on real tax audit cases, the non‐ performed a univariate analysis focusing on compliance with the Albanian tax laws is each independent variable one at a time and used as the measurement of the fraudulent tax at a second stage we have performed a reporting and therefore the tax risk is multivariate analysis based on a binary measured by the amount of the tax evasion logistic regression model. discovered to be performed by the entity. If In the univariate analysis we have the tax evasion amount for each studied entity compared if there is a statistically significant in the sample was found to be low (from 0 – difference in categorizing entities according 50.000 Albanian Lek approximately, 400 to their tax risk and according to the optimal Euro) we have considered the entity risk-free level of each financial ratio. This comparison or minimal-risk and have coded it with 1, is performed independently for both years, whereas if the level of the magnitude of tax 2015 and 2016 initially for the whole data set evasion was found to be high (more than (no categorization according to the size or 50.001 Albanian Lek approximately 400 industry sector) and results are shown in Euro), we have coded that entity with 2. Table 2. (Complete analytical data output can With this set of eighteen be found in Annex 2). For example, for the independent variables and one dependent current ratio we find no statistically

E-ISSN: 2224-3496 304 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

significant difference between the two groups the analysis of tax risk of entities because it of entities categorized based on tax risk (p- does not help the tax auditor to discriminate value>0.05 or sig>0.05). This means that between entities with a high tax risk and those based on a univariate analysis the current with low or free tax risk. ratio is not an important ratio to be used in

No. Category Financial Ratio 2015 Significance 2016 Significance

1 Current Ratio Not important (No Not important (No statistically significant statistically significant difference) difference) 2 Acid test (quick) Ratio Not important Not important 3 Liquid Ratio * Important Ratio Important Ratio (Statistically significant (Statistically significant

Liquidity Ratios difference) difference) 4 Non-Current Liabilities to Total Not important Not important Assets Ratio 5 Total Debt Ratio * Important Ratio Important Ratio (Statistically significant (Statistically significant difference) difference) 6 Times’s interest earned Not important Not important

Debt Ratios 7 Net Profit Margin Not important Not important

8 Profit before Taxes (EBT) Margin Not important Not important 9 Return on Assets ROA Not important Not important

10 Ratios Return on Equity ROE * Important Ratio Important Ratio

Profitability (Statistically significant (Statistically significant difference) difference)

Table 2. Significance of individual financial ratios in categorizing entities as risk free or with a high tax risk. All data set. Source: Primary data of the study. that financial ratios values are highly Based on what we observe from impacted by the sector where the company table 2, based on a univariate analysis for operates. Even by applying the univariate most of the ratios we do not find valuable analysis by industry sector we achieve information contained in their values if they similar results with the analysis done on the are considered one by one in the analysis of whole data set. tax risk; that is except for the Liquid Ratio, After separately analyzing the Total Debt Ratio and ROE that result impact of each ratio on the tax risk based on important in the univariate analysis, the other the optimal and standard values of each ratio, ratios have demonstrated to have almost the we have subsequently continued with another same average values across the group of risk technique, the multivariate analysis through a free entities and the group with high risk. As logistic regression model. As the multivariate a next step for this technique, we have analysis based on the logistic regression does continued the univariate analysis by not make use of the optimal or standardized performing the statistical significance test for values of financial ratios, we have included each different industry sector, having in mind

E-ISSN: 2224-3496 305 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

in this model the ten financial ratios for which As a first step, to avoid the multiple we had a threshold or a criterion that helped collinearity problems, we first excluded from us to categorize them in optimal or non- the model those variables who had a optimal vales. Again, the dependent variable correlation statistically significant at level of is the level or magnitude of the tax audit risk. significance α=0.05. Next, we also excluded Firms with low tax risk (having tax evasion the outliers. Table 3 represents results from lower than 400 Euro) are coded with 1 and Model I, the logistic regression analysis those with high tax risk (having tax evasion based on the sample with all the entities, run higher than 400 Euro) are coded with 2. separately for both years.

2015 2016 Β S.E. Wald Exp(β) Β S.E. Wald Exp(β) Total Debt Ratio -1.314*** .489 7.215 .269 .676 .535 1.598 1.965 NCA Ratio 1.340 .958 1.957 3.820 -2.126** .846 6.314 .119 Inventory Ratio .880 .744 1.401 2.412 -1.548* .823 3.538 .213 Net profit margin -1.548 1.473 1.104 .213 -1.530 1.283 1.422 .217 Revenue .011*** .004 7.148 1.011 .000* .000 3.707 1.000 Constant 1.500*** .490 9.362 4.484 2.151*** .647 11.040 8.596 N 175 183

Table 3: Logistic Regression Analysis, Model I, or Total model for the whole data set 175 & 183 observations, respectively in each of the two years. Source: Primary data of the study. (*p<0.1 **p<0.05 ***<0.01) NCA (Non-Current Assets) ratio is Fixed Assets to Total Assets and Inventory Ratio is equal to Inventory to Total Assets

The regression model for both years, probability to have higher tax risk. The level Model I, includes as explaining variables the of the NCA ratio (p-value>0.1) shows that Total debt ratio, NCA ratio, Inventory ratio, the percentage of investments in fixed assets Net profit margin and the Total Revenues is not statistically significant. We can say the which serve as a variable to control for the same for the inventory ratio and the net profit size of the entities. For 2015, Model I (the margin, they do not result to be statistically total model) is statistically significant χ2 important in Model I for 2015. (DF=5) =16.334, (p-value<0.01), Pseudo R2 For 2016 the Model I (total model) is is 0.152. This means that the explaining also statistically significant χ2 (DF=5) variables make a statistically significant =19.270, (p-value<0.01), Pseudo R2 is 0.174, classification between entities with a high that is for this year the explaining variables risk and entities with a low risk. The can help in classifying between entities with percentage of correctly classifying the high and those with low tax risk. The entities with high risk is 99.3%, whereas the percentage of correctly classifying all the total percentage of correctly classifying all entities both with high and low risk is 85.6%. the entities both with high and low risk is 84.6 Statistically significant variables for 2016 are %. the NCA ratio, Inventory ratios and the We can see that in 2015, the Total Revenues (whereas in the previous year the Debt Ratio (p-value<0.01) and the Revenues NCA ratio and the Inventory ratio were not (p-value<0.01) are statistically significant statistically significant). The 2016 Revenues what confirms the importance of the total are again, as in 2015, an important explaining revenue level and the total debt level in variable (p-value<0.1). determining the entities which have a higher

E-ISSN: 2224-3496 306 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

What we find unusual in the The next step in our multivariate findings of the regression analysis for the analysis is to run logistic regression audited entities is the net profit margin ratio, separately for the entities in the Construction which resulted to be a statistically & Manufacturing sector (Model II), entities insignificant ratio. This may be very well in the Service & Transportation Sector explained by the formula used to calculate (Model III) and entities in the Merchandising this ratio. We believe that this ratio, as sector (Model IV). After clearing the outliers, calculated based on the reported income and in the Construction & Manufacturing Sector not on taxable income is hiding the true we have 51 entities / observation for each “maneuvers” that companies are usually year. For year 2015 the Construction & engaged in, to perform tax evasion. From Manufacturing sector (Model II, Table 4), personal experience of the authors, we know was statistically significant χ2 (DF=5) the profit margin to be one of the first ratios =10.066, (p<0.1), Pseudo R2 is 0.325. The that a tax auditor checks. This fact is well total percentage of correctly classifying all known to everyone, including businesses the entities both with high and low risk is who subsequently try to artificially adjust the 88.2%. For 2015 the only statistically taxable income and bring it to the “right significant variable is the Inventory Ratio amount”. On the other hand, if the entity has which is statistically important at 0.1 a lower than the industry average net profit, significance level. We see that Revenue he again adjusts this ratio to bring it within variable has the same impact in this model as the parameters that he knows are usually with the Model I, but in this case, it is not accepted. As usual mechanism for this statistically significant. Model II for year earnings management practice, companies 2016 is not statistically significant as seen in may regard a portion of their expenses as table 4. non-tax deductible, thereby increasing taxable income and improving the value of this ratio, so that the tax auditor will not detect them. 2015 2016 Β S.E. Wald Exp(β) Β S.E. Wald Exp(β) Total Debt Ratio .556 1.507 .136 1.744 1.132 1.600 .501 3.102 NCA Ratio 2.254 1.809 1.552 9.526 -2.406 1.699 2.007 .090 Inventory Ratio 5.611* 3.224 3.030 273.465 -4.104* 2.273 3.260 .017 Net profit margin -1.217 2.714 .201 .296 -.003 2.149 .000 .997 Revenues .000 .000 2.019 1.000 .000 .000 .730 1.000 Constant -.581 1.372 .179 .559 2.556* 1.411 3.281 12.889

Table 4: Logistic Regression Analysis, Model II, for data for sector 1, Construction & Manufacturing respectively in each of the two years. Source: Primary data of the study. (*p<0.1) NCA ratio is Fixed Assets to Total Assets appendixes of this study. The number of Regarding sector 2, Services and observations (entities) in this sector were Transport, Model III resulted to be respectively 31 for 2015 and 24 for 2016. statistically not significant for 2015 (χ2 In Merchandising sector, Model IV, (DF=5) =0.793, (p-value>0.1) and we we analyzed 101 entities for year 2015 and provide its details only on detailed statistical the results show a statistically significant

E-ISSN: 2224-3496 307 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

model Wald χ2 (DF=5) =19.46, (p- IV for the Merchandising Sector included value<0.01), Pseudo R2 is 0.295. The total 108 entities and was statistically significant percentage of correctly classifying all the χ2 (DF=5) =14.588, (p-value<0.05). The total entities both with high and low risk in the percentage of correctly classifying all the Merchandising sector is 84.2% (Table 5). entities both with high and low risk in this The Total Debt ratio (p-value<0.01) and the year was 85.2%. The ratios that we find Revenues (p-value<0.05) are both statistically significant for year 2016 are the statistically significant for this Model (Table NCA Ratio (p-value<0.05) and the Revenues 5). Other ratios, the NCA ratio, Inventory (p-value<0.1) (table 5). Ratio and the Net profit margin are all not statistically significant. In year 2016, Model 2015 2016 Β S.E. Wald Exp(β) Β S.E. Wald Exp(β) - Total Debt Ratio 1.150 7.740 .041 .517 .586 .780 1.678 3.199** - NCA Ratio 1.332 1.391 .917 3.787 1.256 3.917 .083 2.486** Inventory Ratio -.225 1.033 .047 .798 -1.055 1.057 .996 .348 Net profit margin -.879 1.941 .205 .415 -5.406 4.635 1.361 .004 Revenues .000** .000 4.123 1.000 .000* .000 3.067 1.000 Constant 2.976** 1.042 8.155 19.607 2.112** .881 5.741 8.263

Table 5: Logistic Regression Analysis, Model IV, for data for sector 3, Merchandising in each of the two years. Source: Primary data of the study. (*p<0.1 **p<0.05 ***<0.01) In the Merchandising sector (Model After running the regression IV), the total debt ratio, the NCA Ratio and separately for each sector, from the results of the Revenues are statistically significant, so the individual models (Model II and Model they are in total compatibility with the results IV, because Model III was not statistically from Model I. We conclude that for the significant) we observe that the dependent Merchandising sector, specific risk indicators variables (financial ratios), behave are high levels of total debt and high levels of differently in different sectors and in investments in Fixed Assets. companies with different size (the Revenues is the control variable). In Construction and 5. Main findings and conclusions Manufacturing, inventory was a significant ratio which means that the higher the The aim of this research is to empirically inventory levels of entities in this sector, the study the linkage between financial statement higher is the tax risk of that entity. Revenues ratios and the level of magnitude of tax risk. and other variables are not significant for this For purposes of this study and based on specific sector, while keeping in mind that literature, the level of tax audit risk is revenues were significant for the total set of measured by the magnitude of the tax evasion firms (Model I). Of course, considering the detected in the companies audited in Albania limitation of database we are careful to not during 2017. The independent variables of jump into conclusions, but overall, it seems the study are a list of eighteen ratios and that the amount of Inventory is an important indexes, mainly from the reported financial risk indicator for construction and statements and the filed tax return of the manufacturing companies. entities which were subject of a tax audit in

E-ISSN: 2224-3496 308 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

2017. Out of a total of 1402 entities that were eighteen financial ratios that we initially tax audited in 2017 we randomly selected 183 calculated and included in the dataset. We to include in our sample. We calculated discarded from the model those variables eighteen financial ratios for each entity for which were highly correlated among them two consecutive years, 2015 and 2016 and and, the outliers. We build four different analyzed the respective Tax Audit Reports to models in the multivariate analysis. calculate the magnitude of the tax evasion. Model I, regression for all the entities in the Based on the latter we assigned the entities in sample, despite their sector. the category they belong regarding the level Model II, regression analysis for of tax audit risk. Construction and Manufacturing Sector. We used both univariate and Model III, regression analysis for Services multivariate techniques to analyze the and Transportation Sector. dataset. In the univariate technique we coded Model IV, regression analysis for each of the financial ratio with 0 if its Merchandising Sector. calculated value was close or better than the Results from Model I revealed it optimal value and with 1 if its calculated was statistically significant χ2 (DF=5) value was worse than the optimal value. We =16.334, (p-value<0.01) for 2015 (p- did not include that specific variable in the value<0.01) for 2016. The percentage of univariate analysis if no optimal or threshold correctly classifying all the entities both with value could be identified for it. high and low risk is 84.6 % in 2015 and Consequently, we were left with 10 ratios 85.6% in 2016. Statistically significant ratios only to continue the analysis. for both years and for all the companies are The results of the univariate (1) Total debt ratio, (1) Total Revenues, (3) technique reveal that among the three Non-current assets ratio and (4) Inventory liquidity ratios (current, acid test and liquid Ratio. We notice, quite surprisingly that net ratio) only the liquid ratio is statistically profit margin is not statistically significant in important. Entities with a high level of liquid Model I, which could be explained by high ratio were found to have a low level of tax accounting cosmetics techniques that are risk. Among the solvency ratios we found usually applied to this ratio so that the that only the total debt ratio was significant company would not “get caught” by the tax and disclosed important information on the auditors. tax risk. Entities with high debt ratio have a Model II for the Construction & higher tax audit risk. In the profitability Manufacturing sector for year 2015 was category we find only ROE to be an statistically significant at 0.1 p-value=0.073. important ratio. As a n ext step we applied the For 2015 the only statistically significant univariate analysis separately for each sector, variable is the Inventory Ratio. Model II for but the results were mixed, and no obvious year 2016 was not statistically significant. trend could be identified. Model III, regression analysis for Services Next, we applied the multivariate and Transportation Sector yielded not analysis to our sample, and we used the statistically significant results for both years logistic regression to identify which specific (probably due to fewer entities in this sector). ratio and index (independent variables) is Model IV for the Merchandising sector for linked to tax audit risk (the dependent year 2015 was statistically significant variable). We did not rely on the optimal or (p<0.01) and (p<0.05) for year 2016. The standard values of the ratios, and therefore statistically significant ratios in this model could include in our calculations all the

E-ISSN: 2224-3496 309 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

are the total debt ratio, the Revenues, and the techniques. Other authors could use more Non-Current Assets Ratio. recent information, extended in more Overall, based on results from both countries than Albania only and experiment univariate and multivariate analysis we can in applying innovative techniques that we conclude that the most common variables in mentioned during the literature review, such these models that are linked to tax risk are the as Neural Networks, or Expert Systems. (1) liquid ratio, (2) ROE, (3) total debt ratio, (4) inventory ratio and (5) Non-Current assets ratio. Counter wise to the widely References: spread beliefs that profit margin ratios do [1] Zhuo Zhang, Jia Wang, Financial Model contain information on tax risk, we found no based on Principle Component Analysis relation between these margins and the tax and Support Vector Machine, evasion magnitude probably due to many International Journal of Circuits, ways (and potentially manipulative as well) Systems and Signal Processing, pp. 183- that this ratio is calculated and the low quality 190, Volume 13, 2019. [2] Valbona Cinaj, Manuela Mece, Artur Ribaj, of data where it is based upon. Ilda Kadrimi, The Need for Improvement The above traditional and well- of External Audit Reports of Banks (The known financial ratios that can be easily Case of Banks in Albania which Mainly calculated based on reported financial belong to EU Banks), WSEAS statements only, could indicate quite Transactions on Environment and accurately the high risk of tax audits. Development, ISSN / E-ISSN: 1790- Therefore, we suggest that these ratios could 5079 / 2224-3496, Volume 16, 2020, Art. be included in the daily work procedures of #55, pp. 539-547. the tax auditors, for example in the Internal [3] Albrecht, S. and Romney, M. (1986) ‘Red- Manuals or various trainings. These ratios flagging management: a validation’, could also serve as risk indicators or as red Advances in Accounting, 3: 323–33. [4] Arens, A. and Loebbecke, J. (1994) Auditing: flags by other financial auditors, but still be An Integrated Approach, 6th ed. used with caution due to specifics that each Englewood Cliffs, NJ: Prentice-Hall. company demonstrates in practice regarding [5] Beasley, M. (1996) ‘An empirical analysis of the way these ratios are calculated, especially the relation between board of director in several sectors. composition and financial statement This research has several fraud’, Accounting Review, 71(4): 443– contributions in the field of financial ratios 66. and tax audit because it helps to enrich the [6] Bologna, G., Lindquist, R. and Wells, J. literature database with studies on a specific (1996) The ’s Handbook of field. It also contributes by suggesting Fraud and Commercial Crime. New specific indicators that could be easily York: John Wiley. [7] Beaver W.H. (1966) ‘Financial Ratios As calculated based on reported published Predictors of Failure’, Journal of financial information and that could be , Vol. 4, Empirical helpful in daily procedures of auditors. Also, Research in Accounting: Selected for the purpose of this research, the authors Studies pp. 71-111 have compiled a new database with financial [8] Beneish M. (1999) The Detection of Earnings information and audit reports for many Manipulation, Financial Analysts companies in Albania. Nevertheless, this Journal, 55:5, 24-36, DOI: research has its own restrictions in terms of 10.2469/faj.v55.n5.2296 limited timeframe, database, and application [9] Bell, T., Szykowny, S. and Willingham, J. (1993) ‘Assessing the likelihood of

E-ISSN: 2224-3496 310 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

fraudulent financial reporting: a cascaded [19] Zimbelman, M. F. (1997) ‘The effects of logic approach’, Working Paper, KPMG SAS No. 82 on auditors’ attention to Peat Marwick, Montvale, NJ. fraud risk factors and audit planning [10] Green, B. (1991) ‘Identifying management decisions’, Journal of Accounting irregularities through preliminary Research, 35(5): 75–9. analytical procedures’, unpublished [20] Niskanen J & Keloharju M (2000) Earnings doctoral dissertation, Kent State cosmetics in a tax-driven accounting University. environment: evidence from Finnish [11] Green, B. P. and Choi, J. H. (1997) public firms, European Accounting ‘Assessing the risk of management fraud Review, 9:3, 443-452, DOI: through neural network technology’, 10.1080/09638180020017159. Auditing: A Journal of Practice and [21] Spathis CH. Doumpos M & Zopounidis C Theory, 16(1): 14–28. (2002) Detecting falsified financial [12] Davia, H., Coggins, P., Wideman, J. and statements: a comparative study using Kastantin, J. (1992) Management multicriteria analysis and multivariate Accountant’s Guide to Fraud Discovery statistical techniques, European and Control. New York: John Wiley. Accounting Review, 11:3, 509-535, [13] Hoffman, V. B. (1997) ‘Discussion of the DOI: 10.1080/0963818022000000966 effects of SAS No. 82 on auditors’ [22] J.D. McCalden (2003) A method and attention to fraud risk-factors and audit apparatus for selecting taxpayers to audit planning decisions’, Journal of by a tax collecting entity. Patented Accounting Research, 35(5): 99–104. method. US Patents. [14] Hollman, V. P. and Patton, J. M. (1997) [23] Bay S, Kumaraswamy K, Markus G., Kumar ‘Accountability, the dilution effect and R, Steier D M (2006) ‘Large Scale conservatism in auditors’ fraud Detection of Irregularities in Accounting judgments’, Journal of Accounting Data’ Center for Advanced Research, Research, 35(2): 227–37. International PricewaterhouseCoopers LLP, Auditing Practices Committee (IAPC) Conference Paper, 2006. (2001) The Auditor’s Responsibility to [24] Kearns, Grover S.; Barker, Katherine J.; and Detect Fraud and Error in Financial Danese, Stephen P. (2011) "Developing Statements. International Statement on a Forensic Continuous Audit Auditing (ISA) 240. Model,"Journal of Digital Forensics, [15] Palmrose, Z. (1987) ‘Litigation and Security and Law: Vol. 6: No. 2, Article independent auditors: the role of business 4. failures and management fraud’, DOI:https://doi.org/10.15394/jdfsl.2011. Auditing: A Journal of Practice and 1094. Theory, 6(2): 90–102. [25] Gupta, M and Nagadevara, V. (2007). Audit [16] Schilit, H. (1993) Financial Shenanigans: Selection Strategy for Improving Tax How to Detect Accounting Gimmicks Compliance - Application of Data and Fraud in Financial Reports. New Mining Techniques. Downloaded from: York: McGraw-Hill. https://www.researchgate.net/publicatio [17] Stice, J. (1991) ‘Using financial and market n/254667082_Audit_Selection_Strategy information to identify pre-engagement _for_Improving_Tax_Compliance_- market factors associated with lawsuits _Application_of_Data_Mining_Techniq against auditors’, Accounting Review, ues. 66(3): 516–33. [26] Kallio M, Back B (2011) ‘The Self- [18] Loebbecke, J., Eining, M. and Willingham, Organizing Map in Selecting Companies J. (1989) ‘Auditor’s experience with for Tax Audit’ chapter from book IS material irregularities: frequency, nature, Success Evaluation: Theory and Practice and detectability’, Auditing: A Journal of (pp.347-358), downloaded by Practice and Theory, 9: 1–28. https://www.researchgate.net/publicatio

E-ISSN: 2224-3496 311 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

n/227191297_The_Self- by_using_Financial_Ratios_and_Cluster Organizing_Map_in_Selecting_Compan _Analysis. ies_for_Tax_Audit. [33] Shuli I. Earnings management and the [27] Watrin C, Struffert R, and Ullmann R quality of the financial reporting, (2008), ‘Benford’s Law: an instrument Perspectives of Innovations, Economics for selecting tax audit targets?’ Review and Business, Volume 8, Issue 2, 2011. of Managerial Science 2: 219. [34] Llukani, T. (2013). Earnings management https://doi.org/10.1007/s11846-008- and firm size: An empirical analyze in 0019-9. Albanian market. European Scientific [28] Demir B and Javorcik B (2018) Using Journal, 9(16), 135–143. Benford’s Law to detect tax fraud in [35] Persons, O. (1995) ‘Using financial international trade, Working Paper statement data to identify factors dowanladed by associated with fraudulent financing https://voxeu.org/article/using-benford- reporting’, Journal of Applied Business s-law-detect-tax-fraud-international- Research, 11(3): 38–46. trade [36] Wright, A. and Ashton, R. (1989) [29] Wells, J. 2001. “Irrational Ratios.” Journal ‘Identifying audit adjustments with of Accountancy (August): 80–83. attention-direction procedures’, [30] Grove H. Basilico E (2008) Fraudulent Accounting Review, 64(4): 79–83. Financial Reporting Detection: Key [37] Feroz, E., Park, K. and Pastens, V. (1991) Ratios Plus Corporate Governance ‘The financial and market effects of the Factors, International Studies of SEC’s accounting and auditing Management & Organization, 38:3, 10- enforcement’s releases’, Journal of 42, DOI: 10.2753/IMO0020- Accounting Research, 29: 107–42. 8825380301. [38] Fanning, K. and Cogger, K. (1998) ‘Neural [31] Marghescu R, Dorina & Tähtinen, Minna & detection of management fraud using Back, Barbro. (2010). Using Financial published financial data’, International Ratios to Select Companies for Tax Journal of Intelligent Systems in Auditing: A Preliminary Study. Accounting, Finance and Management, Communications in Computer and 7(1): 21–41. Information Science. 112. 10.1007/978- [39] Daroca, F. and Holder, W. (1985) ‘The use 3-642-16324-1_45. Downloaded by of analytical procedures in renew and https://www.researchgate.net/publicatio audit engagements’, Auditing: A Journal n/49288086_Using_Financial_Ratios_to of Practice and Theory, 4: 80–92. _Select_Companies_for_Tax_Auditing_ [40] Summers, S. L. and Sweeney, J. T. (1998) A_Preliminary_Study. ‘Fraudulently misstated financial [32] Neves Elisabete & Dias, António & Pinto, statements and insider trading: an Carlos & Batista, João. (2015). empirical analysis’, Accounting Review, Signaling Tax Evasion by using 73(1): 131–46. Financial Ratios and Cluster Analysis. [41] Chow, C. and Rice, S. (1982) ‘Qualified Signaling Tax Evasion by using audit opinions and share prices – an Financial Ratios and Cluster Analysis. investigation’, Auditing: A Journal of Working papers. Downloaded by Practice and Theory, 1: 35–53. https://www.researchgate.net/publicatio . n/309187751_Signaling_Tax_Evasion_

E-ISSN: 2224-3496 312 Volume 17, 2021 WSEAS TRANSACTIONS on ENVIRONMENT and DEVELOPMENT DOI: 10.37394/232015.2021.17.30 Rezarta Shkurti, Elena Myftaraj, Elsia Gjika

Contribution of individual authors to the Creative Commons Attribution License 4.0 creation of this scientific article (Attribution 4.0 International, CC BY 4.0)

This article is published under the terms of the Creative  Rezarta Perri Shkurti conceptualized the Commons Attribution License 4.0 idea of the research and was responsible https://creativecommons.org/licenses/by/4.0/deed.en_US for all the  stages of research, analysis, main findings, and of finally writing this paper.  Elena Myftaraj Tomori was responsible for the Statistical Analysis and giving recommendations for the Methodology and Analysis.  Elsia Gjika was responsible for gathering the dataset and giving professional opinion as to the practical implications of the findings of the study.

E-ISSN: 2224-3496 313 Volume 17, 2021