1. Introduction
Small- and medium-sized enterprises (SMEs) play an important role in entrepreneurship and job generation. However, SMEs often have difficulty attracting funding for their activities. To resolve such financial problems, government agencies operate a range of support policies, including credit guarantee programs. A credit guarantee from a government agency helps SMEs procure loans from commercial banks or participate in a pool for P-CBOs (primary collateralized bond obligations). However, when a government-guaranteed SME reports a default, the government agency is obligated to cover the loss, which ultimately burdens taxpayers. From the perspective of stakeholder theory [
1], credit guarantee programs backed by the government broaden the stakeholder base of supported firms to include not only the government agency responsible for the guarantee program but also the taxpayers to whom losses from defaults are transferred [
2].
To minimize the risks for stakeholders, government agencies use an evaluation process to screen SMEs applying for funding programs, based on their tangible and intangible assets. Despite the existing credit scoring model is designed to estimate the default risk by firm’s financial creditworthiness, there is still a very high rate of financial default among recipient SMEs owing to moral hazard. Most moral hazard event can be detected in part by unethical behavior in the accounting process. A high-quality accounting process may represent a low risk of moral hazard because of the reduced degree of information asymmetry between insiders and outside suppliers of capital [
3]. In this context, we believe that a firm’s ethical accounting practices must be assessed during the evaluation process to reduce the risk of default from moral hazard. An additional tool that evaluates a firm’s accounting behavior can be used to complement conventional financial creditworthiness model.
Against this background, there have been attempts to define firms’ accounting ethics and practices with a scorecard system. In 2005, Korea’s Small and Medium Business Administration ran a P-CBO program with a maturity period of three years to support SMEs. To select appropriate SMEs, the program reviewed firms’ basic information and financial status, following a general evaluation process. In addition, for the first time, firms’ accounting ethics were evaluated by using a scorecard with predetermined weights for individual attributes.
Consideration of moral hazard behavior of the supported SMEs is particularly important in Korea. This is because the Korean government continuously expanded its credit guarantee scheme to avoid a temporarily illiquid in SMEs after the Asian financial crisis and the scale of credit guarantee reached to about 6–8% of GDP, which is much higher than other nations—0.1% in the United States and less than 3% in Taiwan. In this situation, such large-scale budget may bring government institution’s indiscriminate supports and the selected SMEs’ careless management. This can increase the burden of stakeholders without efficacy of government supports. In this context, government decided to evaluate firm’s accounting ethics and the scorecard and weights of the attributes that present firm’s accounting behavior were proposed by certified public accountants (CPAs) based on their experience, and these were not updated based the actual data. However, the default rate among P-CBO guaranteed firms turned out to be high, the weights assigned to the individual components of the scorecard needed to be re-tuned to predict default risk in a better manner.
Against this background, this study proposes a complementary credit scoring model based on the accounting ethics of SMEs. The proposed model is developed based on P-CBO data. The proposed accounting ethics scoring model is designed to be used as a complementary screening method for firms already shortlisted based on their credit score of financial creditworthiness and financial information. The suggested model is thus expected to reduce the risk of default due to moral hazard. Hence, our model can prevent stakeholders in government-supported programs, including the public, from making losses.
The logistic regression model is one of the most widely used approaches for developing credit scoring models [
4]. We use a logistic regression model to gain insight into credit scoring based on accounting ethics. The adoption of the proposed ethics-based credit scoring model can contribute to more sound funding than the use of single conventional credit scoring models.
The rest of this paper is organized as follows. In
Section 2, we review previous studies of credit scoring models as well as the attributes of a firm’s accounting behaviors. In
Section 3, we introduce the data and variables used in our proposed model. In
Section 4, our logistic regression analysis is introduced to propose a credit scoring model based on accounting ethics. In
Section 5, we summarize our results and suggest areas of further research.
2. Literature Review
The focus of this study is loan defaults among SMEs associated with moral hazard that can be predicted by accounting practices. Loan defaults due to moral hazard must be taken seriously. This is particularly so with P-CBO programs, which are initiated by government agencies at the burden of taxpayers. To minimize such defaults, the accounting ethics of SMEs need to be assessed along with the conventional credit evaluation. This section reviews previous works of credit scoring models for SMEs and accounting ethics. We examine how credit scoring models for firms have been developed historically and explore aspects of accounting ethics.
2.1. Credit Scoring Models
Many SMEs experiencing capital problems use debt or credit guarantees to improve their financial resources. Traditionally, assessing an applicant’s default risk was a matter of using human judgment based on previous decision experience. However, given the increased demand for such evaluation and the recent advances in computer technology, model-based credit scoring approaches are now used, mainly based on logistic regression [
4,
5] and the neural network model [
6].
Typical approaches of making lending decisions for SMEs—called lending technologies—all consider factors such as the financial status of the business owner and the firm’s assets. Only a few studies have proposed a credit scoring model for SMEs based on non-financial information [
7]. Technological credit scoring models are examples based on non-financial attributes. Sohn et al. [
4] developed the first technology credit scoring model by using a logistic regression based on four non-financial aspects: management ability, technology level, marketability of technology, and profitability of technology. This model was subsequently extended to reflect various practical situations [
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24].
Default stemming from internal fraud can provide larger negative effects on government policies by ultimately providing a burden on taxpayers [
1]. Stakeholder theory explains the complex relationship between a firm and its stakeholders. Credit guarantee programs backed by governments broaden the stakeholder base of supported firms, including both the government agencies responsible for the guarantee program and taxpayers, to whom losses from defaults are transferred [
2]. However, stakeholders’ different interests and complicated connections raise ethical questions about business activities [
25]. Further, accounting ethics have thus far been underutilized for credit scoring.
2.2. Aspects of Accounting Behavior
Blake et al. [
26] provided the case of “creative accounting” in Spain and discussed the ethical issues of manipulating the accounts of businesses. The authors argued that creative accounting has been preferred by firms to mask their financial status even though it is perceived as ethically undesirable.
For the stakeholders of governments’ credit guarantee programs, inaccurate financial reporting is one of the risk factors associated with loan default [
27]. When unethical management behavior occurs during an accounting process, it weakens the financial structure of the company involved, decreases employees’ willingness to work, and can ultimately lead to default [
28]. For example, since the Enron scandal erupted, firm defaults due to moral hazard have led to a general mistrust of accounting, financial reporting, and auditing practices [
29,
30]. Accounting fraud has adverse effects on companies and industries as well as shakes the very foundations of capitalism [
31]. As accounting processes are directly related to the allocation of benefits, there is a high likelihood of making unethical decisions.
Accounting fraud is likely to take place when a firm has a weak internal control system. In particular, many SMEs are exposed to the risk of fraud because of their weak internal control. This is because few SMEs can afford the labor and financial resources needed to conduct external and internal audits. Bruns and Fletcher [
32] pointed out that because the information released by some SMEs having weak internal control may be manipulated or incomplete, banks have difficulty making lending decisions concerning SMEs. The Sarbanes–Oxley Act is now being enforced in the United States, which has increased the need for accounting and financial reporting by small businesses [
33]. As such, SMEs have to resolve issues involved in information asymmetry, which violates stakeholders’ rights and ultimately decreases outside investment [
34].
Unethical behavior within a firm is likely to spoil governance transparency and eventually bring about financial opacity, which is linked to the risk of misappropriation. A credit scoring model based on accounting ethics should estimate the levels of governance transparency and financial transparency that result from business/managerial ethics and misappropriation risk, respectively.
First, in terms of governance transparency, a firm’s internal control structure reflects its degree of business ethics. The risk of unethical business decisions can be reduced when an internal control structure dealing with the firm’s business processes such as internal regulatory systems and clear role assignments is well established. Further, it is much easier to rectify a fault when it occurs in a firm with more governance transparency [
35,
36]. In terms of financial transparency, business routines embedded in the internal control structure can also affect the accuracy of accounting records, which is associated with the risk of misappropriation.
Second, financial transactions provide evidence of financial transparency in practices. Hasumi and Hirata [
37] argued that the transparency of SMEs’ financial statements is an important factor in credit scoring. Higher transparency in financial transactions means less risk of embezzlement. When Patel and Dallas [
38] evaluated a firm’s transparency, they screened financial transactions based on accounting policy factors such as the statuses of accounts, financial statements, and balance sheets, which shed light on ethical accounting behaviors, in the form of an accounting system’s output. Similarly, Simon [
39] found that accounting fraud often occurs through the manipulation of cash balances on balance sheets. For example, to exaggerate the status of the current amounts of cash in hand, a company might issue an accommodation bill without a corresponding business transaction.
Third, Yeh et al. [
40] found that a firm’s relationship with its related party or affiliated company is concerned with both governance and financial transparency because such financial transactions have an issue with wealth exploitation [
41], with subsequent increases in information asymmetry [
42]. Related party transactions (RPTs) involving loans and guarantees negatively influence a firm’s financial transparency. Therefore, to evaluate governance transparency, RPTs should be monitored in terms of their lending behaviors involving related parties, their loans to affiliated companies, and the financial statuses of their affiliated companies.
Finally, the importance of the business and managerial ethics considered by key decision makers is reflected in business practices. Scandals such as those resulting from the actions of Enron and Arthur Andersen, which were driven by unethical behaviors—especially in their accounting processes—demonstrate the necessity of certain business ethics. These types of scandals damage public trust in business, and so ethics regulations that prevent corporate crimes must be imposed. Majority shareholder compensation is another crucial factor in the screening of SMEs’ business practices because, from a structural standpoint, SMEs have an owner-manager structure [
43]. Thus, shareholders are usually the firm’s CEO and parties related to the CEO, meaning that their compensation packages can violate business ethics.
3. Data and Variables
In Korea, P-CBOs are issued to support SMEs that hold competitive technologies but nonetheless require funding support to sustain their businesses. The typical process of evaluating an SME’s application for a P-CBO involves a basic document review and a subsequent credit scoring evaluation. In addition, accounting ethics evaluation procedures have recently been proposed to increase the discrimination power of moral hazard.
Table 1 describes five screening areas and associated variables that reflect such procedures. The data source for this research is the P-CBO program, which evaluates SMEs by using three steps: a basic document review, a financial credit scoring evaluation, and an accounting ethics scorecard. After a document review and financial credit evaluation of the 329 applicant firms, 74 companies were selected. Instead of directly granting the P-CBO to the 74 accepted firms, accounting ethics were additionally assessed to increase the discrimination power for those with high levels of moral hazard. Eventually, the P-CBO was granted to 68 firms that passed all three evaluation processes.
The final evaluation step was performed by a CPA on the basis of a scorecard with 16 elements classified into five screening areas: internal control structure, financial transactions, related parties, business ethics, and “other unethical conduct.” The proposed attributes cover the firm’s aspects in business ethics, managerial ethics, and risk of misappropriation in accounting behavior.
Table 1,
Table 2,
Table 3,
Table 4 and
Table 5 provide details on the individual attributes (denoted by “X” and a number) in the scorecard.
Table 1 presents the variables used to evaluate the internal control structures of SMEs. As the internal control structure represents financial transparency, internal control regulations, role assignments, and the accuracy of accounting are used.
Next,
Table 2 presents the variables of financial transactions to provide evidence of financial transparency. As higher transparency in financial transactions means that managers consider the interests of shareholders, the credit scoring model should include related variables such as the number of accounts, withdrawal methods, evidence of expenditure, and cash in hand. For the X5 and X6 variables, additional demerit points were assigned. For example, if a company was assigned three points for X5 because of one discovered cash-out transaction related to a business deal, and if that company was found to have used its own promissory note rather than a credit card for a commerce purchase transaction, then its final score became one point.
As RPTs are closely related to the risk of wealth exploitation, the P-CBO programs evaluated the status of related parties and their relationships (
Table 3). Loans to related parties, loans to affiliated companies, and the financial statuses of affiliated companies were examined for RPTs.
Table 4 and
Table 5 show the variables for evaluating business ethics and other related factors. Accounting fraud scandals such as the Enron and Arthur Andersen cases are associated with the practices of business ethics in a firm. Therefore, it is necessary to assess a firm’s business ethics. Ethical regulations (X11) cover three types of ethical rules, regarding: (1) whether there are codes of behavior to protect against conflicts of interest with regard to customers; (2) whether there are protection regulations for people who report a colleague’s violations of ethical standards; and (3) whether there are protection (information security) regulations regarding a firm’s trade secrets.
The scale of each attribute is preset based on the CPA’s opinions. CPAs assigned weights between 3 and 9—except X16 (other unethical conduct)—depending on what attributes are associated. As shown in the previous section, the attributes of accounting ethics cover the business and managerial ethics and risk of misappropriation. The CPAs set a weight of 9 for the attributes presenting the risk of misappropriation, while they set 3 and 6 to those of business ethics and managerial ethics, respectively. Most attributes can be evaluated by using the information in a firm’s one-year accounting records; the remaining attributes—mostly the business ethics-related attributes—are scored based on an in-depth interview with the firm’s manager.
The preset weights can be updated when the funding result based on this score is obtained. Thus, we use a logistic regression model to formulate a credit scoring model based on the accounting ethics practices of fund applicant firms and their default. Before using this logistic regression model, it is necessary to eliminate the preset weights assigned by the experts. Hence, we transform the original scale (X) of all the attributes into the same six-point scale as follows:
In this study, we set the scale of the attributes to 6, which is the median of the most preset weights (except X16); the rest of the attributes have weights from 3 to 9. The number chosen as the scale does not affect the default probability to be estimated from the logistic regression. Further, instead of directly applying this to the logistic regression, we take its inverse to reflect the potentially non-linear relationship between the ethics attributes and the log odds ratio of default:
Specifically, by using a logistic regression model, we model the probability of non-default,
, as a function of attributes Z as follows:
where
is the weight corresponding to
estimated by using the maximum likelihood method. Among the 68 firms for which the P-CBO was granted, 22 (32.35%) defaulted before the date of maturity. The average accounting ethics score for the 68 firms was 75.2, while those for the non-defaulting and defaulting firms were 76.7 and 72.1, respectively.
Table A1 presents the descriptive statistics of the defaults as well as the means and standard deviations of each variable X, while
Table A2 provides information on the correlation matrix of ethics attributes X. Moreover,
Table A3 shows the potential for multicollinearity in terms of the transformed independent variable Z, by using variance inflation factors. Values exceeding 20 indicate a problem with multicollinearity. However, no significant multicollinearity problems were found in this study, and thus we conducted the logistic regression with all 16 Z variables.
4. Results of the Logistic Regression
To distinguish non-defaulting SMEs from defaulting ones, we used the reciprocal values (Z1–Z16) of the 16 variables (X1–X16) introduced in
Section 3 for the logistic regression. As displayed in
Table 6, we found five significant variables related to the non-default condition at a significance level of 10%: Z3 (accuracy of the accounting records), Z7 (cash in hand), Z8 (loans to related parties), Z9 (loans to affiliated companies), and Z13 (majority shareholder compensation).
We selected a significance level of 10% because of the relatively small sample size used in our study. A negative estimated coefficient for Z represents a positive association with preventing default, because we conducted a logistic regression with the reciprocal X values. In other words, the variables associated with a higher positive estimated coefficient are associated with a higher risk of default, whereas a lower negative estimated value indicates a lower risk of default. The significant variables Z3, Z8, Z9 and Z13, which have negative estimates, indicate that the original variables (X3, X8, X9 and X13) are positively associated with non-default (
Table 3). In the same vein, X7 is negatively related to the non-default condition.
For SMEs, a high-quality account management system may take a long time to be established. Nevertheless, companies have a responsibility to correctly record the full particulars of accounting results. People interested in a company’s financial statements are stakeholders, such as corporate executives, employees, shareholders, stock market investors, governments (tax authorities), and the media. These groups can have a significant impact on the growth and activities of SMEs. If there is corruption connected to corporate accounting and financial statements, it can lead to mistrust from stakeholders and thus hinder the business activities of the enterprise in question. Therefore, enterprises are responsible for financial transparency in their financial statements. This fact supports the contention that the accuracy of accounting records (X3) has a positive effect on the assumption of a non-default condition.
Increases in loans to related parties (X8) and loans to affiliated companies by firms (X9) are undesirable with regard to business transparency. Generally, related parties are relatives of the CEO or major shareholders of the company, and loans to related parties or affiliated firms may indicate unethical behaviors. If a large amount of money is loaned in this manner, there is a high possibility that misappropriation will occur, along with other ethical problems. For a small business in poor financial health, in particular, providing short-term loans to related parties or affiliated companies can exacerbate cash problems. Majority shareholder compensation (X13) is another important factor when screening SMEs for unethical behavior. SMEs usually have an owner-manager structure, and excessive compensation to a major shareholder raises a red flag during an assessment [
43].
Our logistic regression found that cash in hand (X7) was negatively correlated with the non-default condition. This indicator evaluates the adequacy of the cash balance in hand by checking recent transaction information on a book of original entries. If sudden changes in cash in hand are observed and the status of cash in hand exceeds a suitable level, CPAs assume that the firm has a strong possibility of being engaged in accounting fraud. The CPAs in our study suggested an appropriate level of cash in hand for guaranteed SMEs, and when the cash in hand exceeded this level, they subtracted points. Thus, a company with too much cash would be given a low score for this variable (X7). However, the logistic regression shows that companies with low cash in hand (X7) scores do not tend to default until bond maturity. As Laitinen and Laitinen [
44] found, there is a higher failure risk for firms with less cash flow, suggesting that a company with a low score for this variable holds sufficient cash liquidity, which prevents default arising from low amounts of cash.
Comparing the performance of the proposed accounting ethics-based credit scoring model with the scorecard described in
Table 1,
Table 2,
Table 3,
Table 4 and
Table 5, we found that threshold values varying from 0.355 to 0.405 provided the best prediction accuracy, using jackknife cross-validation. Jackknife cross-validation evaluates the proposed model by forming
N samples by using the “leave one out” procedure. This cross-validation approach is therefore suitable to approximate the proposed model’s error for small sample sizes [
17].
Table 7 compares the classification performance for the currently used scorecard, which has been used to evaluate firms’ accounting ethics, with the proposed ethics-based credit scoring model and the suggested threshold values.
By using our developed credit scoring model based on accounting ethics attributes, the default probabilities of 48 out of 68 companies were predicted correctly. The resulting accuracy rate was 70.59%. The existing scoring model predicted that all 68 companies would not default, but 22 companies (32.35%) eventually did, indicating that its accuracy was only 67.65%. Therefore, the proposed credit scoring model outperforms the existing method. Furthermore, the specificity of the proposed model is 36.4%, while the specificity of the existing model is 0%. Since the specificity is a proportion of identifying default to default, the result shows that the proposed model is more sensitive to detecting default risk than the existing model.
5. Conclusions
Stakeholder theory emphasizes that the interests of a firm and its diverse stakeholders should be fairly considered and addressed. In this context, business ethics in accounting can be emphasized as a firm’s principal responsibility. Since the big accounting ethics scandals of the 2000s, firm transparency has become a particularly important requisite for investors as well as an influential factor in decisions on investments. Unethical accounting management activities occur frequently through weak internal control structures, unethical financial transactions, the lending of large amounts of money to related parties, and improper business ethics. These unethical accounting activities weaken the management conditions of SMEs and can lead to defaults in many cases. Defaults, in turn, damage both firms and their stakeholders. In a P-CBO program, which is backed by the government, the stakeholder base is broadened to include taxpayers. Thus, developing a proper credit scoring model based on accounting ethics is crucial for screening SMEs and realizing investments.
To develop a credit scoring model for accounting ethics, we reviewed previous studies and conducted a logistic regression with P-CBO program data. The factors in our scorecard reflect business transparency and cover both ethical processes and practices pertaining to accounting. Based on our results, we noted that the weights assigned by a CPA do not necessarily serve to bolster the default predictions. Moreover, some of them were insignificant. We found five variables that had significant meanings in relation to predicting non-defaults: the accuracy of accounting records, cash in hand amounts, loans to related parties, loans to affiliated companies, and majority shareholder compensation. These five variables can provide insights and implications for investors making investment decisions. It is essential that SMEs maintain transparent, fair, and ethical accounting approaches and establish methods for overcoming the risks associated with unclear accounting records, loans to related parties and affiliated companies, and excessive compensation to majority shareholders.
The proposed credit scoring model for ethical management in accounting is expected to help define an ethical foundation. It can also be used as a tool to aid financial support decisions concerning SMEs and to reduce the moral hazard associated with financing. When the suggested credit scoring model is used in the accounting process to select ethical SMEs, stakeholders can feel satisfied. As the model can prevent public losses, it has important implications for policies and government programs supporting SMEs. Its implication in the evaluation step can vary depending on the policy of support programs. For example, among the selected SMEs from the conventional evaluation step, the low scoring SMEs (i.e., those rejected by the new model) can be advised to invite external audit during the loan term to get funds, or the scale of the funding support can be adjusted.
For further research, more ethics attributes can be added to cover aspects related to corporate social responsibility. In particular, according to the specific accounting behaviors embedded in individual countries, new attributes describing such unique patterns should be included and their associations with default risk examined. The development of such enhanced models is clearly a worthy subject for further research.