Introduction of the study
Rationale of the study
The stock market serves as a vital platform for trading medium and long-term securities in the modern economy, involving both individual and institutional investors, such as mutual funds and insurance companies These participants engage in buying and selling stocks for various reasons, including profit generation and acquiring control of companies One of the primary advantages of the stock market is its ability to provide a flexible source of capital for business operations Consequently, numerous stock markets have been established since the inception of the New York Stock Exchange (NYSE), significantly contributing to national economies.
The Ho Chi Minh City Stock Exchange (HOSE), established on July 11, 1998, and operational since July 28, 2000, has played a crucial role in the stock market Initially starting with just two listed companies, HOSE has achieved significant growth, boasting around 507 listed companies and a total market value of 365 trillion VND by December 31, 2007 Additionally, the exchange has seen the establishment of three securities investment funds and a diverse range of 366 bonds, with nearly 298,000 investor accounts opened at securities companies.
In recent times, the Vietnam Index, which reflects the performance of Vietnamese companies on the Ho Chi Minh Stock Exchange, has experienced consistent growth, attracting 7,000 foreign investors This period marks a remarkable phase in the development of the Vietnam Stock Exchange.
However, after a rapid and hot growth, the Vietnamese Stock Market in
The year 2008 witnessed a significant downturn, leading numerous publicly traded companies to encounter severe financial challenges The economic crisis had a profound effect, resulting in a 30% reduction in the total profits of all listed firms By the close of 2009, 23 companies were still grappling with the aftermath of this financial turmoil.
Financially distressed companies can significantly harm market participants, including shareholders, creditors, managers, and individual investors To mitigate these risks, investors can identify specific characteristics of financially distressed firms, such as high leverage and low, volatile stock returns Recognizing these warning signs is crucial for both investors and operating enterprises Consequently, numerous qualitative and quantitative studies have been conducted to identify firms at risk of failure.
The two most widely used methods for financial analysis are Multiple Discriminant Analysis and Logistic Regression, which are integral to the development of Altman’s Z Score model (1968) and Ohlson’s model (1980) Pioneering research by Beaver (1966) and Altman (1968) emphasizes the predictive capabilities of models derived from financial ratios Building on this foundational work, the Z Score model has undergone extensive refinement through studies by researchers such as Deakin (1972), Taffler (1985), Goudie (1987), Grice and Ingram (2001), Agarwal and Taffler (2007), and Sandin and Porporato (2007).
Similarly, other studies also have been performed relating to the Ohlson, including Lau (1987), Fauzias and Chin (2002), Boritz, Kennedy, and Sun (2007), Muller, Steyn-Bruwer, and Hamman (2009)
In today's financial landscape, distress management techniques are increasingly utilized for various economic purposes, particularly in credit analysis by financial institutions When customers seek loans, these institutions assess their creditworthiness, and if there are indications of potential financial distress, preventive measures may be enacted This can include rejecting loan applications or requiring borrowers to undertake additional steps to mitigate the risk of default before funds are disbursed.
Determining the most effective technique is a complex question, as each method presents unique advantages and disadvantages The effectiveness of these approaches can vary significantly depending on the timing and context, primarily because predicting an individual's reaction to information is inherently challenging.
Human behavior exhibits both consistent trends and unpredictable elements When analyzing a vast group of investors, such as millions, accurately predicting their collective behavior becomes more challenging than forecasting that of an individual Additionally, research outcomes are influenced by various factors, including the specific characteristics of each stock market and the duration of data collection.
Despite the limited application of financial distress prediction techniques in the Vietnam Stock Exchange, many publicly listed companies have encountered significant challenges Therefore, conducting research to explore the effectiveness of financial ratios in forecasting financial distress is crucial This study holds valuable implications for both private entities and government institutions in evaluating the financial health of firms.
Research objectives and questions
This thesis aims to elucidate the connection between financial ratios and the financial distress experienced by listed companies in Vietnam, building on previous research findings.
- Which is the suitable method to evaluate the impact of financial ratios on the probability of failure ?
- How is the relationship between the financial ratios and the financial distress state of companies?
Structure of the study
This research comprises four chapters: Chapter 1 introduces the overall research content, including the research problem, questions, and objectives Chapter 2 presents a literature review on financial distress prediction Chapter 3 analyzes the collected data and presents the research findings Finally, Chapter 4 discusses the conclusions and implications of the study.
Literature Review
Definition of financial distress
Financial distress is defined in various ways across research, with Dun and Bradstreet (1985) describing it as a disruption in business operations According to Foster (1986), a company experiences financial distress when it faces significant liquidity issues that necessitate critical operational changes Liquidity problems arise when a company's current obligations cannot be met Additional indicators of financial distress include reduced dividend payouts and defaults on debt obligations.
Various studies have identified distinct stages of financial difficulty, including Guthmann and Dougall's (1952) three stages: technical insolvency, unsupportable debt burden, and reorganization Newton (1975) proposed a four-stage deterioration model consisting of incubation, cash shortage, financial insolvency, and total insolvency Additionally, Lau (1987) utilized a five-state model to analyze financial distress, while Somerville (1989) opted for a simpler three-state model.
Financial distress refers to a situation where a company struggles to meet its obligations to creditors, potentially leading to insolvency This necessitates governments worldwide to establish regulations for addressing financial distress in the corporate sector Consequently, there is significant discourse around the legal definitions of corporate failure, which aids researchers in classifying distressed and non-distressed firms For example, in the context of the Malaysian Stock Exchange, financial distress is identified through criteria such as closure under the Companies Act 1965, entering a Scheme of Arrangement and Reconstruction, or undergoing Corporate Debt Restructuring.
Committee; d) Selling the firms’ loans; and e) Restructuring small borrowers
A study in the United Kingdom identifies failed firms based on the regulations outlined in the Insolvency Act of 1986, which provides five primary courses of action: administration, company voluntary arrangement, receivership, liquidation, and dissolution.
In Vietnam, the Law on Bankruptcy, enacted on June 15, 2004, states that a company can be declared bankrupt if it fails to meet its debt obligations However, there is an opportunity for recovery before the High Court officially declares bankruptcy Creditors convene a conference to evaluate and modify the company's rehabilitation plan, which may include various strategies such as mobilizing new capital, altering production commodities, implementing technological innovations, reorganizing the management system, merging or splitting production departments to enhance productivity and quality, selling shares to creditors, and disposing of non-essential assets.
It is uncommon for Vietnamese firms to declare bankruptcy in the High Court due to the complex administrative procedures involved Additionally, gathering information about these companies poses significant challenges Despite operating for over fifteen years, publicly listed companies on the Ho Chi Minh Stock Exchange are not required to file for bankruptcy, even when facing dire financial circumstances.
To enhance the efficiency and credibility of the stock market while safeguarding the interests of directors, intermediaries, and shareholders, the Ho Chi Minh Stock Exchange implemented Decree 04/QD-SGDHCM on April 17, 2009, which amends and adds to various listing regulations In a similar effort, the Hanoi Stock Exchange introduced Decree 324/QD-SGDHN on June 4, aimed at improving its regulatory framework.
In 2010, regulations were established for the listing of securities, categorizing those with unsatisfactory conditions into warnings, control measures, trading suspensions, and delisting Stock Exchanges are responsible for issuing warnings and ensuring full market disclosure regarding these securities The removal of warning signs occurs when listed companies address and resolve the issues that led to their warnings, controls, suspensions, or delistings.
Some underlying conditions of classification pursuant to the law for each case imply the finance nature such as: a) The case for warned stocks:
There is a one-year-overdue debt or a rate of overdue debt higher than 10% equity
There are not enough 100 shareholders holding at least 20% shares of the company
The earning at the same year is negative
The company’s operation is stopped
Listed companies continue to violate the regulations in relation to disclosing the information although being warned
Shares do not trade within 90 days
It is deemed necessary to protect the benefit of investors b) The case for stocks put under control:
Listed companies have not improved situations leading them to being warned
Listed companies violate regulations involved in stocks and the stock market seriously c) The case for delisted shares:
The charter capital decreases to below 80 billion VND
Listed organization’s certificates of business registration or certificates in specialized business are revoked
Shares have not traded in 12 months
Audition organizations have disapproved of or refuse to give idea of listed firms’ latest financial statement
A company listed on the Ho Chi Minh Stock Exchange may face delisting if it reports negative earnings after tax for three consecutive years, resulting in total accumulated losses that exceed its equity as reflected in the latest financial statements.
Stock exchange – the action of delisting The listed firms stop trading when their earnings for two consecutive years are negative
Finding annual financial statements for listed companies can be challenging Financial distress firms are defined as those whose shares have been put under control or delisted, as outlined in Decree 04/QD-SGDHCM dated April 17, 2009, and Decree 324/QD-SGDHN dated June 4, 2010, which amend and add to the regulations for listing.
Ratios in designing models
Foster (1986) emphasizes that utilizing financial ratios is the most effective method for identifying companies facing financial difficulties, as it leverages the consistent patterns between ratios and specific events These ratio models, based on financial statements, reveal distinctions between stable and unstable firms However, careful consideration is essential when interpreting these ratios due to the varying accounting standards that underpin financial reports.
When collecting financial ratios as predictor variables, their selection hinges on their popularity and predictive power demonstrated in prior research This approach is influenced by the absence of theoretical frameworks that establish a causal link between financial ratios and bankruptcy Most evidence supporting this relationship is empirical, as seen in studies by Jones (1987) and others, including Karels & Prakash (1987), Lam (1994), and Wilson.
Wilson and Sharda (1994) emphasize that the advancement of bankruptcy models is closely linked to the selection of economic variables, which enhances predictive accuracy Similarly, Jones (1987) highlights the significance of this approach, noting that numerous studies employing different techniques have yielded consistent ratios This convergence of findings underscores the reliability of the ratios used in bankruptcy prediction.
In 1973, key financial ratios were identified, including return on investment, capital turnover, financial leverage, short-term liquidity, cash position, inventory turnover, and receivables turnover According to Jones (1987), these factors are crucial for economic analysis and interpretation.
Numerous researchers, including Altman, Haldeman, Narayanan (1977), Marais, Patell, Wolfson (1984), and Foster (1986), have developed financial market ratios that provide crucial insights not available in traditional financial statements.
Here is the rewritten paragraph:While Zavgren (1983) cautions that using too many ratios in research can lead to overfitting, Wilson & Sharda (1994) argue that incorporating more ratios can actually improve the accuracy of Neural Network analysis compared to Multivariate Discriminant Analysis This view is supported by Udo (1993), who notes that advances in computing power have made it advantageous to utilize multiple information sources in modeling.
Karels & Prakash (1987) argue that the ratio selections in previous studies did not adhere to the assumptions of Multivariate Discriminant Analysis (MDA) Their research aimed to evaluate whether these ratios fulfill the necessary assumptions of MDA, employing tests to assess normality among the chosen ratios While the ratios do not fully meet the joint normality requirements, their deviations differ from those observed in other studies In comparison to Altman's 1968 research, their ratios demonstrate an overall enhancement in predictive capability.
Karels & Prakash (1987) identified key financial ratios that align with the seven categories established by Pinches et al (1973), which are grounded in experimental research These essential ratios include the working capital ratio, gross profit margin, earnings per share, total debt to total assets, cash flow per share, market value of common stock, asset turnover, sales per cash, and sales per receivables.
One of the important characteristic in relation to the life or the firm is profitability (Lam 1994) The lower these ratios are, the higher the probability of financial distress is
Liquidity ratios are crucial as they indicate a company's ability to meet its obligations without disrupting operations Insufficient liquidity can lead to difficulties in timely debt repayment (Lam, 1994) Notably, the working capital ratio tends to decline when companies face financial challenges.
Companies in poor financial health typically exhibit higher ratios of total debt to total assets (Somerville, 1989) Economic factors such as financial crises, intense competition, and significant interest rate fluctuations can severely impact a company's ability to meet its payment obligations Consequently, leverage levels emerge as a critical component in financial modeling.
Karels & Prakash (1987) highlight that unstable financial conditions in a firm lead to cash flow challenges, while cash flow ratios serve as indicators of a company's potential to generate future cash flows Additionally, there exists a notable indirect relationship between cash flow and the long-term sustainability of dividend payouts A common signal that management lacks confidence in future cash flows supporting dividends is the occurrence of a dividend cut (Somerville, 1989; Lau, 1987).
Activity ratios, including asset turnover, sales per cash, and sales per receivables, are crucial for researchers analyzing company performance According to Zavgren (1985) and Somerville (1989), these ratios tend to be lower for companies in distress, indicating long-term implications Specifically, the sales per receivables ratio serves as a key indicator of a company's ability to recover its debts.
Foster (1986) identified the market price of shares as a crucial indicator of bankruptcy, emphasizing its significance over traditional ratios This preference stems from the observation that financial statements often lack essential information when compared to market data Similarly, Karels & Prakash (1987) concluded that financial statements are less effective in predicting bankruptcy probabilities than market indicators.
Techniques used in financial distress predictions
Numerous estimation techniques have been explored in academic literature to develop prediction models, with Beaver (1966) being one of the pioneers who employed a univariate method to address business complexities However, this approach is criticized for its inadequacy in accurately assessing a company's financial condition due to its overly simplistic nature (Foster, 1986; Jones, 1987; Lam, 1994) Despite these limitations, Beaver's research continues to inspire subsequent studies in the field.
Altman (1968) builds on Beaver's (1966) work by employing a discriminant function with ratios in a multivariate analysis, leading to the widespread use of Multidimensional Analysis (MDA) for distress predictions MDA enhances the limitations of univariate analysis by capturing the multidimensional aspects of a company, thereby avoiding conflicting signals that often arise from univariate methods.
The Multivariate Discriminant Analysis (MDA) has two key limitations that are frequently violated: it assumes that the variables are multivariate normally distributed and that the covariance matrices of predictors across companies are identical (Foster, 1986; Jones, 1987; Lam, 1994) To address the first assumption, Jones (1987) recommends using log and square root transformations, along with the removal of outliers Subsequently, he applies quadratic discriminant analysis to tackle the second assumption effectively.
There are differing opinions regarding the analysis of the quadratic model Altman, Haldeman, and Narayanan (1977) argue that this model is overly sensitive to the derivation sample, resulting in poor classification performance in holdout samples Additionally, while the quadratic structure appears suitable based on statistical data, its validity tests do not align with theoretical expectations (Jones, 1987) Furthermore, Udo (1993) highlights additional issues with the MDA technique, including the impact of autocorrelation and its failure to account for data errors or manage missing values.
Heine (1995) declares that the accuracy rate of model using MDA from
1968 to 1995, specifically Altman (1968) Z score, achieves not less than eighty to ninety percent However, Ohlson (1980) and Karels & Prakash
(1987) add that the prediction probabilities are reliable only if statistical assumptions are not complied in any case
Logit analysis, derived from the logistic cumulative probability function, is a statistical method used to assess classification and prediction accuracy This model incorporates a crucial probability threshold; when a company's probability exceeds this critical level, it indicates a higher likelihood of insolvency.
This method avoids the limiting assumptions of MDA but assumes that the costs associated with type I errors (misclassifying a bankrupt company as non-bankrupt) and type II errors (misclassifying a non-bankrupt company as bankrupt) are equal Furthermore, it presumes that any changes in the independent variables will have a consistent impact.
Unlike MDA, logit analysis poses challenges in correcting for prior probabilities, necessitating the use of a technique known as Weighted Exogenous Sample Maximum Likelihood (WESML) (Zmijewski, 1984) WESML effectively mitigates biases linked to the assumption that type I and type II errors are equal Without this correction, the analysis may yield inaccurate probabilities unless the ratios of failing to stable firms in the overall population align with those in the sample (Jones, 1987) Additionally, adjusting the cutoff score serves to address the costs of misclassification, thereby clarifying the differences between type I and type II errors (Jones, 1987).
Comparing two techniques reveals that neither MDA nor Logit analysis provides substantially better results (Wilson & Sharda, 1994) Tam & Kiang
Research from Somerville (1989) suggests that Logit Analysis outperforms MDA, while Hamer (1983) also finds that Logit Analysis yields slightly more accurate results compared to MDA However, a study conducted in 1992 does not definitively conclude which technique is superior.
Ohlson (1980) highlights the effectiveness of the logit model in addressing the limitations of multiple discriminant analysis (MDA) for predicting corporate failure By utilizing nine carefully selected independent variables, Ohlson calculates the likelihood of failure for industrial firms that were active in trading between 1970 and 1976.
Over the past three years, the US stock exchange has seen 105 firms fail while 2,000 firms remained operational To analyze these outcomes, three predictive models were developed: the first forecasts failure within one year, the second anticipates failure within two years, and the third assesses failure probabilities for either one or two years The likelihood of failure for each firm in these models is determined using a logistic function.
Jones (1987) highlights that Logit analysis models are preferred over MDA due to their robust theoretical foundation for evaluating results Additionally, Harrell and Lee (1985) assert that Logit models remain effective even when the assumptions of MDA are present.
Artificial Neural Networks (ANN) have emerged as a leading method for predicting financial distress, as highlighted by various studies (Charitou and Kaourou, 2000; Tan and Dihardjo, 2001) However, the "black box" nature of ANN, which obscures how it differentiates between failing and non-failing companies, presents a significant limitation (Hawley, Johnson, and Raina, 1990) Additionally, ANN fails to indicate the importance of individual variables in the classification process, further complicating the interpretation of its results.
Hypotheses
The thesis aims to clarify the connection between financial ratios and financial distress in publicly listed companies in Vietnam Notably, the Vietnam Stock Exchange has been established for just over 12 years.
It has been considered as a “young market” One more problem is considered whether these connections are consistent with earlier researches
Thus, some hypotheses to be tested will be introduced in this thesis:
- H1: Earnings per share impacts negatively on the probability of failure
- H2: Cash flow per share impacts negatively on the probability of failure
- H3: Asset turnover impacts negatively on the probability of failure
- H4: Sales per receivables is negatively impacting on the probability of failure
- H5: Working capital is negatively related to the probability of failure
- H6: Sales per cash is negatively related to the probability of failure
- H7: Gross profit margin is negatively impacting on the probability of failure
- H8: Total debt to total assets is positively related to the probability of failure.
Conclusions
Numerous models have been utilized in academic research for predicting financial failure, employing techniques such as multiple discriminant analysis (MDA), logit, and neural networks Although a variety of models exist, many are rooted in the foundational work of Altman (1968) and Ohlson (1980) Recently, logistic regression analysis has gained popularity, becoming the preferred method in a majority of international studies on failure prediction.
By means of the logistic regression technique’s advantage, this study aims to research the financial ratios’ impact on the financial distress among listed companies in Viet Nam.
Research Methods
The model
Logit analysis is a statistical method that extends the Linear Regression model by incorporating one dependent variable and multiple independent variables While it shares similarities with the cumulative normal function, Logit analysis is generally more user-friendly in terms of computation This ease of use is a key reason why Logit analysis is frequently employed as an alternative to the probit model.
Where Y is the state of the company (0 = financial distress and 1= non- financial distress), β is the coefficient and X are the financial ratios calculated, k is the number of explanatory variables
Due to Y being a binary dependent variable rather than continuous, issues like heteroskedasticity and boundary problems arise, where the right side is infinite and the left side is confined to values between 0 and 1 To address these challenges, the logit function is employed.
The dependent variable (Y) represents the logarithm of the odds, which is the ratio of the probability of an event occurring to the probability of it not occurring This model is designed to predict the odds of an event happening within a specific range, rather than estimating probabilities from 0 to 1 At the midpoint of the logistic distribution, where Pi = ẵ, the slope is steepest, indicating that changes in independent variables have the most significant impact on the probability of choosing a particular option Conversely, near the endpoints, small fluctuations in independent variables result in minimal changes in probability (Pindyck & Rubinfeld, 1991) The Logit model relies on the cumulative logistic probability function.
Pi represents the probability of companies being in one of two states (Y = 1 or 0) based on eleven financial ratios The model predicts one state, while the other is represented as 1-Pi To calculate the coefficients (βi) in the logit analysis model, empirical data that includes the actual final state and financial ratio values is utilized The accuracy of the prediction hinges on the correlation between the predicted state (Pi) and the actual state Maximum likelihood estimation is employed to determine the model's parameters It has been established that the influence of variables on a company's state and the sign of each coefficient in the logit function are interdependent For example, the working capital ratio demonstrates this relationship effectively.
(1987) shows a result that a higher working capital ratio leads a higher probability of entering the financially stable state and a lower probability of entering bankrupt state.
Selection of predictor variables
Previous research indicates that there is a lack of a theoretical model for predicting financial distress, leading to the selection of variables primarily based on empirical processes Most studies rely on financial ratios derived from these empirical methods rather than established theoretical frameworks.
This study builds upon the research of Karels and Prakash (1987), utilizing eight of their nine chosen independent variables due to the lack of a suitable set of ratios and challenges in collecting financial statements (Laitinen, 2000) The selected ratios were tested for univariate normality, multivariate normality, and lognormality, revealing some deviation from normality However, the extent of this deviation is relatively lower compared to other research studies, such as Altman's work, which suggests a reasonable basis for analysis despite some limitations.
& Prakash conclude that their ratios enhance the predictive ability
These financial ratios are in five main categories of firms included profitability, liquidity, leverage, cashflow and activity ratios in the model The independent variables are:
1 Working capital is computed by the difference of Current assets and Current liabilities by Total assets (WOCA)
2 Gross profit margin is defined by the difference of Net sales and Cost of Goods sold divided by Net sales (GROPROM)
3 Earnings per share is computed by dividing Net income by number of shares outstanding (EPS)
4 Total debt to total assets is computed by the sum of Current Liabilities and Long Term Debt divided by Total Assets (DEBTTOTAL)
5 Cash flow per share is defined by dividing the sum of Net Income and Depreciation by number of shares outstanding (CASPSHARE)
6 Asset turnover is computed by dividing Sales by Total Assets (ATURNOVER)
7 Sales per cash is computed by dividing Sales by Cash (SALEPERCA)
8 Sales per receivables is computed by dividing Sales by Receivables (SALEPERRE)
The discriminating model utilizes independent variables to predict a company's status, with the importance of each variable determined through logistic regression analysis.
The financial state of a company is categorized into two dichotomous variables: non-financial distress (State 1) and financial distress (State 2) Companies in State 1 are deemed healthy and compliant with listing requirements, while those in State 2 are identified as financially distressed, as detailed in section 2.1 of the literature review.
With the selection of financial ratios as mentioned above, the model’s prediction equation is:
Data set
Every year on December 31, financial statements were sourced from two finance websites, www.cafef.vn and www.cophieu68.com The financial ratios used in the model were derived from this data The sample included companies listed on both the Ho Chi Minh Stock Exchange and the Hanoi Stock Exchange, covering two groups from 2007 to 2011.
Financially distressed firms are characterized by their struggle to maintain stability, and the challenge of supervising this distress has evolved over time Limitations in research data, particularly concerning the number of affected companies, impact the selection of the study period In this analysis, we focus on a group of 28 publicly listed companies identified as financially distressed, categorized into nine distinct sectors.
6 Technology hardware and equipment (1 company)
The second group with the non – financial distress companies that met the following criteria:
- They have not violated the listing requirement that may be lead to situations being warned, put under control, stopped trading and delisted during the research time
- The listed securities were trade on two Stock Exchanges before 2008
- They have enough financial statements during the research time
In the analysis of annual data, listed companies were classified into two categories: financially distressed and non-financially distressed Due to the removal of certain delisted firms from the sample, some lacked four consecutive years of financial data Ultimately, the study compiled a total sample database consisting of 982 non-financially distressed firms and 36 financially distressed firms.
Using Eview software, we analyzed financial ratios as independent variables within a Binary model to explore the relationship between these ratios and the future performance of companies.
Data analysis and Findings
Descriptive Statistics
An essential aspect of data analysis involves examining the distinctions between failed and non-failed companies Table 1 presents the arithmetic mean and standard deviation for eight independent variables across both groups, highlighting key differences in their performance metrics.
One year before failure, six out of eight financial ratios for non-failing companies were notably higher than those of failing companies Specifically, as companies approached failure, metrics such as working capital (WOCA), gross profit margin (GROPROM), and cash flow per share (CASPSHARE) showed a decline in financially distressed firms This trend aligns with previous findings indicating a negative correlation between cash flow and profitability ratios and the likelihood of company failure.
During the failure period, the mean of total debt to total assets (DEBTTOTAL) exhibited an increasing trend, contrasting with healthier firms that maintained lower ratios This clearly illustrates the positive relationship between financial leverage and the likelihood of failure.
Unexpectedly, companies in distress exhibited a higher SALEPERCA (sales per cash) compared to their non-failure counterparts, with the average SALEPERCA for the failure group being twice that of the non-failure group.
Variables Non - failure group Failure group
Mean Std Dev Mean Std Dev
Variables Non - failure group Failure group
Mean Std Dev Mean Std Dev
Correlations
While the regression analysis was performed, one of noticeable problem was the multicolineality regarding the high correlations between the independent variables Table 2 introduces the correlations of the full sample
The highest correlation was between WOCA and DEBTTOTAL although the calculation methods were quite different It suggests that the level of DEBTTOTAL increases as the WOCA decreases
The correlations of 0.56 between Sales per Receivables and Asset Turnover, and 0.51 between Earnings Per Share (EPS) and Cash per Share, indicate significant relationships These strong correlations can be attributed to the similarities in the financial ratios used to calculate them.
Independent variables serve as indicators of distress and are crucial for predicting failure probabilities However, a model that integrates these variables provides a more accurate assessment of distress The following section will explore the optimal combination of various explanatory variables into a unified model.
Regression model
In order to investigate the relationship between the financial ratios and the probability of failure, the next analysis was to run a binominal logistic regression
To address multicollinearity caused by high correlations between two independent variables, it is essential to omit one variable Failing to do so may render both variables insignificant, leading to inconclusive test results.
Based on the analysis of the correlation among eight explanatory variables, eight potential models were identified, each comprising five independent variables after excluding those with high correlation For instance, the model that includes Earnings Per Share (EPS) does not include Cash per Share The results of the logistic regression for these eight models are summarized below.
Table 3 The performance of logistic regression for 8 models
Mo del Variables in model McFadden
SALEPERCA, GROPROM 29.25% 0.2346 EPS, SALEPERRE WOCA
The analysis from Table 3 and Appendices indicates some following results:
Financial ratios such as EPS, ATURNOVER, SALEPERRE, and CASPSHARE negatively influence the likelihood of failure, indicating that lower values of these ratios correlate with a higher probability of financial distress Z-tests confirmed the significant impact of these independent variables, demonstrating a statistically significant relationship at the 5% level Consequently, hypotheses H1, H2, H3, and H4 are accepted.
The findings align with previous research, highlighting the asset turnover ratio (ATURNOVER) as a key indicator of a firm's efficiency in utilizing its assets to generate sales A higher asset turnover indicates greater productivity in revenue generation, leading to increased cash inflows and a reduced risk of financial distress, as demonstrated in the study by Altman and Lavallee (1981).
Sales per receivables (SALEPERRE) measures a company's efficiency in collecting revenue post-sale A higher SALEPERRE indicates quicker revenue collection, enabling the firm to promptly address its debts.
Cash flow per share (CASPSHARE) is inversely related to the likelihood of a company experiencing financial distress, as noted by Westgaard and Van der Wijst (2001) This indicates that higher earnings can significantly reduce a firm's risk of encountering financial difficulties.
A decrease in Working Capital (WOCA) is linked to an increased likelihood of financial distress for companies The analysis demonstrated a significant negative relationship between WOCA and the probability of failure, with a significance level of 0.05 in both Model 5 and Model 6 Consequently, we support the hypothesis H5.
SALEPERCA represents the annual turnover ratio of cash, which tends to be higher for companies experiencing financial distress In Models 5, 6, and 7, a significant positive correlation was found between SALEPERCA and the probability of failure at the 5% significance level, leading us to reject hypothesis H6 This outcome suggests that a high sales-to-cash ratio may indicate insufficient cash reserves, potentially resulting in financial difficulties if further financing is not accessible at reasonable rates.
In our analysis of Model 5 and Model 8, we found that the H7 model remains valid Notably, the results reveal that GROPROM has a significant negative effect on the probability of failure at the 5% significance level, indicating that an increase in GROPROM correlates with a decreased likelihood of failure.
The DEBTTOTAL ratio, analyzed through Model 7 and Model 8, supports H8 at the 5% significance level, indicating that the total liabilities to total assets ratio significantly correlates with a firm's likelihood of experiencing financial distress This finding aligns with previous research conducted by Mohamed et al (2001) and Nur Adiana et al (2008).
Table 3 presents the McFadden R-squared values for each model, which serve as a coefficient of determination This metric is utilized to assess the percentage of variance in the dependent variable that can be accounted for by the independent variables within the model.
In comparison to model 5, where EPS and ATURNOVER variables were substituted with CASPSHARE and SALEPERRE, model 1 demonstrated a significant improvement in McFadden R-squared, increasing from 21.74% to 29.89% Additionally, the Akaike information criterion for model 1 was 0.2325, which is lower than the 0.2582 value observed in model 5, indicating a better fit for model 1.
The analysis of models 3 and 5 revealed a decrease in McFadden R-squared from 29.25% to 21.74% when EPS was substituted with CASPSHARE, indicating that EPS has a more significant impact on the probability of failure compared to CASPSHARE Furthermore, the examination of models 7 and 8 demonstrated that ATURNOVER exerts a stronger influence on the probability of failure than SALEPERRE.
Moreover, the result of logistic regression also demonstrated that the impact of CASPSHARE on the probability of failure seemed to be higher compared to ATURNOVER
The analysis indicates that EPS, cash per share, and asset turnover are the key independent variables significantly influencing the likelihood of failure.
Conclusions
Summary
This study investigates the correlation between various financial ratios and the likelihood of financial distress among companies listed on the Vietnam Stock Exchange, particularly focusing on the year leading up to their potential failure.
Logistic regression analysis reveals that six out of eight financial ratios, including Earnings per Share, Asset Turnover, Sales per Receivables, Cash per Share, Working Capital, and Gross Profit Margin, are negatively correlated with the likelihood of financial failure Notably, Earnings per Share, Asset Turnover, and Cash per Share emerge as the three most significant ratios influencing a firm's financial health Conversely, Total Debt to Total Assets and Sales per Cash are positively associated with the risk of financial distress These findings align with previous research, including Altman's Z-Score model from 1978, reinforcing the importance of profit and activity ratios in assessing financial stability.
The logistic regression analysis reveals that financial distress among listed companies on the Viet Nam Stock Exchange is primarily caused by inefficient operational activities, resulting in losses in earnings per share (EPS), as well as inadequate asset management, which hampers their ability to generate revenue effectively through asset turnover.
Understanding the relationship between financial ratios and a company's condition is essential for investors in Vietnam By analyzing these ratios, investors can gain insights into a company's financial health, enabling them to identify firms facing financial distress This knowledge helps minimize investment risks, particularly when considering stocks that may be underperforming.
Limitation of the research study
One limitation of the study is that the independent variables do not sufficiently explain financial distress, as indicated by McFadden R-squared values hovering around 30% This is partly due to the data being collected only from 2007 to 2011, a period when the number of companies experiencing financial distress was relatively low Although more firms entered financial distress in 2012, gathering relevant financial ratios was not feasible due to the unavailability of their financial statements Consequently, the limited data restricts the ability to identify additional ratios, such as market ratios, that could influence the likelihood of financial failure.
Financial ratios are derived from financial statements, which often present challenges in interpreting accounting standards Additionally, the reliability of these financial statements significantly impacts the outcomes of the analysis, highlighting another limitation in the evaluation process.
The identified limitations prompt the need for additional research to address the gaps in the thesis findings The conclusions drawn regarding the relationships between variables and failure probabilities serve as a foundation for this further exploration It is important to note that each financial market exhibits unique characteristics, leading to varied reactions Consequently, ongoing research into financial ratios, along with the inclusion of macroeconomic factors like inflation, remains essential.
Future research should focus on analyzing data over extended periods, particularly examining the two to three years leading up to financial distress In addition to traditional methods like Logistic Regression and Multivariate Discriminant Analysis, emerging techniques such as Neural Networks have shown promise in predicting financial distress Incorporating these advanced methods will enhance future studies, significantly contributing to the understanding of forecasting failures in Vietnam.
Altman, E.I 1968, 'Financial ratios, discriminant analysis and the prediction of corporate failure', Journal of Finance, vol 23, no 4, pp 589-609
Beaver, W.H 1966, 'Financial ratios as predictors of failure', Journal of Accounting Research, vol 4, no 3, pp 71-111
Boritz, K., and Sun 2007, 'The effect of general price level adjustments on the predictability of financial ratios', Journal of Accounting Research, pp 273-84
Bradstreet 1985, 'Classifying Bankrupt Firms with Funds Flow Components',
Journal of Accounting Research, pp 146-60
Charalambous, C., Charitou, A & Kaourou, F 2000, 'Comparative analysis of artificial neural network models: Application in bankruptcy prediction', Annals of Operations Research, vol 99, pp 403-25
Chin, F.a 2002, 'The differential bankruptcy predictive ability of specific price level adjustments: some empirical evidence', Accounting Review, vol 58, no 2, pp 228-46
Deakin, E.B 1972, 'A Discriminant Analysis of Predictors of Business Failures', Journal of Accounting Research, vol 10, no 1, pp 167-79
Dougall, G.a 1952, 'Predicting Corporate Failure', Accountants Digest, p 3
Foster 1986, 'Current Cost Accounting and the Prediction of Small Company Performance', Journal of Business Finance and Accounting, vol 13, pp 51-70
Goudie 1987, 'Funds Flow Components, Financial Ratios and Bankruptcy',
Journal of Business Finance, pp 595-606
Hamer 1983, 'Methodological Issues Related to the Estimation of Financial Distress Prediction Models', Journal of Accounting Research, vol 22, pp 59 - 82
Heine 1995, 'Mandated accounting changes and debt covenants: The case of oil and gas accounting', Journal of Accounting and Economics, vol 7, pp 39-65
Ingram, G.a 2001, 'Motives underlying the method of payment by UK acquirers: the influence of goodwill‟', Accounting and Business Research, vol 30, no
Jones 1987, 'The Prediction of Small Company Failure: Some Behavioural Evidence for the UK”', Accounting and Business Research vol 65, pp 49-58
Lam 1994, 'Identifying failing companies: a re-evaluation of the logit, probit and DA approaches', Journal of Economics and Business, vol 51, pp 347–64
Lau 1987, 'A Five-State Financial Distress Prediction Model', Journal of Accounting Research, vol 25, no 1, pp 127-38
Lee, H.a 1985, 'Multivariate normality and forecasting of business bankruptcy', Journal of Business Finance and Accounting, vol 14, no 4, pp 573-93
Marais, P.W 1984, 'Failing company discriminant analysis', Journal of Accounting Research, vol 12, no 1, pp 1-25
Muller, S.-B., and Hamman 2009, 'The Non-Submission of Accounts and Small Company Failure Prediction', Accounting and Business Research, vol 73, no
Newton 1975, 'The Probability of Bankruptcy: A Comparison of Empirical Predictions and Theoretical Models', Journal of Banking and Finance, pp 317 - 44
Ohlson, J.A 1980, ' Financial ratios and the probabilistic prediction of bankruptcy', Journal of Accounting Research, vol 18, no 1, pp 109-31
Pinches, M.C 1973, 'Non-Financial Symptoms and the Prediction of Small Company Failure: A Test of the Argenti Hypotheses', Journal of Business Finance and Accounting vol 14, pp 335-54
Porporato, S.a 2007, 'Assessing the probability of bankruptcy', Review of Accounting Studies, vol 9, pp 5-34
Prakash, K 1987, 'Financial Distress Prediction Models: A Review of Their Usefulness', British Journal of Management, vol 2, pp 89 - 102
Sharda, W 1994, 'Predicting Corporate Bankruptcy and Financial Distress: Information Value Added by Multinomial Logit Models', Journal of Economics and Business, pp 269-86
Somerville 1989, 'Failing company discriminant analysis', Journal of Accounting Research, vol 12, no 1, pp 1-25
Taffler 1985 ' Industrial classification in UK capital markets: a test of economic homogeneity', Journal: Applied Economics, vol 17, no 2, pp 291-308
Udo 1993, 'An Empirical Comparison of Bankruptcy Models', The Financial Review, vol 33, pp 35 – 54
Zavgren 1983, 'Predicting bankruptcy resolution', Journal of Business Finance
Method: ML - Binary Logit (Quadratic hill climbing)
Covariance matrix computed using second derivatives
Variable Coefficient Std Error z-Statistic Prob
McFadden R-squared 0.298966 Mean dependent var 0.96334
Akaike info criterion 0.232597 Sum squared resid 28.88697
Hannan-Quinn criter 0.243961 Restr log likelihood -154.351
LR statistic 92.29115 Avg log likelihood -0.11019
Obs with Dep=0 36 Total obs 982
Method: ML - Binary Logit (Quadratic hill climbing)
Covariance matrix computed using second derivatives
Variable Coefficient Std Error z-Statistic Prob
McFadden R-squared 0.298455 Mean dependent var 0.96334
Akaike info criterion 0.232757 Sum squared resid 29.02307
Hannan-Quinn criter 0.244122 Restr log likelihood -154.351
LR statistic 92.13349 Avg log likelihood -0.11027
Obs with Dep=0 36 Total obs 982
Method: ML - Binary Logit (Quadratic hill climbing)
Covariance matrix computed using second derivatives
Variable Coefficient Std Error z-Statistic Prob
McFadden R-squared 0.292518 Mean dependent var 0.96334
Akaike info criterion 0.234624 Sum squared resid 29.0055
Hannan-Quinn criter 0.245988 Restr log likelihood -154.351
LR statistic 90.30054 Avg log likelihood -0.1112
Obs with Dep=0 36 Total obs 982
Method: ML - Binary Logit (Quadratic hill climbing)
Covariance matrix computed using second derivatives
Variable Coefficient Std Error z- Statistic Prob
McFadden R-squared 0.283148 Mean dependent var 0.96334
Akaike info criterion 0.237569 Sum squared resid 29.55923
Hannan-Quinn criter 0.248934 Restr log likelihood -154.351
LR statistic 87.40812 Avg log likelihood -0.11268
Obs with Dep=0 36 Total obs 982
Method: ML - Binary Logit (Quadratic hill climbing)
Covariance matrix computed using second derivatives
Variable Coefficient Std Error z-Statistic Prob
McFadden R-squared 0.217475 Mean dependent var 0.96334
Akaike info criterion 0.258214 Sum squared resid 30.761
Hannan-Quinn criter 0.269579 Restr log likelihood -154.351
LR statistic 67.13468 Avg log likelihood -0.123
Obs with Dep=0 36 Total obs 982
Method: ML - Binary Logit (Quadratic hill climbing)
Covariance matrix computed using second derivatives
Variable Coefficient Std Error z-Statistic Prob
McFadden R-squared 0.210934 Mean dependent var 0.96334
Akaike info criterion 0.26027 Sum squared resid 31.45642
Hannan-Quinn criter 0.271635 Restr log likelihood 154.3505
LR statistic 65.1154 Avg log likelihood 0.124025
Obs with Dep=0 36 Total obs 982
Covariance matrix computed using second derivatives
Variable Coefficient Std Error z-Statistic Prob
McFadden R-squared 0.193098 Mean dependent var 0.96334
Akaike info criterion 0.265877 Sum squared resid 32.44427
Hannan-Quinn criter 0.277242 Restr log likelihood
LR statistic 59.60965 Avg log likelihood
Obs with Dep=0 36 Total obs 982