Abstract

Bond credit rating is a comprehensive evaluation by credit rating agencies on the credit records, financial status and operating results of bond issuing companies. Because of information asymmetry, bond credit rating is influenced by information disclosed by companies through earnings management and information forecasted by analysts as an independent third party. Based on 311 samples of Chinese listed companies which issued bonds for the first time from 2011 to 2017, this paper studied the relationships among company earnings management, analyst forecasts and credit rating of corporate bonds. Our empirical results show that, in order to improve the bond credit rating, company managers do have some earnings management behaviors before bond issuing, and the extent of earnings management is positively correlated with its bond rating. We also find that the forecasted corporate rating by analysts is positively correlated with its bond rating, and analyst forecasts cannot restrain the impact of earnings management on bond rating and does not exclude the collusion with corporate management. Earnings management and analysts forecast and have certain mutual promotion effect on bond credit rating. The conclusions of this paper are conducive to information regulates and stakeholder decisions.

1. Introduction

Information asymmetry and accounting uncertainty are objective existence in capital market, where requires and impels rational companies to manage earnings. When a company plan to issue its bonds, company managers, perhaps, have strong motivation of earnings management for higher bond credit ranking. However, whether the credit rating agencies can identify the earnings management of enterprises, the conclusions drawn by the theoretical circle are not the same. Some scholars have found that enterprises can obtain high bond credit rating through earnings management, however, some scholars point out that credit rating agencies understand the enterprise's earnings management behavior, and give the bond credit rating downgrading penalty.

When companies disclose information about their bonds, securities analysts also verify that information and provide their predictions. As independent professionals who use information to analysis and do research, security analysts play an important role in regulating manager’s behaviors in corporate, protecting minority shareholders, improving investment efficiency and alleviating information asymmetry. But can securities analysts effectively regulate the behavior of corporate managers? Some scholars think that Chinese listed companies will conduct corresponding earnings management behavior in order to cater to analysts’ earnings forecast, while others find that analysts’ forecasts can effectively reduce the degree of earnings management of listed companies through empirical tests. Thus it can be seen, whether there is any earnings management behavior before the bond issue, whether corporate earnings management will improve its bond rating, whether the bond rating results predicted by analysts are accurate, and whether the predicted bond rating modifies the impact of earnings management on bond ratings? Such issues require further study.

Based on the bond rating data of Chinese listed companies from 2011 to 2017, this paper conducted empirical analyses on these problems. Research results show that companies do have earnings management behaviors before bond issuing, and the extent of it is positively correlated with its bond rating. What’s more, analyst forecasts can weaken the impact of earnings management on the rating, because bond rating agencies make comprehensive analyses and judgement according to various relevant information of the target company.

The contribution of this paper is that, we study bond credit rating by the perspective of internal earnings management behaviors, and by the perspective of external analyst forecasts, as well as by the perspective of their synthetical action mechanism respectively. Considering alleviating the influence of information asymmetry, internal governance is linked to external governance in our research. Nowadays, the predication from analysts is increasingly accurate, so analyst forecasts has been widely used, as well as been paid more attention, in corporate strategic decisions and national economic policies.

2. Literature Review

The purpose of this paper is to study the relationships among earnings management, analyst forecasts and corporate bond rating. We summarize the existing research literatures in three categories: (1) earnings management and bond rating; (2) analyst forecasts and bond rating; (3) earnings management and analyst forecasts.

First, related researches on earnings management and bond credit rating. Kisgen ¹ believed that credit rating agencies can measure company’s quality accurately by using the public information and their survey information. Ayers ² further argued that credit rating agencies know clearly the earnings management behaviors in listed companies, and in return, give them downgrade punishment in credit rating and rating adjustments. Besides, Caton et al. ³ pointed out that the listed companies have large amount of earnings management and manipulation before their bonds issuing, whether the first-time or once more, and the bond credit rating is significantly positive in correlation with earnings management behaviors in China’s bond market ^{4, 5}, except the highly indebted companies ⁶.

Second, related researches on analyst forecasts and bond credit rating. Bradshaw ⁷ made a comparison on interpretation ability of analyst ratings among various income model, and found that the bond rating is significantly negative with the fade-rate residual income model and not significantly positive with the sustainable residual income model, while significantly positive with the PEG and LTG models. The interpretation ability of LTG model has increased significantly compared to that of PEG, even the coefficient of LTG is significantly stronger than PEG when PEG and LTG are checked at the same time. Chan and Hameed ⁸ found that analysts who could forecast accurate ratings can indeed predicate a more profitable rating, when accuracy ratings are measured as the absolute value of the difference between actual EPS and predicted EPS. Zhang ⁹ tested the relationship between bond rating and predicted earnings in Chinese companies from the perspective of PE model, and found that the higher the predicted earning, the more likely analysts give a high bond rating.

Third, related researches on earnings management and analyst forecasts. Burgstahler and Dichev ¹⁰ believed that executives of American listed companies have the motivation and behaviors to avoid loss or prevent profit fell through earnings management. Burgstahler and Eames ^{11, 12} found that the closer to the threshold of the analyst’s earnings prediction window, the more significant the phenomenon which listed companies achieve analyst’s performance expectations through earnings management. Zheng and Cai ¹³ also showed that company managers have obvious earning management behaviors because of loss or poor profit, and this would increase the predicting difficulty for the analysts, and therefore, lower the accuracy of the analyst prediction. In addition, Wang ¹⁴ found that there exists the avoidance of unexpected negative earnings in China capital market, and Wei et al. ¹⁵ found that some Chinese listed companies have earnings management behaviors to cater for analysts’ profit forecasts.

Through the review of the previous literature, it can be found that there is still some controversy in the theoretical circle about the relationship between earnings management and bond rating. Some scholars have pointed out that enterprise may improve enterprise profitability through earnings management, to improve the rating of corporate bonds, but some scholars believe that when the earnings management behavior of enterprises is too high, not only will not play a role in promoting the rating of bonds, but will reduce the rating they receive, because the credit rating agencies clearly understand the earnings management of listed companies, and in the credit rating and credit rating adjustment to its downgrade penalty. In addition, domestic and foreign scholars have respectively studied the relationship between internal earnings management and external analysts' forecasts and bond ratings, but the effect of the combination of internal manipulation and external supervision on bond ratings has been seldom studied, this paper will specifically study whether there is earnings management behavior and its impact on bond rating, the impact of analysts' forecast on bond rating and the combined impact of earnings management and analysts' forecast on bond rating. Of course, credit ratings have very important influence on bond prices in bond secondary market ¹⁶.

3. Research Hypotheses

As an important external financing way, bond financing requires credit rating before the bond is publicly issued. From the perspective of bond issuing, bond rating is not only related to the smooth issuance of bonds but also affects the coupon rate and issuance cost of it. For listed companies, it is very important to obtain a favorable initial bond rating; while, for credit rating agencies, profitability is the focus basis of rating. From the principal-agent theory, the interests of the principal and that of agent is not completely consistent. The principal pursuits the maximum shareholder wealth under the minimum agency cost, while agents pay more attention to personal income, On-the-job consumptions and leisure time. Thus, considering the asymmetric information between the Internal managers and the external information users, the managers may conduct earnings management to improve their profitability before their bond issuing ¹⁷. For the enterprises with better performance, they think that the bond rating should be higher, so it seems unnecessary to carry out excessive earnings management. To test the existence of earnings management before the company’s initial issuance of bonds, and the earnings management level of companies with different performance, this paper proposes the following hypothesis:

Hypothesis 1a: the listed companies have earnings management behaviors before the initial issuance of bonds.

If such behaviors exist, then what is the relationship between bond rating and earnings management? Some existing studies point out that earnings management is positively correlated with bond rating. While other scholars pointed out that the rating agencies can identify the company’s earnings management behaviors, and consequently, give rating adjustment to it. And thus, analyst forecasts have an inhibitory effect on earnings management. Although the theory circle is still holding different opinions on the relationship between earnings management and bond rating, the standpoint of positive relevant is more convincing. In order to test the relationship between the both on the condition of excluding the impact of corporate performance, this paper assigns the bond rating into different gradation by scored, and puts forward the following hypothesis:

Hypothesis 1b: when the other conditions are unchanged, earnings management of the listed companies is positively correlated with their bond rating.

Before rating a company, a crucial step for analysts is evaluation of the target company. For appraising a company, analysts will make predictions on the profitability of the company based on the information of past stock price and company current information. For example, the company’s financial statement, competitors, field investigation, industry information and macroeconomic situations. This prediction is the key independent variable for evaluation and rating. Obviously, what we can be sure about the analyst evaluation is the “input variable” (forecast earnings) and the “output variable” (valuation and rating), but how the analyst goes from “input” to “output” is still a “black box”.

Some foreign scholars point out that analyst forecasts are important bases for the evaluation of corporate rating, and is positively correlated with the rating ^{18, 19}. PE model which is believed more accurately by most scholars is often used to test this relationship in analysts’ reports. Besides, we can know from the existing researches that results of analyst forecast are strong supports for investors in selecting and evaluating the target company. According to the signal transmission theory, the market will react to the signals sent by company managers and analyst. Then, what and how analysts predict the role of signaling? In order to test the relationship between analyst forecasts and bond rating, we propose the following hypothesis:

Hypothesis 2: when the other conditions are unchanged, the higher the analyst forecasts is, the higher the bond rating will be.

From the external supervision theory and the efficient market hypothesis, the external third party can play a pre-supervision role in earnings management, but cannot stop it. Furthermore, the respond speed of related information and its authenticity varies, depending on the effectiveness of the market. According to the above statement we know that, from the internal perspective of the listed company, in order to get a higher rating, they may refer to earnings management. That means, earnings management in listed companies can influence its bond rating. What’s more, before giving the final rating, analysts will make earnings forecasts from the perspective of external supervision, and the higher the earnings forecast, the higher the bond rating will be.

Then, what does the combination of internal manipulation and external oversight do to bond rating? Can analyst forecasts, as information from an external third party, affect the internal earnings manipulation? Are analysts predicting results independent and impartial or are they the result of collusion with management? Because there are many imperfections in Chinese bond market, the relevant information is lagging and opaque, and intermediaries such as securities companies are often inextricably linked with bond issuing companies, securities analysts cannot achieve absolute fairness, on the contrary, they are more likely to collude with companies. To test these relationships, we propose the following hypothesis:

Hypothesis 3: the combined effect of earnings management and analyst forecasts can enhance their separate impact on bond rating.

4. Research Design

This paper takes the listed companies in China that issued bonds for the first time in 2013-2016 and thereafter as the initial samples. The financial data observation period is 2011-2017. The financial data, enterprise attributes and analyst forecast data used in this paper are all from the CSMAR (China Stock Market & Accounting Research) Database, and the bond rating records are from the WIND Economic database.

All the rating data come from domestic rating agencies in WIND database. We thinkinternational rating agencies, such as Standard & Poor’s and Moody’s, have formed their own set of standards, but they are not necessarily suitable for China. For example, Huaneng Power International Inc was rated BBB by S&P in 2006, while Chinese CCXI were rated AAA, which is a big gap.The purpose of credit rating industry is to regulate corporate financing and market development internally, and to provide principal protection for domestic companies externally,which is also the concept and practice of many countries, such as Japan, India, Russia and so on.

In addition, the financial and insurance industry need to be eliminated because of particularity of accounting system and financial characteristics; the extreme data may disrupt the normal relationship among the data, all the poor operating results of ST and *ST companies need to be excluded (ST is Special treatment, a company that has lost money for two and three consecutive years is called ST and *ST, whose stock transactions have some policy restrictions). Therefore, we finally obtain 331 companies as the whole sample, including 170 state-owned enterprises and 161 private enterprises.

This paper takes the bond rating of listed companies as the dependent variable. There are four basic methods of bond rating in the market: factor analysis method, multivariate discriminant model method, multiple regression method and analytic hierarchy process. Among them, factor analysis is mainly based on factor score, which is too subjective, and is also difficult to guarantee the fairness and accuracy of rating. The analytic hierarchy process is mainly based on weights to determine the rating, so its classification is relatively rough.This paper uses assigning score method to refine each bond rating, suchasthefollowing:

The bond rating is divided into bond project rating (PR) and corporate entity rating (CR). Corporate overall rating is the overall credit evaluation of the issuing enterprises themselves, which can be regarded as the judgment of the comprehensive solvency of various kinds of debts, while bond project rating is the judgment of the security of a specific debt. And then, scoring method is used to quantify the rating (referred to the existing researches). The ratings of AAA, AA, A, BBB, BB, B, CCC, CC and C were assigned 100, 95, 90, 85, 80, 75, 70, 65, 60, respectively. Where there is “+” or “-”, we will plus 2 points or minus 2 points on the assigned score.

This paper uses the accrual earnings management to measure the independent variable (earnings management). Many researches ^{20, 21} show that the modified Jones model is an ideal method to measure accrual earnings management. Therefore, in this paper, we also use the modified Jones model to measure the accrual earnings management. The total accruals are divided into controllable and uncontrollable accruals. In this paper, total accruals are defined as the difference between net profit and net operating cash flow, that is:

(1)

TA_i,t represents the total accrued profit of enterprise i in year t, NI_i,t represents the net profit of that year, and CFO_i,t represents the net cash flow of operating activities.

We use the Raman and Shahrur modified Jones model ²² to measure the uncontrollable accrualsNDA_i,t:

(2)

Among them, A_i,t-1 represents the total assets from the previous year, ΔREV_i,t represents sales revenue increase, ΔREC_i,t accounts receivable increase, PPE_i,trepresents the net value of fixed assets, ROA_i,t represents return on assets. BM_i,t equals to the market share of tradable shares at the end of the year plus the value of non-tradable shares calculated on net assets plus total liabilities, and then divide total assets. The value of non-tradable shares calculated on net assets equals to the number of non-tradable shares multiply company’s net assets. a₀, a_l, a₂, a₃, a₄ and a₅ from (3) are obtained by the annual regression of all listed companies in China. Controllable accruals (4) are equal to total accruals minus uncontrollable accruals.

(3)

(4)

Referring to the existing studies, this paper downloads the earnings forecast per share (Feps) of each company from the CSMAR economic and financial research database, and use it to represent the analyst forecasts. The mean value of each analyst forecast is taken as the year’s indicator.

Based on previous studies by domestic and foreign scholars, this paper selected corporate size (Size), capital structure (Lev), the revenue growth rate (Growth), return on asset (ROA), operating cash flow ratio (Cashflow), long-term liabilities (LD) and analyst forecasts error rate (Accuracy) as control variables.

The impact of control variables on ratings is expected to be as follows: (1) the logarithm of the total assets of the company is used to indicate the size of the company. It’s expected that the larger corporate size is, the higher the bond rating will be. (2) Asset-liability ratio is used to measure the capital structure of a company. Previous studies have shown that a company with a high liability-asset ratio is more likely to violate its debt contract and is expected to get a lower bond rating. (3) The revenue growth rate is used to measure the growth of the company. The higher the revenue growth rate is, the better the growth of the company will be. However, the company is also likely to conduct positive earnings management. Therefore, it’s double edged. (4) Return on asset is used to measure the profitability of a company. The higher the expected profitability of a company, the higher its bond rating will be. (5) Operating cash flow ratio is used to measure the proportion of operating cash flow to total assets. (6) Long-term debt ratio is used to measure the proportion of long-term liabilities to total liabilities. the higher the ratio is, the greater the probability of default is. (7) Analyst forecasts error rate is used to measure the accuracy of analyst forecasts.

In order to test the influence of earnings management on bond rating, this paper constructs models (5). Since the rating agencies are mainly based on the company’s performance over the past year, the independent variable uses data lagging one year.

(5)

In model (5), β₀ is constant, ε_i,t is the residual. β₁ is earnings management coefficient, which is expected to be positive. The greater its value is, the greater the influence of analyst forecasts on corporate rating and debt rating. According to Standard & Poor’s rating index, it’s expected that control variables of Size, Growth, ROA, Cashflow, LD are positive correlation with the corporate rating and the debt rating, but Lev is contrary to ratings.

In order to test the influence of analyst forecasts on bond rating, this paper constructs models (6) to test the influence of analyst forecasts on corporate rating and debt rating, respectively. In the model (6), the analyst forecast (Feps) is the value of the rating year, and the analyst forecast accuracy variable (Accuracy) is added, which is also an important factor for rating agencies to consider.

(6)

In model (6), β₁ is analyst forecasts coefficient, which is expected to be positive. It shows that the rating agencies will take full account of the analyst’s forecast information. Other control variables have the same meaning and function as the model (5).

This paper constructs models (7) to test the combined influence of earnings management and analyst forecast on bond rating. We introduced the cross-multiplication value of earnings management and analyst forecasts, and tested the comprehensive influence of analyst forecasts and earnings management on corporate rating and debt rating respectively.

(7)

In the model (7), β₁and β₂ are coefficient of earnings management and analyst forecasts. β₃ is the cross-multiplication coefficient of earnings management and analyst forecasts. Given β₃ positive, it means analyst forecasts and earnings management together can strengthen their separate influence on bond rating. if its negative, that means analyst forecasts can, to some extent, restrain the influence of earnings management on bond rating.

5. Empirical Analysis

Referencing to related research (Liu, 2014), this paper first makes a preliminary judgment of the existence of earnings management by analyzing the changes of earnings before and after the credit rating. Table 2 shows the changes of the profitability indicators of the sample near their first rating. To analyzing the profitability of the listed companies before and after the rating, we use return on asset (ROA), return on equity (ROE) and operating profit growth ratio (IG) as the main indicators. From the mean and median value of each indicator, we find that ROA goes down gradually from 2 years before the rating to 1 year after the rating, and in the second year after the rating, the average ROA climbs back up slightly. The mean and median value of ROE keep going down from 2 years before the rating to 2 years after the rating. Specially, for IG, the mean value of it peaks in the year before rating, while in the year of rating, that means the year after the rating, IG value falls to negative. All these statistics show that the listed companies have earnings management behaviors before the first bond rating.

Table 3 gives the descriptive statistical results of all samples. As can be seen from Table 3, the mean value of corporate rating and debt rating is 95.968 and 96.069 respectively, which is between AAA and AA. This indicates that the bond issuer has a high overall credit rating. In addition, the debt rating is slightly higher than the corporate rating, indicating that Chinese listed companies has a high requirement for issuing bonds. The mean absolute value of earnings management is 0.107, which is totally different from zero. The maximum value of the analyst forecasts is 3.210, the minimum value is -0.080, so the average vale is 0.624. There are both positive and negative values, but the positive forecast is significantly higher than the negative value. The average liability-asset ratio is 0.564, with maximum and minimum were 0.872 and 0.203 respectively. This indicates that the liability level of the existing sample is above the average, and they face high financial risks. The standard deviation of the growth index is 0.449, indicating that there is a certain difference in the growth stage of the companies, and thus, may have negative effect on rating’s expectations. The average operating cash flow ratio is 0.029, which is expected to have a positive impact on the rating. Analysts are forecasting an average margin of error of 1.243, and standard deviation of 5.269, indicating that there is a certain differencebetween the analyst forecasts and the real value. So, it’s expected that analyst forecasts have a certain impact on the rating.

Table 4 shows the result of the correlation coefficient of variables. From the correlation coefficient matrix, we can see that the long-term liability ratio and the operating cash flow ratio is significantly positively correlated, and the relationships between company’s growth and operating revenue growth ratio or the net operating cash flow ratio are significantly negatively correlated. Asset-liability ratio and net operating cash flow are negatively correlated with long-term liabilities ratio significantly. The size of the company is negatively correlated with the long-term liability ratio, and is positively correlated with the asset-liability ratio. Analyst forecasts is significantly positively correlated with enterprise size, while analyst forecasts error rate is negatively correlated with the net operating cash flow ratio.

In order to verify the hypothesis 1, this paper adopts the comparative analysis method of earnings management. The first is to judge whether a company has earnings management behavior before issuing bonds by comparing the degree of earnings management before and after issuing bonds, next is to judge whether a company with higher bond rating will have more earnings management.

Earnings management is calculated from the model (1) - (4). In order to test hypothesis la, we classify the absolute value of earnings management as earnings management before rating, earnings management in the rating year, and earnings management after the rating year, and then, compare the mean value of them.

From Table 5, we can find that the extent of earnings management steps down gradually from the year before the rating to the year after the rating. A paired sample test is made on earnings management in the year before rating and in the rating year, and the test results in Table 6 show that there is a significant difference of the earnings management in the 95% confidence interval, indicating that reduction of earnings management in the rating year. What’s more, as profit indicators in Table 2 have pointed out that the ROA and ROE is on the decline, and IG peaks before the rating year and then goes to negative. Therefore, it further illustrates the fact that the listed companies have earnings management behaviors before first rating, and hypothesis 1a is true.

Then, what are the differences in earnings management between companies with different rating? We divide the whole sample into two categories depending on the rating score: 92-95 (including) points and 95-100 points (All the company’s bond rating in the sample is at grade A and above). For corporate entity rating, we have 224 samples in category 92-95 points, and 107 in category 95-100 points. For bond project rating, we have 213 samples in 92-95 points category, and 118 samples in 95-100 points category. We can find the relationships between bond rating and earnings management for different groups from Table 7. There is the same rule whether it’s corporate rating and debt rating, that is, companies with higher ratings have lower earnings management. Thus,it seems that hypothesis 1b could not be verified and is incorrect. This result is identical to the research conclusions of Yang ²³, who argue that rating agencies have the ability of discriminate to identify earnings management behavior of listed companies. But we don’t think so, because the company get a high credit rating mainly due to their good profitability and solvency. Many studies ^{24, 25, 26} show that companies with better performance do not need excessive earnings management, the sectional statistics of ROA and Cashflow in Table 7 also explain this. If we want to test the relationship between bond rating and earnings management, we must exclude the influence of enterprise performance itself, and it’s necessary to carry out multiple regression analysis ²⁷.

To further test the hypothesis1b, this paper carries out multiple regression analysis on model (5), which includes conventional parameters of bond rating agencies as control variables. At the same time, in order to test hypothesis 2 and 3, Multivariate regression analysis of the model (6) and (7) is also needed. We checked the relevant statistical variables of the three models, and find that: (1) the maximum variance inflation factors (VIF) of explanatory and control variables is 1.58, which is far less than the standard value 10, so we believe that there isn’t multicollinearity in this model (The VIF of variables in model (7) are as follows: DA, 1.05; Feps, 1.06; DA*Feps, 1.19; Size,1.43; Growth, 1.04; LD, 1.12; ROA, 1.00; Cashflow, 1.18; Lev, 1.58; Accuracy, 1.03). (2) the Durbin-Watson statistic (D.W) is around 2.13, which shows that there is no autocorrelation. In other words, the sample and explanatory variables of this paper are consistent with the basic assumptions of linear regression.

In order to reflect the influence of property rights, this paper make multivariate regression analysis based on the samples of state-owned enterprises, non-state-owned enterprises and all enterprises respectively, and the regression results of model (5) are shown in Table 8. (1) The adjusted goodness of fit R² can be accepted, and the F statistic is significant at the 1% significance level, which indicates that the model is effective. (2) In full sample regression, whether interpreted variable is CR or PR, the partial regression coefficient of DA is positive, and DA is significant at the 10% level for PR. Corporate ratings may focus on the company’s overall financial situation, including the company's early-stage basis, while bond ratings focus more on the company’s performance in the past year, and management is more motivated to carry out earnings management performance. The empirical results basically confirm the hypothesis 1b. (3) What’s more, for the sample of state-owned enterprises, the symbols and significance of DA are basically the same as that of the total sample. For the sample of private enterprises, CR and PR is significant positive correlation with DA at the 10% level, and the coefficient is larger than the sample of state-owned enterprises, which indicate private enterprises are more likely to conduct earnings management in order to get a better bond rating.

In addition, the empirical results of multiple regression in Table 8 overturn the results ofsegmental comparison in Table 7, because the results in Table 8 consider the company’s own profitability and solvency, which is an important and direct reference for rating agencies. So, the significance of some control variables is also worth analyzing in Table 8, and it can verify the accuracy of bond rating by bond evaluation agencies.

(1) Size has a significant positive impact on both CR and PR, which indicates that the larger the size of the listed company, the higher the bond rating it obtains, and this is in line with the expectation.

(2) Cashflow and ROA are significantly positively correlated with the CR and PR, which is in line with the expectation. According to the significance level, this feature is more obvious in state-owned enterprises.

(3) Lev has a significant negative impact on the CR and PR, which is in line with the expectation. According to the significance level, this feature is more obvious in private enterprises.

(4) LD has a significant positive impact on both the CR and PR, which indicates that the more long-term liabilities an enterprise obtains, the better its external credit will be, and the higher its bond rating will be. This feature is more obvious in state-owned enterprises.

(5) Growth is significantly negatively correlated with the CR at the level of 5% in private enterprises. Growth is characterized by the change rate of main business income, which can be understood as income volatility, representing a certain risk, so it is negatively correlated with bond evaluation, especially in private enterprises.

The regression results of model (6) are shown in Table 9. (1) In full sample regression, whether interpreted variable is CR or PR, the partial regression coefficient of analyst forecast is positive, and significant at the 5% level. The results indicate that analyst forecast has a positive effect on both CR and PR. As an independent third-party organization, information released by analyst is often considered objective, and the rating agencies will make full use of the information. That is, the higher the analyst forecasts are, the higher the bond rating of the listed company, Hypothesis 2 is validated. (2) Whether interpreted variable is CR or PR, analyst forecasts coefficient is positive and significant at the 5% level in the sample of state-owned enterprises. Inthesampleof private enterprise, analyst forecasts coefficient is alsopositive and significant at the 1% level, and the coefficient is larger than the sample of state-owned enterprises. Namely, analyst forecasts have a bigger impact on bond rating in private enterprises. This may be the reason that the information asymmetry of private enterprises is more serious, and rating agencies prefer to get information from analysts. (3) There is no significant positive or negative correlation with the Accuracy, indicating that the accuracy of analysts’ past forecasts has not been considered by rating agencies, which also had proved by Yao ²⁸. The symbols and significant of other control variables are basically the same as those of model (5).

Table 10 shows the multiple regression results of hypothesis 3. The relevant statistics indicate that the model (7) is effective.

(1) The coefficient of DA is positive in the whole sample, which is like hypothesis 1b, shows that rating agencies only use company performance information and can’t identify earnings management behavior. Analyst forecasts are still significantly positive, which implies analyst forecasts have greater influence on bond rating. The cross-multiplication coefficient of earnings management and analyst forecasts is positive but not significant, which shows to some extent that hypothesis 3 exists and is probably correct.

(2) In state-owned enterprises, DA is significantly positively correlated with the CR, and analyst forecasts is significantly positively correlated with the PR. Cross-multiplication coefficient is positive, especially significant positive for CR. In private enterprises, the analyst forecasts coefficient is significantly positive, and the cross-multiplication coefficient is significantly positive, indicating that the combined effect of analyst forecasts and earnings management in private enterprises strengthens the impact of each on the bond rating. hypothesis 3 was confirmed.

(3) From the control variables, ROA and Cashflow are significant. It is easy to see that the rating agencies prefer cash flow to net profit in the state-owned enterprises and just reversed in the private enterprises. Of course, earnings management of profits is much more convenient than cash flow ²⁹.

Understandably, the earnings management of enterprise management is to get a better bond rating, but whether the analyst as an independent third party can objectively analyze and expose the earnings management behavior of enterprise management? So, By analyzing the influencing factors of earnings management, we can find some important information to explain the hypothesis 3. Therefore, taking earnings management (DA) as dependent variable in model (6), the regression results are shown in the Table 11. Earnings management is significantly positively correlated with analyst forecasts, whether in total sample or subsample. Since the analyst’s forecast is based on the company’s profitability, solvency, operation and growth in the previous year, it should reflect the real situation without earnings management.

To explain the empirical results, on the one hand, the results of analysts forecast may become the goal of the company, which will urge company management to carry out earnings management by various ways to achieve the results of analyst forecast; on the other hand, it does not exclude the collusion between analysts and the management, that is to say, the results of analyst forecast are obtained through consultation with the company management, so the analysts do not undertake evaluation function as an independent and impartial third party.

The Table 11 also shows what kind of companies prefer earnings management, from which we can find the ways of earnings management. Growth is significantly positively correlated with earnings management, while Lev and Cashflow are just reversed. So high income growth gives earnings management opportunities, but debt rate and cash flow cannot be manipulated. ROA is negatively correlated with earnings management but not significant, which shows that companies with strong profitability do not need excessive earnings management, and the motivation of profit manipulation is weak, which is consistent with the previous analysis.

Dueto the dependent variable (bond rating) is a discrete variable, and by referring to the existing literature, this paper uses the ordered logit regression method to test the robustness of three assumptions in this paper.

To be specific, since all the samples are rated with A, AA, AAA, or add “+”, “-” to them, we will divide A, AA, AAA rating samples into 1, 2, 3 groups, and then quantify the rating level to 1, 2, 3. In corporate entity rating (CR), we have 7 samples in the first group, 236 samples in the second group, and 88 samples in the third group. For bond project rating (PR), we have 6 samples in the first group, 238 samples in the second group, and 87 samples in the third group.

Table 12 shows the robustness test results (Table 12 does not list control variables, and the coefficients, symbols, and significance of the control variables are consistent with those above). No matter it’s in the whole sample or the sample of state-owned enterprises and private enterprises, no matter it’s model (5) or (6), earnings management and analyst forecasts are significantly positive correlated with CR and PR, analyst forecasts can enhance their separate impact on bond rating.

There is no theoretical endogeneity in the empirical research. Firstly, the company’s related variables lag one year (the annual report was published before April next year), and the analyst forecast is based on the financial data of the previous year of the company. Secondly, the rating agencies have multiple ratings for the same bond, and we use the last rating result, which is given based on company’s financial data from the previous year.

6. Conclusions and Suggestions

This paper studies the relationship between earnings management, analyst forecasts and corporate bond credit rating, and the empirical results show that: (1) Before the initial issuance of bonds, listed companies have serious earnings management behavior to obtain higher bond credit rating. Although companies with strong profitability have a lower degree of earnings management, excluding the profitability factors, the higher the degree of earnings management, the higher the bond rating, the bond rating agencies cannot effectively identify the company’s earnings management behavior. (2) The information of analyst forecast is an important reference for bond rating agencies. The higher the analysts forecast, the higher the bond rating. Moreover, the impact of analysts forecast on the corporate entity rating (CR) is greater than that of bond project rating (PR), and the impact on private enterprises is greater than that of state-owned enterprises. Bond rating agencies prefer profit information of state-owned enterprises and cash flow information of private enterprises. (3) Analysts forecast have a strong positive correlation with earnings management, which cannot represent the fairness of external third parties to a certain extent, and does not exclude the collusion with corporate management. Earnings management and analysts forecast and have certain mutual promotion effect on bond credit rating.

Bond investors should have their own independent judgment, not blindly follow the conclusions of securities analysts, which cannot really play the role of fairness and supervision as a third party, and have a certain degree of collusion with the management of enterprises. Forecast information isn’t a personal view of analysts because of its strong influence on bond investors and rating agencies. Governmental subdivision and industry organizations must enact regulations to curb inappropriate and incorrect opinions on bond transaction. Analysts and their organizations that seriously mislead investors need to be severely punished including fines and cancellation of qualifications. If an analyst is found to have conspired with the company management, such conduct constitutes a crime and is subject to criminal law sanctions.

Bond rating agencies should not only have certain professional qualifications, but also fulfill their rating tasks with due diligence. Especially, they should not conspire with bond issuing companies. The biggest scandal of credit rating industry in 2018 was the suspected high-priced sales credit rating by Dagong Global Credit Rating (During the period from November 2017 to March 2018, Dagong Global Credit Rating directly demanded consulting fees while providing credit rating services for the company. For example, after two working days Xinguang Company paid a high consulting fee, Xinguang Debt rating has been quickly adjusted from AA to AA+), which had a serious warning, rectifying within a time limit, and suspending the business related to the debt financing instrument market for one year. National Association of Financial Market Institutional Investors (NAFMII) has also announced that it will accept S&P Credit Rating (China) and enter the interbank bond market for the registration of bond rating business.

Fortunately, National Development and Reform Commission (NDRC) has carried out the credit evaluation of the principal underwriters of corporate bonds and credit rating agencies in 2019. The evaluation indexes are divided into credit behavior index, business ability index, and expert & institution evaluation index. The evaluation results will be published on the “Credit China” website and included in the credit files of enterprise bond intermediaries. According to the evaluation results, the main underwriters and rating agencies are classified and managed, and the punishment measures of trustworthiness incentive and dishonesty are implemented.

Of course, it is also necessary to establish a corporate credit evaluation system, improve accounting standards and corporate governance mechanism, which are foundation for containing corporate profit manipulation ^{30, 31}.

Acknowledgements

The author would like to thank the instructions of Professor De-ren XIE (Tsinghua University) and Professor Zhao-guo ZHANG (Huazhong University of Science and Technology). This work is supported by the National Social Science Foundation (No. 13BJY018) and Enterprise Management Research Center of CTBU (No. QYGLTD201802).

References