國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
VII. THE EMPIRICAL RESULTS
Table 1 shows that each model has different goodness of fit with adjusted R squared and t-ratio at p<0.05 level of confidence. The data is collected from CSMAR database and X, Y, Z represent independent variables for each models. I conducted regression test with between data of panel data. As you can see from Table 1, the result of the modified Jones model (C9, B, D, E,K) and Yoon (2006) model as well as The new model (C1, C4, C5, C7, F) describe higher goodness of fit (Adjusted R squared>50%) with more than one independent variable is statistically significant at p<0.05 level of confidence.
‧
Regression tests for comparative analysis of alternative models. The results are based on a sample of 3326 firm-years.
Modified Jones model Yoon (2006) model The new model
Industry n 𝑋1 𝑋2 𝑋3 adjR2 𝑌1 𝑌2 𝑌3 adjR2 𝑍1 𝑍2 𝑍2 adjR2
Ho: the each independent variable is not significant.
H1: the each independent variable is significant.
If t-ratio implies that Ho holds mean that independent variable does not have significant power to explain the dependent variable.
Since the defree of freedom is more than 29, the coefficient variable will be significant with 2.045>t-ratio at p<0.05 level of confidence.
‧
independent variable. If the model eliminates the overlapped effect among independent variables, it will provide the pure magnitude of EM. Therefore, the correlation effect on dependent variable needs to be tested by VIF, which allows to observe whether the model includes multicollinearity or not. If the model shows multicollinearity, independent variables are highly correlated. Table 2 indicates that Yoon (2006) model has higher possibility of multicollinearity, but other models from Table 2. In addition this finding is also stated in the Yoon (2006). They argued that multicollinearity is caused by the high correlation between 𝒀𝟏 and 𝒀𝟐.Table 2
Comparative VIF test for overall regression test for alternative models.
Modified Jones model
In Table 3 I implemented correlation test for Yoon (2006) model and the result points out that there is a high positive correlation between 𝒀𝟏 and 𝒀𝟐. This result tells us that Yoon (2006) model includes overlapped effect, therefore, it is not reliable for detecting EM in China.
‧
company’s financial data is retrieved from CSMAR. The total number of North America listed companies are 2834 and they were classified into 14 industries and 1663 Chinese listed companies were classified into 17 industries. The classification code is different between GICS and CSMAR. This study uses the simulation to specify industry with group code and business type. Figure1 shows that bar 1 represents the change of accounting item in 2012 and bar 2 represents the change of accounting item in 2013. According to the graph, the change rate of receivable and payable is similar between China and North America, but the change of inventory shows higher rate in China, compared to North America. The change of impairment cannot be compared because the impairment test employs different method in both countries. North America companies uses IFRS which does not allow them to reverse the price unless the industry has unique characteristic. For example, the inventories are held by producers of agricultural and forest products, minerals and mineral products. Therefore, the sign of impairment appears to be negative on the financial statement of North American companies and much less companies appropriates impairment loss. While North America does not show enough data on impairment, the change of impairment dramatically change in Chinese companies during 2011 to 2013.
Table 3
Correlation test for Yoon (2006) model
(obs=3326)
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Figure 1
Time Series for the variable change of accounting items, each variable is standardized by Revenue. Year 1 represents the change from 2011 to 2012 and Year 2 represents the change from 2012 to 2013. Sample consists of 2834 firm-year selected from GICS and
1663 firm-year selected from CSMAR.
North America ΔAccount Receivable China ΔAccount Receivable
North America Δ Account Payable China Δ Account Payable
North America Δ Impairment loss and gain
Only (200/total 2834) is available from North America database. Their impairment is attributed to the
goodwill.
China Δ Impairment loss and gain
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
North America Δ Inventory China Δ Inventory
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Since the modified Jones model and Yoon (2006) model do not consider direct effect of impairment and inventory, the model is susceptible to catch the unique characteristic of accounting items of Chinese companies. Table 4 and Table 5 display the regression result from three accrual-based model. The modified Jones model shows better goodness of fit in (C5, D, E, K) in 2012. Yoon (2006) model and the new model shows better goodness of fit in (C1, C4, C6, C7, C8, F) in 2012. The modified Jones model shows better goodness of fit in (C7, C9, B) in 2013. Yoon (2006) model shows better goodness of fit in (C1, C4, C5, B, H) and The new mode shows better goodness of fit in (C1, C4, C6, C7, C8, F) in 2013. At a first sight it looks like Yoon (2006) model and the new model is the same result as Table 1. However, the regression was tested in 2012 and 2013, separately. Table 6 and 7 indicates that every model includes multicollinearity in 2012 and 2013. As the result, modified Jones model, Yoon model and the new model are required to being omitted the collinearity to obtain enhanced result of DA. C1, C2, C4, C5, C6, C7, C8, C9, F, J from the new model and D from the Modified Jones model in 2012 as well as C1, C4, C5, C7, A, E, F, H, J from the new model and C8, C9 from the modified Jones model in 2013 is used to calculate DA. Consequently, Hypothesis 1 (The modified Jones model is not effective detecting EM in China) is rejected and Hypothesis 2 (Yoon (2006) model is not effective detecting EM in China) is not rejected and Hypothesis 3 (The new model is not effective detecting EM in China) is rejected. These results tell us that the modified Jones model and the new models are able to catch unique characteristic of NDA when we apply them for particular industries and year in China.
However, the use of them depends on what type of data we use. For example, Yoon (2006) model needs to be tested with panel data due to the multicollinearity. Also, there should exist other accounting items which can enhance the model explanatory power in each model.
‧
Comparison tests for EM based on alternative models to measure DA. Each model estimated coefficient which is extracted from Chinese listed-companies in 2012. Sample of 1663 firm-year is collected from
CSMAR.
Modified Jones model in2012 Yoon (2006) model in 2012 The new model in2012
I N 𝑋1 𝑋2 𝑋3 adj.
P-value < 0.1, 0,05, 0.01 level of confidence is denoted by *, **,***, respectively.
Table5
Comparison tests for EM based on alternative models to measure DA. Each model estimated coefficient which is extracted from Chinese listed-companies in 2013. Sample of 1663 firm-year is collected from
CSMAR.
Modified Jones model in2013 Yoon (2006) model in 2013 The new model in2013
I N 𝑋1 𝑋2 𝑋3* adj.
P-value < 0.1, 0,05, 0.01 level of confidence is denoted by *, **,***, respectively.
‧
Variable Influence Factor represents the level of multicollinearity. If the VIF>1/(1-R²) the multicollinearity is statistically significant on the independent variable. Test is conducted in 2012.
The sample of 1663 firm-year is collected from CSMAR database.
Modified Jones model in 2012 Yoon (2006) model in 2012 The new model in 2012
I n 𝑋1 𝑋2 𝑋3 1 significantly no-relationship. Variable Influence Factor (VIF) shows how multicollinearity has increased the instability of the coefficient estimates (Freund 2000: 98). Put differently, it tells you how “inflated” the variance of the coefficient is, compared to what it would be if the variable were uncorrelated with any other variable in the model. Klein (1962) suggests alternative criterion that if R exceeds R2 of the regression model. In this vein, if VIF is greater than 1/(1-R²) or a tolerance value is less then (1-R²), multicollinearity can be considered as statistically significant.
‧
Figure 2 reports the bar graph of DA. The vertical line denotes the magnitude of DA and horizontal line denotes the name of industry. Also, Figure 2 shows that the DA bar between SOE and NSOE looks similar in 2012 and 2013. The data table beneath the graph allows to see the median of DA (C1, C2, C6, F, J) in 2012 and (C4, C5, C8, E, F, E) in 2013 are greater than NSOE. The most interesting here is that the bar location is opposite during two periods. For instance, bar graph in 2012 is asymmetric with bar graph
Table7
Variable Influence Factor represents the level of multicollinearity. If the VIF>1/(1-R²) the multicollinearity is statistically significant on the independent variable. Test is conducted
in 2013. The sample of 1663 firm-year is collected from CSMAR database.
Modified Jones model in2013 Yoon (2006) model in 2013 The new model in 2013
I n 𝑋1 𝑋2 𝑋3 significantly no-relationship. Variable Influence Factor (VIF) shows how multicollinearity has increased the instability of the coefficient estimates (Freund 2000: 98). Put differently, it tells you how “inflated” the variance of the coefficient is, compared to what it would be if the variable were uncorrelated with any other variable in the model. Klein (1962) suggests alternative criterion that if R exceeds R2 of the regression model. In this vein, if VIF is greater than 1/(1-R²) or a tolerance value is less then (1-R²), multicollinearity can be considered as statistically significant.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
in 2013. Although only (C1, C4, C5, C7, C8, F) can be compared between two periods, the asymmetric results can be seen these industries.
Figure 2
Each Graph represents the level of DA for SOE and NSOE. Sample is retrieved from IPO companies financial statement.
‧
According to the Table 8, the PPE in 2012, LTD and AR in 2013 is statistically significant on DA. However, the regression results of SOE holds multicollinearity in 2013.
Therefore, it should be retested to eliminate multicollinearity. The goodness of fit for SOE in 2012 is quite high, which is indicated by the adjusted-R squared (98%). In addition, the regression result of the adjusted-R squared (17.5%) of SOE in 2013 is comparatively better than NSOE.
Table 8
The results of difference test regressing DA on accrual and non-accrual items.
Sample is retrieved from IPO companies’ financial statement.
SOE NSOE SOE NSOE
(10906.293) (57407.122) (1.629e+09) (6.767e+10)
Nh 165 210 N 152 197
1/(1-R²) 55.56 1.16 1.21 1.09
adj. R2 0.980 0.102 adj. R2 0.175 0.044
OAR = other account receivable, AP = account payable, OAP = other account payable, IMP = impairment loss and gain, INV = inventory, STD = short-term debt, PPE = property plant and equipment, and LTD = long-term debt.
*All variable are standardized by revenue, except for LTD.
*LTD is standardized by Liability.
*P-value < 0.1, 0,05, 0.01 level of confidence is denoted by *, **,***, respectively.
‧
results, table 4 tested correlation. Table 9 shows PPE is positively correlated with DA at (99%) in 2012 and Table 5 shows LTD and AR is negatively correlated with DA at (-42%) and (-27%) in 2013. Compared to NSOE, SOE controls DA by PPE, AR and LTD. This result of LTD is consistent with Cheng (2015), but only 64 SOEs appropriate LTD in 2013 and also the regression test includes multicollinearity. Additionally, short-term debt is positively correlated with payable in 2012 and 2013 and LTD is positively correlated with Account Receivable in 2013.Correlation test on DA for SOE in 2013.
DA AR OAR AP OAP IMP INV STD PPE LTD
Table 11 reports that LTD and AR is statistically significant at p<0.05 level of confidence for SOE without multicollinearity. Table 12 shows that LTD (-49%) and AR
‧
(-51%) are negatively correlated. On the contrary, difference test for NSOE does not show better result. Coefficient of LTD positively related with DA for NSOE. Moreover, independent variables are not statistically significant. It may be because independent variables are not directly correlated with DA. For the additional information about LTD, 34 companies gained LTD and 25 companies paid LTD from 2012 to 2013. Other 8 companies does not change LTD from 2012 to 2013. It suggests that Hypothesis 4 (SOE has less incentives to manage earnings because of bank loans.) is rejected in 34 companies because they borrow money and decreased the magnitude of DA in 2013.
Table 11
P-value < 0.1, 0,05, 0.01 level of confidence is denoted by *, **,***, respectively.
Table 12
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
VIII. CONCLUSIONS AND IMPLICATIONS
In this paper we found that the specific accrual-based model needs to be applied for each year and each industry. In other words, the model sensitively reflects the unique features of each industry as well as the change of accounting item’s value. The above mentioned problems should be carefully taken care of in order to accurately detect EM.
Therefore, panel data analysis may not be feasible for EM study because it will dismiss the discretional change of accounting item’s value. And hence, two different models such as the modified Jones model and the new model performed better in particular industry and year. Especially, the new model showed more appropriate performance, compared to other accrual-based model. Consequently, I found that SOE has conducted stronger EM in pre-IPO market during 2011 to 2013 by manipulating with PPE, the flow of AR and LTD. In addition, only 34 companies reported that they lent loans from bank and other 25 companies paid loans back from 2012 to 2013.
These findings can be implicated as following three points. First, the new model is not theoretically verified, but the use of appropriate model conducts different results from prior researches. 10We thus need to suggest a new solid model specification to clarify the EM in China. Secondly, Chinese SOE undergoing IPO process manipulates earnings because they are not exposed under strict audit and monitor. The magnitude of EM between SOE and NSOE is not different, therefore, SOE camouflages earnings by manipulating accrual items. Finally, Chinese central bank appreciated interest rate after 2010 and it caused higher deposit rate as well as lower investment rate. Therefore, I assume that some SOE needed to increase their earnings to obtain bank loan before they enter IPO market.
‧
[2] DeFond, M. L., & Jiambalvo, J. (1994). Debt covenant violation and manipulation of accruals. Journal of accounting and economics, 17(1), 145-176.
[3] Dechow, P. M., Sloan, R. G., & Sweeney, A. P. (1995). Detecting earnings management. Accounting review, 193-225.
[4] Chaney, P. K., & Lewis, C. M. (1995). Earnings management and firm valuation under asymmetric information. Journal of Corporate Financ1995
e, 1(3), 319-345.
[5] Bernard, V. L., & Skinner, D. J. (1996). What motivates managers' choice of discretionary accruals?. Journal of Accounting and Economics, 22(1), 313-325.
[6] Freud, R. J., & Littell, R. C. (2000). SAS system for regression. Sas Institute.
[7] McNichols, M. F. (2001). Research design issues in earnings management studies.
Journal of accounting and public policy, 19(4), 313-345.
[8] Baltagi, B. (2008). Econometric analysis of panel data. John Wiley & Sons.
[9] Hsiao, C. (2003). Analysis of panel data, 2nd. Cambridge: Cambridge University Press. Kose, MA, ES Prasad and ME Terrones (2003), Financial Integration and Macroeconomic Volatility, IMF Staff Papers, 50, 119-142.
[10] Roychowdhury, S. (2006). Earnings management through real activities manipulation. Journal of accounting and economics, 42(3), 335-370.
[11] Yoon, S. S., Miller, G., & Jiraporn, P. (2006). Earnings management vehicles for Korean firms. Journal of International Financial Management & Accounting, 17(2), 85-109.
[12] Wang, Q., Wong, T. J., & Xia, L. (2008). State ownership, the institutional environment, and auditor choice: Evidence from China. Journal of accounting and economics, 46(1), 112-134.
[13] Ball, R., & Shivakumar, L. (2008). Earnings quality at initial public offerings.
Journal of Accounting and Economics, 45(2), 324-349.
[14] Hujino Yutaka, (2009). New problems and present situation of the accrual-based model. Saint Paul’s University Economic Study volume 62.3, p95-p112 (藤野裕. (2009).
裁量的会計発生高推定モデルの現状と新たな課題: モデルが仮定する条件の現 実妥当性について. 立教経済学研究, 62(3), 95-112.)
[15] Berger, A. N., Hasan, I., & Zhou, M. (2009). Bank ownership and efficiency in China: What will happen in the world’s largest nation?. Journal of Banking & Finance, 33(1), 113-130.
‧
initial public offerings: a re-examination. Rock Center for Corporate Governance Working Paper, (23).[17] Chen, H., Chen, J. Z., Lobo, G. J., & Wang, Y. (2010). Association between borrower and lender state ownership and accounting conservatism. Journal of Accounting Research, 48(5), 973-1014.
[18] Aaker, H., & Gjesdal, F. (2010). Do Models of Discretionary Accruals Detect Actual Earnings Management via Inventory? A Comparison of General and Specific Models. A Comparison of General and Specific Models (June 24, 2010).
[19] Benmelech, E., Kandel, E., & Veronesi, P. (2008). Stock-based compensation and CEO (dis) incentives (No. w13732). National Bureau of Economic Research.
[20] Islam, M. A., Ali, R., & Ahmad, Z. (2011). Is modified Jones model effective in detecting earnings management? Evidence from a developing economy. International Journal of Economics and Finance, 3(2), 116.
[21] Paul, R. K. (2006). Multicollinearity: Causes, Effects and Remedies. IASRI, New Delhi.
[22] Alareeni, B., & Aljuaidi, O. (2014). The modified jones and yoon models in detecting earnings management in Palestine Exchange (PEX). International Journal of Innovation and Applied Studies, 9(4), 1472.
[23] Shen, Z., Coakley, J., & Instefjord, N. (2014). Earnings management and IPO anomalies in China. Review of Quantitative Finance and Accounting, 42(1), 69-93.
[24] Cheng, C. A., Wang, J., & Wei, S. X. (2015). State ownership and earnings management around initial public offerings: Evidence from China. Journal of International Accounting Research, 14(2), 89-116.
[25] Tetsuo Hujita, (2014). Financial progress and challenges in China. The Japan Research Institute, United, Pacific Rim Business Information 14, No.55. （藤田哲雄.
(2014). 中国の金融改革の進展と課題. Rim: 環太平洋ビジネス情報, (55), 62-79.）
[26] Cai, Y., & Zhang, L. (2015, May). The relation between equity incentives and earnings management in China. In Information Engineering and Education Science:
Proceedings of the International Conference on Information Engineering and Education Science (ICIEES 2014), Tianjin, China, 12-13 June, 2014 (Vol. 5, p. 29). CRC Press.
[27] PwC China Assurance Partner, Doing business and investing in China. Retrieved
Feb 15 2006, from:
http://www.pwccn.com/webmedia/doc/634940150734265198_iic_full.pdf. (Global Annual Review 2015 (PWC). p146-p160)
[28] Xu, W. Chinese domestic IPO over-issuance.
‧
Historical flow of Financial deregulation in China Hujita, (2014)
:Panel A
Reform of Chinese financial market 1980 Separation of finance and government
1984 Renmin bank introduce the deposit reserve ratio operation
1985 Credit Creation Plan formulated by National planning commission, Ministry of Finance and Renmin Bank
1995 Central Bank and Commercial Bank law was enacted
① Ensuring the dependence of monetary policy
② Eliminating influence of the local government on the blanch of Renmin Bank
※ Monetary policy was supervised by government officials
※ Big four Bank was separated from Renmin Bank
※ Mostly loan was available for SOE 1995~1997 30%-50% loan was regarded as bad debt
1998 270billion RMB was injected for Four big commercial bank
1999~2000 1.4trrilion RMB bad debt was separated into four asset management company (AMC)
2001 Entering in the WTO
① Chinese government’s agenda shows more liberalization of interest rates
② Less restriction on ownership takeovers and mergers and acquisitions
③ Greater freedom of operational and geographical scope in the banking industry
※①➁③ explained in the paper of Berger (2009)
2002 Big four bank prompted demutualization and management structure, corporate governance and financial status was improved
2003 Accepting foreign financial institution for fund management and international competitiveness
2005~2007 Big three banks were listed on stock market
2010 Chinese agricultural bank was listed and world largest IPO was conducted in this year
~2013 More than 3 % of difference between lending ratio and deposit interest ratio
2013 Lending ratio deregulation
‧
Transportation, postal service F F01 F03 F05 F07 F09
F11 F19 F21 73 5 4 9
Information Technology G G81 G83 G85 G87 98
Wholesale and Retail trades H H01 H03 H11 H21 135 7 4 11
Real estate J J01 J05 J09 151 1 1 2
Public Facilities, public service K K01 K21 K30 K32
K34 56 4 0 4
Total Number 1663 213 209 422
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The industry classification and comparison between CSMAR and GICS Panel C
CSMAR Compustat GICS
Industry Group Name Industry Name
C1 3010,3020 Food, staples retailing, Beverage, Tobacco
C2 1510 Materials 151050 Paper and forest products
C4 2520 Consumer Durable & Apparel 252030 Textiles, Apparel and
Luxury goods
C5 1010 Energy 101010,101020,
151010 Oil, gas and chemical
C6 4510,4520,4
530
Software, Services, technology hardware, equipment, Semiconductors and equipment
C7 1510 Material 151020,151030 Construction Materials
and Packaging
C8 2510 Automobiles and components
C9 3510,3520 Health care equipment, pharmaceutical biotechnology and life sciences
B 1510 Material 151040 Metals and Mining
F 2030 Transportation
G 5010 Telecommunication Services
H 2550 Retailing
J 4040 Real Estate
K 2530 Consumers services 253010 Hotels, Restaurants and
Leisure