Descriptive statistics and Correlation

920 Specifically, Median for DM is bigger than median for RM. These statistics is to provide that managers are more likely to manage DM comparing to RM in Korea. Panel 2 of Table 4-1 provide that coefficient of correlation between RM and DM is significant negatively at 1 level. Specifically, RM is positively correlation with the factors of DMBig4 and AbnDA t-1 . Also, DM is positively correlation with the DK_SCORE which is factor for RM. Incentive variables, NP, is positively correlation with both RM and DM. Table4-1 Descriptive Statistics and Correlation on the Main Variables Panel 1: Descriptive Statistics 기술통계량 variable n mean std min median max RM 1,842 0.0005 0.083 -0.2991 0.0008 0.304 DM -0.001 0.093 -0.344 0.0030 0.328 Big4 0.437 0.496 0.000 1 Abnda t-1 -0.0007 0.203 -1.030 0.0003 0.959 DK_SCORE 0.284 0.497 0.000 1 ISLE_S 1.058 1.758 0.001 0.589 22.865 NP 0.145 0.353 0.000 1 LEV 1.051 9.228 0.034 0.7277 25.232 SIZE 19.271 1.441 16.161 19.021 24.874 Panel 2: Correlation on the Main Variables Correlation variable DM Big4 AbnDA t-1 DK_SCORE ISLE_S NP1 LEV SIZE RM -0.340 0.097 0.198 -0.0004 0.164 0.241 -0.022 0.241 DM 0.021 -0.361 0.053 0.025 0.227 -0.074 0.040 Big4 0.0265 0.009 0.126 0.049 -0.010 0.216 Abnda t-1 -0.006 0.002 -0.073 0.007 0.010 DK_SCORE -0.029 -0.017 -0.337 -0.057 ISLE_S 0.098 -0.005 0.591 NP -0.006 0.139 LEV 0.015 Footnote 1 , , are significant at the 10, 5, 1 level , respectively Definition of variables RM: Real activity management, that is abnormal CFO estimated using Roychowdhury2006 DM: discretionary accruals estimated using adjusted Jones model1991, 1995 BIG4: Dummy variable if firm which is audited BIG4 audit company is 1, otherwise 0. Abnda t-1 : Abnormal accruals, 921 DK_SCORE: Dummy variable that equal 1 if the firms K_SCORE is less than median of full samples K_SCORE, and otherwise 0 ISLE_S: The percentage of the companys sales to the total sales of its industry. NP1: Dummy variable, that is, it is 1 if reported earnings before managing earnings is positive, otherwise 0 . LEV: Current debtcurrent asset Panel 2 of Table 4-1 provide that coefficient of correlation between RM and DM is significant negatively at 1 level. Specifically, RM is positively correlation with the factors of DMBig4 and AbnDA t-1 . Also, DM is positively correlation with the DK_SCORE which is factor for RM. Incentive variables, NP, is positively correlation with both RM and DM. Table 4-2 provides the means for main variables to compare to whether earnings management is different RM from DM base on firms characteristics. In the case for firms which are possibility to avoid reported losses, mean of RM is 0.008 and DM is 0.009. However, firms which report losses are -0.028mean of RM and -0.043mean of DM. Also, means of RM and DM for firms which have been audited by Big4 audit company are 0.008 and -0.001. Means of RM and DM for firms which have been audited by Non-Big4 audit company are 0.008 and -0.001. Table 4-2 Comparison of RM and DM based on Firms Characteristics DM RM Avoidable firms reported losses 0.009 0.008 Reported losses -0.043 -0.028 Audited by Big4 audit firm -0.001 0.008 Audited by Non-Big4 audit firm 0.001 0.002 Table 4-3 provide comparing to earnings management between RM and DM base on industries. Table 4-3 Comparison of RM and DM base on classified industries 922 Classified industries DM RM Food and Kindred Products 0.0022 -0.0018 Textiles, Except Sewn Wearing 0.0016 0.0073 Wood and Paper Except Furniture 0.0002 0.0081 Chemical and Chemical Production 0.0051 0.0061 Rubber and Plastic Production -0.0059 -0.0104 Other Non-metallic Mineral -0.0047 0.0035 Basic Metals -0.0048 -0.0032 Fabricated Metal Production 0.0095 -0.0009 Other Machinery and Equipment -0.0084 0.0194 Other Transport Equipment 0.0021 0.0083 Electronic Components 0.0059 0.0078 Medical, Precision and 광학 -0.0079 -0.0000 Sale of Motor Vehicles and Motorcycles 0.0027 0.0128 Retail Trade, Except Motor Vehicles -0.0086 -0.0021 Others -0.0041 0.0093 4.3 Hypotheses Analysis 4.3.1 Analysis results for amounts of earnings management To test whether managers make RM and DM decisions simultaneously or sequentially, we conduct the Hausman test by regressing DM on the exogenous variablesthe factors of DM, incentives, and control variables, the instrument for RMthe predicted value from the first stage regression, and the actual RM. If RM is determined after DM, then the coefficient on the instrumental variable of RM should equal zero. Table 4-4 reports the results of the Hausman tests for Model 1. Hausman tests reject the exogeneity of RM in the DM regressionswith p-value ranging from 0.001. Also, Hausman test reject the exogeneity of DM in the RM equations. Which means DMor RM is correlated with RMor DMs error term. These results indicate that RM and DM are determined simultaneously. Table 4-4 Hausman Test for Simultaneity versus Sequentially of RM andor DM Modeln=1,842 RM DM Endogenous Coefficientp-value Coefficientp-value variables: DM -0.0750.001 RM -0.0780.0002 923 P_DM -0.8870.001 P_RM -0.9620.0001 Hausman test 1 st -stage adj. R 2 56.64 54.04 2 st -stage adj. R 2 70.46 59.23 p-value for Hausman stat. 0.0001 0.0001 Definition of Variables: DM: discretionary accruals RM: abnormal CFO P_DM or P_RM: the predicted value form the first-stage regression Given the finding of the simultaneously of RM and DM, we use the recursive regressive Model 2 to test H2-H4. The results are reported in Table 4-5. In the Table 4-5, Big4 and Abnda t-1 are factors for DM management activity, DK_SCORE and ISLE_S are factors for RM management activity. H2 predicts that in both models, while Big4 is negatively related with DM, positively related with RM. Consistent with H2, Table 4-5 shows that RM is significant positively related with Big4 at 10 level. However, DM is insignificant related with Big4, suggesting that firms are reluctant to manager earnings using DM when firms have audited by Big4 audit firm. Also, tests of H3 are indicated by the coefficient estimates for Abnda t-1 in both models. If firms used DM activity much more in previous year, they may be constrained using DM in next year. Thus, we expected that managers are more likely to manage earnings using RM activity. Abnda t-1 in RM model is significant positively at 1 level as coefficient t-value is 0.09212.81. While in DM model is significant negatively at 1 as coefficientt-value is -0.166- 21.63. The results are to support H3. One of the factors for RM, DK_SCORE, is dummy variable that equals one if the firms K_ SCORE is smaller than median of full samples K_ SCORE and zero otherwise. DK_ SCORE in RM model is significant negatively at 1 level as coefficientt-value is -0.00311.69 . While in DM model is significant negatively at 1 as coefficientt-value is 0.00310.26 . The results are to support H4. Also, NP and LEV as incentive variables are expected positive coefficient. In the