Parte II: Teoría de Sistemas
4. Críticas al enfoque luhmanniano
In the following, I extend the univariate insights above with multiple regressions. I estimate Eq. (2) using pooled OLS, cluster standard errors by firm and year (Gow et al. 2010), and present the results in Table C.8. In column 1 I show the estimation with only firm related variables. Accruals are insignificantly related to Log(TA) and negatively related to TLTA, consistent with single-step regressions in Chen et al. (2018). Both current and future employee growth is positively related to accruals. Lagged and leaded cash flows are negatively related to accruals, and current cash flows are positively related to accruals, consistent with Larson et al. (2018). Accruals are positively related to lagged ROA, consistent with Kothari et al. (2005).
In column 2 I show the effect of criminal executives, and find that firms run by criminal executives are more prone to use earnings management (captured by the coefficient of the interaction between %CrimEXEC and NEW_FIN). These results are consistent with Davidson et al. (2015). Then, in column 3 I include the effect of employees. In this regression, the effect of
168
executives is muted and becomes insignificant, potentially due to high correlation between
%CrimEXEC and %CrimEMPL. However, the effect of criminal employees is highly statistically and economically significant: one standard deviation of %CrimEMPL is associated with an increase in ROA of 0.4 percentage points (0.112*(0.0262+0.0099)).
Figure C.2: time-series properties of DACC per executive/workforce group
This figure shows the time-series properties of DACC by time preceding and following NEW_FIN=1. Ranges denote 95% confidence intervals. All continuous variables are winsorized at the 1 and 99 percent level. CrimEXEC indicates that the majority of executives have a criminal record. CrimEMPL indicates that the workforce is relatively criminal, and takes the value one when the percentage of employees with a criminal record is above the within-year median. 1/1 is an indicator taking the value one if CrimEXEC=1 and CrimEMPL=1. 1/0 is an indicator taking the value one if CrimEXEC=1 and CrimEMPL=0. 0/1 is an indicator taking the value one if CrimEXEC=0 and CrimEMPL=1. 0/0 is an indicator taking the value one if CrimEXEC=0 and CrimEMPL=0.
In column 4 through 6 I show the results of the four groups based on CrimEXEC and
CrimEMPL, respectively. Relative to each of the other three groups, I consistently find that firms with both criminal executives (CrimEXEC=1) and criminal employees (CrimEMPL=1) are associated with income-increasing DACC when the firm issues new finance (captured by the slope on 1/1*NEW_FIN). The effects are statistically significant (two-tailed tests, p-values in the range <0.001;0.022) and are economically significant, as DACC is used to increase ROA by 0.82-1.25 percentage points, or about 11.4-17.4% of the sample mean ROA (0.0082/0.075 ; 0.0125/0.075).
169 The results provide empirical evidence for H3 and corroborate the results of the univariate analyses above: earnings management is positively associated with firms in which both executives and employees are relatively criminal. A notable finding is that within firms with criminal executives, the positive association with earnings management is driven by firms who also employ criminal employees.
Table C.8: Discretionary accruals, criminal executives and criminal employees
(1) (2) (3) (4) (5) (6) OPACCi,t OPACCi,t OPACCi,t OPACCi,t OPACCi,t OPACCi,t
Proportion of criminal executives and criminal employees
%CrimEXECi,t 0.0006 0.0004
(0.62) (0.40)
%CrimEXECi,t*NEW_FINi,t 0.0064** 0.0047
(1.98) (1.51)
%CrimEMPLi,t 0.0099*
(1.78)
%CrimEMPLi,t*NEW_FINi,t 0.0262***
(2.90)
CrimEXEC / CrimEMPL
1/1i,t 0.0022* 0.0002 0.0014
(1.90) (0.13) (0.56)
1/1i,t* NEW_FINi,t 0.0125*** 0.0082*** 0.0108**
(4.29) (2.64) (2.30)
1/0i,t 0.0008 -0.0012 Base level
(0.45) (-0.58)
1/0i,t* NEW_FINi,t 0.0016 -0.0026 Base level
(0.31) (-0.57) 0/1,t 0.0020 ** Base level 0.0012 (2.26) (0.58) 0/1i,t*NEW_FINi,t 0.0042* Base level
0.0026 (1.79) (0.57) 0/0i,t Base level -0.0020
**
-0.0008 (-2.24) (-0.46) 0/0i,t*NEW_FINi,t Base level
-0.0042* -0.0016 (-1.79) (-0.31) NEW_FINi,t 0.0063** 0.0052** 0.0010 0.0031 0.0073** 0.0047 (2.47) (2.12) (0.46) (1.49) (2.30) (0.81) Firm controls log(TA)i,t 0.0002 0.0002 0.0003 0.0003 0.0003 0.0003 (0.47) (0.52) (0.73) (0.62) (0.62) (0.61) TLTAi,t -0.0682*** -0.0683*** -0.0686*** -0.0686*** -0.0686*** -0.0686*** (-10.45) (-10.46) (-10.50) (-10.54) (-10.54) (-10.54) STD_ROAi,t -0.0175 -0.0175 -0.0180 * -0.0175 -0.0175 -0.0175 (-1.59) (-1.60) (-1.66) (-1.59) (-1.59) (-1.59) PPEi,t -0.0088*** -0.0090*** -0.0099*** -0.0095*** -0.0095*** -0.0095*** (-2.87) (-2.96) (-3.28) (-3.20) (-3.20) (-3.20)
Discretionary accruals controls
EMPLGRi,t 0.0612*** 0.0613*** 0.0611*** 0.0612*** 0.0612*** 0.0612***
170
EMPLGRi,t+1 0.0862*** 0.0862*** 0.0862*** 0.0861*** 0.0861*** 0.0861***
(25.39) (25.24) (25.80) (25.70) (25.69) (25.69) EMPLGRi,t*NOAi,t-1 0.0084 0.0082 0.0077 0.0078 0.0078 0.0078
(0.65) (0.64) (0.60) (0.61) (0.61) (0.61) OPCFi,t-2 0.0353*** 0.0353*** 0.0354*** 0.0354*** 0.0354*** 0.0354*** (12.38) (12.39) (12.35) (12.38) (12.38) (12.38) OPCFi,t-1 0.0750*** 0.0750*** 0.0750*** 0.0750*** 0.0750*** 0.0750*** (15.71) (15.72) (15.79) (15.74) (15.74) (15.74) OPCFi,t -0.7088*** -0.7088*** -0.7089*** -0.7088*** -0.7088*** -0.7088*** (-96.77) (-96.48) (-96.22) (-96.36) (-96.35) (-96.33) DumOPCFi,t 0.0039*** 0.0039*** 0.0038** 0.0038*** 0.0038*** 0.0038*** (2.59) (2.61) (2.57) (2.61) (2.61) (2.61) DumOPCFi,t*OPCFi,t -0.1648*** -0.1648*** -0.1652*** -0.1650*** -0.1650*** -0.1650*** (-14.13) (-14.13) (-14.13) (-14.15) (-14.15) (-14.15) OPCFi,t+1 0.0915*** 0.0915*** 0.0915*** 0.0915*** 0.0915*** 0.0915*** (20.62) (20.65) (20.56) (20.60) (20.60) (20.60) OPCFi,t+2 0.0485*** 0.0485*** 0.0486*** 0.0486*** 0.0486*** 0.0486*** (16.64) (16.61) (16.58) (16.59) (16.60) (16.52) ROAi,t-1 0.2777*** 0.2774*** 0.2769*** 0.2768*** 0.2768*** 0.2768*** (18.04) (18.01) (18.11) (18.18) (18.18) (18.18) Intercept 0.0743*** 0.0741*** 0.0732*** 0.0737*** 0.0757*** 0.0745*** (12.15) (12.20) (11.70) (12.04) (12.56) (11.35) Industry FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES
N 50,396 50,396 50,396 50,396 50,396 50,396 Adjust R. sq. 0.8185 0.8186 0.8187 0.8187 0.8187 0.8187 This table shows the results of estimating Eq. (2), i.e. how criminal executives and criminal employees are related to accruals. %CrimEXEC is the proportion of the workforce with a criminal record. %CrimEMPL is the proportion of the executives with a criminal record. CrimEXEC indicates that the majority of executives have a criminal record. CrimEMPL indicates that the workforce is relatively criminal, and takes the value one when the percentage of employees with a criminal record is above the within-year median. 1/1 is an indicator taking the value one if CrimEXEC=1 and CrimEMPL=1. 1/0 is an indicator taking the value one if CrimEXEC=1 and CrimEMPL=0. 0/1 is an indicator taking the value one if CrimEXEC=0 and CrimEMPL=1. 0/0 is an indicator taking the value one if CrimEXEC=0 and CrimEMPL=0. Firm controls and discretionary accruals controls are defined in appendix. Standard errors are clustered by firm and year (Gow et al. 2010). t statistics in parentheses. ***, **, * Represent significance levels at 0.01, 0.05, and 0.10, respectively (two-tailed test). All continuous variables are winsorized at the 1 and 99 percent level.