4. ANÁLISIS Y RESULTADOS
4.1. FASE DEFINIR
4.1.2. FASE MEDIR
4.1.2.3. Recopilación de datos para establecer las líneas base
Finally, this study will identify earnings management phenomenon through related party transactions between controlling shareholders and IPO subsidiaries. Firstly this study will examine the relation between the size of related party transactions
and contemporaneous operating performance of IPO firms in each of the three investigation periods (the pre-IPO period, the IPO year and the post-IPO period). The reported operating performance is regressed on the two RPT variables: operating RPTs and net loans. In the post-IPO period, the average of operating performance of the four post-IPO years is regressed on the average of operating RPTs and the average of net loans of the four post-IPO years. The OLS cross-sectional models for testing the RPT-based earnings managemen! hypothesis are specified as follow:
ROA^ = -I- * Net __ Loari^4- * Operolingterns^ (7 9)
CFO^ = + /?, Net _ Loan^ + pj Operating _ items, 4-g, (7.10)
ROA, = Po -h (3, '^Net_loon. + P, * Operating_items, + p^ *Size, + p^ Age,
+ P^ * Capitol_Qxpendilii7'e, + p^, * Government_siibsidy, ^ ^ + P j Ownership_type. + p^ * Ownership_concentration,
+ * Board_C om position, + 8,
CFO, = Pu 4- Pi * Net _loan, + P, * O perating_ items, + P^ Size, +■ p , * Age, + P^ Capital_Q x^enditure, + P^, ^ Government_siibsid}
+ Pi Ownetsship_typ(i, + A Ownership_concentration
+ P<) * Board_Composition, + s.
where ROA, and CFO/ denote EBITDA and Cashflow from operations scaled by lagged (-1) total assets respectively in a given period / (/ = -1, 0, 1). Met /oan, represents the difference between loans by controlling shareholders to listed subsidiaries and loans by listed subsidiaries to controlling shareholders scaled by lagged (-1) total assets in a given period / (/ = -1, 0, 1). Operating items;,
assets in a given period i (/ = - 1 , 0 , 1).
In formula (7.11) and (7.12), a set of control variables have been introduced.
Firms' SizBj is measured as the beginning-year total assets, and Age, is measured as the difference between the establishment year and the IPO year.
Capital Expenditure! and Government Subsidy; denote asset-scaled capital investment (adjusting for depreciation charges) and asset-scaled government subsidy. Ownership Type; is the type of the controlling shareholder, which is a dummy variable taking the value of 1 if the controlling shareholder is ultimately owned by the state at the end of the year; 0 otherwise. Ownership Concentration,
and Board Compositiom represent the proportion of ownership held by the controlling shareholder and the proportion of directors representing the controlling shareholder in the board at the end of the IPO year respectively.
In Table 7-2, a Pearson inter-correlation test between each two of the independent variables is also conducted. The tables show that the correlation results between operating RPTs and ownership and governance variables are strongly significant. The correlation result implies that IPO firms with high concentrated ownership structure and board composition are likely to conduct more RPTs with their controlling shareholders. The high correlation of these variables gives rise to a concern of muiticoiiinearity between control variables with explanatory variable (operating RPTs). However, as discussed earlier, to simply remove the three ownership and governance variables is not a good decision, because omitting a variable in the regression may substantially bias the true value of estimated parameters, and the error term is inflated. Further, muiticoiiinearity between control variables with explanatory variable is not a research problem, even though the control variables will have a strong effect on the explanatory variables.
models (model 7.11 and model 7.12). Although such regressions ignore the | contemporaneous correlations among variables, and can lead to biased standard I
i errors (but not biased coefficient estimates), this specification provides regression | I coefficients that allow an easy interpretation of the economic significance. I Furthermore, a similar version of regression models has also been specified in ! ! prior literature (Ritter, 1991; Teoh et al., 1998). However, it is unlikely to assess the : relative importance of these variables in explaining the variance of the dependent
variable, and it should be beard in mind that the control variables would have an • effect on explanatory variables.
7.5 Summary
This chapter presents a brief introduction on the methodology of regression analysis. Then, it presents the research design to test each of the three hypotheses:
First of all, a regression analysis is introduced to test managerial ownership dispersion hypothesis. The change in operating performance from before to after the IPO is regressed on the contemporaneous change in managerial ownership and a set of control variables.
Secondly, in order to examine discretionary accruals around the IPO (accrual-based earnings management), modified Jones (1991) model is employed to separate discretionary accruals from non-discretionary accruals. So, the movement of discretionary accruals over time indicates that managers are manipulating accounting accruals for income-reporting objectives. It is not necessary to bring in a regression model to investigate the effect of discretionary accruals adjustments on reported earnings of IPO firms, because discretionary accruals is a component of reported earnings.
Finally, another regression analysis is introduced to test RPT-based earnings management hypothesis: the operating performance is regressed on the contemporaneous amounts of operating RPTs and net loans, and a set of control variables. Such regressions ignore the contemporaneous correlations among variables and can lead to biased standard errors (but not biased coefficient estimates), but this specification provides regression coefficients that allow an easy interpretation of the economic significance.