Research Model t+y t t t t t t

576 of future operating performance. So that, the lower BTM ratio, the higher the future operating performance.  OIROA variable is considered as a control variable because it is convinced that rate of operating income return on firm‘s total asset in year t affects future operating performance. The better operating performance in this period, the better operating performance in next period.  D03 and D04 variables are dummy variables of observation period to account for the adjusted mean difference of dependent variables between years in research period. 3.2 Variable Definition 3.2.1 Dependent Variable The future operating performance dependent variable is measured by Operating Income Return on Assets OIROA ratio and Cash Flows Return on Assets CFROA ratio. OIROA

i,t

= Operating Income

i,t

TA

i,t

Where: Operating Income

i,t

: operating income before interest and yearly tax on i firms t year CFROA

i,t

= Operating Cash Flows

i,t

TA

i,t

Where: Operating Cash Flows

i,t

: cash flows from operating activities on i firms t year

3.2.2 Independent Variables

a. Real activities manipulation through selling, general, and administrative expense This manipulation is measured by log abnormal value of selling, general, and administrative expense ABNSGA which known by regressing the Gunny‘s β005 estimation model which referring to Anderson et al.‘s 577 2004 model. Consistent to Anderson et al. 2004, the model uses variable log spesifications and deflating all variables with SGA in a year before to control heteroskedasticity and scaling effect. The estimation model is as mention below: log SGA

i,t

= α + α 1 log S

i,t

+ α 2 log S

i,t

x Sdown SGA

i,t-1

S

i,t-1

S

i,t-1

Where : SGA

i,t

= selling, general, and administrative expense plus advertising or promotion expense on i firms t year S

i,t

= net sales Sdown = dummy variable will be equal to 1 if net sales in t period is decreasing compared to net sales in t-1 period and 0 if otherwise b. Real activities manipulation through production cost This manipulation is measured by abnormal value of production cost ABNProdCost Dechow et al., 1998; Roychowdhury, 2006. ProdCost

i,t

= α + α 1 1 + α 2 S

i,t

+ α 3 ∆S

i,t

TA

i,t-1

TA

i,t-1

TA

i,t-1

TA

i,t-1

Where : ProdCost

i,t

= the total of production cost COGS +∆INV on i firm t year ∆S

i,t

= the difference in net sales c. Real activities manipulation through gain on asset sales This manipulation is measured by abnormal value of gain on asset sales ABNGAIN which known by regressing the Gunny‘s β005 estimation model as below. + α 3 log S

i,t-1

+ α 4 log S

i,t-1

x Sdown + S i,t-2 S i,t-2 + α 4 ∆S

i,t-1

+ TA

i,t-1