Research Methodology

4.2.1. Regression analysis of equity liquidity

In this section, I examine the effect of earnings management on equity liquidity by using regression models. I use cross-sectional regression method to see whether the earnings management reduces equity liquidity. In order to provide a baseline regression against which subsequent test can be compared, I estimate equation (5) by simple ordinary least squares estimation (OLS) regression.

1 ,

i i i

PSP = +α β ADCA +ε (5) From previous discussion, I hypothesize more earnings management affects the percentage spread larger. When the insider can easily manipulate reported earnings, the information asymmetric between insider and outsider is large. So I expect the coefficient on β is positive. ₁

Previous cross-sectional studies of spreads suggest that a number of spread determinants should be controlled for in the empirical analysis; the strongest and most consistent findings are that relative spreads are negatively related to share price, trading volume, and competition in the market to provide liquidity, and positively related to return volatility and the adverse selection problem confronting specialists (Welker 1995). Following Welker (1995), the return volatility (SDRET), dollar value of volume (DOLVOL) and stock’s closing price (CLP) are added to the cross-sectional regression model as control variables.

The further regression model is as follows:

1 2 3 4

lnPSP_i = +β β ADCA_i +β lnDOLVOL_i+β lnCLP_i+β lnSDRET_i+ε_i, (6)

According to Welker (1995), I take the natural log of percentage bid-ask spread to control for heterosecdasticity. At the same time, the dollar value of volume (DOLVOL), the return volatility (SDRET) and stock’s closing price (CLP) are added to the cross-sectional regression model as control variables. In order to smooth the data, we take log value on dollar value of volume (DOLVOL), the return volatility (SDRET) and stock’s closing price (CLP). Finally, according to previous studies, we predict that the stocks with large return volatility will have wider spreads, and the stocks having larger price and daily dollar volume will have smaller spreads. So we expect the coefficient onβ₂ and β₃ are negative; the others coefficient on β₁ and

β4 are positive.

4.2.2. Two-Stage Least Squares (2SLS) Regression

However, the multiple regression models assume that absolute value of discretionary current accruals (ADCA) is exogenous variables. On the other hand, if earnings management is endogenous variable; our result will suffer from an endogeneity problem (omitted variables bias).Omitted variables appear when we would like to control for one or more additional variables but usually because of data unavailability, we cannot include them in regression model. At the same the time, it means ADCA is correlated with the disturbance. Hence we cannot expect ordinary least squares estimation (OLS) to consistently estimate any coefficient on independent variables. In this section, we use two-stage least squares (2SLS) regression to solve this problem. We address this concern by using market to book ratio (MBRT) as instruments for the earnings management variables (ADCA). Park and Shin (2003) point that the market to book ratio (MBRT) has been significant influences the level of earnings management in empirical result. The higher growth opportunities (MBRT) will result in the higher earnings management.

4.2.3. Simultaneous equation model

In this section, we consider the percentage bid-ask spread (PSP) and earnings management (ADCA) are simultaneously determined; our results may suffer from simultaneous equation bias. Accordingly, a system of equations in which both percentage bid-ask spread (equation 7) and earnings management (equation 8) are treated as an endogenous variable is estimated. In this viewpoint, I apply the Three Stage Least Squares (3SLS) to estimate parameters in a set of simultaneous equations.

The simultaneous equation set shows below：

0 1 2 3 4

lnPSP_i =β +β lnADCA_i+β lnSDRET_i+β lnDOLVOL_i+β lnCLP_i+ε_i, (7)

0 1 2 3 4

lnADCA_i =λ λ+ lnPSP_i+λ lnLEV_i+λ lnSIZE_i+λ lnMBRT_i+ω_i, (8)

In the simultaneous equation set, the endogenous variables are bid-ask spread (PSP) and earnings management (ADCA) and the instrument variables are SDRET,DOLVOL,CLP,LEV, SIZE and MBRT. In order to smooth the data and to get elasticity coefficients from the estimated equation, we take log value on all variables.

According to Park and Shin (2003), the financial leverage (LEV), firm’s size (SIZE) and growth opportunities (MBRT) are added to another regression model (Equation B4) as control variables. Highly financial leverage (LEV) firms may be less able to manipulate reported earnings because they have been monitored closely by lenders. Therefore, if the lender monitoring effect prevails, then earnings management will decrease with financial leverage (LEV). Firm size (SIZE) may be capturing the firm’s information environment. The level of disclosure information will more transparent with firm size. Big firms are less likely to hide the behavior of manipulate earnings than small firms. However, Zmijewski and Hagerman (1981) suggest that political costs increase with firm size and with firm risk. Managers of large and/or

high-risk firms, therefore, have greater incentives to exploit the latitude in accounting to reduce these political costs. Besides the positive theory implications of firm size, but the size of the firm may also be capturing the firm's information environment and could thus have a negative relationship with the level of managed accruals. Therefore, I do not predict the sign of the relationship between firm size (SIZE) and the level of earnings management. I consider that a firm’s growth opportunities as a potential determinant of earnings management. Park and Shin (2003) reported that firms with high growth opportunities may need to overinvest intentionally in current assets in anticipation of future sales growth. The temporary overinvestment in current assets can lead to a positive relationship between absolute of discretionary current accruals (ADCA) and growth opportunities (MBRT). So, we expect the growth opportunities (MBRT) have a positive relation with earnings management.