RESEARCH METHODOLOGY Sampling Procedures and Data Source

Taiwan’s accounting standard for income taxes was promulgated on June 30, 1994 and effective for financial statements issued after December 31, 1995. Therefore, our research period is ranged from 1995 to 2002. We delete firms owned by the government and firms in the banking, insurance, mutual fund, brokerage industries because of their special industrial characteristics and accounting⁹. Moreover, we also eliminate outlier data of 1% in each variable after screening out inappropriate firms by above rules.

Data sources of this study are from: (1) retrievals from the Taiwan Economic Journal (TEJ) for financial figures and financial ratios; (2) hand-collected based on SFI on Line of Anyan’s data bank for valuation allowance of DTA, loss carryforwards, significant customers and contingencies that are disclosed in annual and/or quarterly financial reports. Total firm-year data are 1,398 as shown in Table 1.

Table1: Sample Selection Process and Sample Data Statistics

Total firm-years on valuation allowance of DTA from 1995 to 2002. 3,037 Deduct：

Firm-years with fiscal year end other than December 31. (16) Firms-years without complete data related to deferred income taxes. (355) Firms-years without complete disclosure in financial statements. (458) Firms-years with missing variables and outlier data. (377)

Final test sample 1,831

Empirical Models

Based on the previous discussions on our research hypotheses, we employ a Tobit regression model to examine determinants associated with the amount of the valuation allowance of DTA for verifying the positive and negative evidence in paragraph 17 of Taiwan’s new accounting for income tax.

9 Generally speaking, utility enterprises regulated by the government have less earnings management incentives. Firms in the banking have no deferred tax account, and firms in the mutual fund and trust institution have no income tax expense. These firms are deleted from our sample.

ALLOWit＝

β

₀+

β

₁FUTURE_it+

β

₂MAJOR_it+

β

₃GROWTH_it+

β

₄AVEROA_it +

β

₅STRATEGY_it+

β

₆OPEB_it+

β

₇ORIGIN_it+

β

₈MARKET_it +

β

₉ABERR_it+

β

₁₀CUMLOSS_it+

β

₁₁DISTRESS_it+

β

₁₂CONTIN_it +

β

₁₃NOLCF_it+

ν

_it (1) In the above regression, the dependent variable is the ratio of valuation allowance of DTA to deferred tax assets. And its independent variables depend on individual positive and negative evidence. Definitions of the dependent variable and independent variables are shown as following.

Dependent variable

ALLOW_it = Valuation allowance of DTA / Deferred tax assets. (Note: “Deferred tax assets” here refers to the amount before deducting the valuation allowance of DTA. That is, the “gross amount” of deferred tax assets.)

Independent variable

FUTURE_it ＝future reversals of existing taxable temporary difference.

＝Deferred tax liability / Deferred tax assets

MAJOR_it ＝the major customers for the past three years, it is a dummy variable which is coded as 1 if a firm reports any significant customers for three consecutive years, and 0 otherwise.

GROWTH_it ＝a firm’s future earning growth, it is a dummy variable which is coded as 1 if EPS_it

≥

EPS_it−1

≥

EPS_it−2, and 0 otherwise.

AVEROA_it ＝average return of earnings before tax and interest on total assets for three fiscal years.

＝(ROA_it+ROA_it−1+ROA_it−2)/3, where, ROA_it=EBIT_it/ASSET_it, EBIT_itis earnings before tax and interest, and ASSET_itis total assets.

STRATEGY_it＝tax-planning strategies. Considering the data availability, we assume the proxy of tax-planning strategies is the effective tax rate (ETR).

When pre-tax income is positive, ETR is calculated as income tax expense divided by pre-tax net income. But when pre-tax income is negative and current tax payment is positive (zero), ETR is assigned to one (zero).

OPEB_it ＝DTA derived from the Other Post-Employment Benefits temporary difference / Deferred tax assets.

ORIGIN_it ＝DTA derived from other original temporary difference / Deferred

tax sassets. (Note: “ORIGIN” here also refers as DTA less OPEB and loss carryforwards).

MARKET_it ＝Equity market value / Equity book value, where deferred tax is not included in the book value of equity. It is the proxy for an excess of appreciated value over the tax basis of the entity’s net assets.

ABERR_it ＝Interaction which can be measured by CUMLOSS_it

∗

D_it, where, CUMLOSS _it is a dummy variable which is coded as 1 if cumulative net income is negative for the three-year period ending with the most recent reporting period, and 0 otherwise. D_it is also a dummy variable which is coded as 1 if a net loss incurred in only one of the preceding three years, and 0 otherwise.

All of the nine variables are positive evidence that a firm will create taxable income in the future. Their expected coefficients are negative.

CUMLOSS_it ＝Cumulative losses in recent three years, it is a dummy variable which is coded as 1 if cumulative net income is negative in recent three years, and 0 otherwise.

DISTRESS_it ＝A firm has negative cash flows (operating income or net income) in three successive years, it is a dummy variable which is coded as 1 if a firm has such situation, and 0 otherwise.

CONTIN_it ＝A firm’s material contingency, it is a dummy variable which is coded as 1 if a firm has such situation, and 0 otherwise.

NOLCF_it ＝DTA derived from loss carryforwards¹⁰ / Deferred tax assets.

All of the four variables are negative evidence that a firm will not create taxable income in the future. Their expected coefficients are positive.

In determining the factors of valuation allowance of DTA, different firms may have different propensities towards valuation allowance, even though they are not found to have the valuation allowance. Since many firms’ valuation allowances of DTA are zeros, we adopt a Tobit model to construct our regression. Equation (1) can be transformed into equation (2). The dependent variable yi we observe is:

⎪⎩

Dependent variable yi is the valuation allowance of DTA recorded in financial

10 Wu and Lin (2003) indicate that the association between investment tax credits and valuation allowance of DTA is positive but not significant. Such result is not consistent with their hypothesis. The possible reason is that tax incentive effect of investment tax credits was reduced after the integrated taxes system was implemented in Taiwan in 1998. Therefore, when we reexamine the determinants of valuation allowance of DTA, the component for investment tax credits is not included.

statement. y_i^∗ is a latent variable and represents a firm’s exact propensity to the valuation allowance. Assume:

),

observed. It is a latent variable of valuation allowance. For example, even though both firms A and B are observed to have no valuation allowance (y_A＝y_B＝0), firm A may still have a higher propensity toward valuation allowance (y^∗_A＞y^∗_B). Once business conditions are changed, y^∗_A may exceed zero at a higher speed than that of y^∗_B, so that

yA＞0 will be observed while y_B is still equal to zero.

Based on the explanations of the Tobit model, we suggest the following four equations derived from equation (2) and (3) used to examine the economic implications among the variables (Greene, 1995、2000；Lin, 2000；Chen, 2002；Chen et al., 2002).

E(yi｜xi, y_i>0)=

Where,

φ

(z) and

Φ

(z) are standard normal density function and cumulative standard normal distribution function, respectively; xij is the value of the j^th variable in the i^thobservation vector (the i^th firm) ； and z = β′ /x_i σ, the standard normal deviation.

Equation (4) only reflects situation where y_i> 0. Given x = xi, it is the expected value of valuation allowance for those firms having valuation allowance (y_i> 0), while firms without valuation allowance (y_i＝0) are excluded. In such data characteristic, the traditional ordinary least squares regression will lead to biased and inconsistent

β ⁾

(the

estimator of

β

). This is because the second term (

) (

) (

z z

Φ

σ φ

) in equation (4) is ignored.

Given x = xi, equation (5) represents the marginal effect of a unit increase in the j^th variable ofx_ijvector on the expected value of y_ifor those firms having valuation allowance. It only considers firms with y_i> 0, and ignores the fact that the change in x_ij may cause some firms’ propensity for valuation allowance to increase from y ≤^∗_i 0 to y_i^∗

＞0, i.e., to make a firm without valuation allowance becoming a firm having valuation allowance as situations change.

In addition, equation (6) represents, given x = xi, the overall expected value of valuation allowance not only for the y_i> 0 firms, but also for the y_i＝0 firms. Equation (7) is the partial derivative function of equation (6). We use equation (7) to examine the effect of a unit increase in the independent variable on the marginal increase in the valuation allowance for both firms already having valuation allowance, and those firms having no valuation allowance but would subsequently having it if situation changes.

Therefore, equation (7), instead of equation (5), plays the main role to examine the determinants of valuation allowance in this study since it carries more relevant economic implications. As for the effect of dummy variable on dependent variable, we use equation (8) to compute this marginal effect (Greene, 1995、2000；Lin, 2000):

) 0 ,

( ) 1 ,

(

y x dummy

=

E y x dummy

=

E _{i i}

－

_{i i} (8)

V. ANALYSES OF EMPIRICAL RESULTS

在文檔中 The Effect of Dismissal Threat on Auditor Independence (頁 40-44)