EMPIRICAL RESULTS Sample Selection and Data Sources

On July 1, 2004, the Financial Accounting Standards Committee of the Accounting Research and Development Foundation of the Republic of China issued SFAS No. 35, Accounting for the Impairment of Assets. This accounting standard came effective for financial statements ending on and after December 31, 2005, with early adoption permitted. Therefore, Taiwan’s listed firms are required to adopt SFAS No. 35 in the first quarter of 2005. First of all, we examine the determinants of adoption timing of SFAS No.

35 and the amount of assets impairment losses, respectively, for listed firms in Taiwan.

Panel A of Table 1 illustrates our sample selection process. From the updated quarterly financial data of listed firms in the TEJ Finance database (May 2005), we find 341 listed firms in Taiwan recognizing assets impairment losses in their 2004 annual reports or in the first quarter of 2005. However, the TEJ database doesn’t include firms

that made long-term equity investments accounted by the equity method and reduce their investment income due to assets impairment losses recognized by investee companies.

We believe that such long-term equity investor companies should be included in our sample since these investor companies have significant influence on major decisions of investee companies, such as the adoption timing of SFAS No. 35 and the amount of assets impairment losses recognized. After checking data disclosed on the Taiwan’s Market Observation Post System, we include additional 36 firms with long-term equity investment having the stated nature in our sample.

However, we exclude 3 firms (Walsin Lihwa, Zinwell and Universal Technology) that have assets impairment losses in both 2004 annual report and the 2005 first quarterly report. In addition, 10 firms in the banking industry are also excluded because the accounting of this regulated industry is substantially different from accounting for firms in other industries. Our final sample firms with asset impairment losses consists of 364 listed companies, including 105 firms recognizing assets impairment losses in their 2004 annual reports (the “early adopters” of SFAS No. 35) and 259 firms recognizing impairment loss in the first quarter of 2005 (the “non-early adopters”).

In testing Hypothesis 1 (Investigating the determinants of the adoption timing of SFAS No. 35), the experiment group is those 105 firms recognizing assets impairment losses in their 2004 annual reports, and the matched control sample consists 105 firms that adopted SFAS No. 35 in the first quarter of 2005 in the same industry and of similar market value (size). In total, we have 210 observations for testing the Hypothesis 1.

However, when testing Hypothesis 2 (Exploring the determinants of the amount of assets impairment losses), the experiment group is those 364 firms recognizing assets impairment losses in either their 2004 annual reports or 2005 first-quarter reports, and the matched control sample consists another 364 firms that did not recognize assets impairment losses in the same period, in the same industry and of similar market value (size). The final sample size for Hypothesis 2 is 728 firms.

In order to further analyze the determinants of the amount of assets impairment losses for each of the five types of long-lived assets (i. e., long-term equity investment, fixed assets, goodwill, identifiable intangible assets and other assets), we collect respective impairment data from footnotes of 2004 annual reports and 2005 first-quarter reports in the financial report database of the Market Observation Post System. The event dates of assets impairment losses disclosure and earnings announcement are retrieved from financial announcement dates in the financial market events data of the TEJ Firm Database.

Panel A of Table 1 illustrates the industry distribution of firms with assets impairments. We find sample firms are concentrated in the electronics industry (49.73% = 181/364), which is consistent with our hypothesis proposing that SFAS No. 35 will have more effect on the electronics industry. In addition, the proportion of firms with assets impairments in the Taiwan Stock Exchange (= 20.45% = 245/1198) is close to that in the GreTai Securities Exchange (OTC) (19.50% = 119/610).

Table 1: The Sample Selection and Industries Compositions Panel A：Sample Selection Process

Firms with impairments in the first six months of 2005 in TEJ database 341 add: Firms reducing their investment income resulting from impairment loss on long-term

equity investments accounted for in the equity method 36 less: Firms with impairment disclosures in both their 2004 annual reports and 2005

first-quarter reports (3)*

Firms in the Banking industry ( 10)

Final sample: Firms with assets impairments 364

Early adopters (impairments disclosed in the 2004 annual reports) 105**

Non-early adopters (impairments disclosed in the 2005 first-quarter reports) 259 Matched sample: Firms without assets impairments (for Hypothesis 2) 364

* Three firms (Walsin Lihwa, Zinwell and Universal Technology) disclosed assets impairment losses in both their 2004 annual reports and 2005 first-quarter reports, as shown in the TEJ Database.

** The control group for Hypothesis 1 consists of firms that adopted SFAS No. 35 in the first quarter of 2005.

Panel B：Sample Compositions by Industries, Exchanges, and Timing of Adopting SFAS No. 35 Exchange Adopting SFAS No. 35

Industries Number Percentage

TSE-Listed firms(n=1198)

OTC-Listed firms(n=610)

Early Adopters

Non-Early Adopters

Cement 7 1.92% 7 0 1 6

Foods 12 3.30% 11 1 3 9

Plastics 9 2.47% 8 1 0 9

Textiles 30 8.24% 25 5 9 21

Machinery 11 3.02% 8 3 2 9

Wire & Cable 10 2.75% 10 0 3 7

Chemicals 15 4.12% 8 7 2 13

Glass 4 1.10% 4 0 3 1

Paper & Pulp 3 0.82% 3 0 0 3

Steel & Iron 14 3.85% 12 2 8 6

Rubber 2 0.55% 2 0 0 2

Automobiles 1 0.28% 1 0 0 1

Electronics 181 49.73% 99 82 53 128

Construction 27 7.42% 19 8 8 19

Transportation 4 1.10% 3 1 2 2

Tourism 2 0.55% 1 1 0 2

Wholesale 8 2.20% 6 2 5 3

others 24 6.59% 18 6 6 18

TOTAL 364 100.00% 245 119 105 259

Descriptive Statistics of the Sample

Table 2 provides descriptive statistics for firms with assets impairment disclosures.

In panel A of Table 2, we show the extent of impairment losses for early adopters and non-early adopters of SFAS No. 35. Among 364 sample firms, 105 (28.85%) are early adopters and 259 (71.15%) are non-early adopters. The mean of impairment losses per share (0.7427) and the mean of assets-deflated impairment loss (0.0318) for early adopters are significant higher than that of non-early adopters (0.3536 and 0.0168 respectively). It seems that an early adopter has incentives to “clean the deck” through recognizing a large amount of impairment losses.

Types of long-lived assets subject to revaluation under SFAS No. 35 include long-term equity investments accounted for by the equity method, fixed assets, goodwill, identifiable intangible assets, and other assets. Panel B of Table 2 shows descriptive statistics of these five types of impairment losses. The total percentage of firms recognizing these five types of impairment losses is over 100% (=541/364), which implies some firms recognize more than two types of impairment losses. Furthermore, we also find the rankings in percentage and extent of the types of impairment losses are the same. The top three types of assets suffering impairment losses are: fixed assets (31.61%), long-term investments (27.17%) and other assets (26.43%). In addition, the impairment losses per share and assets-deflated impairment losses for these three assets are also substantially higher than other types of long-lived assets, such as goodwill and identifiable intangible assets.

Table 3 shows the relationship between the timing of adopting SFAS No. 35 and earnings performance. For 105 early adopters, 65 (61.9%) have pre-impairment net losses and 40 (38.1%) have pre-impairment net income. For 259 non-early adopters, on the contrary, 121 (46.7%) have pre-impairment net losses and 138 (53.3%) have pre-impairment net income. The χ² value of the contingency table is 6.8957 (p < 0.01) and categorical φ coefficient is 0.1376. These statistics show that a firm’s earnings performance significantly affects its adoption time of SFAS No. 35. Firms with poor earnings performance would take “big baths” by recognizing impairment losses in hope that their future earnings will improve. Therefore, their managers early adopt SFAS No.

35 in 2004 annual report to clean up assets impairment losses.

Panel A of Table 4 presents descriptive statistics for sample firms with and without assets impairment losses, respectively. Univariate analyses show that, irrespective of other factors, firms with assets impairment losses have significantly higher ∆MGT but significantly lower BATH and MTB. It seems that management’s reporting motivations (changes in top management, taking big baths) and a firm’s operational factor (the market-to-book ratio) are significantly different between impairment observations and non-impairment observations.

Table 2: Descriptive Statistics for Sample Firms with Impairment Disclosure Panel A：The Frequency and Amounts of Impairment Losses

Adoption of Impairment loss per share¹ Deflated impairment loss² SFAS No. 35 Number Percentage Max. Min. Mean Median Max. Min. Mean Median Early adopters 105 ³ 28.85% 7.9868 0.0197 0.7427 0.4087 0.3120 0.0008 0.0318 0.0167 Non-early adopters 259 ⁴ 71.15% 4.0574 0.0000 0.3536 0.1521 0.1944 0.0000 0.0168 0.0065

Total 364 100.00%

Panel B: The Frequency and Amounts of Different Types of Impairment Losses

Types of Early adopters

Non-early

adopters Total Impairment loss per share¹ Deflated impairment loss² Impairment

loss n % n % n % Max. Min. Mean Median Max. Min. Mean Median

Long-term

investment 44 41.90% 103 39.77% 147 27.17% 6.0800 0.0000 0.1225 0.0000 0.2121 0.0000 0.0049 0.0000 Fixed assets 55 52.38% 116 44.79% 171 31.61% 4.0574 0.0000 0.1840 0.0000 0.1944 0.0000 0.0086 0.0000 Goodwill 14 13.33% 21 8.11% 35 6.47% 2.1748 0.0000 0.0286 0.0000 0.0849 0.0000 0.0014 0.0000 Identifiable

intangible

assets 13 12.38% 32 12.36% 45 8.32% 2.1129 0.0000 0.0287 0.0000 0.0998 0.0000 0.0014 0.0000 Other assets 43 40.95% 100 38.61% 143 26.43% 3.4240 0.0000 0.1019 0.0000 0.2957 0.0000 0.0048 0.0000

169 ³ 372 ⁴ 541

1 Impairment losses per share = impairment losses / outstanding shares

2 Deflated impairment losses = impairment losses / total assets at the beginning of the quarter

3 Early adopters consist of 105 firms. Because some firms recognize multiple types of impairment losses, the total observations for early adopters are 169.

4 Non-early adopters consist of 259 firms. Because some firms recognize multiple types of impairment losses, total observations for non-early adopters are 372.

5 Majority of firms recognize only some types of impairment losses. Therefore, the minima and medians of impairment loss per share and assets-deflated impairment losses are 0’s.

Table 3: Contingency Table for Timing of Adopting SFAS No. 35 and Earnings Adoption Timing

of SFAS No. 35

Pre-impairment Net Loss

Pre-impairment Net Income

Total

Early adopters 65 (61.9%) 40 (38.1%) 105 (100%) Non-early adopters 121 (46.7%) 138 (53.3%) 259 (100%)

Total 186 178 364

1 Pre-impairment net loss and pre-impairment net income refer to pretax net income and net loss before assets impairment losses.

Table 4: Sample Statistics and Variable Definitions Panel A：Summary Statistics for Whole Sample (n = 728)

Impairment observations (n = 364)

Non-impairment

observations (n = 364) Test statistics Variable Mean Median Mean Median t-value Z-value Reporting Motivations:

BATH -0.0103 0.0000 -0.0061 0.0000 1.87* 1.69*

SMOOTH 0.0165 0.0000 0.0128 0.0000 1.04 0.21

∆ MGT 0.3542 0.0000 0.1853 0.0000 5.24*** 4.88***

FIN 0.0026 0.0000 0.0028 0.0000 -0.11 0.59

Operational Factors:

∆ INDROA -0.0062 -0.0078 -0.0061 -0.0078 -0.37 -0.32

ELEC 0.5000 1.0000 0.5000 1.0000 0.00 0.00

RET -0.0212 -0.0475 -0.0012 -0.0114 -1.32 -1.60

∆ SALE -0.0166 -0.0112 -0.0178 -0.0125 0.18 0.00

∆ OCF -0.0019 -0.0040 -0.0080 -0.0088 1.00 1.22

MTB 1.0951 0.9009 1.2886 1.0912 -3.44*** -3.70***

Panel B：Summary Statistics of Impairment Observations (n=364) Early-adopters

(n = 105)

Non-early adopters

(n = 209) Test statistics Variable Mean Median Mean Median t- value Z-value Reporting Motivations:

BATH -0.0283 0.0000 -0.0031 0.0000 -4.54*** 7.08***

SMOOTH 0.0034 0.0000 0.0209 0.0000 4.01*** -3.99***

∆ MGT 0.3552 0.0000 0.3551 0.0000 0.24 0.24

FIN 0.0057 0.0000 0.0014 0.0000 1.26 0.15

Operational Factors:

∆ INDROA -0.0061 -0.0105 -0.0062 -0.0078 0.18 -3.47***

ELEC 0.5048 1.0000 0.4942 0.0000 0.26 0.26

RET -0.0058 -0.0065 -0.0322 -0.0583 1.24 1.64

∆ SALE 0.0157 -0.0014 -0.0288 -0.0144 4.00*** 4.68***

∆ OCF 0.0279 0.0079 -0.0141 -0.0107 4.09*** 4.64***

MTB 1.0922 0.8667 1.0972 0.9325 -0.06 -0.76

*, **, and *** indicate statistical significance levels of 10%, 5% and 1%, respectively.

Variable definitions:

BATHit : the proxy for taking big baths = (ΔE^it < median of the “unexpected negative earnings”, then BATHit =ΔE^it ; otherwise, BATHit = 0.) SMOOTHit : the proxy for income-smoothing = (ΔE^it > median of the “unexpected negative earnings”, then SMOOTHit =ΔE^it; otherwise, SMOOTHit = 0.) ΔMGTit : equals 1 if firm i changes its top management (defined as CEO, chairman of the board or CFO) from year t-1 to t; and 0 otherwise. FINit : total amounts of firm i’s issuance of equity capital and corporate bonds in quarter t, divided by firm i’s total assets at the end of quarter t-1. Δ INDROAit : the median of growth rate of return on assets from quarter t-1 to t in firm i’s industry. ELECit: equal 1 if firm i at quarter t belongs to the electronics industry, and 0 otherwise. RETit : firm i’s quarterly stock returns from quarter t-1 to t. ΔSALEit : firm i’s change in sales from quarter t-1 to t, divided by total assets of firm i at quarter t. ΔOCFit : firm i’s change in cash flows from operations from quarter t-1 to t, divided by total assets of firm i at quarter t. MTBit : equal 1 if firm i's market-to-book ratios at quarter t is below 1, and 0 otherwise.

Panel B of Table 4 partitions the impairment observations into early-adopters and non-early adopters. Univariate analyses show that, irrespective of other factors, early-adopters of SFAS No. 35 have significantly lower reporting motivations in BATH and SMOOTH. They also have significantly higher operational factors in ∆SALE, ∆OCF, and significantly lower operational factor in ∆INDROA. Overall, the majority of management’s reporting motivations and a firm’s operational factors are significantly different between early-adopters and non-early adopters. Since univariate analyses of Table 4 do not consider the effects of other independent variables, we further employ the multivariate analyses in the following section.

Table 5: Logistics Regression: The Timing of Adopting SFAS No. 35 ( n=210)

i i

i i

i

i BATH SMOOTH MTG FIN INDROA

ADOPT

= α

₀

+ α

₁

+ α

₂

+ α

₃

∆ + α

₄

+ α

₅

∆

i i

i RET SALE

ELLC

+ + ∆

+ α

₆

α

₇

α

₈

+ α

₉

∆

OCF_i

+ α

₁₀MTB_i

+ ε

_i (1)

Variable Predicted Sign Coefficients (χ² value in parentheses)

Intercept 0.2630 (0.32)

Reporting Motivations:

BATH － -41.6901 (16.37) ***

SMOOTH ＋ -10.7116 (1.48)

∆ MGT ＋ 0.2134 (0.37)

FIN － 6.5466 (1.17)

Operational Factors:

∆ INDROA － -1.4359 (0.01)

ELEC ＋ -0.2897 (0.44)

RET － 1.0880 (1.55)

∆ SALE ＋ 13.2301 (15.48) ***

∆ OCF ＋ 11.6091 (12.60) ***

MTB － -0.3651 (1.68)

Likelihood ratio 83.1915 ***

Concordant 85.5%

a.*, **, and *** indicate statistical significance levels of 10%, 5% and 1%, respectively.

b.Variables definitions:

ADOPTi: equals 1 if firm i recognizes impairment losses in 2004 annual report, and 0 otherwise. BATHi : the proxy for taking big baths = (ΔEⁱ < median of the “unexpected negative earnings”, then BATHt =ΔEⁱ ; otherwise, BATHi = 0.) SMOOTHi : the proxy for income-smoothing = (ΔEⁱ > median of the “unexpected negative earnings”, then SMOOTHt

=ΔEⁱ; otherwise, SMOOTHi = 0.) ΔMGTi : equals 1 if firm i changes its top management (defined as CEO, chairman of the board or CFO) from year t-1 to t; and 0 otherwise. FINi : total amounts of firm i’s issuance of equity capital and corporate bonds in quarter t, divided by firm i’s total assets at the end of quarter t-1. ΔINDROAi : the median of growth rate of return on assets from quarter t-1 to t in firm i’s industry. ELECi : equal 1 if firm i at quarter t belongs to the electronics industry, and 0 otherwise. RETi : firm i’s quarterly stock returns from quarter t-1 to t. ΔSALEi: firm i’s change in sales from quarter t-1 to t, divided by total assets of firm i at quarter t. ΔOCFi : firm i’s change in cash flows from operations from quarter t-1 to t, divided by total assets of firm i at quarter t. MTBi : equal to 1 if firm i's market-to-book ratio at quarter t is below 1, and 0 otherwise.

Multivariate Analysis: Hypothesis One

Table 5 presents the results of logistics regression for exploring the determinants of the timing of adopting SFAS No. 35 for listed firms in Taiwan (Hypothesis 1). The experiment group consists of firms that early adopted SFAS No. 35 in their 2004 annual report (n = 105), and the matched control group consists of firms that adopted SFAS No.

35 in the first quarter of 2005 (n = 105). The concordant of the logistics regression is 85.5%.

We find that, among the reporting motivations, only taking big baths (BATH) (estimated coefficient = -41.61901, p < 0.01) is significantly negative, as predicted.

Income smoothing (SMOOTH) and other reporting motivations are insignificant. These results reflect that firms which experienced poor earnings performance are apt to early adopt SFAS No. 35. Remarkably, among operational factors, only Sales Growth (ΔSALE) (estimated coefficient = 13.2301, p < 0.01) and cash flow growth (ΔOCF) (estimated coefficient = 11.6091, p < 0.01) are significantly positive, as predicted. These results show that when determining the timing of adopting SFAS No. 35, management will consider whether a firm can endure the impact of impairment losses. Only if sales growth and operational cashflows growth are high enough will a firm early adopt SFAS No. 35.

Multivariate Analyses: Hypothesis Two

Table 6 presents the results of tobit regressions in examining the determinants for amounts of assets impairment losses for early adopters of SFAS No. 35 (Hypothesis 2).

The experiment group consists of 105 firms that early adopted SFAS No. 35 in their 2004 annual report. The matched control group consists of 105 firms without impairment losses in their 2004 annual report, which are matched with our experimental sample firms in terms of market value (size) and industry. The explanatory power of tobit regression is 17.42%. We find the amounts of assets impairment losses for early adopters are only significantly related with management’s reporting motivations, such as taking big baths (BATH) (estimated coefficient = -0.1724), income smoothing (SMOOTH) (estimated coefficient = 0.4836) and changes in top management (ΔMGT) (estimated coefficient

=0.0219), as predicted. Firms would early adopt SFAS No. 35 to recognize large impairment losses in the period of unexpectedly poor earnings performance to improve future earnings performance or to have the restoration flexibility of impairment losses in the future period (BATH). Firms also may recognize impairment losses in the period of unexpectedly good earnings performance to smooth income (SMOOTH). In addition, new managers would early adopt SFAS No. 35 to recognize a large impairment loss to increase firms’ earnings performance in the future period (ΔMGT). Our results are consistent with Zucca and Campbell (1992) who documented that managers recognizing impairment losses aim at taking big baths or smoothing income.

However, firms’ operational factors do not affect the amount of assets impairment decision for early adopters in Taiwan. The results in Table 6 indicate managers of early adopters have significant reporting motivations, but insignificant operational factors, for the amount of assets impairment decision. Managers have incentives to “clear the deck”

of impaired assets and signal “the worst period has already passed and future performance can be improved.” Managers also have incentive to smooth earnings in order to increase their bonuses for the next year. However, the amount of assets impairment decision for

earlier adopters is not significantly associated with firms’ operational factors, industrial technology, changes in environment, firms’ past operational condition or assets usage conditions.

To examine the above results in depth, we also run separate tobit regressions in Table 6 for each of the five asset types (i.e., long-term equity investment, fixed assets, goodwill, identifiable intangible assets and idle assets) for early adopters. Our empirical results show that the amounts of assets impairment decision for early adopters are driven primarily by management’s reporting motivations, especially when top management is changed (∆ MGT). However, firms’ operational factors do not play significant roles in the amount of impairment decisions for respective assets types. The explanatory power of regression for long-term equity investment model (20.40%) is the highest one among five assets, and that of goodwill model (4.01%) is the lowest one. In addition, impairment losses from long-term equity and fixed assets for early adopters of SFAS No. 35 are driven primarily by taking big baths (BATH) negatively and changes in top management (ΔMGT) positively. However, impairment losses from idle assets for early adopters are driven primarily by smoothing income (SMOOTH) positively and changes in top management (ΔMGT) positively. Nevertheless, early adopters recognizing impairment losses from goodwill and intangible assets are only driven by changes in top management ( Δ MGT) positively, but unaffected by other proxies for management’s reporting motivations (estimated coefficients of BATH and SMOOTH are insignificant). Hence, managers of early adopters would not manage earnings through impairment losses from goodwill or identifiable intangible assets. However, managers have motivations to “clear the deck” of impaired fixed assets and impaired long-term investment for improving future earnings. In addition, managers may recognize impairment losses for idle assets in order to smooth earnings and increase their bonuses in the following year.

Table 7 presents the results of tobit regressions in examining the determinants of the amount of assets impairment decision for non-early adopters (Hypothesis 2). The experiment group consists of 259 non-early adopters of SFAS No. 35 in the first quarter of 2005. And, the matched control sample consists of 259 firms without impairment losses in the first quarter of 2005, which are matched with our firms in the experiment group in terms of market value (size) and industry. The explanatory power of tobit regression is 11.04%. Different from early adopters of SFAS No. 35, non-early adopters consider both reporting motivations and operational factors in determining the amounts of assets impairment.

Among reporting motivations for non-early adopters, the income-smoothing proxy (coefficient of SMOOTH = 0.0296) and the top management changes (coefficient of Δ MGT = 0.0075) are significant and consistent with our prediction. The taking big baths variable (BATH) is no longer significant as with early adopters. We infer that firms with unexpectedly poor earnings performance would have adopted SFAS No. 35 early in the fourth quarter of 2004 (evidenced in Table 5). Among operational factors for non-early adopters, stock return performance RET is significant and consistent with prediction, which implies the worse a firm’s past stock price performance has been, the more likely its management will recognize the assets impairment losses. However, the sales growth variable, ΔSALE, is significant and contrary to our prediction. Firms with higher sales

growth may face more technological innovations and, therefore, may need to recognize more assets impairments.

Table 6: Tobit Regressions: Determinants for the Amounts of Assets impairment Losses for Early Adopters (n=210)

it it

it it

it

it BATH SMOOTH MGT FIN INDROA

WOTA

= α

₀

+ α

₁

+ α

₂

+ α

₃

∆ + α

₄

+ α

₅

∆

Generic Separate Tobit regressions

Variable

Predicted

Sign Tobit

Regression Long-term

investment Fixed assets Goodwill Intangible

assets Idle assets

∆INDROA － -0.0666 0.3915 -0.0021 0.0571 0.0931 -0.6063

(-0.12) (1.57) (-0.01) (0.25) (0.64) (-1.72)*

a.*, **, and *** indicate statistical significance levels of 10%, 5% and 1%, respectively.

b.t-value in parentheses c.Variable definitions:

WOTAit : firm i’s pre-impairment loss at quarter t, divided by firm i’s total assets at the end of quarter t-1. BATHit : the

WOTAit : firm i’s pre-impairment loss at quarter t, divided by firm i’s total assets at the end of quarter t-1. BATHit : the

在文檔中 The Effect of Dismissal Threat on Auditor Independence (頁 74-88)