Financial Reporting Changes and Internal Information Environment: Evidence from SFAS 142

Transcription

1 Singapore Management University Institutional Knowledge at Singapore Management University Research Collection School Of Accountancy School of Accountancy Financial Reporting Changes and Internal Information Environment: Evidence from SFAS 142 Young Jun CHO Singapore Management University, Qiang CHENG Singapore Management University, I-Hwa YANG Singapore Management University, Follow this and additional works at: Part of the Accounting Commons Citation CHO, Young Jun; CHENG, Qiang; and YANG, I-Hwa. Financial Reporting Changes and Internal Information Environment: Evidence from SFAS 142. (2014). Research Collection School Of Accountancy. Available at: This Conference Paper is brought to you for free and open access by the School of Accountancy at Institutional Knowledge at Singapore Management University. It has been accepted for inclusion in Research Collection School Of Accountancy by an authorized administrator of Institutional Knowledge at Singapore Management University. For more information, please

2 Financial Reporting Changes and Internal Information Environment: Evidence from SFAS Introduction In this paper, we examine the effect of a change in a firm s external financial reporting on its internal information environment. Several recent studies suggest that changes in financial reporting have real consequences such as managers investment and capital allocation decisions (e.g., Graham et al. 2011; Shroff 2012; Cho 2013). These papers show that changes in financial reporting improve firms investment and internal capital allocation efficiency. However, it is unclear how the changes in financial reporting affect managers real decisions. Prior research provides two non-exclusive arguments. First, some studies argue that external reporting affects agency costs and thus real decisions. For example, Hope and Thomas (2008) and Cho (2013) provide evidence consistent with this argument. Second, Shroff (2012) argues that due to the changes to external reporting, managers must collect more information in order to comply with the new standards. Through this process, managers are likely to obtain new information and such new information helps them to make better and more informed real decisions (i.e., the information hypothesis). However, there is no direct evidence on this mechanism. The purpose of our study is to provide direct evidence on this information hypothesis. To test the information hypothesis, we focus on the introduction of Statement of Financial Accounting Standards (SFAS) No. 142, Goodwill and Other Intangible Assets, by the Financial Accounting Standards Board (FASB). Effective with fiscal years beginning after December 15, 2001, the adoption of this standard changed the accounting for goodwill dramatically. Prior to SFAS 142, accounting for goodwill was governed by APB 17 and SFAS 121. APB 17 requires the amortization of goodwill. SFAS 121 provides specific 1

3 guidelines on rules for the impairment of long-lived assets, including goodwill. Specifically, goodwill is reviewed for impairment with its related assets when circumstances indicate that the carrying amount may not be recoverable, but the test is not conducted on a regular basis. In contrast, SFAS 142 removed goodwill amortization and required firms to perform a twostep impairment test on an annual basis. In step one, the firm determines whether the fair value of the reporting unit for which the goodwill resides is lower than its current book value. If so, the firm proceeds to step two to compare the unit s recorded goodwill to the implied goodwill (i.e., the excess of the unit s fair value over its net assets excluding goodwill at fair value). If the recorded goodwill is lower, the firm records an impairment loss. Unlike other settings where information was readily available to managers but not required for disclosure before the new standard became effective (e.g., SFAS 131 on segment reporting), the adoption of SFAS 142 provides a good setting to test the information hypothesis because prior to the adoption of the new standard, managers do not necessarily have the information about the fair value of the reporting units. Because of the lack of specific FASB guidance and managers incentives to delay recognition of bad acquisition decisions, managers are not likely to have actively conducted what is similar to the impairment test required by SFAS 142 prior to its adoption. 1 Under SFAS 142, managers are mandated to assess the fair value of reporting units on a regular basis, and therefore, they have to gather more information to perform the test. To the extent that managers obtain new private information in applying SFAS 142, the adoption of the new accounting standard is likely to improve managers information sets and the firm s internal information environment. 2 As such, the adoption of SFAS 142 is an ideal setting to study how a change 1 Both anecdotal and empirical evidence suggests that managers can avoid timely recognition of goodwill impairment (Hayn and Hughes 2006; Ramanna and Watts 2011). For example, Steel writes about the delayed impairment of goodwill from Tata s acquisition of Corus, a British steelmaker, in the Economist (May 14 th, 2013). The impairment occurred six years after the acquisition. 2 Prior research on SFAS 142 provides mixed evidence on its consequences (Bens et al. 2011; Li and Sloan 2012; Chen et al. 2013; Ramanna and Watts 2011). However, it is important to note that our hypothesis merely 2

4 in external reporting requirement affects internal information environment. We use a difference-in-differences research design to isolate the effect of SFAS 142. The treatment group is a set of firms that reported goodwill throughout our sample period. The control group consists of firms that never reported goodwill during the same time period and, therefore, were not affected by the SFAS 142 adoption. The difference-in-differences analysis is effective in eliminating the bias that likely arises from differences in firm characteristics between the treatment and control groups and also from economy-wide confounding events or time trends. Following Goodman et al. (2013), who use management forecasts to infer the information available to managers for making internal capital budgeting and investment decisions, we use management earnings forecast errors to capture the quality of the firm s internal information environment. Consistent with our hypothesis that managers information sets improved from their efforts to comply with SFAS 142, we find that compared to control firms, firms affected by SFAS 142 provide more accurate earnings forecasts in the post-sfas 142 period than in the pre-sfas 142 period. Our result suggests that the annual assessment of the fair value of reporting units required under SFAS 142 induces managers to collect new information about the reporting units of the firm, which they presumably did not have previously, thereby allowing them to better forecast the firm s future performance. Next, we investigate how the effect varies with the effectiveness of firms internal and external monitoring mechanisms. Feng et al. (2009) document a positive association between internal control weakness and management forecast errors, suggesting that firms with poor internal controls are more likely to rely on erroneous internal management reports to generate earnings forecasts. Therefore, we conjecture that the effect of SFAS 142 is stronger for firms requires SFAS 142 to be a change to the firm s accounting and does not require the standard to be an improvement to external financial reporting. We argue that managers obtain new information during the goodwill impairment tests and whether that information is provided in financial reporting in a timely fashion or not is a separate issue (as with other financial reporting issues, such as earnings management). 3

5 with weak internal controls if the new reporting requirements lead firms to conduct additional tests and more rigorous valuations of goodwill. In addition, prior research finds that firms with stronger board independence are less likely to engage in fraud and are more likely to provide frequent and accurate earnings forecasts, consistent with boards of directors performing a monitoring role in reviewing the firm s information acquisition effort and disclosure policy (Dechow et al. 1996; Ajinkya et al. 2005). Similarly, several studies find that audit committee financial expertise is associated with better financial reporting quality (Krishnan and Visvanathan 2008; Dhaliwal et al. 2010). Therefore, if firms with weak governance characteristics, as measured by board independence and audit committee financial expertise, acquired less information (necessary to estimate future cash flows and fair value at each reporting unit) prior to the SFAS 142 adoption, then these firms should experience a bigger improvement in their internal information environment following the SFAS 142 adoption. Consistent with our predictions, we find that the reduction in forecast errors observed among treatment firms in the post-sfas 142 period are significantly greater for firms with internal control weakness, lower board independence and fewer financial experts on their audit committees. Given the well-documented association between institutional ownership and analyst following and disclosure quality (Lang and Lundholm 1996; Ajinkya et al. 2005), we also expect and find that firms with lower institutional ownership and lower analyst following experience a greater improvement in forecast accuracy in the post-sfas 142 period. The cross-sectional evidence increases our confidence in attributing our main finding to the effect of the treatment, i.e., information spillover under SFAS 142. In addition, if our argument is correct, we would expect to find stronger results for firms with more goodwill (relative to their total assets) because these firms managers have to spend more time in valuing the reporting units and the uncovered information is relatively 4

6 more important in forming managers earnings forecasts. Similarly, if managers expect an impairment of their goodwill, they are likely more diligent in collecting the information and conducting the impairment test, hence leading to a bigger improvement in the firm s internal information environment. Consistent with these conjectures, we find that the results are stronger for firms with greater amounts of goodwill relative to total assets and for firms with a higher likelihood of goodwill impairment, as measured by a decline in stock or accounting performance during the year. Finally, we conduct a falsification test and find that our results are not observed around a pseudo-financial reporting shock, ensuring that our finding is attributed to the adoption of SFAS 142, not by time trends or fundamental heterogeneity between treatment and control firms. Our study contributes to the recent stream of literature that documents the real consequences of financial reporting by providing shedding light on the underlying mechanism through which changes in financial reporting can affect firm behavior (e.g., Graham et al. 2005; Hope and Thomas 2008; Graham et al. 2011; Shroff 2012; Cho 2013). We extend this research by showing that a change in financial reporting requirements expands the manager s information set and improves the firm s internal information environment, leading to better forecasts of future firm performance. Our results indicate that a change in external reporting system can improve managers real decisions not only by reducing information asymmetry between managers and outside stakeholders, as established in prior literature, but also by helping increase new information internally available to managers. Our study also contributes to the line of literature that examines the consequences of the SFAS 142 adoption. To date, this body of research has focused on managers discretionary timing of goodwill impairment and investors responses to these charges (Ramanna and 5

7 Watts 2011; Li and Sloan 2012; Chen et al. 2013). These studies suggest that managers have incentives to delay the recognition of goodwill impairment charges and that investors do not see through firms overstated goodwill balances. We extend this stream of research by documenting an unintended information spillover effect of SFAS 142 on firms internal information environments. This finding should be of interest to regulators, practitioners, and researchers, as the merits of SFAS 142 and the move toward fair-value accounting are still under considerable debate. The remainder of the paper proceeds as follows. The next section discusses the institutional features of goodwill accounting and develops hypotheses. Section 3 describes the sample selection and variable definitions. Sections 4 and 5 report the main empirical results and additional analyses, respectively. Section 6 concludes the paper. 2. Institutional Background and Hypothesis Development Prior to SFAS 142, APB Opinion 17 and SFAS 121 addressed how goodwill and other intangible assets should be accounted for after they are recognized in the financial statements. APB Opinion 17 views goodwill as an asset with a finite life and, therefore, directs it to be amortized by systematic charges to income over the period estimated to be benefited. The period of amortization should not, however, exceed forty years (AICPA 1970, para. 9). In addition, it states that a company should evaluate the periods of amortization continually to determine whether later events and circumstances warrant revised estimates of useful lives. If estimates are changed, the unamortized cost should be allocated to the increased or reduced number of remaining periods in the revised useful life but not to exceed forty years after acquisition (AICPA 1970, para. 31). SFAS 121 governs accounting for the impairment of long-lived assets, certain identifiable intangibles, and goodwill related to those assets before the adoption of SFAS 6

8 142. Under SFAS 121, firm should review long-lived assets and certain identifiable intangibles for impairment whenever events or changes in circumstances indicate that the carrying amount of an asset may not be recoverable (FASB 1995, para.4). Goodwill is reviewed together with the related assets (i.e., those acquired in the same transaction of business combination that creates goodwill) if the related assets are being tested for recoverability. Under SFAS 121, an asset is considered unrecoverable if its carrying value is lower than the sum of expected future cash flows. An impairment loss is calculated as the amount by which the carrying amount of the asset exceeds the fair value of the asset. The introduction of SFAS 142 dramatically changed the accounting for goodwill. It removes goodwill amortization and requires firms to perform a two-step impairment test, which includes calculating the fair value of reporting units on an annual basis. Exhibit 1 provides an illustration of the two-step impairment test under SFAS 142 (Comiskey and Mulford 2010, Table 2, p.750). In step one, firms estimate the fair values of its reporting units and compare them with the carrying values. The purpose of this step is to determine if there is any potential for impairment. In Exhibit 1, Collectors Universe, Inc. has two reporting units, GCAL and AGL, and both have carrying values that exceed the fair value. This suggests that for Collectors Universe, there is a potential for impairment of goodwill of both the GCAL and AGL reporting units. Therefore the firm proceeds to step two, in which it determines if the potential for the impairment established in step one is realized. More specifically, in step two, the firm calculates the implied goodwill as an excess of the fair value of its reporting unit over the fair value of the reporting unit s net assets. In a normal transaction of business combination, such an excess is generally recorded as goodwill. The firm recognizes an impairment loss if the recorded goodwill (book value) is greater than its implied goodwill. In Exhibit 1, GCAL has $8,166,000 as its recorded goodwill but its implied goodwill amounts only to $899,000, and therefore determines its impairment as $7,267,000. Similarly, AGL 7

9 determines its impairment as $1,797,000. [Insert Exhibit 1] Some differences exist between the old standards (i.e., APB Opinion 17 and SFAS 121) and the new standard (i.e., SFAS 142), and these differences can enhance firms internal information environments. First, under SFAS 142, firms are required to perform an annual assessment of the fair values of their reporting units. Compared to the passive amortization of goodwill under APB Opinion 17, fair-valuing assets and liabilities at the reporting unit level requires much more involvement by the management, as managers must spend additional time and effort to collect additional information for the appraisal of assets and liabilities that generally do not have readily available market prices. To obtain a reasonable estimate of the fair value, firms need to estimate the expected future cash flows generated from the assets and the risks associated with those cash flows. 3 Second, although APB Opinion 17 required firms to evaluate the useful lives of goodwill continually, it provided little guidance on when or how to revise the useful lives. Therefore, before the adoption of SFAS 142, managers are less incentivized to develop a systematic approach for valuing goodwill and identifying its remaining useful life. 4 While SFAS 121 requires firms to review their assets for potential impairment, firms are not required to conduct the recoverability test on a regular basis. 5 However, under SFAS 142, goodwill should be tested for impairment on an annual basis. Moreover, testing must be performed even between annual assessments on an interim basis if an event occurs or 3 Comiskey and Mulford (2010) review 10K filings for a sample of US firms and find that the most frequently used methods in estimating the fair value of the reporting units are the present value of future cash flows and market multiples, or a weighted average of the two. They argue that both methods involve significant managerial discretion, and firms are generally cautious with the disclosures of their fair value estimates because changes in estimates and assumptions could materially affect the determination of fair value and/or goodwill impairment (Beacon Enterprise Solutions Groups 2008). 4 Consistent with goodwill amortization amounts providing little information content, Jennings et al. (2001) find that earnings before goodwill amortization explain significantly more of the distribution of share price than earnings including goodwill amortization. 5 Riedl (2004) examines write-offs of long-lived assets before and after SFAS 121 and finds that write-offs incurred after the adoption of SFAS 121 are more likely to be associated with big baths and manager reporting incentives, rather than economic factors. 8

10 circumstances change that would more likely than not reduce the fair value of a reporting unit below its carrying amount (FASB 2001, para. 28). The above discussion implies that managers information sets are likely to be improved following the adoption of SFAS 142. As discussed in Shroff (2012), managers are likely to become better informed after a change in accounting standards, because managers must invest time and effort to collect additional information to comply with the new rule. This change in the manager s information set has a positive spillover effect and can improve the firm s investment efficiency. This argument is referred to as the information hypothesis. It thus follows that a firm s internal information environment is improved following SFAS 142. While internal information environment is not directly observed by researchers, an improvement in internal information environment is likely to manifest as an increase in managers forecast accuracy. In particular, Goodman et al. (2013) examine whether the quality of management earnings forecast is associated with capital budgeting decisions, because managers use similar forecasting skills in forecasting earnings and return to investments. They conclude that earnings forecasts, which are used for external reporting, can be used to infer the firm s internal information quality. Following Goodman et al. (2013), we use the quality of management earnings forecasts as a proxy for the quality of the firm s internal information environment. Since management must rely on internal reports to generate these external forecasts, greater management forecast accuracy (negative of forecast error) is likely to indicate higher internal information quality (Feng et al. 2009). This discussion leads us to our first hypothesis: H1: Ceteris paribus, management forecast accuracy is higher for treatment firms following the adoption of SFAS 142. The effect of SFAS 142 on management forecast accuracy is likely to vary across firms depending on the quality of the firm s internal information environment. For example, firms with better corporate governance may already have valued their goodwill on a regular basis 9

11 prior to SFAS 142, as prior research suggests that corporate governance characteristics and internal controls are positively associated with financial reporting quality. Ajinkya et al. (2005) find that firms with greater board independence and more institutional investors provide forecasts with higher quality. Feng et al. (2009) show that firms with internal control weaknesses provide forecasts with greater error. Similarly, several studies find that audit committee financial expertise is associated with better financial reporting quality (Krishnan and Visvanathan 2008; Dhaliwal et al. 2010). Therefore, if firms with stronger internal and external monitoring mechanisms already collected the information about the reporting units prior to the adoption of SFAS 142, then they should experience a smaller improvement in their internal information environment following the adoption of SFAS 142. This leads us to our second hypothesis: H2: The positive association between the adoption of SFAS 142 and management forecast accuracy is stronger for firms with weak monitoring mechanisms. 3. Research Design 3.1 Data and Sample Given that SFAS 142 becomes effective for fiscal years starting after December 15, 2001, the calendar year of the adoption (as in financial statement dates) is 2002 for December fiscal year-end firms, and 2003 for non-december fiscal year-end firms. To test for the effect of SFAS 142, we focus on the six-year period around the adoption of SFAS 142, in which the pre-sfas 142 period is for December fiscal year-end firms and for non-december fiscal year end firms. The post-sfas 142 period covers the following three years accordingly. To select our sample, we begin with the Compustat /CRSP merged dataset. The treatment group consists of firms which report goodwill in their financial statements throughout the six-year sample period (as recorded in Compustat), while the control group 10

12 consists of firms which do not report any goodwill at any point in time during our sample period (i.e., firms with zero or missing balance of goodwill as reported in Compustat). Firms that report goodwill in one year but do not report in another year are excluded from the sample to increase the power of the test. We then merge this sample with the management forecast data from First Call. We use forecast error as a proxy for a firm s internal information environment (Goodman et al. 2013). Excluding firms that do not issue earnings forecasts during our sample period or whose forecast errors are not defined, our sample consists of 2,460 firm-years, 1,767 from treatment firms and 693 from control firms. 3.2 Regression Model We use a difference-in-differences design to isolate the effect of SFAS 142 on firms earnings forecast errors by addressing the impact of the difference between treatment and control firms and time trend. Given that treatment firms have goodwill while control firms don t, the former is likely to differ from the latter fundamentally in firm characteristics. For example, because treatment firms engaged in acquisitions that created goodwill, they are likely to be larger, more diversified, more profitable, and face higher growth opportunities relative to control firms. More specifically, we estimate the following regression: Forecast Error i,t = β 0 + β 1 Post142 + β 2 Treatment + β 3 Post142 Treatment + β 4 InstOwn i,t-1 + β 5 NumAnal i,t-1 + β 6 Disp i,t-1 + β 7 RetVol i,t + β 8 Size i,t-1 + β 9 EqIss i,t + β 10 Lit i,t + β 11 MTB i,t + β 12 ROA i,t + β 13 Ret i,t + β 14 NumSegs i,t + β 15 Loss i,t + β 16 Horizon i,t + β 17 Surprise i,t + Industry Dummies + Year Dummies + e i,t (1) Forecast Error is the error in managers forecast of the firm s current fiscal year earnings, averaged across all forecasts issued in the period between the earnings announcement of the previous fiscal year and the current fiscal year-end. (That is, we drop pre-announcements.) Forecast Error of each earnings forecast is calculated as the absolute value of the difference between management earnings forecast and actual earnings, scaled by 11

13 the stock price at the beginning of the fiscal year. Post142 is an indicator variable that equals one for the post-sfas 142 period, and zero otherwise. Treatment is an indicator variable that equals one for firms in the treatment group, and zero for control firms. Post142 Treatment is a product of Post142 and Treatment. Hypothesis H1 implies a negative coefficient on the coefficient on this interaction term. Equation (1) controls for the effect of various firm and forecast characteristics. InstOwn is institutional investors ownership, measured as the number of shares held by institution investors as reported in Thomson Reuters Institutional (13f) Holdings database divided by the total number of shares outstanding. Firms not covered by 13f institutions are assumed to have zero institutional ownership. NumAnal is the number of unique analysts who issue earnings forecast for the firm as reported in the IBES database. Firms not covered by the IBES are assumed to have zero analyst coverage. Disp is the dispersion of analyst forecasts, measured as the standard deviation of analyst earnings forecasts divided by the absolute value of mean analyst earnings forecast, where we use the latest forecast issued by each analyst before the fiscal year-end. RetVol is return volatility, measured as the standard deviation of the firm s daily stock returns over the fiscal year. Size is firm size, measured as the natural logarithm of total assets. EqIss is an indicator variable for firms that issue additional equity during the fiscal year as reported in Security Data Corporation s (SDC) Global New Issues database. Lit is an indicator variable for firms in the industries subject to high litigation risk (i.e., SIC codes being within , , , , , and ). MTB is the market-to-book ratio, measured as the market value of equity divided by book value of equity. The firm s market value of equity is calculated as the number of shares outstanding multiplied by the closing price at the fiscal year-end. ROA is return on assets, measured as income before extraordinary items divided by the beginning-of-period total assets. Ret is size and market-to-book adjusted annual return over the fiscal year. Numseg is 12

14 the number of operating segments. Loss is an indicator variable that equals one if the firm s income before extraordinary items is less than zero, and zero otherwise. We expect forecast error to be negatively associated with InstOwn, NumAnal, Size, EqIss, Lit, ROA, and Ret, and positively associated with Disp, RetVol, Numseg, and Loss. Horizon is the number of days between managers forecast date and fiscal year-end, averaged across all forecasts issued in the period between the announcement of earnings of the previous fiscal year and the current fiscal year-end. Surprise is the average difference between management earnings forecasts and prevailing consensus analyst forecasts, scaled by the stock price at the beginning of the year. Prevailing consensus analyst forecasts are calculated using the forecast issued by each analyst before the managers earnings forecast dates. We expect forecasts issued earlier in the fiscal period to have greater error. Similarly, forecast errors are likely to be positively associated with the analyst surprise 3.3 Descriptive Statistics Panel A of Table 1 presents the descriptive statistics on the variables used in our study. The means of Forecast Error are and for treatment and control groups, respectively. The difference in Forecast Error between the two groups is not significantly different from zero. Note that firms in the treatment group are fundamentally different from those in the control group because treatment firms should have experienced acquisitions. As indicated in Table 1, treatment firms have significantly higher institutional ownership, have higher growth opportunities, have better performance, and have more segments. It is thus importantly to control for the impact of these firm characteristics in the analyses. Panel B of Table 1 shows correlation coefficients between variables. The correlations of Forecast Error with control variables are generally consistent with the findings of prior literature. [Insert Table 1] 13

15 4. Empirical Results Table 2 presents the regression results. In Model 1, we estimate the regression using the full sample, and in Model 2 using a shorter window covering two-year period immediately around SFAS 142 adoption (i.e., one year before and one year after the SFAS 142 adoption). In both Models 1 and 2, the coefficients on Post142 Treatment is significantly negative (p=0.03 and 0.00 respectively), indicating that treatment firms experienced a greater decrease in managers forecast errors than did control firms after the adoption of SFAS 142. Our finding is consistent with SFAS 142 improving internal information environment as a result of information spillover by assessing the fair value of reporting units on an annual basis. [Insert Table 2] Table 3 presents the results for the cross-sectional analyses. Hypothesis H2 implies a greater decrease in forecast errors for firms with weak internal monitoring after the adoption of SFAS 142. In Panel A of Table 3, we examine the role of internal controls, where MW equals one if the firm is identified to have material weakness in internal control, and zero otherwise. We use Doyle et al. s (2007) data of material weakness in internal control from August 2002 to November 2005, the data disclosed under Sections 302 and 404 of the Sarbanes-Oxley Act of Considering that our sample period starts from 1999, we assume that firms that disclosed internal control weakness suffered the problem not only in the year it was disclosed but also in preceding three years before the disclosure. The variable of interest in Panel A is the three-way interaction of MW Post142 Treatment. Results show that in both Models 1 and 2, both the coefficient on Post142 Treatment and the coefficient on MW Post142 Treatment are significantly negative. These results suggest that treatment firms on average experience a significant decrease in forecast errors after the SFAS 142 adoption, and the effect is significantly greater for firms with internal control 6 We thank Weili Ge for sharing the data at 14

16 weakness than for firms without internal control weakness. In Panel B of Table 3, we focus on the financial expertise of audit committee. We use the data from Corporate Library to identify the background of audit committee members, and calculate the proportion of the committee members who have accounting or finance background. Because our data is available only in the post-sfas 142 period for most our sample firms, we assume that the expertise of audit committee is stable over time in our sample period. Under this assumption, for each firm, we calculate the average of such a proportion across the sample years available, and use the averaged value as a firm-specific (time-constant) measure of audit committee expertise. WeakAC is one if the averaged value of the proportion is less than 50%, and zero otherwise. The variable of interest in Panel B is the three-way interaction of WeakAC Post142 Treatment. As reported in both Models 1 and 2, the coefficient on three-way interactions is significantly negative, suggesting that the decrease in forecast errors is greater for firms with weak audit committees than for firms with strong audit committees. In Panel C of Table 3, we examine the independence of boards of directors. More specifically, we define WeakBoard to take a value of one if the proportion of independent directors is lower than two thirds of total number of board members, and zero otherwise. To determine the independence of boards of directors, we use data collected from Compact Disclosure. In Panel C, the variable of interest is the three-way interaction of WeakBoard Post142 Treatment, and the coefficient is significantly negative in both Models 1 and 2. Overall, these results are consistent with firms with weak internal monitoring experiencing a greater improvement in internal information environment after the adoption of SFAS 142 than firms under strong internal monitoring. [Insert Table 3] In a similar vein, in Table 4 we investigate the impact of external monitoring such as 15

17 institutional investors and financial analysts. In Panel A of Table 4, we define LowInst to take a value of one if the firm s InstOwn (i.e., institutional ownership) is lower than or equal to 50%, and zero otherwise. The variable of interest in Panel A is the three-way interaction of LowInst Post142 Treatment, and the coefficient is significantly negative in both Models 1 and 2. In Panel B of Table 4, we focus on analyst coverage and we define LowAnalyst to take a value of one if the firm s analysts following is less than or equal to seven (i.e., the sample median of NumAnal), and zero otherwise. The variable of interest in Panel B is the three-way interaction of LowAnalyst Post142 Treatment. The coefficient is significantly negative in both Models 1 and 2. Results in Table 4 generally support the idea that the adoption of SFAS 142 enriches internal information (through its requirement of assessing fair value at each reporting unit on an annual basis) to a greater extent for firms under weaker external monitoring compared to firms under stronger external monitoring. 7 [Insert Table 4] 5. Additional Tests 5.1 The extent of managers effort in collecting information for goodwill impairment test To the extent that managers can exercise some degree of discretion in applying accounting standards, the effect of SFAS 142 on internal information environment would likely differ depending on managerial incentives to review the value of goodwill. In Panel A of Table 5, we focus on the size of goodwill relative to total assets. Managers are likely to spend more time and effort in reviewing their goodwill as the intangible is more important for firm assets. Therefore, for firms with more goodwill, the information spillover effect is likely to be greater. In Panel A, we define LargeGoodwill to take a value one if the ratio of goodwill 7 In Table 4, we do not include the main effect of LowInst (in Panel A) and LowAnalyst (in Panel B) because the regression already controls for InstOwn and NumAnal. 16

18 to total assets is greater than sample median (9.5%), and zero if less than sample median. Given that goodwill is zero for control firms, this analysis focuses only on treatment firms. The coefficients on Post142 is significantly negative in both Models 1 and 2, suggesting that firms with small-sized goodwill (relative to total assets) experience an improvement in internal information environment as a result of annual assessments of fair value required under SFAS 142. The coefficient on Post142 LargeGoodwill is negative in both models, although it is insignificant at the conventional level in Model 1, consistent with the notion that the effect of SFAS 142 on information spillover is larger as goodwill is more important for firm assets. In Panel B of Table 5, we focus on the ex-ante likelihood of goodwill impairment. If managers believe that the goodwill is closer to impairment, they likely invest more time and effort in collecting information. While firms are required to assess the fair value of reporting units on an annual basis, if an event occurs or circumstances change that would more likely than not reduce the fair value of a reporting unit below its carrying amount (FASB 2001, para. 28), firms should carry out the impairment test even between the annual assessments. Prior literature suggests that a decline in market value likely affects the firm s decision to perform the interim test (Dharan 2009; Comiskey and Mulford 2010; Chen et al. 2013). In Panel B, we define ImpairLikely as one if the firm s MTB (the market-to-book ratio) decreases over the year and is less than one at the end of the year, and zero otherwise. The coefficient on Post142 is significantly negative in both Models 1 and 2, suggesting that SFAS 142 created information spillover for firms with no triggering events (through annual tests). Importantly, the coefficient on Post142 ImpairLikely is also significantly negative in both Models 1 and 2. These results suggest that firms facing a higher chance of goodwill impairment (as indicated by a decline in market value) spend more time collecting information to calculate the fair value and, therefore, experience a greater improvement in 17

19 internal information environment. In Panel C, we use an alternative measure of ImpairLikely. Instead of relying on stock prices, we use the change and level of ROA to measure the likelihood of goodwill impairment because market values may not be a reliable indicator of asset impairment, as argued by some practitioners. 8 Specifically, we redefine ImpairLikely to take a value of one if the firm s ROA decreases and it is smaller than zero for the year, and zero otherwise. As in Panel B, both coefficients on Post142 and on Post142 ImpairLikely are significantly negative, suggesting that the annual test of goodwill impairment leads to an improvement in internal information environment, and such an improvement is greater for firms with high exante likelihood of goodwill impairment. 9 [Insert Table 5] 5.2 Falsification test In Table 6, we examine whether our result holds around a pseudo-financial reporting shock. The purpose of this analysis is to ensure that our result is attributed to the treatment (i.e., the adoption of SFAS 142), not to time trends or fundamental heterogeneity between treatment and control firms. For this purpose, we define PseudoPost to take a value of one if the firm-year observation is in the 4 th to the 6 th year after SFAS 142 adoption, and zero if it is in the 1 st to the 3 rd year after SFAS 142 adoption. Same as our main analysis, Treatment takes a value of one if the firm reports goodwill throughout the six-year period as recorded in Compustat, and zero if it does not report goodwill at any point in time during the period. (As 8 For example, Emtec s market capitalization was less than the total shareholders equity, which is one factor the company considered when determining whether goodwill should be tested for impairment between annual tests. The company doesn t believe that the reduced market capitalization represents a goodwill impairment indicator as of August 31, 2008 (EMTEC, Inc., 2008). The management of Alcoa Inc. also believes the company s forecasted cash flows constitute a better indicator of the current fair value of Alcoa s reporting units than the current pricing of its common shares (Alcoa Inc. 2009). Then firms are likely to view their goodwill as being closer to impairment and pay more attention to its valuation when their accounting (or cash flow) performance has been deteriorating. 9 In an untabulated analysis, we do not find evidence that firms whose impairment tests resulted in actual goodwill impairment experienced a greater improvement in internal information environment than firms that did not impair goodwill, suggesting that information spillover occurs mainly in the first step of the impairment test, not in the second step. 18

20 before, firms reporting good will in some years but not in other years are deleted to increase the power of the test.) The research design is the same as in Table 2 otherwise. As reported in Table 6, the coefficient on PseudoPost Treatment is not significantly different from zero in both Models 1 and 2. These results increase our confidence that our findings documented in this study are attributable to the adoption of SFAS 142. [Insert Table 6] 6. Conclusion In this paper, we examine the effect of a change in a firm s financial reporting on its internal information environment by investigating the impact of SFAS 142 adoption on the accuracy of management earnings forecasts, a proxy for the quality of internal information environment. While prior studies document a link between financial reporting changes and real consequences such as investments (e.g., Graham et al. 2011; Shroff 2012; Cho 2013), there is limited research on the mechanism behind this link. In our study, we suggest that information spillover is one of the underlying mechanisms. We argue that a mandatory change in external reporting system induces managers to acquire new information and, therefore, improves managers information set and internal information environment, which in turn can result in more efficient investment. The adoption of SFAS 142 is an ideal setting to test this effect because it requires firms to perform a two-step impairment test, which includes calculating the fair value of reporting units on an annual basis, inducing managers to collect information they otherwise would not have. Using a difference-in-difference research design, we find that firms affected by SFAS 142 experienced a greater increase in their forecast accuracy, consistent with an improvement in the firm s internal information environment following the adoption of SFAS 142. We also document that firms with weaker corporate governance mechanisms experience a greater 19

21 improvement in their forecast accuracy. The findings in this paper should be of interest to standard setters, practitioners, and researchers, as they suggest that a change in financial reporting for external purposes can have a positive information spillover effect on the firm s internal information environment. 20

22 References AICPA, 1970, Accounting Principles Board (APB) Opinions No. 17: Intangible Assets. Ajinkya, B., S. Bhojraj, and P. Sengupta, 2005, The association between outside directors, institutional investors, and the properties of management earnings forecasts, Journal of Accounting Research 43: Alcoa Inc., 2009, Form 10-K, Annual report pursuant to section 13 or 15(d) of the securities exchange act of Bens, D., W. Heltzer, and B. Segal, 2011, The information content of goodwill impairments and SFAS 142, Journal of Accounting, Auditing, and Finance 26: Chen, W., P. Shroff, and I. Zhang, 2013, Consequences of booking market-driven goodwill impairments, Working Paper, University of Minnesota. Cho, Y. J., 2013, Segment Disclosure Transparency and Internal Capital Market Efficiency: Evidence from SFAS No. 131, Working Paper, Singapore Management University. Comiskey, E., and C. Mulford, 2010, Goodwill, triggering events, and impairment accounting, Managerial Finance 36: Dechow, P., R. Sloan, and A. Sweeney, 1996, Causes and consequences of earnings management: An analysis of firms subject to enforcement actions by the SEC, Contemporary Accounting Research 13: Dhaliwal, D., V. Naiker, and F. Navissi, 2010, The association between accruals quality and the characteristics of accounting experts and mix of expertise on audit committees, Contemporary Accounting Research 27: Dharan, B., 2009, Preparing for goodwill impairments, Law360, February 5, Doyle, J., W. Ge, and S. McVay, 2007, Accruals quality and internal control over financial reporting, The Accounting Review 82: EMTEC, Inc., 2008, Form 10-K, Annual report pursuant to section 13 or 15(d) of the securities exchange act of FASB (Financial Accounting Standards Board), 1995, Statement of Financial Accounting Standards No. 121: Accounting for the impairment of long-lived assets and for long-lived assets to be disposed of. FASB (Financial Accounting Standards Board), 2001, Statement of Financial Accounting Standards No. 142: Goodwill and other intangible assets. Feng, M., C. Li, and S. McVay, 2009, Internal control and management guidance, Journal of Accounting and Economics 48: Goodman, T., M. Neamtiu, N. Shroff, and H. White, 2013, Management forecast quality and capital investment decisions, The Accounting Review, Forthcoming. Graham, J., C. Harvey, and S. Rajgopal, 2005, The economic implications of corporate financial reporting, Journal of Accounting and Economics 40: Graham, J., M. Hanlon, and T. Shevlin, 2011, Real effects of accounting rules: Evidence from multinational firms investment location and profit repatriation decisions, Journal of Accounting Research 49: Hayn, C., and J. Hughes, 2006, Leading indicators of goodwill impairment, Journal of Accounting, Auditing, and Finance 21: Hope, O., and W. Thomas, 2008, Managerial empire building and firm disclosures, Journal of Accounting Research 46:

23 Jennings, R., M. LaClere, and R. Thompson, 2001, Goodwill amortization and the usefulness of earnings, Financial Analysts Journal 57: Karamanou, I., and N. Vafeas, 2005, The association between corporate boards, audit committees, and management earnings forecasts: An empirical analysis, Journal of Accounting Research 43: Krishnan, G., and G. Visvanathan, 2008, Does the SOX definition of an accounting expert matter? The association between audit committee directors accounting expertise and accounting conservatism, Contemporary Accounting Research 25: Lang, M., and R. Lundholm, 1996, Corporate disclosure policy and analyst behavior, The Accounting Review 71: Li, K., and R. Sloan, 2012, Has goodwill accounting gone bad?, Working Paper. Ramanna, K., and R. Watts, 2011, Evidence on the use of unverifiable estimates in required goodwill impairment, Review of Accounting Studies, 17: Riedl, E., 2004, An examination of long-lived asset impairments, The Accounting Review 79: Shroff, N., 2012, Corporate investment and changes in GAAP, Working Paper, MIT. 22

24 Appendix: Variable Definition Forecast Error is the error in managers forecast of the firm s current fiscal year earnings, averaged across all forecasts issued in the period between the earnings announcement of the previous fiscal year and the current fiscal year-end. (That is, we drop pre-announcements.) Forecast Error of each earnings forecast is calculated as the absolute value of the difference between management earnings forecast and actual earnings, scaled by the stock price at the beginning of the fiscal year. InstOwn is institutional investors ownership, measured as the number of shares held by institution investors as reported in Thomson Reuters Institutional (13f) Holdings database divided by the total number of shares outstanding. Firms not covered by 13f institutions are assumed to have zero institutional ownership. NumAnal is the number of unique analysts who issue earnings forecast for the firm as reported in the IBES database. Firms not covered by the IBES are assumed to have zero analyst coverage. Disp is the dispersion of analyst forecasts, measured as the standard deviation of analyst earnings forecasts divided by the absolute value of mean analyst earnings forecast, where we use the latest forecast issued by each analyst before the fiscal year-end. RetVol is return volatility, measured as the standard deviation of the firm s daily stock returns over the fiscal year. Size is firm size, measured as the natural logarithm of total assets. EqIss is an indicator variable for firms that issue additional equity during the fiscal year as reported in Security Data Corporation s (SDC) Global New Issues database. Lit is an indicator variable for firms in the industries subject to high litigation risk (i.e., SIC codes being within , , , , , and ). MTB is the market-to-book ratio, measured as the market value of equity divided by book value of equity. The firm s market value of equity is calculated as the number of shares outstanding multiplied by the closing price at the fiscal year-end. ROA is return on assets, measured as income before extraordinary items divided by the beginning-of-period total assets. Ret is size and market-to-book adjusted annual return over the fiscal year. Numseg is the number of operating segments. Loss is an indicator variable that equals one if the firm s income before extraordinary items is less than zero, and zero otherwise. Horizon is the number of days between managers forecast date and fiscal year-end, averaged across all forecasts issued in the period between the announcement of earnings of the previous fiscal year and the current fiscal year-end. Surprise is the average difference between management earnings forecasts and prevailing consensus analyst forecasts, scaled by the stock price at the beginning of the year. Prevailing 23