Corporate Diversification and Overinvestment: Evidence from Asset Write-Offs*

Transcription

1 Corporate Diversification and Overinvestment: Evidence from Asset Write-Offs* Gil Sadka and Yuan Zhang November 10, 2008 Preliminary and incomplete Please do not circulate Abstract This paper documents that diversified firms are more likely to overinvest than undiversified firms. Prior research on inefficient investment in diversified firms has been criticized on the measurement errors in measuring the unobservable marginal investment opportunities. Using an ex post measure of overinvestment that does not rely on investment opportunity set asset write-offs, we find that diversified firms are more likely to incur tangible write-offs as well as goodwill write-offs and that the extent of the write-offs increases as firms become more diversified. In addition, we show that the diversification discount is related to subsequent asset write-offs, suggesting that diversified firms trade at a discount because investors understand the overinvestment and correctly anticipate future write-offs. * We thank Stephen Penman for helpful comments. All errors are our own. ** Gil and Yuan are from Columbia University. The authors s are gs2235@columbia.edu, and yz2113@columbia.edu.

2 Corporate Diversification and Overinvestment: Evidence from Asset Write-Offs 1. Introduction Prior studies (e.g., Wernerfeld and Montgomery, 1988; Lang and Stulz, 1994; Berger and Ofek, 1995) document that more diversified firms trade at a discount compared with more specialized firms. The effect of diversification on market value is known as the diversification discount, usually measured either as Toibin s q (e.g., Lang and Stulz, 1994) or as the difference between the value of diversified firm and an imputed value based on undiversified firms (e.g., Berger and Ofek, 1995). 1 These studies attribute the diversification discount to inefficient investment. For example, Jensen (1986) hypothesizes that mangers of firms with free cash flows tend to engage in empire building. Specifically, these managers tend to overinvest in other industries to increase the size of the firm to gain power and prestige (see Jensen and Murphy, 1990) and other personal benefits instead of distributing the cash flows. 2 Other theories (e.g., Stulz, 1990; Rajan, Servaes and Zingales, 2000) suggest that diversification results in an inefficient internal capital market, which leads to the diversification discount. However, recently there have been two major criticisms of this literature. First, some researchers have challenged the existence of diversification discount itself or the notion that diversification destroys firm value. For example, researchers have attributed the diversification discount to the endogenous nature of the diversification choice (Villalonga, 1 Another strand of evidence on the value destroying effects of diversification comes from short- and long-term event studies (e.g., Agrawal, Jaffe, and Mandelker, 1992). 2 Among these private benefits are risk diversification (Amihud and Lev, 1981 and Mansi and Reeb, 2002) and entrenching mangers as they invest in projects that fit their particular skill sets (Shleifer and Vishney, 1989). 1

3 2004a) or to the artifact of poor segment data (Villalonga, 2004b). Other papers have partially attributed the diversification discount to the discount in either the target firm (Graham, Lemmon, and Wolf, 2000) or in the individual firm before it undertook a diversifying merger (Chevalier, 2004), suggesting that it is not the diversification itself that destroys value. Second, the diversification discount research that examines investment efficiency has largely relied on the association between capital expenditure and proxies for investment opportunities to document overinvestment and/or underinvestment (e.g., Rajan, Servaes, and Zingales, 2000). However, this method has been criticized on the basis that the proxies used for investment opportunities (e.g., Tobin s q) do not capture the marginal investment opportunities, which is the underlying theoretical construct that should be used in these analyses. In particular, Whited (2001) corrects for measurement error in q and finds no evidence of inefficient allocation of investment. In this paper we employ a measure of overinvestment that does not rely on measures of the investment opportunity set. Specifically, we use asset write-offs as an ex post measure of overinvestment. We first examine whether diversified firms are more likely to have asset write-offs. In further analysis, we test whether the incidence as well as the magnitude of asset write-offs is correlated with the diversification discount as measured in prior literature. There are several advantages of using asset write-offs to measure overinvestment. First, it is a direct measure of overinvestment and does not rely on estimations based on the unobservable marginal investment opportunity. Second, unlike prior research that focuses mainly on capital expenditure in investigating investment inefficiency, focusing on write-offs allows us to examine all types of overinvestment in assets, including overinvestment in mergers or acquisitions. Finally, write-offs could also capture cases where the price paid for 2

4 the asset was fair yet the management of the asset was lacking, resulting in a permanent decline in value. This type of investment could be caused by managers empire building incentive as they invest in industries that are beyond their knowledge or expertise. It is important to note, however, that our measure of investment inefficiency cannot address underinvestment and that its nature of being a measure of extreme overinvestment implies that it is not a comprehensive measure of all overinvestment. Based on a sample of firms in , we find that diversified firms are significantly more likely to write off tangible assets than undiversified firms. Specifically, our univariate findings suggest that diversified firms are about 17% more likely to record a tangible write-off than undiversified firms, and that this difference is mainly driven by firms with at least three business segments. In multivariate regression analysis, we find that while diversified firms are more likely to record tangible write-offs than undiversified firms, the magnitudes of the write-offs are not significantly larger. However, both the likelihood and the magnitude of the tangible write-offs increase with respect to the number of business segments, suggesting firms with greater degree of diversification drives the effects of diversification. In other words, as a firm becomes less specialized, it not only becomes more likely to overinvest, it also overinvests to a greater extent in tangible assets. We extend our analysis by studying goodwill write-offs as well. We show that diversified firms are also more likely to make bad acquisition decisions and/or overpay for targets. Specifically we document that diversified firms are more likely to record goodwill write-offs compared to undiversified firms. Further, diversified firms also report higher goodwill write-offs than undiversified firms. Both the likelihood and the magnitude of the goodwill write-offs also increase with the number of business segments. 3

5 In further analysis, we link our findings to the diversification discount documented in prior research. We follow Berger and Ofek (2005) and find that the diversified firms in our sample have a diversification discount of about 18-20% relative to the undiversified firms. We show that this diversification discount is significantly associated with subsequent writeoffs. This result suggests that diversified firms trade at a discount because investors expect future write-offs due to overinvestment. Overall, our findings are consistent with the hypothesis that diversified firms are more likely to invest inefficiently and consequently trade at a discount. Finally, we examine whether corporate governance plays any role in mitigating the overinvestment in diversified firms. We find some evidence that among diversified firms, those with higher percentage of independent directors report significantly lower goodwill write-offs. We do not, however, find that independent directors play any role in curtailing overinvestment in tangible assets among diversified firms. Our paper makes the following contributions to the literature. First, prior research that examines asset write-offs focuses more on the determinants of the magnitude of write-offs (e.g., Francis, Hanna and Vincent, 1996). Ours is one of the first studies that employ asset write-offs as a direct and ex post measure of overinvestment, which has important implications in documenting corporate investment inefficiency. Second, the literature on corporate diversification has largely focused on the relation between capital expenditure and Tobin s q to document investment inefficiency. However, as discussed above, given the problems in using Tobin s q to proxy for marginal investment opportunity, the evidence on whether diversified firms indeed engage in inefficient investment has been inconclusive. Using asset write-offs to directly capture the extent of overinvestment, we contribute to the 4

6 literature on diversification discount by providing additional and more direct evidence that diversified firms are more likely to overinvest. Third, we extend the research on investment inefficiency of diversified firms by showing that diversified firms are also more likely to engage in bad acquisitions. Graham, Lemmon, and Wolf (2002), among others, attribute the diversification discount to the discount in the target firms. However, even if the target firms are traded at a discount before the acquisitions, the acquiring firms should not pay above value. Finally, our results on the role of corporate governance, specifically, of independent directors, are consistent with Byrd and Hickman (1992) who find that bidding firms with an independent board have significantly higher announcement-date abnormal returns for their tender offers than other bidders. While our results suggest that more independent boards are more likely to make major corporate decisions in the best interests of shareholders, they also suggest that such a role seems to be limited to acquisition decisions as opposed to other investment decisions. The remainder of the paper proceeds as follows. Section 2 develops hypothesis. Section 3 discusses the sample and descriptive statistics. Section 4 examines the relation between diversification and asset write-offs and Section 5 examines the relation between the diversification discount and asset write-offs. Section 6 analyzes the role of independent directors in mitigating overinvestment in diversified firms. We conclude in Section Hypotheses Development 2.1. Background and Literature Review Diversification Discount and Inefficient Investment 5

7 A long standing line of research in the finance literature has shown that diversification reduces firm value. For example, using segment-level data over , Berger and Ofek (1995) show that diversified firms have values that average 13% -15% below the sum of the imputed stand-alone values of their segments (see also Lang and Stulz 1994; Comment and Jarrell 1995). 3 This literature also examines the possible sources for the diversification. Most of this literature has attributed the diversification discount to inefficient allocation of capital expenditures across divisions. For example, Lamont (1997), Shin and Stulz (1998), and Rajan, Servaes, and Zingales (2000) find that internal capital markets in diversified firms transfer funds across divisions in a suboptimal manner. Some researchers focus specifically on overinvestment (e.g., Jensen 1986; Stulz 1990; Berger and Ofek, 1995), while others also examine under-investments (e.g., Ozbas and Scharfstein, 2008). In measuring investment efficiency (or lack thereof), this literature has largely relied on the association between capital expenditure and proxies for investment opportunities to document inefficiency in investments (e.g., Berger and Ofek 1995; Rajan, Servaes, and Zingales, 2000). 4 However, this method has been criticized on the basis that the proxies used for investment opportunities (e.g., Tobin s q in Shin and Stulz, 1998; sales growth in Lamont, 1997) do not capture the marginal investment opportunities, which is the underlying 3 Some researchers have challenged the existence of a diversification discount or the notion that diversification destroys value. For example, Villalonga (2004a) points out that diversification is an endogenous choice, which depends on the investment opportunity set (see also Campa and Kedia, 2002). Using a selection model, he finds that diversification does not destroy value. Further, Villalonga (2004b) finds that the diversification discount is an artifact of poor segment data. Using establishment-level data, he shows that diversified firms trade at a premium, not discount. Other papers have partially attributed the diversification discount to the discount in either the target firm (Graham, Lemmon, and Wolf, 2000) or in the individual firm before it undertook a diversifying merger (Chevalier, 2004), suggesting that it is not the diversification itself that destroys value. 4 One exception is Schoar (2002) which documents that firms that diversify experience a decline in productivity measured using plant-level data. The decline in productivity is apparent in the incumbent plant. 6

8 theoretical construct that should be used in these analyses. As explained in Whited (2001), marginal investment opportunity is usually defined as the firm manager s expectation of the present discounted value of the future marginal product of capital. As has been widely discussed in the corporate finance and macroeconomics literatures, observable measures of Tobin s q may diverge substantially from unobservable marginal q (Whited 2001). Whited (2001) corrects for measurement error in q and finds no evidence of inefficient allocation of investment. There also have been other papers that have challenged the inefficient investment explanation for the diversification discount. For example, Maksimovic and Phillips (2002) find evidence in a large sample of U.S. manufacturing plants that it is the equilibrium distribution of comparative advantage across firms in an industry, and not agency costs that gives rise to the discount. To summarize, given the problems associated with using the Tobin s q to proxy for the unobservable marginal investment opportunities, researchers still debate whether the diversification discount is attributable to inefficient investment Write-Offs and Overinvestment In this subsection, we discuss why and how the write-offs can be considered as a direct and ex post measure of overinvestment, and briefly summarize impairment accounting for long-lived assets and goodwill. When firms recognize an asset write-off, there are at least two possibilities regarding the initial investment. First, the firm may have invested in negative net-present-value (NPV) projects, which clearly indicates overinvestment. Second, when the firm initially made the 7

9 investment, it may be a positive NPV project based on the firm s assessment. However, as time went by, the firm may realize that they optimistically estimated the present value of the project by either over-estimating the future cash flows or under-estimating the inherent risk. While a single case may simply reflect the underlying nature and difficulty in making estimation in investment decisions, systematically doing so implies inefficiency, and particularly, overinvestment by the managers. 5 Accordingly, we use the recognition of writeoffs in the financial statements as a direct and ex post measure of overinvestment and examine the investment inefficiency in diversified firms. Firms started to provide write-offs information separately in 2001 following SFAS 144 Accounting for the Impairment or Disposal of Long-Lived Assets (FASB 2001b) for long-lived assets and SFAS 142 Goodwill and Other Intangible Assets (FASB 2001a) for goodwill and other intangible assets. Our information on write-offs is obtained from Compustat (data380 for tangible write-offs and data368 for goodwill write-offs). According to Compustat, tangible write-offs include impairments and write-downs/write-offs for assets other than goodwill, while goodwill write-offs include impairments and write-downs/writeoffs for goodwill. Presumably, the tangible write-offs reported by Compustat thus include write-offs not only from long-lived assets (i.e., property, plant, and equipment), but also from other assets such as prepaid expenses or inventory. However, in the following discussion, we focus on the long-lived assets as they are the main scope of SFAS 144 and should account for the majority of the tangible write-offs. Tangible write-offs 5 Write-offs on projects previously considered as NPV can also be due to natural disaster or macro-economic conditions that are out of managers control. However, such write-offs should not have systematic correlations with the diversification status of a firm. 8

10 According to SFAS 144, a long-lived asset shall be tested for recoverability whenever events or changes in circumstances indicate that its carrying amount may not be recoverable. Impairment for the long-lived asset is the condition that exists when the carrying amount of a long-lived asset exceeds its fair value and an impairment loss shall be recognized only if the carrying amount of a long-lived asset is not recoverable and exceeds its fair value (FASB, 2001b). This suggests that when a company recognizes impairment, it acknowledges that a previous investment in long-lived assets has been permanently impaired. We note, however, that the write-offs are at best an under-estimate of the extent of overinvestment by managers for at least two reasons. The first reason relates to the requirement of recognizing write-offs specified by SFAS 144 (FASB, 2001b). Specifically, SFAS 144 requires firms to (i) recognize an impairment loss only if the carrying amount of a long-lived asset is not recoverable from its undiscounted cash flows expected to result from the use and eventual disposition of the asset and (ii) measure an impairment loss as the difference between the carrying amount and fair value of the asset (i.e., discounted cash flows). To the extent that the carrying value of the asset is above its fair value but below the undiscounted cash flows, even though the asset itself is already impaired, the firm will not recognize impairment and write off the assets. Thus, such cases of overinvestment will not be subsequently recognized in write-offs. Second, it can take years for a company to realize that some of their investments are overinvestments and the overinvestment in one year may be reflected in write-offs over several years in future. Thus write-offs in a particular year may only reflect a small fraction of the total overinvestment. 9

11 Goodwill write-offs Unlike tangible write-offs, which reflect previous overinvestment in tangible assets, particularly capital expenditures, goodwill write-offs reflect previous inefficient investment decisions in mergers or acquisitions, suggesting that the reporting firm may have over-paid for the target firm. Under SFAS 142 (FASB 2001a), Goodwill of a reporting unit shall be tested for impairment on an annual basis and between annual tests 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. The first step of the goodwill impairment test, used to identify potential impairment, compares the fair value of a reporting unit with its carrying amount, including goodwill. If the carrying amount of a reporting unit exceeds its fair value, the second step of the goodwill impairment test shall be performed to measure the amount of impairment loss, if any. The second step of the goodwill impairment test, used to measure the amount of impairment loss, compares the implied fair value of reporting unit goodwill with the carrying amount of that goodwill. 6 If the carrying amount of reporting unit goodwill exceeds the implied fair value of that goodwill, an impairment loss shall be recognized in an amount equal to that excess. Again, the goodwill write-offs are at best under-estimates of the firms overinvestment in the form of mergers and acquisitions. The above-mentioned reason related to the lag in time for recognizing write-offs applies to goodwill impairment as well. Further, as specified by SFAS 142, the loss recognized for goodwill impairment cannot exceed the carrying amount of goodwill. Thus, to the extent that the firm not only overpaid for goodwill, but also 6 The implied fair value of goodwill shall be determined in the same manner as the amount of goodwill recognized in a business combination is determined. 10

12 overpaid for the fair value of the underlying net assets, the goodwill write-offs will not reflect the total over-payment, although that part can potentially be recognized in subsequent tangible write-offs. To summarize, we believe that tangible write-offs and goodwill write-offs provide ex post revelation of firms overinvestment in either tangible assets or in mergers and acquisitions. Further, these write-offs only provide a lower boundary estimation of the degree of prior overinvestment. In this paper, we use the ex post write-offs to examine whether diversified firms are indeed more likely to over-invest than non-diversified firms Hypotheses Development Diversification and write-offs Our first hypothesis is based on theories that suggest that diversified firms will make less efficient investments (see Martin and Sayrak, 2003 for a good summary). First, diversified firms are less specialized, which suggest they may not be as efficient in making investment decisions. Second, managers may engage in empire building activities, growing the size of the firm to gain power and prestige, while not necessarily maximizing shareholder value (e.g., Jensen, 1986; Jensen and Murphy, 1990). Third, more diversified firms may have a less efficient internal capital market, resulting in an inefficient allocation of resources (e.g., Rajan, Servaes, and Zingales, 2000). Based on these theories we expect diversified firms to overinvest more frequently than undiversified firms. Therefore, we expect that diversified firms are more likely to incur write-offs than undiversified firms. Directly following this hypothesis, if diversification results in less efficient investments, we expect the inefficiency to rise as the firm becomes more diversified. For 11

13 example, if diversified firms are less specialized and less focused, then more business segments would reduce specialization further and result in even more inefficient investments. Hence, we expect investment efficiency to decline with the number of segments. In the context of our empirical study, we expect the likelihood of a write-off to increase with respect to the number of segments. Our first hypothesis can be stated as follows: H1a: More diversified firms are more likely to report higher tangible write-offs. While prior studies primarily focus on overinvestment in capital expenditures (e.g., Berger and Ofek, 1995), our setting allows us to separately identify overinvestment in mergers and acquisitions. We hypothesize that inefficient investment decisions would also result in inefficient mergers and acquisitions. We expect that more diversified firms are more likely to over pay for target firms. Therefore, we expect that more diversified firms are more likely to incur goodwill write-offs compared to more specialized firms. Accordingly, our corresponding second hypothesis is: H1b: More diversified firms are more likely to have higher goodwill write-offs. Write-offs and diversification discount If investors are rational they should expect diversified firms to be more likely to invest inefficiently. To the extent that investors correctly expect these firms to overinvest, they would value the diversified firms at a discount. In the context of our empirical study, we expect that the estimated diversification discount (e.g., Berger and Ofek, 1995) would be positively associated with future tangible as well as goodwill write-offs. In sum, if diversified firms are more likely to overinvest and investors are rational, we hypothesize that: H2a: The diversification discount is correlated with future tangible write-offs. H2b: The diversification discount is correlated with future goodwill write-offs. 12

14 Role of independent directors Our final hypothesis examines the role of independent directors, if any, in mitigating the overinvestment in diversified firms. The board has the legal authority to ratify and monitor managerial initiatives, evaluate the performance of top managers, and reward or penalize that performance (Byrd and Hickman 1992). One of boards important responsibilities is to monitor the firm by approving major corporate decisions such as mergers and acquisitions. Prior research has provided evidence that firms with a majority of independent directors make major corporate decisions in the best interests of shareholders (Masulis, Wang, and Xie, 2007). While independent directors may be able to mitigate overinvestment in general, their incentives to monitor are expected to be particularly strong for diversified firms because of the potential empire-building incentives of corporate managers. Bacon (1985) suggests that independent directors may be particularly adept at monitoring acquisition. The objectivity and business acumen of independent directors is particularly important in monitoring the acquisition process when managers empire-building ambitions or executive pride conflict with shareholders interests (Byrd and Hickman, 1992). Consistently, Byrd and Hickman (1992) find that bidding firms with an independent board have significantly higher announcement-date abnormal returns for their tender offers than other bidders. Therefore, to the extent that independent directors are able to curtail overinvestment, we expect the following: H3a: Independent directors mitigate the degree of tangible write-offs for diversified firms. H3b: Independent directors mitigate the degree of goodwill write-offs for diversified firms. 13

15 3. Sample and Descriptive Statistics Our sample period is from 2001 to 2006, as firms started to provide tangible or goodwill write-offs information in 2001 following SFAS 144 and SFAS 142. We obtain segment data from Compustat segment file following Burger and Ofek (1995). Specifically, we delete firmyears with any segment in the financial industry (SIC ) or with missing SIC code. We require that the segment have non-missing sales information in the segment file and that the total sales across all segments of a firm-year be within 1% of the total sales reported in the Compustat annual data to ensure the quality of the segment information. Further, to avoid undue influence of small firms, we require the firm to have total sales greater than $20million. This procedure yields a sample of 16,195 firm-years with non-missing write-off and segment data from 2001 to 2006, with the number of individual firms ranging from 2,312 in 2006 to 2,925 in Our multivariate analyses use samples of smaller sizes depending on the availability of control variables. We use two measures of diversification. DIVERSE i,t is a dummy variable that equals 1 if the firm operates in more than one business segment and 0 otherwise. NSEG i,t is the number of individual business segments with unique 4-digit SIC code and it captures the degree of diversification. We also have two measures for both tangible write-offs and goodwill writeoffs. Specifically, for tangible write-offs, D_TWO i,t is a dummy variable that equals 1 if the firm has positive tangible write-offs, 7 and 0 otherwise. TWO i,t, on the other hand, measures the magnitude of tangible write-offs relative to total assets and is calculated as tangible writeoffs divided by the sum of tangible write-offs and total assets. If the firm does not report positive write-offs, TWO i,t is set to zero. The goodwill write-off variables are measured 7 Compustat presents an impairment loss as a negative number. In this paper, we take the absolute value of the negative write-offs in measuring our write-off variables. 14

16 analogously: D_GWWO i,t is a dummy variable that indicates positive goodwill write-offs, and GWWO i,t, calculated as goodwill write-offs divided by the sum of goodwill write-offs and goodwill, measures the magnitude of goodwill write-offs relative total goodwill. If the firm does not report goodwill or goodwill write-offs, GWWO i,t is set to zero. To mitigate the influence of extreme observations, we winzorize TWO i,t and GWWO i,t at 99% of our sample. Table 1 presents the percentage of diversified firms and the percentage of firms with tangible write-offs and goodwill write-offs respectively over our sample period. Overall, 42% of our sample firm-years have more than one business segment, with the percentage slightly increasing from 2001 to On the other hand, about 16% of our sample firms have positive tangible write-offs and 5% have positive goodwill write-offs. We present in Table 2 the descriptive statistics of the write-off variables. Given that there are less than 50% of our sample firms report positive tangible or goodwill write-offs, the medians of all four variables are zero. The average TWO i,t is 0.003%, suggesting among all of our sample firms, the average tangible write-off is about 0.3% of total assets before the writeoff. This seems relatively small. On the other hand, the average goodwill write-off for the full sample is about 2.3% of the pre-write-off goodwill. Given that the majority of firms do not report write-offs, in Panel B, we present the descriptive statistics of the write-off variables for firms with positive tangible or goodwill write-offs only. Among these firms, the average tangible write-off is about 2.1% of total assets and median is about 0.8%. We also report the raw numbers in million dollars. On average, these firms report a tangible write-off in the amount of 17.3 million dollars, and the median is about 4.2 million dollars. Among firms that report positive goodwill write-offs, the average goodwill write-off is 41.8% of goodwill and the median is 34.1%, suggesting that 15

17 when firms do recognize goodwill write-offs, the magnitude is usually substantial relative to the recognized goodwill. The absolute value of these goodwill write-offs has a mean of 36.7 million dollars and a median of 12.3 million dollars, also appearing to be sizable. We also compare the frequency and magnitude of write-offs between firms with a single segment and firms with multiple segments. Results of the univariate analyses are presented in Table 3. We focus on both a dummy variable indicating whether the firm is multi-segment and hence diversified (DIVERSE i,t ) and the number of individual business segments (NSEG i,t ). Both of these variables are measured at the beginning of the year. Panel A presents the means and medians of the four write-off variables for single-segment (i.e., undiversified firms) and all multi-segment (i.e., diversified) firms separately. We first note that even though all these variables have medians of zero for the two groups of firms, the medians of diversified firms are all statistically different from those of undiversified firms based on Wilcoxon tests. The following discussion focuses on the means. Overall, 14.9% of undiversified firms report tangible write-offs while 17.5% of diversified firms report goodwill write-offs. The magnitude is on average about 0.3% of pre-write-off assets for both groups of firms. On the other hand, 4.4% of undiversified firms report goodwill write-offs, in comparison to 6.9% of diversified firms. The magnitudes are 2.2% and 2.4% of goodwill respectively. The differences are only significant for the indicator variables (i.e., D_TWO i,t and D_GWWO i,t ), but not for the magnitude variables (i.e., TWO i,t and GWWO i,t ). Panel B presents the same statistics of the four write-off variables by number of segments. The results are generally consistent with those presented in Panel A. The medians of multi-segment firms are mostly statistically different from those of single-segment firms based on Wilcoxon tests. As to means, for tangible assets, the percentages of firms reporting 16

18 write-offs (D_TWO i,t ) are similar for firms with one or two segments (14.9% and 14.4%), and the differences are not statistically significant. The percentages of firms reporting write-offs are significantly higher for firms with three or more segments (ranging between 19.5% and 23.4%) than for single-segment firms. The magnitude of the tangible write-offs (TWO i,t ), however, appear to be similar across different number of segments, measuring at about 0.3% of pre-impaired total assets and the differences from single-segment firms are generally insignificant. For goodwill write-offs, there is generally a clear pattern that the probability of reporting goodwill write-offs increases monotonically with the number of business segments of a company. For single-segment firms, about 4.4% of these firms write down goodwill. For multiple-segment firms, this percentage increases to 5.7% for two-segment firms, 7.6% for three-segment firms, and 9.5% for firms with at least five segments, and all these percentages are statistically different from that for single-segment firms. The magnitude measure of goodwill write-offs, GWWO i,t, generally shows a similar pattern, with the mean being 2.2% of pre-impairment goodwill for single-segment firms, and % for firms with between two and four segments. For firms with at least five segments, this variable becomes smaller at 1.4%. However, the difference in the magnitudes of goodwill write-offs is generally insignificant between single-segment firms and multi-segment firms. Overall, the univariate results presented in Table 3 provide preliminary support to our hypothesis by showing that diversified firms are significantly more likely to report tangible as well as goodwill write-offs. The effects of diversification on the average magnitude of these write-offs, on the other hand, are generally weaker and insignificant. 17

19 4. Diversification and Write-offs In this section we use a multivariate regression model to test whether the relation between diversification and write-offs (tangible write-offs as well as goodwill write-offs) shown in Table 3 are robust to different control variables, such as the book-to-market ratio, size, and capital expenditure. As specified in Hypothesis 1, we test whether diversified firms are more likely to incur a write-off as well as whether diversified firms incur larger write-offs. Our tests employ both a dummy variable indicating a diversified firm as well as the number of individual business segments. Our control variables include the book-to-market ratio (BM i,t ) as a proxy for growth and investment opportunity. We also control for size using the natural log of total assets (LOGTA i,t ). We expect that firms with poor performance are more likely to report asset writeoffs. Therefore, to control for operating performance, we include earnings increase, calculated as change in earnings deflated by beginning period total assets (EG i,t ), and operating margin, calculated as earnings deflated by revenue (MARGIN i,t ). We also include an indicator variable that receives the value of 1 if the earnings change is negative and zero otherwise (DEC i,t ) and an indicator variable that receives the value of 1 if the earnings is negative and zero otherwise (LOSS i,t ) 8. In addition to operating performance, we expect that stock performance, measured as cumulative abnormal returns (CAR i,t ), is also negatively correlated with the probability of reporting asset write-offs. Finally, since write-offs are more likely to occur in firms that invest more, we control for capital expenditures (CAPX i,t ), which should also capture growth potential of a firm. All stock control variables (BM i,t and LOGTA i,t ) are measured at the beginning of the year while all flow control variables are measured over the current year. 8 Hayn (1995) and Basu (1997) suggest that firms reporting losses are more likely to report asset write-offs. 18

20 In addition to these control variables, we include year dummies as well to control for variations in write-offs across our sample period. 9 For all of our regressions, we adjust for the correlation among observations by clustering the standard errors at the firm level. Specifically we employ the following regression model: Dep Var i,t =y i year dummies + b 1 Diversification Variable t-1 + b 2 BM t-1 + b 3 LOGTA t-1 + b 4 EG t + b 5 MARGIN t + b 6 DEC t + b 7 LOSS t + b 8 CAR t + b 9 CAPX t + i,t (1) The dependent variables are our measures of either tangible or goodwill write-offs. In estimating the regression, there are two alternative diversification variables: DIVERSE i,t-1 indicating diversified firms and NSEG i,t-1 for number of business segments; both are measured at the beginning of the year. Diversification and tangible write-offs We begin our regression analysis by testing whether diversified firms are more likely to incur a tangible write-off. We estimate two versions Equation (1) with two different dependent variables. In the first version, the dependent variable is D_TWO i,t, the dummy variable indicating tangible write-offs. We use logit estimates in this regression. The results are reported in Table 4 Panel A. Our findings are consistent with diversification discount as diversified firms are more likely to incur tangible write-offs. The coefficient on DIVERSE t-1 is positive (0.116) and statistically significant at the 5% level. In addition, the probability of a write-off increases with the number of segments. The coefficient on NSEG t-1 is positive (0.077) and statistically significant at the 1% level. In Panel B of Table 4 we report the second version of Equation (1), where TWO i,t is the dependent variable. We use OLS estimates for this regression. The coefficient on 9 For example, Jorgensen, Li and Sadka (2008) show that there was a significant increase in goodwill write-offs in the period of with the introduction of SFAS

21 DIVERSE t-1 is insignificant, suggesting that while diversified firms are more likely to incur a write-off (as shown in Panel A), the magnitude of write-offs does not vary significantly between diversified firms and undiversified firms. The coefficient on NSEG t-1, on the other hand, is statistically significant at the 5% level, probably because the effects of more diversified firms (e.g., firms with at least three business segments). However, the coefficient does not appear economically significant; for every additional business segment, the average write-off for a firm goes up by only about 0.023% higher with respect to total assets. Considering that the average write-off at undiversified firms is about 0.335% of total assets, this result suggests that a diversified firms with three business segments on average report about 14% (0.023x2/0.335) higher of write-offs than undiversified firms. Diversification and goodwill write-offs We proceed to study whether diversified firms are more likely to incur a goodwill write-off. Table 5 Panel A reports logit estimates for Equation (1) using D_GWWO i,t as the dependent variable and Panel B reports OLS estimation using GWWO i,t as the dependent variable. In Panel A, our findings are consistent with the hypothesis that diversified firms are more likely to engage in inefficient acquisitions resulting in a higher likelihood of incurring a goodwill write-off. The coefficient on DIVERSE t-1 is positive (0.467) and statistically significant at the 1% level. Consistently, the coefficient on NSEG t-1 is positive (0.191) and statistically significant at the 1% level. These findings suggest that diversified firms are more likely to incur a goodwill write-off and that this likelihood increases with the number of segments. In contrast to tangible write-offs, our findings in Panel B suggest that diversified firms incur larger goodwill write-offs than their undiversified counterparts. The coefficient on 20

22 DIVERSE t-1 is positive (0.475) and statistically significant. The coefficient on NSEG t-1 is positive (0.152) and significant at the 10% level. These results suggest that the magnitude of goodwill write-offs is greater for diversified firms than undiversified firms and that this magnitude generally increases as firms become more diversified. 10 In sum, our findings in Table 4 and Table 5 provide evidence that the likelihood of asset (tangible or goodwill) write-offs is significantly related to the degree of diversification and that the magnitude of asset write-offs also increases as the firm gets more diversified. These results are consistent with the theories that diversified firms are more likely to overinvest and that diversification destroys value. 5. Diversification Discount and Write-offs In this section, we examine how the diversification discount documented in prior research (e.g., Berger and Ofek, 1995) is related to tangible write-offs as well as goodwill write-offs. While the results in Section 4 already provide evidence that diversified firms are more likely to overinvest than undiversified firms, linking the write-offs to diversification discount as measured in prior research would lend support to the existence of the diversification discount and the investment inefficiency being an explanation for the diversification discount. We first examine whether there is indeed a diversification discount for our sample firms. We measure the excess values of our sample firms relative to the industry medians following Berger and Ofek (1995). Specifically, for each firm in our sample, we calculate its 10 As a sensitivity test, in examining goodwill write-offs, we also add into Equation (1) the ratio between goodwill and total asset at the beginning of the year to control for the existence and magnitude of the goodwill. Our results are robust. 21

23 total capital as the book value of debt plus the market value of equity. We estimate each segment of multi-segment firms capital by multiplying its sales with the median capital-sales ratio among all single-segment firms in our sample with the same four-digit SIC code. We then calculate the imputed value of the multi-segment firm by adding this estimated capital across all of its segments. The excess value of the firm is calculated as the log of the ratio between the firm s capital and its imputed value. A positive excess value suggests a premium while a negative excess value suggests a discount. Table 6 presents the descriptive statistics of the excess value at the beginning of our sample years (EXCESS i,t-1 ) for one-segment and diversified firms separately. We show the excess value at the beginning of each year of our sample periods. For the undiversified (i.e., one-segment) firms, the average excess value is -4.3% and the median, as expected, is 0. The distribution is relatively symmetric around zero. In contrast, the average excess value for the diversified firms is -22.5%, and the median is -20.7%. The distribution is considerably negatively skewed. Both the differences in means and in medians between undiversified and diversified firms are statistically significant, and the magnitude is similar to prior research (e.g., 13-15% in Berger and Ofek, 1995). Thus, we are able to replicate the existence of diversification discount shown in prior research for our more recent sample. That is, diversified firms are valued at a discount of about 18-20% relative to the imputed values based on their one-segment counterparts. We next examine whether the diversification discount measured at the beginning of the year is correlated to the subsequent tangible and goodwill write-offs. Specifically, we employ the following regression model, following Berger and Ofek (1995): EXCESS i,t-1 =y i year dummies + g 1 Write-Off i,t + g 2 LOGTA t-1 22

24 + g 3 MARGIN i,t-1 + g 4 CAPX i,t-1 +µ i,t (2) We use four alternative measures of write-offs, namely, the indicator and magnitude measures of tangible and goodwill write-offs respectively. Panel A of Table 7 reports the results for tangible write-offs and Paned B reports results for goodwill write-offs. Our findings suggest that diversified firms are trading at a discount in part because investors expect these firms to incur write-offs in the near future. In other words, diversified firms are traded at a discount because investors expect diversified firms to make more inefficient investments. For example, the coefficients on the indicator variables for tangible and goodwill write-offs are both negative ( and , respectively) and statistically significant. These coefficients suggest that the incidence of the tangible and goodwill writeoffs can explain about 5.8% and 8.4% of the roughly 20% diversification discount, which appears economically significant. Consistently, a larger future write-off is associated with a lower valuation. The coefficient on TWO i,t and GWWO i,t are both negative and statistically significant. Overall, we are able to show that both the incidence and the magnitude of subsequent write-offs are significantly correlated with the diversification discount. These results are consistent with the existence of diversification discount and also provide support to prior research that attributes the discount to inefficient investment. 6. Diversification and Write-offs: The Role of Independent Directors Finally, we test whether independent directors can mitigate overinvestment in diversified firms. Our regression model is similar to Equation (1), but with two additional variables. First, we include the percentage of independent directors (IND i,t-1 ) measured at the beginning of the 23

25 year. In addition, we include an interaction term of IND i,t-1 and DIVERSE i,t-1. If independent directors improve the efficiency of investment at the diversified firms, we expect the magnitude of their asset write-off to decline. In terms of our regression model, we expect the coefficient on the interaction term described above to be negative. The dependent variables for tangible and goodwill write-offs are TWO i,t and GWWO i,t, respectively, capturing the magnitude of the write-offs. Table 8 presents the results of our estimation. The coefficient on the interaction term on diversification and percentage of independent directors is insignificant when the dependent variable is tangible write-offs (TWO i,t ) but is significantly negative when the dependent variable is goodwill write-offs (GWWO i,t ). These results suggest that independent directors are generally unable to mitigate overinvestment in tangible assets, but can significantly curtail overinvestment in mergers or acquisitions. In addition, the effects of independent directors also appear to be substantial in magnitude in Panel B where the dependent variable is GWWO i,t. Specifically, the coefficient on DIVERSE i,t-1 is and the coefficient on the interaction term is These coefficients suggest that a 10% increase in the percentage of independent directors translates into approximately a 13% decline in coefficient on DIVERSE i,t-1. We also estimate the logit version of this regression with dependent variable being D_TWO t and D_GWWO t respectively. The results are generally weak. Overall, our results suggest that while the role of an independent board in mitigating overinvestment is relatively small, an independent board does significantly decrease the extent of overinvestment in acquisitions or mergers by diversified firms. 24

26 7. Conclusion This paper studies whether diversified firms are more likely to overinvest. Our proxy for overinvestment is an ex post measure future write-offs of tangible assets and goodwill. Our findings are consistent with the implications of the diversification discount insofar as diversified firms are more likely to report write-offs (both tangible and goodwill). In addition, our findings suggest that diversified firms trade at a discount partly because investors correctly anticipate future write-offs. Finally, we document that independent directors can significantly reduce the extent of diversified firms to overinvest through mergers and acquisitions. 25

27 References Agrawal, A., Jaffe, J.F., Mandelker, G.N., 1992, The post-merger performance of acquiring firms: A re-examination of an anomaly, Journal of Finance, 47, Amihud, Yakov, and Baruch Lev, 1981, Risk reduction as a managerial motive for conglomerate mergers, Bell Journal of Economics, 12, Bacon. Jeremy, 1985, The role of outside directors in major acquisitions and sales, The Conference Board Research Bulletin, 180. Basu, Sudipta, 1997, The conservatism principle and the asymmetric timeliness of earnings, Journal of Accounting and Economics, 24, Berger, Philip G., and Eli Ofek, 1995, Diversification s effect on firm value, Journal of Financial Economics, 37, Berger, Philip G., and Eli Ofek, 1999, Causes and effects of corporate refocusing programs, Review of Financial Studies, 12, Byrd, John W., and Kent A. Hickman, 1992, Do outside directors monitor managers? Evidence from tender offer bids, Journal of Financial Economics, 32, Campa, Jose Manuel and Simi Kedia, 2002, Explaining the diversification discount, The Journal of Finance, 57, Chevalier, Judith, 2002, What do we know about cross-subsidization? Evidence from merging firms, Advances in Economic Analysis & Policy, 4 (Issue 1 Art. 3). Comment, Robert and Gregg A. Jarrell, 1995, Corporate focus and stock returns, Journal of Financial Economics, 37, Financial Accounting Standards Board, 2001a, SFAS 142 Goodwill and other intangible assets. Financial Accounting Standards Board, 2001b, SFAS 144 Accounting for the impairment or disposal of long-lived assets. Francis, Jennifer, J. Douglas Hanna, and Linda Vincent, 1996, Causes and effects of discretionary asset write-offs, Journal of Accounting Research, 34, Graham, John R., Michael L. Lemmon, and Jack G. Wolf, 2002, Does corporate diversification destroy value? The Journal of Finance, 57,