Dismantling internal capital markets via spinoff: effects on capital allocation efficiency and firm valuation

Journal of Corporate Finance 11 (2005) 253 275 www.elsevier.com/locate/econbase Dismantling internal capital markets via spinoff: effects on capital allocation efficiency and firm valuation Chris R. McNeil a, William T. Moore b, * a Black School of Business, Penn State Erie, Erie, PA 16563, USA b Moore School of Business, University of South Carolina, Columbia, SC 29208, USA Received 1 January 2002; received in revised form 1 April 2003; accepted 1 June 2003 Available online 27 August 2003 Abstract We investigate the linkage between changes in firm value and changes in capital allocation efficiency resulting from dismantling internal capital markets via spinoffs. We find no evidence of wholesale misallocation of capital pre-spinoff. On the average, excess value increases following spinoffs. Furthermore, changes in excess value are positively linked to changes in capital allocational efficiency following spinoff. We find that spinoff announcement returns are greater (smaller) when the parent allocates capital to the unit to be spun off in a seemingly less (more) efficient manner. Divested division capital expenditures move toward industry levels after spinoff, regardless of their relative investment opportunities. D 2005 Elsevier B.V. All rights reserved. JEL classification: G31; G34 Keywords: Capital budgeting; Restructuring 1. Introduction Spinoffs and other forms of divestiture dismantle existing internal capital markets; that is, the agents and mechanisms within a firm that influence the intra-firm allocation of capital and the monitoring of capital productivity. There are opposing views regarding the influence of these markets on firm value. Potential benefits include more effective monitoring, alleviation of capital budgeting constraints, and winner picking (see, * Corresponding author. Tel.: +1-803-777-4905; fax: +1-803-777-6876. E-mail address: aardvark@moore.sc.edu (W.T. Moore). 0929-1199/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/s0929-1199(03)00060-9

254 C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 e.g., Alchian, 1969; Stein, 1997). Alternatively, internal capital markets may reduce incentives for division managers to exert effort, allow managers of poorly performing divisions to expropriate rents from other divisions or otherwise misallocate capital by subsidizing poorly performing divisions(e.g., Meyer et al., 1992; Scharfstein and Stein, 2000). 1 We investigate the effects of dismantling internal capital markets via spinoff on the efficiency of capital allocation and firm value, and we hope to add a few more findings to what is quickly becoming a vast body of knowledge. Gertner et al. (2002) find that the association between capital expenditure levels and investment opportunities of divested divisions strengthens post-spinoff when the parent and divested division operate in unrelated industries and, furthermore, that this effect is more evident in the case of spinoffs with higher announcement returns. Burch and Nanda (in press) document that post-spinoff improvement in excess value (EV) is greater when the spinoff reduces the diversity of investment opportunities among the parent s business segments. Ahn and Denis (in press) document a positive relationship between change in excess value and change in investment efficiency. Our study contributes to the literature on spinoffs and internal capital markets in the following ways. We begin by examining divested divisions to determine whether their capital investment behavior changes. There is little evidence of systemic misallocation of capital to these divisions before spinoff. On the average, relative to their respective industries, these divisions spend about the same post-spinoff. However, those that were receiving capital subsidies while housed in parent firms decrease spending, while those that were being rationed by parents tend to increase expenditures, regardless of relative investment opportunities. Both efficient investment patterns (i.e., subsidizing high q and rationing low q divested divisions) and inefficient investment patterns (i.e., rationing high q and subsidizing low q divested divisions) tend to go away after spinoff. We then measure changes in the allocational efficiency (AE) of capital expenditures before and after spinoffs at the combined firm level (i.e., parent plus divested division). In some cases, we find major improvements, and in some, we observe deterioration, but the average change is near zero. However, changes in allocational efficiency appear to be linked to pre-spinoff levels of efficiency. That is, firms that we categorize as inefficient before the spinoff appear to become more efficient in the aftermath. Our data also reveal that firms judged to be efficient allocators pre-spinoff tend to become less efficient afterward, thus, some spinoffs result in deterioration of allocational efficiency. Next, we turn to effects on firm value first measured by changes in excess value, where excess value is the difference in market value of the combined firm and a portfolio of stand-alone firms operating in the same segments (Berger and Ofek, 1995). We find that median changes in excess value are positive (as do Burch and Nanda, in press), and that they are positively linked to changes in allocational efficiency of capital expenditures (consistent with Ahn and Denis, in press). Finally, we find that spinoff announcement returns, a second measure of change in value, are related to the parent s policy of investment in the division to be divested. We 1 Additional theoretical developments concerning internal capital markets include Williamson (1981), Gertner et al. (1994), Rajan et al. (2000) and Maksimovic and Phillips (2002).

C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 255 find that breaking up internal capital markets affects firm value in the anticipated ways; that is, announcement returns are greater (smaller) when the parent allocates capital to the unit to be spun off in a seemingly less (more) efficient manner. Taken together, our results indicate that parents show no general tendency to misallocate capital before spinoff. For those that allocate capital unwisely before spinoff, dismantling internal capital markets has positive value effects on average, as expected. However, it also appears that many parents were creating value through their allocation of capital before spinoff. For these parents, breaking up internal capital markets that function efficiently results in less favorable value consequences, controlling for other effects of spinoffs. We acknowledge the bias arising from firms self-selecting into our spinoff sample. This bias does not disturb evidence linking value effects and pre-spinoff capital investment policies. However, the fact that firms choose to divest units via spinoff does mean that those with well functioning internal capital markets are less likely to appear in the sample, while those with poorly functioning internal capital markets are more likely to appear. This study proceeds as follows. In Section 2, we provide additional background on internal capital markets and spinoffs. Section 3 contains a description of our sample. Sections 4, 5 and 6 contain analyses of capital expenditures, allocational efficiency, and value effects, respectively. In Section 7, we present concluding remarks. 2. Background 2.1. Internal capital markets The extant literature presents mixed evidence concerning the influence of internal capital markets on firm value. For example, Berger and Ofek (1995) and Denis and Thothadri (1997) provide evidence linking inefficient investment to the diversification discount. Shin and Stulz (1998) find that divisions (of conglomerates) are treated alike when they should not be (p. 533). 2 Scharfstein (1998) shows that high (low) Tobin s q manufacturing divisions of diversified firms tend to invest less (more) than their standalone industry peers, and that this seemingly inefficient investment behavior is more pronounced when management ownership holdings are small. Rajan et al. (2000) document that capital allocation efficiency declines as the diversity of investment opportunities among a firm s business segments increases, and that greater diversity of investment opportunities is associated with lower firm value. Another group of studies, however, supplies evidence of efficient investment by multidivisional firms. Hubbard and Palia (1999) and Billet and Mauer (in press) find that an internal capital market enhances value when it overcomes financial constraints. Hubbard and Palia (1999) document that acquisition announcement returns for acquirers in the 2 In their sample of diversified firms, Shin and Stulz (1998) find that investment in the smallest segment is no less sensitive to other segment cash flows when the smallest segment has the best prospects relative to the other segments in the firm.

256 C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 1960s were greater when financially unconstrained bidders acquired financially constrained targets. Similarly, Billett and Mauer (in press) show that capital subsidies are associated with greater excess value when the subsidy is provided to segments that would likely be capital constrained as stand-alone firms. 3 Using plant-level data and a measure of investment opportunities other than Tobin s q, Maksimovic and Phillips (2002) provide evidence that multi-divisional firms allocate capital in an efficient manner. Finally, a third group of studies questions the methodologies employed in some works that find evidence of internal capital market inefficiency. Chevalier (2000) and Whited (2001) propose that using investment opportunities of stand-alone firms within an industry as proxies for business segment investment opportunities produces measurement error that can lead to biased results. Both of these studies find that evidence of internal capital market inefficiency is diminished when they employ a proxy other than stand-alone industry q for investment opportunities. 4 Thus, the question of how efficiently multi-divisional firms are able to utilize their internal capital markets is a matter of much debate. Spinoffs offer a useful setting in which to explore this question further. A spinoff occurs when a firm (i.e., parent) distributes to its existing shareholders on a pro-rata basis the ownership of a new firm or firms (i.e., divested division(s)), created from a portion of the parent s assets. The spinoff distribution represents a partial dismantlement of the parent firm s internal capital market in that after the spinoff the divested division determines its level of capital spending independent of its former parent. We can observe the capital expenditure levels for the divested division and the parent s other operations before and after this partial dismantlement. Furthermore, a spinoff distribution does not directly change firm level capital resources since spinoffs are non-cash transactions, unlike many other forms of divestiture. 2.2. Spinoffs and internal capital markets Positive average wealth effects associated with spinoff announcements are well documented. Cumulative average abnormal returns reported in prior studies range from 2.6% to 3.6%, conducted over various sampling periods and using different event windows. 5 A number of explanations have been offered for this wealth effect, including, for example, increased focus and the reduction of negative synergies, wealth transfer and the reduction of asymmetric information. 6 Miles and Rosenfeld (1983) and Daley et al. (1997) discuss the possible connection between parent firm investment policies and valuation effects of spinoffs. Miles and 3 Billet and Mauer also find, however, that the transfer of capital away from segments with better investment opportunities is associated with lower excess value. 4 In the related literature concerning the conglomerate discount, Graham et al. (2002) and Campa and Kedia (2002) provide evidence that it is not appropriate to benchmark business segment value against that of stand-alone firms. 5 See, for example, Hite and Owers (1983), Schipper and Smith (1983), Daley et al. (1997), Krishnaswami and Subramanian (1999) and Gertner et al. (2002). 6 Other explanations that have generally received less attention include: greater market completeness (Miles and Rosenfeld, 1983); removal of legal, regulatory or tax constraints (Schipper and Smith, 1983); divergence of opinion (Kudla and McInish, 1988); capital structure (John, 1993); and tax timing (Mauer and Lewellen, 1990).

C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 257 Rosenfeld suggest that managers of parent firms may be unable to replicate the role of financial markets (p. 1598), in which case spinoffs would improve the allocation of capital. Daley et al. (1997) posit that spinoffs create value by bonding managers to limit cross-subsidization of poorly performing operations. They note that divestiture serves as the ultimate bond not to subsidize a division. Daley et al. (1997) find no evidence that supports this bonding argument and also note that capital expenditure levels do not change post-spinoff at the combined firm level (i.e., comparing capital expenditures of the parent pre-spinoff to the combined capital expenditures of the parent and divested division post-spinoff). Focusing on the divested divisions, however, Gertner et al. (2002) find that the association between capital expenditure levels and investment opportunities for divested divisions strengthens post-spinoff when the parent and divested division operate in unrelated industries, and that this effect is more evident in the case of spinoffs with greater announcement returns. Burch and Nanda (in press) show that excess value increases when a spinoff reduces the diversity of investment opportunities among a parent s business segments, consistent with the model of Rajan et al. (2000). Ahn and Denis (in press) find that change in excess value following spinoff is positively associated with change in measures of investment efficiency and also provide evidence of underinvestment in higher q segments pre-spinoff. Billett and Mauer (2000) document a positive association between tracking stock announcement returns and internal capital market value, indicating that the market reacts more (less) favorably to the sustainment of a well (poorly) functioning internal capital market. In this study, we seek to add additional findings about internal capital markets and spinoffs. We examine capital allocation efficiency before spinoff, changes therein after spinoff and the association between allocational efficiency and spinoff value effects. First, capital spending in the divested divisions is examined. Spending in these units is likely to experience the greatest effect from dismantling the internal capital market. We look for evidence of efficient/inefficient investment pre-spinoff by examining patterns in capital spending in the divested divisions before and after divestiture. Second, we examine measures of investment efficiency for the parent firms, using a metric from Rajan et al. (2000), and, separately, for the divested divisions. Finally, we examine whether spinoff value effects are related to capital allocation efficiency. 3. The sample Our sample consists of a group of firms (parents) that spin off a portion of their assets (divested divisions) to form new, independent firms over the period 1980 through 1996. The initial list of firms comes from Security Data Corporation (SDC) and the Center for Research in Securities Prices (CRSP). 7 The final sample consists of 158 divisions, 153 spinoff events (on five occasions, a parent in one announcement commits to spin off two 7 CRSP denotes spinoffs distributions in their dividend file with one of the following codes: 3762, 3763, 3764 and 3765.

258 C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 divisions) and 139 parents (14 parents announce spinoffs on two separate occasions) after applying the following filters: (1) the parent owns at least 95% of the division prior to the spinoff; (2) the parent retains no more than 5% of the division post-spinoff; (3) the CRSP data base contains data for the parent; (4) the Compustat data base contains data for the parent and the division; (5) the division s principal line of business does not include the financial or insurance sectors; (6) the division is not divested as a limited partnership or royalty trust; (7) the spinoff is not the product of a merger or acquisition; and (8) the parent is headquartered in the US. The spinoff distributions are widely dispersed across the 17-year sample period. The greatest frequency of spinoffs contained in the sample occurs in 1996 (11% of the total sample). The primary lines of business of the parents and divisions cover 41 and 43 industries, respectively, as classified at the two-digit SIC code level. There is little clustering across industries with only one industry accounting for 11.1% and 9.49% of the sample for parents and divisions, respectively. Table 1 contains summary statistics for the parents and divested divisions. Parents are typically large firms with mean (median) book values of assets over US$4.3 billion (US$650 million) and sales of over US$3.9 billion (US$1 billion). On the average (at the median), the divisions are also sizeable operations with book value of assets of nearly US$590 million (US$156 million) and annual sales of over US$688 million (US$176 million). Table 1 Descriptive statistics Mean Median Minimum Maximum Panel A. Parents Total assets in book value (US$ millions) 4374.27 650.81 7.23 198,598.00 Annual sales (US$ millions) 3950.66 1002.39 5.68 152,172.00 Tobin s q 1.69 1.40 0.68 5.39 Panel B. Divested divisions Total assets in book value (US$ millions) 589.26 156.70 0.04 8786.50 Annual sales(us$ millions) 688.76 176.45 0.23 9960.10 Tobin s q 1.63 1.25 0.15 9.75 Proportion of assets divested (book value) 0.25 0.23 0.01 0.88 Division q parent q 0.06 0.19 4.50 6.95 Panel A contains descriptive statistics measured pre-spinoff for a sample of parent firms that complete a spinoff from 1980 through 1996. Panel B provides statistics for the divested divisions. The sample consists of 153 spinoff events involving 158 divisions (on five occasions, a parent in one announcement commits to spin off two divisions), and 139 parents (14 parents announce spinoffs on two separate occasions). The descriptive statistics incorporate parent data for each spinoff announcement. Tobin s q after spinoff is calculated as the ratio of market value of equity plus book value of assets minus book value of common equity to book value of assets.

C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 259 Parents typically divest a sizeable portion of their total assets. The mean (median) ratio of book value of divested assets to book value of parent assets is 25% (23%). On the average, Tobin s q for divested divisions is slightly lower than that of parent firms. 4. Capital expenditures of divested divisions We begin our analysis by examining industry-adjusted capital expenditures of the divested divisions before and after spinoff. The results indicate that parents materially influence capital investment in their divested divisions pre-spinoff. However, there is little evidence of endemic inefficient investment. Cases of inefficient investment are predominantly those of subsidizing low-q divisions. In examining divested division investment, we calculate averages pre-spinoff (over years 3, 2) and post-spinoff (over years + 1, + 2) of the divested division s industryadjusted ratio of capital expenditures to beginning assets (ICAPRATE). 8 An average value has the desirable properties of capturing investment levels over time and reducing the influence of lumpy capital expenditures. We exclude year 1 to reduce the possibility that our measures reflect a change in investment in anticipation of the spinoff. DICAPRATE denotes the average post-spinoff measure of industry-adjusted capital expenditures minus the average pre-spinoff industry-adjusted capital expenditures for a divested division. We report both mean and median values of pre-spinoff, post-spinoff and changes in industryadjusted capital expenditures in Table 2, but we focus on median values because there is some skewness in the data. Before the spinoff, capital investment in the full sample of divested divisions is statistically indistinguishable from industry levels. The median pre-spinoff level of industry-adjusted capital expenditures is 0.0089 (as shown in the first column), not statistically different from 0. After the spinoff, when the divested divisions are operating independently of their former parents, there is some statistical evidence that the divested divisions invest less than industry levels. The median post-spinoff industry-adjusted capital expenditure is 0.0063, significant at the 5% level, indicating that the divested divisions spend about 8.6% less than the average of stand-alone firms operating in the same industry (computed as 0.0063/0.07314 where 0.07314 = median for the industries for the same time periods). However, average DICAPRATE does not statistically differ from zero for the full sample (median = 0.0049). We also see that of the 131 divested divisions, 63 or roughly half (48.1%) experience an increase in their industry-adjusted ratio of capital expenditures. Therefore, for the divested divisions, there is no significant across-the-board increase or decrease in capital investment associated with spinoff. When we look more carefully at capital expenditures in conjunction with divested division investment opportunities as measured by q, an interesting pattern emerges. 8 As in Rajan et al. (2000), beginning assets equal book value of assets minus capital expenditures plus depreciation, and the industry capital expenditure ratio is the asset weighted average of stand-alone firms at the three digit SIC level. When there are fewer than five stand-alone firms at the three-digit SIC level, we move to broader SIC levels to compute an industry average. We define a stand-alone firm as a firm that reports at least 95% of its sales in one segment.

260 C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 Table 2 Industry-adjusted capital expenditures for divested divisions Category Full sample (1) High q, subsidized (2) Low q, rationed (3) High q, rationed (4) Low q, subsidized Pre-spinoff ICAPRATE Mean 0.0405 0.1194** 0.0462*** 0.0399*** 0.1882** Median 0.0089 0.0666*** 0.0361*** 0.0210*** 0.0581*** Post-spinoff ICAPRATE Mean 0.0284 0.2377 0.0163* 0.0104 0.0144 Median 0.0063** 0.0020 0.0154*** 0.0118 0.0034 Change (DICAPRATE) Mean 0.0121 0.1183 0.0299*** 0.0503** 0.1738** Median 0.0049 0.0677*** 0.0190*** 0.0341** 0.0496*** No. change > 0 63 2 43 12 6 Proportion change > 0 (%) 48.1 11.8 69.4 75.0 16.7 N 131 17 62 16 36 Industry-adjusted ratio of capital expenditures to beginning assets (ICAPRATE) for divested divisions before and after spinoff. Industry adjustments are made using the asset weighted average of stand-alone firms, where industries are defined at the three-digit SIC level when at least five industry firms are available. Pre-spinoff ICAPRATE is the average of years 3, 2 relative to spinoff. Post-spinoff ICAPRATE is the average of years + 1, + 2 relative to spinoff. DICAPRATE = post-spinoff ICAPRATE minus pre-spinoff ICAPRATE. In the columns labeled (1) (4): high (low) q indicates that division q > ( < ) industry q; and subsidized (rationed) indicates that pre-spinoff ICAPRATE > ( < ) 0. Asterisks indicate the significance from t tests (Wilcoxon Rank Sum tests) for whether the mean (median) differs from 0. * Significant at the 0.10 level. ** Significant at the 0.05 level. *** Significant at the 0.01 level. Combinations of pre-spinoff capital subsidy/rationing and divested division q constitute four alternative policies of investment in divested divisions followed by a parent: (1) relatively high q, capital subsidy; (2) low q, capital rationing; (3) high q, capital rationing; and (4) low q, capital subsidy. 9 Policies (1) and (2) are deemed efficient while (3) and (4) are inefficient. To distinguish between high q and low q and between capital subsidy and capital rationing, divested division q and capital expenditures are compared against industry averages. Divested division q is calculated at the earliest possible date after spinoff, with industry q measured as of the same year. We note that our classifications are based only on whether capital flows in a correct direction relative to industry levels (i.e., to or from a divested division based on its relative investment opportunities), because optimal levels of investment are unobservable. In Table 2, we present industry-adjusted capital expenditures for each investment policy in columns labeled (1) through (4). Of the 131 divested divisions for which we have sufficient data, the majority are low q relative to their respective industries (98, about 75% of the sample). Of these 98, 62 (63%) 9 We calculate q as the market value of equity minus book value of equity plus book value of assets, all scaled by book value of assets. Whited (2001), Chevalier (2000) and Gertner et al. (2002), among others, construct q in a similar fashion. Chung and Pruitt (1994) show that a similar estimate of q explains over 95% of the Lindenberg and Ross (1991) q.

C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 261 experience capital rationing pre-spinoff. Of the high-q divisions, roughly half (17 of 33) receive capital subsidies. Thus, wholesale inefficient investment by the parents is not evident. Roughly 60% of the parents (computed as (17+ 62)/131) allocate capital to their divested divisions in a seemingly efficient manner when we benchmark divested divisions against their industry averages. The predominant form of inefficient investment is subsidizing low-q divisions (36 divisions are low q and subsidized, 16 divisions are high q and rationed). 10 We see that subsidized (rationed) divisions experience a statistically significant and material decrease (increase) in industry-adjusted capital expenditures following spinoff, regardless of their relative q. For each policy category, divested division capital investment moves toward the industry average and away from the pre-spinoff pattern of investment. Pre-spinoff, each policy category has investment levels that differ from the respective industry average (by construction), while post-spinoff, only policy category (2) expenditures differ from the industry. Median DICAPRATE is significant for each of the four categories of investment policy, indicating that there is a substantial change in capital investment in divested divisions following spinoff. Both efficient and inefficient investment patterns tend to dissipate after spinoff. Thus, spinoffs appear to vary in their impact on investment efficiency for the divested divisions. Looking more closely at each policy category provides additional insight. Divested divisions in category (1) received substantial capital subsidies pre-spinoff as evidenced by a median industry-adjusted level of 0.0666, which compares to a median industry level of 0.0778 for the same industries. Thus, category (1) divisions have investment levels that are about 86% higher than the industry (0.0666/0.0778) before spinoff. After spinoff, capital investment falls to industry levels, even though the divisions have higher q than the industry average. The post-spinoff median industry-adjusted capital expenditures are not statistically different from zero, and median DICAPRATE is 0.0677 (significant at the 1% level), approximately equal to the median level of pre-spinoff subsidy. These divisions experience the greatest absolute median change in ICAPRATE of the four categories. Following spinoff, only 2 of these 17 divisions (11.8%) increase capital spending on an industry-adjusted basis. The category (2) divisions are rationed both before and after the spinoff, however, the rationing is less severe afterward (pre-spinoff median = 0.0361, post-spinoff median = 0.0154, both statistically different from 0). Median DICAPRATE is 0.0190, significant at the 1% level, representing a 22.9% increase relative to median industry levels (0.0190/0.0828). Following spinoff, 43 of the 62 divisions (69.4%) experience an increase in industry-adjusted capital investment. Category (3) divisions, which were rationed pre-spinoff even though they have high q, experience an increase in industry-adjusted capital spending following spinoff, as expected. Median DICAPRATE equals 0.0341, representing a 44.9% increase relative to industry levels (0.0341/0.0760). Of the 16 divisions, 12 (75%) show an increase in 10 Similar patterns of investment policy are documented when we classify investment policy by comparing divested division q and capital expenditures against parent q and industry capital expenditures adjusted for the parent s propensity to invest.

262 C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 industry-adjusted expenditures. However, the increased spending only brings these divisions back to the industry norm (post-spinoff median ICAPRATE = 0.0118, not statistically different from zero). Finally, category (4) divisions (low q but subsidized) experience a decrease in industryadjusted capital investment following spinoff such that their post-spinoff investment returns to industry levels. Median DICAPRATE is 0.0496, representing a 66.9% decrease compared to the industry median ( 0.0496/0.0741). Post-spinoff median ICAPRATE is 0.0034, not statistically different from 0, even though these divisions have low q. Of the 36 divisions, 30 (83.3%) experience a decrease in industry-adjusted capital spending. These results contrast somewhat with those of Ahn and Denis (in press). They find evidence of under-investment in high q segments prior to spinoff, while we find that the predominant source of inefficient investment is over-investment in low q divested divisions, but little evidence of wholesale misallocation of capital to these divisions in the pre-spinoff period. The differences in these findings can be at least partially attributed to the differing methodologies of the two studies. Ahn and Denis examine capital expenditures for all of a parent s segments and use industry median q for single segment firms to proxy for the q of each of a parent s segments. We examine divested division capital expenditures and compute divested division q using divested division market values of equity. Thus, one explanation for the differing conclusions of the two studies is that parent s tend to divest low-q divisions, allowing the parents to better concentrate resources on their remaining higher q segments. We find that about 75% (63%) of the divested divisions are low q, based on comparing the divested division s q to the q of its industry (parent). 5. Allocational efficiency 5.1. Assessing efficiency We now turn to examination of capital allocation efficiency at the combined firm level (parent plus divested division) as well as at the divested division level. At the combined firm level, we adopt a measure of allocational efficiency from Rajan et al. (2000), then we define two measures of efficiency of investment in divested divisions. The combined firm measure is denoted as allocational efficiency and is identical to that used by Rajan et al. (2000, p. 66): " AE ¼ Xn I j BA j ðq j qþ I j SS BA j¼1 j BA SS Xn I j w j I!!#, j SS BA j j¼1 j BA SS BA; ð1þ j where BA j = book value of assets for segment of j as of the beginning of the year = identifiable assets capital expenditures + depreciation; q j = asset-weighted average q for stand-alone firms operating in the same industry as segment j, where q = market value of equity plus book value of assets minus book value of equity, all scaled by book of value assets; q = asset-weighted average q of the firm s business segments; I j = capital expendi-

C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 263 tures for segment j; I ss j /BA ss j = asset-weighted average ratio of capital expenditures to beginning assets for stand-alone firms operating in the same industry as segment j; w j = proportion of firm s investment in segment j as measured by BA j /BA; and BA= book value of assets of the parent firm as of the beginning of the year. Throughout this study, all industry measures are computed at the three-digit SIC level as in Rajan et al. except that we move to a broader SIC level when there are fewer than five stand-alone firms at the three-digit SIC level. We define a stand-alone firm as a firm that reports at least 95% of its sales in one business segment. The term I j BA j I j SS P n BA SS j¼1 w jðþ in Eq. (1) is the adjusted investment ratio in Rajan j et al.(2000). It is the industry- and firm-adjusted investment ratio for a given segment j. The measure AE will be positive to the extent the firm invests in relatively high-q divisions and disinvests in low-q divisions. The measure will be smaller or even negative if the reverse of either case is true. Rajan et al. refer to Eq. (1) as a relative measure of capital allocation efficiency because it is assessed with respect to he firm s q (q ) and the firm s relative investment, P n j¼1 w jði j =BA j Ij SS =BA SS j Þ. AE is measured pre-spinoff and post-spinoff for each firm. 11 Data for the divested division are included in the calculation of AE after as well as before the spinoff distribution. The differences (post-spinoff minus pre-spinoff) are calculated to form the change in efficiency (DAE). Examining the change in allocational efficiency reduces problems arising from benchmarking business segments against single segment firms. Following Rajan et al., we calculate allocational efficiency only for firms with no business segments operating in the financial service industries. By construction, none of our sample s divested division primary operations are from the financial services industry, however, a number of parents report a segment of this type. In addition, we require that a parent, inclusive of its divested division, report business segments operating in at least two distinct three-digit SIC industries. 12 We also estimate division-specific measures of investment efficiency for divested units. Relative investment efficiency (IE REL ) for a divested division is defined as: IE REL ¼ðq DIV q PAR Þ I DIV I DIV SS BA DIV BA SS Xn w j DIV j¼1 I j I j SS BA j BA SS j!! : ð2þ In Eq. (2), q DIV and q PAR are Tobin s q ratios for the division and parent, respectively. The ratio of capital expenditures to assets of the division is I DIV /BA DIV. The weighted average ratio of capital expenditures to assets for stand-alone firms in the same industry SS /BA SS DIV. as the divested division is I DIV Eq. (2) reflects efficiency relative to capital allocations for all division of the parent firm. We also estimate a measure that applies only to the divested divisions. This measure 11 Pre-spinoff is defined as 2 years before the spinoff, or 3 years if data are not available in year 2. Postspinoff is defined as 1 year following spinoff, or 2 years if the data are not available in year + 1. We do not use data from year 1 for two reasons. First, we want to avoid the possibility that our pre-spinoff measurement of allocational efficiency reflects a change in investment in the division to be divested as management anticipates the spinoff distribution. Second, a number of firms do not report business segment level data for the division to be divested in 1 year, a fact that also supports our first stated reason. 12 A firm operating in only one industry segment would have an AE of zero.

264 C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 reflects the difference in q for the division ( q DIV ) and the average q of single-segment firms in the same line of business q SS DIV. The difference is scaled by the difference between capital expenditure ratio for the division (I DIV /BA DIV ) and the average ratio for singlesegment firms in the same industry (I SS DIV /BA SS DIV ), and the result is termed absolute investment efficiency (IE ABS ):! IE ABS ¼ðq DIV q SS DIV Þ I DIV I DIV SS BA DIV BA SS : ð3þ DIV Positive (negative) values of IE REL and IE ABS reflect efficient (inefficient) pre-spinoff investment in the divested division. That is, the measures are positive (negative) when the parent subsidizes a divested division that has relatively good (poor) investment opportunity or when the parent rations a divested division that has relatively poor (good) investment opportunities. The spinoff distribution permits the use of divested division s observed equity market value in the calculation of IE REL and IE ABS rather than the industry q typically used for a division within a multi-divisional firm. Although the division s q cannot be observed until after the spinoff, we use this q in lieu of a pre-spinoff, industry median q because it avoids the measurement error problems of using an industry q that Whited (2001) points out, and more generally, because we view division post-spinoff q as a better measure of a division s pre-spinoff investment opportunities. In support of this, we note that there is substantial variation in q among stand-alones. For example, for the Compustat sample of stand-alones in the 1980s, the median standard deviation and range of q at the four-digit SIC level are 0.9132 and 2.8920, respectively. Furthermore, IE ABS requires a measure of q other than an industry proxy. For IE REL and IE ABS, we measure parent q and industry q in the same year that the divested division q is measured. For the capital expenditure components of IE REL (2) and IE ABS (3), we take an average of years 3 and 2 as in our examination of industry-adjusted capital expenditures. Again, we exclude year 1 to reduce the likelihood that our measure reflects a change in the parent s policy of investment in anticipation of the spinoff. In calculating AE, IE REl and IE ABS, we winsorize the top value of the ratio of capital expenditures to beginning assets fop business segments at one. This moderates the influence of a few extreme observations in the cross-sectional analysis. Winsorization does not materially change the univariate results. The cross-sectional results without winsorization are also not materially different from those presented after high leverage or high influence observations are excluded. Cross-sectional results using data not winsorized are available from the authors upon request. In the cross-sectional analysis, we check the sensitivity of results to high leverage and high influence observations. 5.2. Findings In Table 3, we describe the efficiency measures AE, IE REL, and IE ABS. We also report tests of differences in efficiency pre- and post-spinoff (DAE). There is no evidence of wholesale inefficient allocation prior to spinoff. Mean and median pre-spinoff measures of allocational efficiency for the parent (AE) and for the divested division (IE REL and IE ABS ) are all statistically indistinguishable from zero.

C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 265 Table 3 Measures of investment efficiency AE IE REL IE ABS A. Pre-spinoff measures Mean 0.00030 ( 0.15) 0.00843 (0.53) 0.00407 ( 0.33) Median 0.00003 ( 0.54) 0.00072 (0.93) 0.00225 (1.22) Minimum 0.05492 0.98471 1.24570 Maximum 0.09486 1.24186 0.60923 N 88 122 137 B. Post-spinoff measures Mean 0.00074 ( 0.41) Median 0.00011 (0.26) Minimum 0.09661 Maximum 0.06105 N 88 C. Change DAE Mean 0.00044 ( 0.21) Median 0.00003 ( 0.21) Minimum 0.08446 Maximum 0.05601 N 88 Allocational efficiency of investment of parent firms (AE) is measured according to Eq. (1) in the text. Postspinoff calculations treat the parent and its divested division as a combined entity. Investment efficiency of divested divisions is measured on a relative (IE REL ) and an absolute basis (IE ABS ) based on Eqs. (2) and (3). In parentheses are t statistics for tests of mean = 0 and z statistics for tests of median = 0. Mean values of DAE are also insignificantly different from zero at reasonable levels (t = 0.21) as are median differences (z = 0.21). Thus, on the average, parent firms allocational efficiency does not change following spinoffs. This results suggests, as does our analysis of divested division capital expenditures, that spinoffs are not commonly motivated by a desire to improve the allocation of capital, unless most firms are unable to realize intended improvements. Spinoff may yet influence allocation efficiency but in a manner that varies across firms. While the average (mean or median) change in allocational efficiency is not distinguishable from zero, we identify a pattern between DAE and the level of efficiency before the spinoff. The mean change in allocative efficiency is 0.0066 for parent firms that are deemed inefficient before spinoff (AE < 0), and the mean of DAE is 0.0077 for parent firms that were relatively efficient (AE>0). The difference in sample means is significant at the 0.01 level (t = 3.73). Thus, spinoff is associated with an improvement (decline) in allocational efficiency at the combined firm level when a parent invests inefficiently cefficiently before spinoff. 6. Valuation effects We next examine whether the valuation effects of spinoffs are related to allocational efficiency, using two alternative value measures. One is the familiar 3-day cumulative

266 C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 abnormal return surrounding the spinoff announcement. The second is the change in excess value, where excess value is the difference in the value of a firm and the value of a portfolio of pure-play firms that operate in the same industries as the firm in question, as in Berger and Ofek (1995). 13 6.1. Cumulative abnormal return The market model is employed to estimate abnormal returns associated with the announcements. The market model estimation period consists of 255 trading days ending 20 days prior to the spinoff announcement. The CRSP value-weighted index is used as the market index for the market model estimation. The estimated spinoff announcement returns reported in Table 4 are similar to those reported in previous studies. The cumulative average abnormal return for the 3-day window around the spinoff announcement (days 1, 0 and + 1, where day 0 is the announcement date) has a value of 3.53% (median of 3.48%), with 72.37% of the announcing firms having positive estimated abnormal returns. The range of CAR (cumulative abnormal return) is 16% to 20%, thus while spinoffs, on the average, lead to gains, there is substantial variation in returns experienced by individual firms. 6.2. Excess value Our sales-based measure of excess value (Berger and Ofek (1995)) is defined as: EV ¼ lnðmv=ivþ ð4þ IV ¼ Xn j¼1 S j ðmv=sþ j ; ð5þ where MV = market value of firm = sum of market value of equity and book values of debt and preferred stock; IV = sum of segment-imputed values; S j = annual sales of segment j; and (MV/S) j = ratio of market value to sales for the median single-segment firm in industry j. We also measure an asset-based gauge of EV, where S j in Eq. (5) is replaced by the book value of assets of segment j of the firm, and MV/S in Eq. (5) is interpreted as the ratio of market value of assets to book value of assets for the median single-segment firm in industry j. 13 Campa and Kedia (2002) and Villalonga (2002) question the validity of the excess value measure as constructed in Berger and Ofek (1995). Both studies point to the endogeneity of diversification, that the divisions of diversified firms may be fundamentally different from their stand-alone counterparts to which the excess value measure compares them. These issues surrounding excess value are of lesser concern in our study. Our main focus is on the change in excess value, not the diversification discount. Taking the difference in excess value should eliminate systematic bias in the measure. In addition, we examine spinoff announcement returns as an alternative measure of value change.

C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 267 Table 4 Cumulative Average Abnormal Return (CAAR) for spinoff announcements CAAR a 3.53% b (14.06)*** Median a 3.48% c (6.26)*** Proportion positive 72.37% Cumulative Average Abnormal Returns for a sample of 153 firms that complete a spinoff from 1980 through 1996. CAAR is for a 3-day window (t = 1, 0, + 1) relative to announcement date (t = 0). Results are based on 152 announcements because one firm has missing returns during the event window. Announcement dates are verified using Lexis-Nexis. For announcements prior to 1985, the announcement dates are also verified using the Wall Street Journal Index and F&S s Predicast. Abnormal returns are estimated using the market model with Scholes Williams betas and the value-weighted CRSP index. The estimation period consists of 255 trading days ending 20 trading days before each spinoff announcement. For firms with multiple stock issues, abnormal returns are estimated using a value-weighted portfolio approach. a Based on n = 152 observations. b The t statistics in parentheses incorporates the Boehmer et al. (1991) adjustment for event-induced variance and the Mikkelson and Partch (1988) adjustment for measurement error in market model parameters. c The test statistic is from Corrado (1989). *** Significant at the 0.01 level. We measure sales-based and asset-based EV 2 years before the spinoff distribution for each parent because many parents do not report segment level data for the division to be divested for the fiscal year prior to the spinoff distribution. EV is measured again 1 year after the spinoff distribution, treating the parent and divested division as if they were still one entity. 14 The differences between pre- and post-spinoff excess values are denoted DEV. Sales-based excess value and change in excess value are described in Panel A of Table 5 for those observations where we can calculate excess value before and after the spinoff. Pre-spinoff, the parents trade at a substantial discount of about 0.23 on the average. The mean change in sales-based excess value (DEV) is 0.0452, and is not significant. The median change of 0.1096 is marginally significant (0.107 level). The discrepancy between the mean and the median is due to one extreme observation (DEV = 3.39). With this observation omitted, we find the mean of DEV rises to 0.0815, significant at the 0.10 level, and the median rises to 0.1271 (also significant at the 0.10 level). We conclude that sales-based excess value improves on average following spinoffs (we note the caveat that these data are noisy). However, the proportion of firms that trade at a discount remains nearly unchanged, from 68.75% pre-spinoff to 64.46% post-spinoff. In Panel B of Table 5, we describe the asset-based measure. Pre-spinoff, parents are valued at a discount, on the average, of about 0.11. Post-spinoff, the discount averages about 0.04, a noted improvement. The mean change is 0.0782, significant at the 0.05 level (t = 2.03). The median change is 0.0590, significant at the 0.02 level (Wilcoxon 14 We calculate EV in year 3 (year + 2) when it cannot be calculated in year 2 (year + 1). As in Berger and Ofek (1995), EV is not calculated for firms that report business segments operating in the financial services industries. Segment sales and segment assets are grossed up and down so that the respective sum equals the firm s total sales or assets.

268 C.R. McNeil, W.T. Moore / Journal of Corporate Finance 11 (2005) 253 275 Table 5 Excess value EV pre-spinoff EV post-spinoff DEV A. Sales-based excess value Mean 0.2327 ( 3.93)*** 0.1875 ( 2.87)*** 0.0452 (0.76) Median 0.2737 ( 3.64)*** 0.1433 ( 2.45)** 0.1096 (1.62) Minimum 1.9135 2.7151 3.3986 Maximum 1.1781 1.2000 1.6944 Proportion trading at a discount 68.75 64.46 N 96 96 96 B. Asset-based excess value Mean 0.1149 ( 2.72)*** 0.0367 ( 0.84) 0.0782 (2.03)** Median 0.0716 ( 2.63)*** 0.0357 ( 0.49) 0.0590 (2.34)** Minimum 1.5888 1.4633 1.0560 Maximum 1.0102 1.1181 1.1623 Proportion trading at a discount 59.80 53.61 N 97 97 97 EV is defined in Eq. (4). It is measured for each firm 2 years before spinoff. The change in EV is the difference between the value assessed in year + 1 relative to the spinoff and the measure pre-spinoff. Post-spinoff calculations treat the parent and its divested division as combined entity. Sales-based excess values are in Panel A and asset-based measures are in Panel B. Test statistics appear in parentheses. ** Significant at the 0.05 level. *** Significant at the 0.01 level. Z = 2.34). We conclude that asset-based excess value increases on the average following spinoffs. Again, however, the proportion of firms trading at a discount changes little, from 59.80% to 53.61%. Perhaps surprisingly, the correlation between spinoff announcement returns (summarized in Table 4) and change in excess value is small and not statistically significant for either measure of excess value. Similarly, however, Billet and Mauer (2000) find no statistically significant association between tracking stock announcement returns and their measure of excess value, and Graham et al. (2002) find no association between takeover announcement returns and change in excess value. Spinoff announcement returns and change in excess value appear to contain different information. 6.3. Valuation effects and capital allocation efficiency We now turn to tests of whether changes in value are related to changes in investment efficiency. We estimate linear models of CAR and sales- and asset-based measures of DEV. For the model of CAR, we include the following control variables. TAX = 1 if the spinoff distribution is taxable and 0 otherwise. The size of the divested division relative to its parent (SIZE) measures the materiality of the event to the parent and its stockholders. UNRELATED = 1 if the parent and divested division s primary operations are in different industries at the two-digit SIC level and thus indicates a focus increasing spinoff.