Corporate Focus and Discontinued Operations

Corporate Focus and Discontinued Operations Richard A. Lord Department of Accounting and Finance Feliciano School of Business Montclair State University Yoshie Saito Department of Accounting Strome School of Business Old Dominion University A manuscript submitted for presentation at the 2017 Multinational Finance Society January 15, 2017 Abstract We test three propositions from the Corporate Focus Hypothesis to determine whether discontinue operations affect current and future operating and financial performance. Firms that discontinue operations are more diversified, less profitable, face higher financial constraints, and have weak marketbased performance. Enterprises that discontinue negative-valued operations are larger, more heavily leveraged and make higher payouts. While stock market performance for those making positive-valued divestitures are weaker than for control firms, they are superior to those announcing negative-valued disposals. Following a divestiture, diversification is sharply reduced, there is clear improvement in market-based performance, cash holdings increase and other financial constraints are eased. JEL Codes: G34, L25 Keywords: Discontinued Operations, Corporate Focus, Firm Performance

Corporate Focus and Discontinued Operations Abstract We test three interrelated propositions from the Corporate Focus Hypothesis to determine whether discontinue operations result in more effective use of economic resources that affect future operating and financial performance. We find that firms that discontinue operations have more diverse operations, are less profitable, face higher financial constraints, and have weak market-based performance. Enterprises that discontinue negative-valued operations are larger, more heavily leveraged and make higher payouts to shareholders. While stock market performance for the firms making positive-valued divestitures are weaker than for the control sample, they are superior to those announcing negative-valued disposals. Following a discontinued operation, corporate diversification is sharply reduced, there is a clear improvement in the market-based performance of the firms, cash holdings are bolstered and other financial constraints are eased. JEL Codes: G34, L25 Keywords: Discontinued Operations, Corporate Focus, Firm Performance 1. Introduction We examine decisions to discontinue operations, and the effects of these actions on subsequent firm performance. Much prior research has concentrated on the reasons for a broad set of corporate divestitures. But, these have usually relied on small samples of firms that make public announcements of these disposals. Discontinued operations, which are reported on the income statement, represent a significant class of divestitures, which are intended to signal important strategic shifts in corporate operations. These are major decisions that require the approval of the board of directors. Although reports are still somewhat infrequent, they are a separate line item on the financial statement, which permits assessment of a large sample of divestitures. In the past, research on discontinued operations was rare because of a perception of their transitory nature (Fairfield, Sweeney and Yohn 1996; Burgstahler, Jiambalvo, and Shevlin 2002; Gu and Chen 2004). Recent studies by Barua, Lin and Sbaragila (2010) and Curtis, McVay and Wolfe (2014) in accounting journals examine earnings management and persistence 1

associated with these actions. But, further investigation is needed on the reasons for the choices and their ensuing effects. In this study, we seek to fill these gaps. Prior research on the Corporate Focus Hypothesis (John and Ofek 1995 and Comment and Jarrell 1995) suggests that overly wide diversification will strain managerial abilities, which leads to poor operating performance and financial constraints. Earlier studies on divestitures generally rely on small samples of enterprises that publically announce major disposals. We use a much larger sample collected from the report of discontinued operations during the period between 1973 and 2012 and present evidence that these items represent a significant strategic disposal decisions. In addition, we test for an interrelationship among the three elements of the Corporate Focus Hypothesis. Finally, we extend the prior work to see the degree of improvement in corporate performance following the discontinuations. We use logistic regression models to analyze whether the elements of the Corporate Focus Hypotheses explain managerial disposal decisions reported as discontinued operations, and to determine differences between firms that announce negative and positive-valued divestitures. We test three propositions to determine whether these announcements represent significant shifts in strategy that result in more effective use of economic resources that affect future operating and financial performance. Our results provide strong support for the Corporate Focus narrative. Firms that divest units clearly have far more diverse operations than those that do not. The enterprises that discontinue operations tend to be less profitable, face higher financial constraints, and to have weak market-based performance measures. We find some notable differences between enterprises that discontinue negative and positivevalued operations. The former are smaller, make low capital investments, and have scanty cash holdings. The latter are larger, more heavily leveraged and make higher payouts to shareholders. While stock market performance for the firms making positive-valued divestitures are weaker than for control firm, they are superior to those announcing negative-valued operations. 2

There is some prior work on firm performance following a divestiture. Desai and Jain (1999) present evidence that stock returns and operating cash flows improve for a small sample if firms that make focus-increasing spin-offs. We examine a much broader array of performance variables, and show that there is clear improvement in many important operating and financial performance measures in the years after discontinuations, which suggest that managers are using firm resources more effectively. The extent of corporate diversification is sharply reduced, and there is a clear improvement in the accounting and market performance of the firms. Their financial constraints are also eased to some extent. There is strong evidence that their cash holdings are bolstered by the divestitures. In addition, leverage at the firms that discontinue an operation does not rise as much and their capital expenditures do not fall as much as for the control sample. However, despite the relative improvement for most of the operating factors, after the divestiture all of the measures for firms that discontinue an operation still are inferior to those for the control sample. So, they do not completely catch-up with the average enterprise after the disposals. We make several important contributions to the literature on divestitures. Discontinued operations are important decisions involving major shifts in corporate strategy. By concentrating on this item, we can collect a much larger and clear-cut sample than many other previous studies. Because the reported items also have positive and negative signs, we provide some insight into the contrast between firms that dispose of assets of different relative quality. We also present evidence on the extent of improvement in corporate performance following the divestitures. The remainder of our paper is organized as follows. First, we provide some background information on discontinued operations. In the following two sections, we develop our hypotheses and describe the variables used to test them. In the next two parts, we present univariate statistics, and the results from the logistic regression specifications. Then, we assess firm performance and operating and financial characteristics before and after the reports of discontinued operations. In the final section we summarize our findings. 3

2. Discontinued Operations If a firm decides to divest a unit, it may be sold, spun-off or abandoned. Under what conditions may it be booked as a discontinued operation, and what value is recorded? According to SFAS No. 144 (implemented in 2002), the discontinued operation must be a separate major line of business or geographical area of operations; and that can be distinguished operationally and for financial reporting purposes. The parent firm can have no significant continued involvement in the divested unit. It may be disposed of in its entirety or piecemeal. 1 Generally, activities such as closing facilities, abandoning products or even product lines, and changes in the size of the work force in response to market forces should not be booked as discontinued operations (but there are tests to allow these if their impacts are significant). When a firm discontinues an operation it must record the item on the income statement, where it appears below income from continuing operations to remind analysts that these cash flows differ from those generated by ongoing activities. The item reported consists of three components: (1) the profits or losses generated by the unit in the operating year, 2 (2) the capital gain or loss on the sales or disposal of the unit, and (3) the tax effects. Consider a simple hypothetical example given in Spiceland, Sepe and Nelson (2011, pp. 182). A firm decides to discontinue a unit that suffered an operation loss of $4.20 million. The asset has a book value of $12.00 million, and it is sold for $14.00 million, which is a $2.00 million capital gain. So, the gross value of the disposal is a $2.20 million loss. The firm s income tax rate is 40%, so the total value reported under discontinued operations is a loss of $1.32 million. Therefore, reports of discontinued operations give a slightly blurred picture of the details of the profitability of the disposed unit or the selling price. But, by definition they do represent major corporate 1 The reporting is more complex if the firm decides to dispose of the unit in the next operating period (or periods), or if the discontinuation will take place over several years. In these cases there will be reports during a number of accounting cycles. In addition, enterprises occasionally report revisions to the original amount announced in later years. For an interesting, and reasonably clear practical example, see the discussion in the 2012 Annual Report by Becton, Dickinson and Company on the sale of their Biosciences Division. 2 This means that the earnings on the unit are moved out of earnings from continuing operations to allow investors and analysts to compare them with reported earnings in future years. 4

restructurings that require approval of the board of directors, so they do indicate an important change. The positive and negative signs of the items also provide some indication of the relative quality of the disposed assets. 3. Hypotheses We analyze the effect of the three interrelated components of the Corporate Focus Hypothesis, suggesting that firms that are diversified beyond the capabilities of their managers, will perform poorly, and face financial constraints, which makes them more likely to divest a unit. We advance three propositions on these elements of the Corporate Focus Hypothesis. Concentrating on these items also allows us to employ a much wider sample than previous studies of divestiture choices. We include a broad range of explanations for disposal decisions, and test for interactions between these factors. We posit that firms that discontinue operations should have unusually diverse operations. It is difficult for the top-level management team to make appropriate decisions for many unrelated operating divisions. Kaplan and Weisbach (1992), Comment and Jarrell (1995) and John and Ofek (1995) find that tightening the corporate focus of overly diversified firms is a primary reason to divest assets. Therefore, our first hypothesis is: H1: Broadly diversified firms are more likely to discontinue an operation. The earlier literature also suggests that poorly performing firms are more liable to divest operations. John, Lang and Netter (1992) and Berger and Ofek (1999) and Kruse (2002) find that weak performance is a common reason for corporate restructuring to increase focus. Lang and Stulz (1994) present evidence that firms with widely diverse operations tend to have low values of Tobin s Q. These finding imply that discontinued operations should contain information about important strategic managerial decisions that reveal inappropriate past diversification decisions. Thus, the second hypothesis is that: H2: Poorly performing firms are more likely to discontinue an operation. 5

Financially constrained enterprises are also often pressed to sell assets to meet other corporate obligations. Lang, Poulsen and Stulz (1995), Allen and McConnell (1998), Kruse (2002) and Bates (2005) find that heavily leveraged firms frequently deal assets to raise capital. Similarly, Schlingemann, Stulz and Walking (2002) show that enterprises holding low levels of cash are often forced to make asset sales to finance new investments. Therefore, the third hypothesis is: H3: Financially constrained firms are more likely to discontinue an operation. There is a strong possibility that the three elements of the Corporate Focus Hypothesis are intertwined. So we also test for interactions between these effects on the choice to discontinue an operation. 4. Variable Descriptions In this section we describe our categorical measures of discontinued operations used in the logistic regression models, and twenty explanatory variables to examine our hypotheses on the motives for firms to dispose of a unit. 4.1 Classification of Discontinued Operations Many firms fail to report discontinued operations. Because this is the primary variable of interest, we set all of the observations on Compustat when the item is not reported to a value of zero. In our logistic regression specifications, we employ two definitions of our dependent variable, DO-CHOICE. The first classification is the binary case; firms that announce discontinued operations and those that do not. In the second classification, DO-CHOICE is tertiary depending upon whether a reported divestiture has a positive or negative value. Earlier studies in accounting by Black, Carnes and Richardson (2000) and Dechow and Ge (2006) also examine differences between positive and negative reports of discontinued operations and special items. 4.2 Corporate Focus The extent of diversification is at the heart of the Corporate Focus Hypothesis. Scherer and Ravenscraft (1984) estimate a Herfindahl Index (HERF) to measure the degree of diversification using the 6

sum of the squared values of the ratio of unit sales for each division to total firm revenues. If the measure has a value of one, the firm is completely undiversified, as the value of the index decreases the entity is more operationally diverse. We employ the value of the index in the year before the discontinuation. This proxy is used in earlier studies by John and Ofek (1995), Comment and Jarrell (1995), Berger and Ofek (1999), Desai and Jain (1999) and Dittmar and Shivadasani (2003). 4.3 Operating Performance We employ eight proxies for various elements of corporate performance. We calculate three market-based performance measures. The first, is the firm s equity beta estimated in the year before the divestiture decision (BETA). We surmise that the divesting enterprises should have higher risks. Second, is the Sharpe Ratio (SHARPE), which is a measure of equity return for the risk taken. This is the ratio of the annualized return on the firm s stock over the prior three years to the annualized standard deviation of the daily returns over the same period. 3 We expect that enterprises that discontinue operations to have low Sharpe Ratios. Our other market-based measure of firm performance is the average of the firm s Tobin s Q (TOBINQ) over the prior three fiscal years. 4 This is a proxy for the market participants assessment of firm valuation. Entities with low relative share prices should be more likely to discontinue an operation. 5 We employ several accounting based measures of corporate performance. First, managers and financial analysts focus heavily on corporate sales expansion. Therefore, following Berger and Ofek (1999), Bates (2005) and Colak and Whited (2007), we include revenue growth over the prior three years (REV-GROW). Because the distribution of sales growth rates is highly skewed, we add one to the raw value and take the logarithm of this sum. We expect enterprises with lower growth are more likely to discontinue an operation. Negative revenue growth should be a particularly potent sign of weak 3 We winsorize both the annualized returns and the standard deviations of these returns at both the largest and smallest 1% of observations 4 We winsorize TOBINQ at the largest 1% of observations. 5 However, there is an interesting contrary proposition. Nanda (1991), and Nanda and Narayanan (1999) suggest that undervalued enterprises should consider divesting profitable units to force markets to properly value the firm s assets. Krishnaswami and Subramaniam (1999) argue and find that spin-offs enhance value because they can reduce information asymmetry by conveying the profitability and operating efficiency of the divested unit. 7

performance. Therefore, we employ an approach suggested by Berger and Ofek, incorporating a one/zero dummy variable set to one if revenue growth for the prior three years is negative (NEG-GROW). We expect this dummy variable to be positively related to the probability of a divestiture. We examine two widely used measures of earnings; the gross profit margin (GPM), and the net profit margin (NPM). We also estimate the closely related operating cash flows (CF), defined as the ratio of net income plus depreciation to revenues. For all three of these measures we take the average over the prior three years. 6 The distributions of all of these profitability variables are highly skewed, with many extreme observations, especially of large negative values. To smooth the variation in the series, we subtract the ratios from one and take the logarithm of this difference. 7 Then, take the negative value of this value to ensure the sign is the same as the original. We expect the accounting-based measures of corporate performance to be negatively correlated with the probability of discontinuing an operation. 4.4 Financial Constraints The third factor in the Corporate Focus Hypothesis is that financial constraints can hinder operating and financial performance. Firms with difficulty in raising capital may need to sell assets to raise funds to invest in new opportunities or fulfill impending financial obligations. We examine seven proxies for corporate constraint. First, Berger and Ofek (1999) use one/zero dummy variables set to one when profitability is negative. We create three such dummy variables when the lagged values of gross profits (NEG-GPM), net profits (NEG-NPM) and cash flows (NEG-CF) are negative. These dummies should be positively related to the probability of discontinuations. 8 The fourth measure of constraint is the average of the total-debt-to-asset ratio over the three years before the discontinuation (LEV). 9 Lang, Poulsen and Stulz (1995), Allen and McConnell (1998), Kruse (2002) and Bates (2005) find that firms 6 These three profitability measures are highly correlated, so in the logistic regression models we only include the cash flow measure. 7 Taking the logarithm of the differences results in the loss of all observations where the ratio is less than -100%. 8 Again, because the correlation between these measures, we only present the results of the logistic specification using NEG-CF as the measure of financial constraint. 9 We winsorize LEV at the largest 1% of observations. 8

that employ high leverage often lack the flexibility to obtain new funding, so they may sell assets to raise capital. 10 The fifth proxy is the average ratio of cash to total assets (CASH) over the prior three years. 11 Schlingemann, Stulz and Walking (2002) find that firms with low cash-holdings are more likely to sell assets to support new investment. We also look at two other components of cash flows that are not reported on the income statement. Following Bates (2005) we calculate the average of the ratio of capital expenditures to revenues over the prior three years (CAPEX). 12 The second cash flow measure is total payouts to shareholders (PAYOUT). To compute this ratio, we first find the net stock repurchase from shareholders; the difference between repurchases and issuances. If this value is negative (issuances are higher than repurchases) we set the value to zero. Then we take the average of the ratio of dividend payments (to both common and preferred stockholders) and net repurchases, to revenues over the prior three years. 13 Greater capital expenditures and payouts to equity holders create calls on cash flows that leave less for other required outlays. So, it seems that the companies that discontinue operations should have higher capital expenses and payouts. But, Bates (2005) finds that firms that spin-off operations tend to make low historical capital expenditures. 4.5 Interrelationship of Factors in the Corporate Focus Hypothesis To analyze whether the choice to discontinue an operation is affected by the interrelationship between corporate focus, firm performance and financial constraints, we employ a method suggested by Opler and Titman (1993). We create a dummy variable set to one for firms with below average values for its (Fama-French) industry sector and in the lagged fiscal year for both the Herfindahl Index and free cash flows 14 (LOHERF-LOFCF), and regress this against the probability of divesting a unit. We produce two 10 On the other hand, Colak and Whited (2007) find no evidence of high financial leverage for firms that divest units. 11 We winsorize CASH at the largest 1% of observations. 12 We winsorize CAPEX at the largest 1% of observations. 13 We winsorize PAYOUT at the largest 1% of observations. 14 In this case we use free cash flows, which is defined as net income, plus depreciation, minus capital expenditures and cash dividends. 9

similar dummies for enterprises that have below-average Herfindahl Indices and both low measures of Tobin s Q (LOHERF-LOQ) and unusually high financial leverage (LOHERF-HILEV). 15 These three dummy variables should be positively related to the probability of discontinuing an operation. 4.6 Firm Size Firm size often has important effects on corporate decisions. Large entities have easier access capital markets. Because annual revenues are highly skewed, we employ an adjusted measure of lagged sales suggested by Aggarwal and Samwick (1999) as our proxy for size. We estimate the cumulative density function for firms listed on the major North American stock exchanges for the fiscal year. 16 This variable, CDF-REV, will take on a value between zero and one. For example, if a firm is larger than 80% of the observations in the annual sample, CDF-REV will take a value of 0.80. 5. Sample Data and Summary Statistics There are numerous studies of announcements of divestitures, asset sales or spinoffs. However, few corporations actually publicly disclose decisions to dispose of units, so these earlier works must rely on small samples. In order to collect a larger number of observations, we examine announcements of discontinued operations in the annual income statement, which are relatively plentiful. 5.1 Preliminary Unconstrained Sample We collect the required accounting data from 1973 through 2012 from the annual Compustat database. The information needed to calculate the Herfindahl Index of corporate diversification is taken from the Historical Segments database in Compustat. This data is central to our study, but is only 15 In an OLS regression, multiplicative cross-product terms are commonly used to test for such interactions. But, Ai and Norton (2003) note that in logistic models there are difficulties in interpreting the coefficient estimates on crossproduct terms. 16 First, we collect data on annual revenues for all of the firms on Compustat that trade on the nine major North American stock exchanges for each fiscal year. We then calculate the mean and standard deviation of the variable for each year, and estimate the cumulative density function for this distribution. Then, for each observation we estimate the portion of the annual sample that has a lower value of revenue under the CDF for that year (CDF-REV). 10

available from 1976, which limits the beginning of the sample period to 1977. Data to estimate beta and Sharpe Ratio are taken from CRSP. 17 Because of the required lags to estimate some of our independent variables, a firm must have observations for at least four consecutive years to be included in the sample. Of course, we eliminate an entity where any of our explanatory variables are missing. We also remove financial firms (industries 45 through 48 in the Fama and French fifty industrial classifications). In the preceding section, we describe adjustments to some of the data series and cases where we winsorize variables that result in the loss of some other observations. The preliminary unconstrained sample consists of 98,728 firm-year observations, for 10,399 firms during the years 1977 to 2012. Out of this sample there are 12,463 reports of discontinued operation, about 12.60% of the observations. 5.2 Final Matched Sample Comparing the firms that discontinue an operation to all others listed on COMPUSTAT could give a misleading picture of how they differ. Therefore, in earlier studies Berger and Ofek (1999), Krishnaswami and Subramaniam (1999), and Dittmar and Shivadasani (2003) select a matched sample for comparison. We follow this approach in our analysis of the three interrelated elements of the Corporate Focus Hypothesis. Using this method assures that we controlling for industry and year effects, which include the condition of the economy and markets. Colak and Whited (2007) also note that using a matched sample can reduce endogeneity problems. Our choice of the matching firms is based on four criteria. The observation must be from the same Fama-French industry sector and fiscal year as the entity that discontinues an operation. Next, to assure that the control firms do not also make divestitures in the proximate period, we eliminate all 17 In estimating the beta and the Sharpe Ratio, we carefully match the dates to the actual corporate fiscal year. We use daily returns for that year taken from the CRSP database. To calculate beta we regress the daily returns for the firm onto the value-weighted index. Any entity that has less than 100 observations in a fiscal year is dropped. 11

potential matches that announce a discontinued operation in the prior or subsequent three years. 18 Then, to control for size, the lagged value of total assets must be within 20% of the divesting enterprise. This yields 9,879 pairs of matched firms, for a total sample size of 19,758 firm-year observations. 5.3 Matched Sample Univariate Statistics Table 1 contains the univariate statistics for seventeen performance variables for the entities that discontinue an operation compared to matched firms. Descriptions of the variables are also provided in an appendix at the end of the paper. These statistics provide the preliminary test of our three propositions concerning the elements of the Corporate Focus Hypothesis. The first three columns of the table contain the median, mean, and the standard deviation of the explanatory variables for the sample of enterprises that discontinue an operation. The next three columns show the same three parameters for the control group of firms that do not make a divestiture. The last two columns contain test-statistics for the differences between the samples of firms that announce discontinued operation and those that do not. The first is a nonparametric Wilcoxon Rank Test, and the second a T-test based on the difference of the means between the two subsamples. 19 The differences between the variables in the two samples are all highly significant, and they bear out our conjectures concerning the elements of the Corporate Focus Hypothesis. First, the lagged Herfindahl Indices (HERF) are significantly smaller for the firms that discontinue an operation. This confirms that the divesting entities are considerably more widely diversified (a lower index number represents a more diverse enterprise). In fact, more than half of the control firms are completely undiversified (the median index value is one). Performance is clearly weaker for the firms that discontinue an operation. They have significantly higher betas (BETA), and lower Sharpe Ratios (SHARPE) and Tobin s Q (TOBINQ) than the control 18 We treat observations after 2009 slightly differently. We do not eliminate these potential matching firms if the lack of data on announcements of discontinued operations in the following three years is a result of the end of sample period in 2012. 19 Because the variance often fluctuates dramatically between the categories, we use the Satterthwaite approximation of the T-value. 12

firms. The divesting companies obviously have much lower revenue growth (REV-GROW) over the preceding three years, and, over the same period, all three of the measures of profitability and cash flows (GPM, NPM and CF) are significantly smaller. Most of the relevant explanatory variables suggest that enterprises that announce discontinued operations face greater financial constraints. As expected, they have higher financial leverage (LEV) and lower cash holdings (CASH) over the prior three years. Their odds of having negative profits and cash flows in the prior years (NEG-GPM, NEG-NPM and NRG-CF) are obviously far greater than for the control sample. On the other hand, they make lower payouts to shareholders (PAYOUT) and capital investments (CAPEX) than the control firms. Finally, the enterprises that discontinue an operation are slightly smaller, with lower lagged revenues (CDF-REV). This is even though the size of assets is one of the matching factors. 6. Logistic Regression Specifications and Results In this section we first present our basic empirical model. We then show parameter estimates for two specifications, based on different categorical dependent variables. Finally, we discuss the results of some alternative tests used to check the robustness of our findings. 6.1 Logistic Model Specification To test the effects of the explanatory variables on the probability of discontinuing an operation in a multivariate setting we estimate binary multinomial logistic (logit) regression models of the general form: DO-CHOICE t = α 0 + Φ 1 HERF t-1 + β 1 BETA t-1 + β 2 SHARPE t-1,t-3 + β 3 TOBINQ t-1,t-3 + Δ 1 LOHERF-LOQ t-1 + Σ 1 REV-GROW t-1,t-3 + Σ 2 NEG-GROW t-1 + Σ 3 CF t-1 + θ 1 NEG-CF t-1 + Δ 2 LOHERF-LOFCF t-1 + θ 2 LEV t-1,t-3 + Δ 3 LOHERF-HILEV t-1 + θ 3 CASH t-1,t-3 + + θ 4 PAYOUT t-1,t-3 + θ 5 CAPEX t-1,t-3 + λ 1 CDF-REV t + ε 13

where the coefficient Φ 1 shows the effect of corporate diversification, the β i of market-based performance measures, the Σ i of accounting-based performance variables, the θ i of the financial constraints, the Δ i of the interaction terms, and the λ 1 on firm size. We are interested in how the firms that discontinue an operation differ from the control sample. But, we also wish to shed some light on the differences between those that announce negative and positive-valued discontinuations. Therefore, we estimate two specifications. In the first, DO-CHOICE t is binary depending upon whether or not a firm discontinues an operation. In the second model, the values of DO-CHOICE t fall into three discrete classes; whether the entity does not announce a discontinued operation, and then whether the reported value of a divestiture is positive or negative. This specification highlights the effects of the independent variables on the probability of announcing a discontinued operation of the two different signs. 6.2 Results for the Binary Logistic Regression Specification The dependent variable in the binary logistic regression (DO-CHOICE) is sorted into two classes based on whether or not a firms announces a discontinued operation, so the estimated parameters represent the differences for entities that report from those that do not. The coefficient estimates and standard errors are given in the first two columns. Because the estimates from a logit regression are interpreted differently than those from an OLS model, we follow the common practice of calculating marginal effects (at the mean) and economic effects. These are given in the third and fourth columns. The marginal effects show the mean change in the dependent variable for a one-unit change in the explanatory factor. The economic effect is the product of the average marginal effect and the standard deviation for the independent variable for the entire sample. These represent the percentage change in the probability that an entity will announce a discontinued operation if the independent variable changes by one standard deviation. Economic effects are not relevant for dummy variables, which can only have values of zero and one, in this case the marginal effects show the power of the relationship. There is no 14

direct equivalent to r 2 in a logit model, but the estimates of the maximum rescaled r 2 is over 20% for this specification. The results of the binary logit model are shown in Table 2. Descriptions of the variables are, again, provided in the appendix. Once more, the findings provide persuasive support for our three propositions regarding the interrelated components of the Corporate Focus Hypothesis. There is strong evidence that firms with lower values of the lagged Herfindahl Index(HERF), which implies greater corporate diversification, are more likely to discontinue an operation. This suggests that when corporations become too widely diversified there are compelling motives to divest units to bring the operations into sharper focus. The economic effect shows that this has a very powerful influences on the decision; if the Herfindahl Index is one standard deviation lower, there is a 6.32% greater chance the firm will discontinue an operation. The results for the firm performance variables also confirm our hypotheses. The market-based measures are all weaker for the entities that discontinue an operation than for the control sample; they have high beta measures (BETA), and low Sharpe Ratios (SHARPE) and Tobin s Q (TOBINQ). The economic effects suggest that a one standard deviation change in beta and Tobin s Q results in more than a 3% greater probability of discontinuing an operation, and of 2% for the Sharpe Ratio. The coefficient estimates for this specification suggest that historical cash flows (CF) are lower for the firms that discontinue an operation, but the adjusted measure of long-term revenue growth (REV- GROW) is not statistically significant. 20 Clearly, entities with negative revenue growth over the prior three years (NEG-GROW) are more than 11% more likely to discontinue an operation. Most of our proxies confirm that the enterprises that dispose of an asset face more financial constraints than the control sample. Two of these factors have a very powerful impact on the decision. The marginal effect suggests that firms with negative cash flows in the previous year (NEG-CF) are 17% more liable to discontinue an operation. The divesting firms also have much higher financial leverage 20 However, the sensitivity analyses that we discuss below, cast some doubt on the robustness of both of these results. 15

(LEV); the economic effects shows that for a one standard deviation change in leverage, the probability of announcing a discontinued operation increases by almost 6%. For the other three measures of financial constraint, the economic effects suggests the relationships are not as robust. But, the divesting firms hold lower levels of cash (CASH), payout (PAYOUT) more, and commit less to capital expenditures (CAPEX) than the control firms. We also include three dummy variables, following the methodology of Opler and Titman (1993), to test for interactions among corporate diversity, operating performance, and financial constraints. The parameter estimates for these dummies confirm that there are important interactions between the three elements of the Corporate Focus Hypothesis, and suggest that these strongly influence the decision to discontinue operations. In particular, firms with low Herfindahl Indices and below (industry) average Tobin s Q (LOHERF-LOQ) are over 9% more likely to execute a divestiture. Highly diversified enterprises with low free cash flows (LOHERF-LOFCF) are about 4% more liable to discontinue an operation, and these with wide diversification and high financial leverage (LOHERF-HILEV) are almost 3% more apt to do so. We also confirm that the entities that discontinue an operation are smaller (CDF-REV), and the economic effect suggests this relationships is fairly robust. The binary logistic regression results provide very strong support for our hypotheses concerning the three interrelated elements of the Corporate Focus Hypothesis. The firms that discontinue operations are significantly more diversified. They are less profitable, and their market-based performance measures are weak. There is also fairly convincing evidence that they face greater financial constraints. The interactive terms confirm that these factors are interrelated, and they have a significant impact on the choice to discontinue an operation. 6.3 Results for the Tertiary Logistic Regression Specification The results of the tertiary logistic regression are given in Table 3. Here the dependent variable (DO-CHOICE) is sorted into three discrete classes depending on whether a firms announces a negative or 16

positive-valued discontinued operation, or does not divest a unit. The latter is, again, the control group, so the two parameter estimates for each explanatory variable represent how firms that discontinue an operation of the two different signs differ from those that do not. As above, the maximum rescaled r 2 for this specification is over 20%. Firms that discontinue either negative or positive-valued operations are both far more diversified, and appear to have lower cash flows than the control sample. 21 They are also much more likely to have negative revenue growth and cash flows, and the marginal effects, again, show that these two factors have a very strong impact on the disposal choice. However, there are distinct differences in the market-based firm performance measures (BETA, SHARPE and TOBINQ). They are all markedly poorer for the enterprises that discontinue negativevalued operations. In particular, the Sharpe Ratios for the entities that book positive-valued divestitures are not significantly different from the control firms. There is a distinct contrast in historic revenue growth for the two types of divesting firms. Growth is higher than for the control sample for the entities that discontinue negative-valued operations, and it is below average for those the dispose of positive-valued units. This may explain why REV-GROW is not significant in Table 2. 22 The interactive term LOHERF-LOFCF also shows that the enterprises that divest negative-valued operations are also far more likely to be in the quadrant with high diversification and low cash flows (5% as compared to 2.5% for firms with positive-valued discontinuations). There are also interesting differences based on the sign of the divestiture in the proxies for financial constraint. Entities that discontinue negative-valued operations are much smaller, they make fewer capital expenditures and have lower cash holdings. Their payout policy is not significantly different from the control firms. On the other hand, the enterprises that discontinue positive-valued operations 21 The sensitivity analyses below cast some doubt on the latter result. 22 The sensitivity analyses again raise some questions about the robustness of these results. 17

make significantly higher payouts and employ greater debt. While, their cash holding and capital expenditures are not significantly different from the control sample. So, the contrast between the two groups provides some new insights. The firms that discontinue negative-valued operations are smaller, with lower cash holdings, low capital investment and strikingly weak stock market performance. The enterprises that divest positive-valued units are larger, more heavily levered, and make greater payouts. They still have relatively high measures of beta and low values of Tobin s Q, but their Sharpe Ratios are on par with the control firms. 6.4 Sensitivity Analyses To test the robustness of our results to model specification, we conduct several sensitivity analyses in four areas. Most of these have little effect on the results. But, there are some differences when we adjust the value of the explanatory variables to their industry-year median values and use the unconstrained sample in the models. The coefficient estimates associated with the adjusted historical revenue growth (REV-GROW) and cash flows (CF) are clearly the most sensitive to the specification. First, we substitute the gross and net profit margins (GPM and NPM), in-turn, along with their corresponding dummy variable for negative values (NEG-GPM and NEG-NPM) for the cash flow measures (CF and NEG-CF). These substitutions have no substantial effects on the results for either our binary or tertiary logit specifications. Second, we include industry and year fixed effects both separately and combined. These additions have little effect on the outcomes in either specification. Third, we adjust most of the explanatory variables by subtracting the median value for their Fama-French sector for the fiscal year. We do not adjust the Herfindahl Index and the five dummy variables. These changes have some effects on the parameter estimates. In the binary specification, the coefficient on REV-GROW is positive and significant (it is not significant in Table 2), and that on CF is insignificant (it is negative in Table 2). In the tertiary model, both estimates on REV-GROW are again significantly positive (that for negative-valued divestitures is negative in Table 3), and those on CF are 18

not significant. In addition, the Sharpe Ratio is positive and significant for firms discontinuing positivevalued operations (it is insignificant in Table3). Finally, we estimate both models using the unconstrained sample, which has important implications for the characteristics of the control firms. In both cases, the coefficient estimates on our proxy for firm size (CDF-REV) are significantly positive, meaning the entities that divest a unit are larger than this control sample (the results were opposite using the matched sample). In the binary specification, the relationship between REV-GROW is negative and significant, and that on CF is insignificant. In the tertiary model, the coefficient on REV-GROW for negative-valued divestitures is not significant, and that for capital expenditures (CAPEX) for enterprises that dispose of positive-valued unit is significantly negative (it is not statistically significant in Table 3). We believe that using the matched sample is the best approach to analyze the difference between two groups of firms. But, even when we use the unconstrained sample, our major results still shine-through. Overall, the results of these sensitivity tests support our major findings. The firms that discontinue operations are more widely diversified, financially constrained, less profitable, and perform poorly in equity markets. This is especially true for entities that discontinue negative-valued operations. 7. Firm Performance Following Discontinued Operations If discontinued operations relay important information about strategic shifts of resources in response to weak past performance, it is important to test whether corporate performance improves after these divestitures. There is earlier work on improvements in stock returns and operating cash flows by Desai and Jain (1999) following spin-offs. But, our setting allows us to employ a much larger sample and to test a wider array of performance measures. Table 4 contains data on averages of eleven of the performance variables for three-year periods before and after a divestiture (note that this differs from the presentation of three of the variables, HERF, BETA and CDF-REV; above we employ a single year lag in the logistic regression models). For this analysis, we also use the matched sample, but we select only the pairs where both firms report all of the 19

required variables for the three years before and after the announcement. This means that all observations after 2009 are lost. 23 This results in a sample of 5,586 enterprises that discontinue an operation and an equal number of control firms, for a total of 11,172 firm-year observations. In Table 4 there are two rows for each variable; the top row is the data for the entities that discontinue an operation, while the second is for the control firms. The first three columns contain the median, mean and standard deviation for the variables for the three years before the divestiture. The same statistics for the average values for three following years are given in the next three columns. In the last two columns, we present test-statistics showing the differences in the changes between samples and through time. These are again Wilcoxon Rank tests and T-tests. For each variable the test statistics in the first row (DO Bef-Aft) show the differences between the period following the divestiture and the period before for the firms that discontinue an operation. A positive value for these statistics indicates that the average values are higher after the announcement than before. The test in the second row (DO-No DO Chng) are for the differences between the after and before averages for the firms that discontinue an operation compared to the control sample. For these statistics, a positive value shows that the measure increased more (or decreased less) for the entities that divest a unit than the control firms between the two periods. 7.1 Changes in Corporate Focus Obviously, the difference in the change in corporate focus between the two matched samples through time is dramatic. There is a clear decline in the Herfindahl Indices (HERF) for the control firms in the later years, implying that their diversification increases sharply. But, the opposite is plainly true for the divesting enterprises; the indices rise very significantly, which indicates a sharp and sustained tightening of corporate focus after discontinuing the operation. However, after the change, the average divesting entities remain slightly more diverse than the corresponding control firm. Nevertheless, this still 23 Also, because we employ three lagged values of the Herfindahl Index, all observations before 1979 are unavailable. 20

provides strong corroborating evidence that reducing overly wide diversification is a powerful and persistent motive to dispose of an operation. 7.2 Changes in Corporate Performance and Firm Size The test statistics show that beta (BETA) increases significantly after an entity discontinues an operation. But, this is also true for the control sample, and there is no evidence that the increase between the two groups differ. So, divestitures do not seem to have an unusual effect on relative systematic risks over time. There is a steep and significant increase in the Sharpe Ratio (SHARPE) following a discontinuation. 24 By contrast, the average ratio clearly declines for the control sample. So discontinued operations are followed by distinct improvements in stock returns for the risk taken. There is not significant evidence that Tobin s Q (TOBINQ) for the firms that discontinue an operation changes after the divestiture, but it does decline for the typical control firm. These results confirm that there is a relative improvement in the subsequent performance of the market-based variables for the firms that discontinue an operation. But, despite the progress, their average measures of Tobin s Q and the Sharpe Ratio are still lower than for similar non-divesting firms. Revenue growth (REV-GROW) declines significantly for divesting firms in the three years after the announcement. But, it falls by even more for the control firm. So, again, the enterprises that discontinue an operation perform relatively better after the divestiture. Revenues (CDF-REV) for the typical control firm increase sharply in the later period as compared to the earlier. While the relative size of the enterprises that discontinue an operation, decline after the actions. The results for the two accounting measures of profitability (GPM and NPM) are somewhat similar. There is no evidence that they improve for the entities that discontinue an operation. But, at least the Wilcoxon Rank tests suggest that they do much better than the control sample, where both measures 24 Further analysis, not tabulated, shows that returns and their standard deviations both increase significantly following a discontinued operation. Clearly, the rise in returns eclipses that in risk for the typical divesting firm. 21

of earnings typically decline slightly. The two tests for differences show that cash flows (CF) improve after a divestiture, and that this increase is more pronounced than for control firms. This is similar to Bates (2005), who also finds considerable improvements in cash flow after spin-offs. However, after the divestitures, all of the average profitability and cash flow figures are still comparably lower than for the control sample. 7.3 Changes in Financial Constraints Financial leverage increases significantly after a firm discontinues an operation. But, at least the Wilcoxon test provides some evidence that this change is less pronounced than the increase in debt by the typical control firm in the subsequent period. Cash holdings by the divesting firms increase significantly after the announcement. While they are essentially unchanged for the control sample. So, the firms that discontinue an operation seem to pad their cash account with at least some of the proceeds from the sale. Payouts by firms that discontinue an operation increase significantly after the action, but the rise is no greater than for the typical control firm. On the other hand, capital expenditures decline for both the divesting and control firms. But, the T-test shows that the decrease in mean capital spending is significantly less pronounced for the entities that discontinue an operation. Our results provide some interesting similarities, but also contrasts, with Bates (2005) study of focus-increasing spin-offs. He finds that the two most common actions after the sale is reduction of debt followed by cash retention. Increasing internal investment and payouts to shareholders are less common. We find that for firms that discontinue operations leverage, cash holdings and payouts to equity holders all increase. Capital expenditures decline, but by less than for the control firms. However, despite the relative improvements that we document in financial leverage, cashholdings, capital expenditure investments and corporate payouts, the firms that discontinue an operation still lag behind the control sample in these measures of financial constrain and accounting performance measures after the divestitures. 22