Corporate Restructuring and The Likelihood of Emergence from Financial Stress

Transcription

1 Corporate Restructuring and The Likelihood of Emergence from Financial Stress Gregory D. Kane* Associate Professor University of Delaware Department of Accounting and Management Information Systems Newark, Delaware (302) Frederick M. Richardson Professor Virginia Polytechnic Institute and State University College of Business Department of Accounting Blacksburg, VA USA (540) *Corresponding Author Revision: 5/00 1

2 Corporate Restructuring and The Likelihood of Emergence from Financial Stress ABSTRACT Companies experiencing financial stress can attempt to mitigate financial stress through restructuring of the company's asset base. Management may choose to expand through activities involving investment in assets in hopes of increasing sales and/or reducing fixed costs per unit produced (sold). Management may also choose to contract the asset base in order to eliminate and/or reduce investment in unprofitable or risky ventures, improve liquidity, reduce earnings volatility, and reduce the need for operating capital. At issue is the question of how observed changes in the investment base affect the likelihood of emergence from a stressed condition. In this paper, we report on the results of an empirical study that examined this question. We find that, when higher (lower) earnings accompany expansion of the asset base, the likelihood of emergence increases (decreases) as compared to the impact of the same level of earnings for firms that are not expanding. Thus, expansion enhances the impact of the earnings signal. Similarly, contracting firms are more likely to emerge when relatively lower earnings accompany contraction. For contracting firms, however, the lower earnings resulting from mitigation partially cancels the normally positive impact of the unconditional current earnings signal (i.e., the relatively higher the earnings, the greater the likelihood of emergence from stress). Thus, in both cases, the knowledge of expansion and contraction appear to be incrementally useful in explaining emergence from corporate financial stress. 2

3 INTRODUCTION In this paper, we report on the results of an empirical study in which we investigated the association between growth and contraction of a company's asset base and the likelihood of emergence from stress rather than entering into corporate failure. Accounting-based statistical models effectively explain and predict (at rates greater than chance) corporate failure only when financial stress is an apparent precursor to the event (Hopwood, McKeown and Mutchler, 1994). In most cases, however, model accuracy substantially degrades when trying to predict failure more than one year in advance. This is a major problem because much of the loss experienced by claim holders often occurs earlier. One reason why predictive accuracy is so poor earlier than one year before failure is that standard statistical models are relatively static in design (Altman, 1970; Richardson and Davidson, 1984). 1 Consider the case in which a company experiences financial stress. 2 Managers who assess potential negative results from such stress will more than likely take restructuring actions in an effort to mitigate its effects. If the actions are effective, companies tend to improve, and the probability of failure declines accordingly. On the other hand, the actions could be ineffective, but nonetheless expend scarce resources. In such cases, the stress will be exacerbated rather than relieved, thereby increasing the likelihood of corporate failure. In either case, a dynamic process (i.e., the attempt by management to mitigate the stress) has significant influence over the likelihood that a firm will successfully emerge from financial stress. This study focuses on the association between two potentially mitigating actions (growth or contraction activity), and the likelihood that either action will lead to emergence from stress, thereby reducing the risk of corporate failure. Obviously, a crucial element in a firm s ability to survive and avoid failure is its ability to manage and mitigate stress when it occurs. Stress is a much more common 3

4 phenomenon than bankruptcy. Many causes of corporate stress are arguably chaotic in nature (e.g., recession, changes in technology, consumer preferences, and so forth) and/or unpredictable. As such, to the extent that they are unpredictable, infrequent, and/or unusual, such causes result in an initial shock that is at least partly unavoidable from the perspective of the company. When faced with stress that is at least partly unavoidable, a company risks the possibility of failure. Three important issues emerge for debt and equity holders: (1) the degree to which a firm is sensitive to the initial shock; (2) the level of liquid resources available to cushion such a shock and thus stave off failure; and (3) what occurs following the initial shock of the stress. In this paper, we focus on processes that relate to the third issue. In particular, a stressed company may take some restructuring action that relieves the initial stress and returns the firm to an unstressed condition. Restructuring can result in contraction or expansion of the firm's business activities. If the restructuring takes the form of a growth action, companies experiencing financial stress attempt to return to profitability and/or positive liquidity through expansion of the asset base. Such actions should help these companies increase sales and/or reduce fixed costs per unit produced (sold). Concurrently, margins, profitability, and ultimately operating cash flows should improve, thus increasing the likelihood such companies will emerge from a stressed condition. Alternatively, stressed companies may contract by shrinking the asset base (including the elimination of unprofitable or risky ventures) in order to improve liquidity, reduce earnings volatility, and reduce the need for operating capital. The type of action that companies employ to mitigate financial stress may provide useful information to investors interested in assessing the likelihood that such firms will successfully emerge from a stressed condition. Our findings are as follows. When higher (lower) earnings accompany expansion, the likelihood of emergence increases (decreases) as compared to the impact of the same level of earnings for firms 4

5 that are not expanding. Thus, expansion enhances the impact of the earnings signal. This finding supports our prior expectations, and suggests a greater likelihood of persistence or even growth in earnings as the result of expansion. On the other hand, contracting firms are able to emerge when relatively lower earnings accompany contraction. The lower earnings may reflect, at least in part, restructuring actions such as segment discontinuations, labor cutbacks, and other such efforts. For contracting firms, the decreased earnings resulting from restructuring partially cancel the normally positive impact of the unconditional current earnings signal. Thus, as with expansion, the relatively higher the earnings, the greater the likelihood of emergence from stress. In any event, in both cases, expansion and contraction appear to be more informative when earnings are affected by these conditions. Net of earnings effects, only contraction appears to be informative about the likelihood of emergence. This indicates that contracting firms with no associated earnings effects will be less likely to emerge from a stressed condition in the following year. We suggest that the expansion/contraction signal is thus conditional with respect to its implications. If earnings effects accompany the signal, an effective restructuring action is underway, thereby increasing the likelihood of emergence from a stressed condition. If, however, the signal is not accompanied by earnings effects, either (1) an ineffective restructuring action is underway, (2) the mitigating effect of the action (to the extent it is sufficient to cause emergence) is lagged by more than one year, or (3) the expansion/contraction is not helpful in mitigating financial stress. A second prior expectation was that, absent a near-term bottom line effect, the costs of the restructuring effort--to the extent they weaken short-term profitability and consume scarce liquid resources--could actually aggravate the initial stress and thereby reduce the likelihood of emergence in the year to come. As noted, our empirical results support this prior expectation for contraction without earnings effects. The 5

6 results we obtained for expansion, however, do not support this prior expectation--controls for expansion without associated earnings effects are insignificantly related to emergence from stress for the sample we examined. The remainder of this paper is organized as follows. In the next section we provide background and theoretical development of our hypotheses. Remaining sections present the design and methodology used in the empirical investigation followed by presentation of results, conclusions, and suggestions for future research. BACKGROUND Argenti (1976) has studied how and why corporate failure occurs. He proposes three principal types of performance patterns leading to corporate failure, which he describes as follows: (1) fledgling companies for which the "trajectory", or firm performance, never rises above a level he labeled as "poor"; (2) companies that shoot up to very high levels of performance before crashing down; and (3) companies with performance that partially collapses, followed by a relatively longer "plateau" period of sub-par performance, and then rapid decline into insolvency. A number of other researchers have also suggested that the processes resulting in corporate failure can be characterized by a few principle types (see Mutchler, 1985; Menon and Schwartz, 1987; McKeown, Mutchler, and Hopwood, 1991; and Hopwood, McKeown, and Mutchler, 1993). These studies have documented the occurrence of at least two primary failure processes: a relatively rapid, unexpected failure in which financial stress, as proxied by accounting numbers, is not evident (hereafter, "unstressed"); and another process of relatively long duration in which financial stress is evident (hereafter, "stressed"). Hopwood, McKeown, and Mutchler (1994) tested accounting-based statistical prediction models conditioned on these two process types. 6

7 As might be expected, they found that standard accounting-based statistical models can only significantly explain and predict (at rates greater than chance) the "stressed" type of corporate failure. Kane, Richardson, and Graybeal (1996) extended the research on the role of financial stress in the occurrence of corporate failure by examining recession-induced stress and its impact on the ability of standard statistical models to predict the likelihood of corporate failure. They posit two possible effects. First, the effects of recession-induced stress are likely to mimic signals given by failing firms (e.g., lower profitability, deteriorating liquidity, and so forth). As a result, by controlling for recession-induced stress, failure prediction models are better able to distinguish between the effects of recession and a true failure process, thereby improving failure prediction. Second, they argue that some firms mitigate recession-induced stress by anticipating business slowdowns and taking corrective action (e.g., reducing inventory, tightening lending criteria, bolstering liquidity, etc.) These actions reduce default risk in recessionary environments. Accounting information can detect, and thus proxy for, this risk-reducing activity, thereby improving failure prediction. They report evidence that supports both explanations. CORPORATE RESPONSES TO FINANCIAL STRESS Beaver's cash reservoir theory (Beaver, 1966) suggests that the predictable type of corporate failure (i.e., where stress is evident prior to the prediction period) includes manifestations of a positive feedback process that can cause the company s resources to reach depletion and trigger failure. The process is "positive" in the sense that the factors causing the initial stress (i.e., net losses and/or a lack of sufficient liquid resources) are reinforced and amplified. Such a process could unfold as follows. Initially, some event exogenous to the firm (e.g., recession, technological advance on the part of a competitor, etc.) acts to stimulate a condition of financial stress, such as nonprofitability and/or 7

8 insufficient working capital. This initial stress drains the firm's resources and creates more stress, which in turn leads to additional loss of resources. Greater losses can also occur because financial stress increases the risk of default on stakeholder claims. Stakeholders (customers, vendors, creditors, stockholders) react to the increased risk by doing business elsewhere or by demanding a premium (e.g., a sales discount for customers). The positive feedback cycle that ensues amplifies the initial stress. After sufficient repetitions of this process, the initial stressed condition can worsen considerably. The process can continue indefinitely. Alternatively, at some lower point, a new equilibrium develops that reflects a chronic stress condition. The situation nonetheless beckons for intervention. Management may take restructuring actions that are designed to reduce or stop the outflow of resources associated with the resource-draining condition that resulted from the initial stress. Successful actions will increase the company's liquidity, profitability, and efficiency, thereby mitigating financial stress. Third party stakeholders, perceiving the reduced risk, will lower the risk premium (or sales discount) and/or return and become customers and suppliers once again. Thus, a negative (dampening) feedback process ensues and stress is further mitigated. As the process repeats, the likelihood of emergence from a stressed condition increases more and more. Managements, in initiating such restructuring actions, logically would make optimal choices if their companies were rationally managed. This means that, regardless of what decisions are ultimately made in response to ongoing stress, they should be aimed at mitigating financial stress and increasing the likelihood of emergence therefrom. Unfortunately, the pressure to act decisively under stressed conditions may cause managers to react based on limited information that, upon further reflection and with superior information (such as that afforded by hindsight), is not contextually appropriate. Even if such actions are fully informed, however, they may still be insufficiently effective to significantly improve 8

9 the likelihood of emerging from a stressed condition. A related issue surrounds the secondary effects any restructuring action might cause, even if these efforts are initially effective at relieving the stressed condition. For example, consider the way a restructuring action may be perceived by third-party stakeholders. The company may lay off employees, withdraw from unprofitable markets, suspend plans to modernize, and/or otherwise break previous commitments to various stakeholders. If stakeholders view these actions as risk-reducing with respect to their claims on the resources of the company, they will be inclined to assent when asked to improve the stressed firm's profitability by reducing any previously negotiated risk premium (or sales discount). This would in turn increase the likelihood of the company's emergence from a stressed condition. On the other hand, if the new action proves to be ineffective, it may be perceived as a signal that (1) stakeholder claims and contractual agreements are now more likely to be dishonored and/or (2) the attempts to mitigate stress are aggravating the initial stress condition. We conclude, then, that restructuring actions, depending on their impact on key factors such as profitability and liquidity, will increase or decrease the perceived risk inherent in future claims third-party stakeholders have. Thus, because of direct and secondary effects, a restructuring action will have an overall positive (negative) effect and, by decreasing (increasing) the premium (or sales discount), increase (decrease) the likelihood of emergence from corporate failure. As noted above, our focus is on two potentially mitigating responses to financial stress: expansion and contraction of the asset base. Consider the case of a company experiencing financial stress. Such a company has a number of options, including voluntary liquidation, appeals for forgiveness of some or all debt claims, and declaration of either Chapter 7 or Chapter 11 bankruptcy. Prior to utilizing these options, however, the company can seek operational remedies. First, the company's management could employ growth strategies aimed at expanding volume and capacity. For these 9

10 strategies to be successful, two crucial exogenous factors must be present. Market prices for the firm's products and services must exceed the "all-in" costs of new investments that the firm has made to expand volume and capacity. Additionally, there must be sufficient external demand to absorb the company's output at prices that maintain a positive net present value for the new investment projects. The all-in cost of a company, on a per-unit basis, will likely be influenced by the degree of operating leverage (i.e., the relative mix of fixed and variable costs). Some companies, because they have relatively high levels of fixed costs, will be particularly sensitive with respect to profitability across varying levels of productive activity (i.e., the all-in cost per unit will change relatively more). As a result, such companies will be more motivated to use a growth strategy than will other companies in which reductions in unit volume have a much smaller impact on the bottom line, thanks to high levels of variable costs. A second possible response to the occurrence of financial stress is for management to reduce the scope of operations. By contracting, companies hope to restore bottom line profitability, improve the liquidity of their investment structure, and reduce operating capital needs. Contraction can be accomplished by cutting expansionary expenditures such as research and development, withdrawing from marginal or high risk business lines, reducing overhead expenses, selling off non-operating assets such as a corporate headquarters, and increasing operating efficiency through increased labor output demands, shutdown of old obsolete facilities, and so forth. Companies that would be particularly attracted to a contraction strategy include those that have relatively greater proportionate levels of variable costs. Such companies can expect that any attempt to grow through a business downturn will have a smaller payback and thus be more difficult to effect. Focusing on reduced costs (reduction of 10

11 the fixed cost component, together with better variable cost efficiencies) will lower the breakeven point and permit a relatively more effective return to profitability, even at reduced production levels. The other issue that affects the choice of expansion versus contraction is the question of variability of earnings. For example, the better alternative may be contraction if a company is involved in relatively risky ventures that are more likely to produce extreme variability in earnings performance in its efforts to achieve an increased return. In this case, selling off or otherwise dispatching risky investments will improve the chances of emerging from financial stress (i.e., positive profitability and working capital) even while expected returns for the new asset mix are declining. This explanation for choice of contraction is implied by two of the few theories of corporate failure that have been offered: the cash reservoir theory offered by Beaver (1966) and the gambler s ruin theory of Wilcox (1971, 1976). Conversely, companies with a less risky investment mix may be less inclined to choose contraction in response to financial stress. Numerous other factors drive the choice of operational response to financial stress. First, product cycle factors could drive such a choice. For example, firms that produce microchips, when faced with stress, would logically choose expansion to avoid inventory obsolescence. Alternatively, companies that produce basic staples would prefer contraction because their product cycles are at the mature stage. Another example factor is firm size. Companies with minimal levels of accumulated assets have little to lose and thus would be more willing to gamble via expansion. Tradition and political factors also come into play. Some companies have a long history over which they have gained expertise in acquisition. Other companies face regulatory and political difficulty if large numbers of workers are laid off, and would logically choose expansion to avoid the costs associated with a decision to contract. 11

12 Obviously, companies faced with the occurrence of financial stress can simply do nothing. If the financial stress is predicted to be temporary in nature and available resources are expected to be sufficient to pay claims and avoid failure, a firm could choose to not react, or to react in ways that are minimal and involve little or no major change in strategy. Of course, this choice is also the default possibility. If, for example, companies do not perceive that a stressed condition exists, they may not react. Similarly, they may perceive the occurrence of stress but management inefficiency may obstruct their ability to react (e.g., because of poor internal communication or legal restraints on managers ability to react). HYPOTHESIS DEVELOPMENT AND METHODOLOGY The explanations and reasons just presented suggest that controls for the choice of restructuring action (contraction or expansion) may be useful in predicting whether companies can successfully emerge from financial stress and thereby lower default risk. This issue is the subject of our empirical investigation. Because there has been little prior research related to the prediction of financial stress, as opposed to bankruptcy, we assume the processes are related and, as a benchmark, invoke the corporate failure prediction model invoked by Hopwood, McKeown and Mutchler (1994; hereafter, HMM). Our primary criterion for model selection is that the model tested is unbiasedly capable of predicting financial stress, as defined herein, at a rate significantly better than chance alone can explain. The variables included in the HMM model are as follows: (1) net income/total assets (NITA), (2) current assets/total assets (CATA), (3) current assets/current liabilities (CACL), (4) cash/total assets (CASHTA), (5) current assets/sales (CASALE), (6) long-term debt/total assets (LTDTA), and (7) the natural logarithm of firm sales (LSALES). 3 LSALES is included as a control for firm size. Most of these 12

13 variables have well known priors in their role as predictors of corporate failure, and we propose, by extension, financial stress. Net income over total assets is a proxy for profitability--the greater the profitability, the greater the likelihood of emergence from stress. Similarly, the greater the liquidity, as proxied by the current ratio, cash/total assets, and current assets/total assets, the better the chances of emergence from stress. On the other hand, a relatively large ratio of current assets/sales implies a buildup of inventory and an attendant decline in sales and operating income, which can hurt a company's chances of emerging from stress. The greater the leverage a company exhibits, the more difficult it could be to emerge from a stressed condition. Finally, larger companies have more resources available to aid in throwing off a stressed condition. All of these variables are placed into a non-linear model (the logistic regression model) and tested for their contributions to improving the probability of emerging from corporate stress (the dependent variable). We used the model above to test two hypotheses, stated below in null form: (H1) (H2) The HMM variables, described above, are not significantly associated with emergence from corporate stress. Net contraction or expansion of the investment base is not associated with the likelihood of emergence from stress. A company's operational response to financial stress logically could manifest in numerous functional areas, including marketing, credit lending, fixed investment activity, and financing actions. In this research, we focused on only one of these areas- fixed investment activity. We chose this area because (a) it could be operationalized separately from the resulting outcome, and (b) to the extent that such changes were effective, they should be useful as a signal of future emergence from financial stress. When companies expand, they acquire assets with the hope they will yield future economic benefits in excess 13

14 of their costs. To the extent such a strategy is successful, there should be an association of such activity with future emergence from stress. When a company contracts, assets are generally sold or otherwise disposed of. Again, to the extent that such changes result in future economic benefits to the firm through lower costs, improved profitability, and so forth, contraction of the investment base should be useful as a signal of increased likelihood of emergence from financial stress. Further, in the case of contraction, even the intent to reduce the investment base may be reflected in net investment activity. In cases in which sales and restructuring efforts are anticipated, management can control balance sheet presentation by permanently writing down affected asset carrying values and/or reclassify them as part of discontinued operations. Thus, fixed investment activity may also serve as a useful signal of management's future intention to contract, thus increasing its value as a signal of future emergence from financial stress. The notions of growth and contraction can be conceptualized in both the absolute and relative sense. In absolute terms, growth (contraction) would be measured as any positive (negative) change in the investment base. In relative terms, on the other hand, growth (contraction) is operationally defined as positive or negative change in the investment base, as compared to the investment base of a comparison firm or group of firms. 4 The absolute notion of growth and contraction appealed to us because it is simple in nature and relatively easy to operationalize. Unfortunately, it also suffers from two significant disadvantages. First, company managements do not make decisions to expand or contract in isolation. Such choices necessarily take into account the business environment in which such firms operate. Indeed, within a dynamic business environment, one can argue that there is no absolute definition of expansion and contraction--i.e., contraction and expansion are a matter of relativistic perception. From the 14

15 perspective of growth, a slower growing company, while perhaps still experiencing positive changes in its investment base, appears to be shrinking in relative size and force. Although such a company may be expanding in the absolute sense, it is contracting in the relativistic sense. Thus, a decision to grow less than others facing similar conditions may be interpreted as a choice to contract in relative terms. Because an absolute measure of change in the investment base does not account for this relativistic perception, it is arguably misspecified as a proxy for corporate action, at least in environments where dynamic change is constantly occurring. Second, an absolute measure of expansion and contraction presents serious and inherent methodological difficulties. Examples of difficulties associated with observations pooled across time include scaling effects from confounding factors such as inflation; time-dependent confounding events that are macro in nature, such as recession and expansion; and other intertemporal factors that induce cross-sectional dependencies in the data. For these reasons, we chose to utilize in this research a relative definition of expansion and contraction of the investment base. To accomplish this, we first measured the percentage change (increase or decrease) in property, plant and equipment (PPE) from its average for the previous two periods. We then focused on the major methodological issue in the use of a relative notion of change in the investment base--selection of a benchmark for the comparison group. Lacking any theoretical guidance on what might be best, we chose to model strategic choice entirely from the relativistic perspective, with as few assumptions on the function of strategic choice as possible. This was accomplished via a ranking methodology. We ranked the sample companies from lowest to highest growth rate, and grouped the resulting ranked sample into thirds. Indicator variables were then constructed to surrogate for expansion and contraction in separate logistic regressions. For 15

16 the expansion run, companies in the highest one-third were given a nominal value of 1 and all other observations were coded 0. For the contraction run, companies that fell into the lowest one-third were given a nominal value of 1 and all other observations were coded 0. The resulting dummy variables were added to the HMM model to proxy for differences in intercept caused by relative contraction and expansion activity. In addition, to allow for the slopes of the individual variables to be different for relative contraction and expansion activity, we added slope dummy variables composed of each independent variable multiplied by the value of its intercept dummy (1 or 0). The rank-based procedure just described has the advantage of being relativistic. That is, rank transformation is useful because it creates variables that are naturally relative in nature, and it reduces the information demand to that of an ordinal set. Of course, the choice of groupings of the ranked observations is arbitrary. To provide some assurance that our results are not driven by any arbitrary method of grouping, we tested a number of different groupings to determine if the results we report herein appear robust to the grouping choice. 5 Another methodological problem is that some firms may choose to contract some parts of their business operations while expanding other parts. As a result, to the extent that such actions are offsetting, the reported net changes in fixed investment activity will not detect restructuring activity. This mis-specification of restructuring activity would likely bias our results towards insignificant findings. Our prior expectations were that changes in property, plant and equipment (i.e., net fixed asset investment activity) should be significantly associated with the likelihood of emergence from corporate financial stress. This is based on the presumption that expansion and contraction are activities that companies take to mitigate the effects of financial stress. The model contains a profitability variable (the ratio of net income to total assets) that may also detect whether a company will emerge from stress to 16

17 the extent that the mitigating actions affect earnings in the current year. On the other hand, controls for contraction and expansion may better control for the source of profitability and the likelihood of earnings persistence. Furthermore, if companies undergo contraction or expansion to re-establish a non-stressed state quickly, and these actions are (on average) successful, these restructuring actions may act as costly signals to third parties that management believes stress can be mitigated (i.e., positive profitability and/or working capital). This feedback signal may then contribute to actual emergence and have future positive effects on profits and liquidity as these third parties (investors, creditors, customers, and suppliers, for example) increase or maintain business with the company. Sample The sample consisted of companies on the New York Stock Exchange, American Stock Exchange, and NASDAQ exchange that exhibited one of the symptoms defining stress at the balance sheet date during the years The sample is divided into two subsamples--one consisting of 2,789 companies that no longer met a definition of stress one year later, and a second consisting of 9,309 companies that continued to exhibit stress one year later. Those observations with missing data at the second measurement point were screened out. The accounting data were obtained from the COMPUSTAT tapes and directly from annual reports. To find which predictor variables, together with changes in fixed investment, were individually and significantly different one year after parent companies were diagnosed as stressed, we conducted Student's t-tests on the means of each. The results of these t-tests are exhibited in Table 1. Note that all predictor variables are significantly different from their previous amounts (although CASHTA is only marginally so). PPECH (change in property, plant and equipment) is marginally different. Collectively, these descriptive results suggest that choice of mitigating action is relevant to prediction of the likelihood 17

18 of emergence from stress. To provide additional confidence in interpreting the signs of coefficient estimates, we also investigated correlations among model variables. Table 2 presents a correlation matrix of these variables. Pearson correlations are listed above the diagonal and Spearman rank correlations are listed below the diagonal. No that, with the exception of (1) CATA with CACL, (2) CASALE with CACL, and (3) CASHTA with CACL, none of the correlations depicted (Pearson or Spearman) exceed 0.50, suggesting that interpretation of coefficient signs can be interpreted with some confidence. *** TABLES 1 AND 2 ABOUT HERE*** Tests of Improvement in Explanatory Power To test the incremental difference in explanatory power, we examined the reduction in deviance caused by the addition of the new variables. The test statistic is λ( β / β ) = 2log L( β) 2LogL ( β ), (1) where λ( β / β ) 1 2 is defined to be the change in deviance between the full model, labeled β, and the reduced model β 2 (i.e., the contribution of the new conditioning variable set β 1 ), and L is the maximum likelihood estimator for the indicated models. The statistic has an approximate chi-square distribution with k degrees of freedom, where k is the incremental difference in parameters between any two models. 18

19 RESULTS Table 3 presents the results of our test of the first hypothesis, i.e., that the chosen failure prediction model is useful to the prediction of continuing/emerging from a stressed condition. The -2logL statistic gives a relative measure of goodness of fit, such that smaller values indicate more desirable models. The chi-square statistic reflects the reduction in deviance that occurs when additional predictor variables are added to the model. This latter statistic has an approximate chi-square distribution under the null hypothesis that all model predictor variables are zero. In our case, the chi-square statistic of is significant with p < for the model we chose to distinguish continuing stress from emergence from stress. This supports our first hypothesis: the model is significantly associated with the likelihood of emerging from stress. The signs of the coefficient estimates in Table 3 are consistent with our prior beliefs that proxies for profitability (NITA), liquidity (CATA, CACL), and firm size (LSALES) are each positively associated with emergence from stress. Proxies for leverage (LTDTA) and efficiency in managing current assets (CASALES) are negatively associated with emergence from stress. The coefficient for proportion of cash held (CASHTA) is insignificant, suggesting that cash gives little or no additional information for assessing the likelihood of emergence from stress beyond that detected by the broader liquidity variable, CACL. ***TABLE 3 ABOUT HERE*** Table 3 further presents the results of our tests of the second hypothesis. Recall that in testing this hypothesis, we added controls under two approaches. The simplest approach (but the more restricted) is to add a nominal variable for the condition of interest, contraction or expansion. This 19

20 allows the intercept to be different for contraction and expansion. The more complex approach is to allow the slopes of the hyperplanes as well as the intercepts to be different across conditions. Construction of this unrestricted model is accomplished by multiplying each variable observation by its condition control (1 or 0) and adding it to the model. The restricted model is descriptively useful in answering the very basic question of whether, ceterus paribus, knowledge of expansion or contraction is informative to investors in predicting the likelihood of emerging from corporate stress. The unrestricted model, on the other hand, better speaks to the theoretical development we presented earlier--i.e., it can distinguish between investment activity changes that are associated with reported results and actions that are not so related. We report first on the restricted case and then the unrestricted case. The restricted models are shown in Table 3, column 3 for expansion and column 5 for contraction. The change in deviance for the addition of an expansion control is calculated by subtracting the 2logL statistics in columns 3 and 2. Note that the difference between these two statistics is 3.1, which is insignificant (p < 0.100). Furthermore, the expansion indicator variable, EXP, is insignificant. This implies that expansion as a restructuring activity, used alone in a model and with no consideration given to the impact of expansion on reported results, is not informative to investors seeking knowledge about the likelihood of emergence from financial stress. On the other hand, subtracting the 2logL statistics in columns 5 and 2 yields a difference of 7.9, which is significant at p < This is mirrored by the significant coefficient on the contraction indicator variable, CON. Thus, restructuring through contraction gives information related to the likelihood of emergence from stress. The sign of the coefficient is negative, however, suggesting (in the unconditional sense) that contraction is ineffective as a mitigating action and may actually aggravate stress and worsen the odds of emergence from stress. 20

21 Results for the unrestricted models are shown in columns 4 and 6 of Table 3. Note that neither indicator variable, CON nor EXP, is significant in the unrestricted model. Nonetheless, results indicate that knowledge of a restructuring response with respect to the investment base is robustly useful to investors. Furthermore, the effect on reported results of the restructuring action appears to play a significant role in how such information should be interpreted. First, when full controls (both intercept and slope) are added for expansion, the change in deviance is 19.1, which is significant at p < Adding full controls for contraction yields an even more significant change in deviance of 60.4 (p < 0.001). Thus, when models are unrestricted, the ability of both to detect and explain emergence from a stressed condition is significantly improved in comparison to the unconditional base model shown in column 2 of Table 3 (i.e., without controls). Second, inspection of the significant slope variables for the unrestricted models yields interesting results. The interaction of expansion with NITA (column 4) is both positive and significant (coefficient of 1.707), suggesting that, when expansion of the investment base is associated with profitability, profitability is likely to be increasing over time. This signal may be interpreted to suggest that expansion that increases earnings, whether intentional or not, acts to mitigate financial stress and increases the likelihood of emergence from a stressed condition. The interaction of contraction with NITA (column 6) gives an opposite picture. While still significantly associated with the likelihood of emergence from stress, the size of the NITA effect is considerably reduced (by the EXP*NITA coefficient of ). The interpretation is that when companies contract to mitigate financial stress, the initial impact on earnings is often negative. Such a company will face short-term effects such as lost economies of scale, restructuring charges related to contraction, charge-offs and write-downs, losses on the sales of assets, and so forth. Management carries out the contraction, however, in an effort to provide relief from 21

22 financial stress by ultimately improving the level of profitability and liquidity. As a result, because the contraction tends to reduce earnings initially, the relation between current and future earnings is inverted. Other observations can be made based on the data expressed in Table 3. Note that for the case in which companies contract, the intercept is significantly modified (at p < 0.01) for the restricted case but not for the unrestricted case. This result implies that knowledge of contraction, after controlling for reported effects such as the impact on earnings, is insignificantly related to the likelihood of emergence from stress. This suggests that ineffective contraction--i.e., without earnings effects--may aggravate stress, at least in the near term, and weaken the likelihood of emergence in the year to come. On the other hand, when companies attempt to mitigate stress by expanding, knowledge of that expansion is insignificantly associated with emergence from stress. We speculate that this difference emerges because of the relative timing and strength of the contraction vs. the expansion signal. Under generally accepted accounting principles, particularly as the result of applications within the standards of conservatism, contractive actions are more likely to be recognized immediately. For example, such actions are typically accompanied by the establishment of reserves for future asset sales, costs associated with reductions in the labor force, plant shutdowns, and other restructuring actions. Disclosures of impairment when assets are written down and reclassifications when operations are discontinued furnish further signals of contraction. Expansion, however, involves recognition of new assets and liabilities, as opposed to the removal of old assets and liabilities. Furthermore, expansion implies longer-term investment activity, with longer periods over which new costs will be allocated and expensed. Therefore, expansion actions may take longer to be recognized in the financial reports. These differences in timing of the two signals suggest a difference in their power. Such observations 22

23 constitute speculation on our part, however, and as such are beyond the scope of the present inquiry. They may nonetheless prove fruitful for future research. We next address the liquidity variables in the unrestricted models in Table 3. Note that the interaction of contraction with CATA (coefficient of ) is negative and significant with respect to its association with emergence from financial stress one period in advance. This result is consistent with prior expectations derived from the same argument as that presented above for the interaction with NITA. When companies contract, they use up liquid resources in hopes of improving profitability and liquidity in future periods, thus relieving financial stress. Hence, if contracting companies exhibit low levels of liquidity, it signals they are taking actions to relieve stress and will be more likely to emerge than companies that are not taking such actions. On the other hand, all interactions of expansion activity with liquidity measures are insignificantly associated with emergence from stress. As with our above discussion of the intercept term, we speculate that the accounting and reporting for expansion under GAAP explain this result. Finally, a few of the other interactions in our unrestricted models are significantly associated with emergence. These are included in the model for completeness. As opposed to earnings and liquidity effects, however, we have no prior expectations concerning the relationship between an action to mitigate financial stress and these other factors (e.g., leverage and firm size). Any interpretations of these results are thus purely speculative in nature. We do offer the following comments, in the hope that they may be helpful in guiding future research. Leverage, as measured by LTDTA, is directly and significantly associated with emergence from stress (coefficient of 0.811) when expansion is chosen as the mitigating action, but negatively associated with emergence (coefficient of ) when contraction is chosen. This appears to say that expanding firms are able to take on additional debt and use it to aid 23

24 their recovery; the additional earnings can be used to service the new debt. On the other hand, additional debt for contracting companies, for which the influence of earnings on emergence is markedly reduced, simply increases the fixed payments they are facing and becoming increasingly unable to service. Although size does not appear to play a part when using expansion to emerge from stress, it does contribute when contracting. The significantly positive coefficient (0.083) on CON*LSALES implies that larger firms can more appropriately use contraction as a means of emergence than can smaller firms. CONCLUDING REMARKS In this paper, we have shown that publicly available reporting information is useful in estimating the likelihood of emergence from financial stress, a condition that is precursor to the occurrence the types of corporate failure that can be predicted using standard statistical models. We focused on two types of reporting information: data on profitability, liquidity, leverage, size, and associated ratios that have shown to be useful in predicting corporate failure; and relative changes in the fixed investment base. We find that both sets of reporting information are useful in estimating the likelihood of emergence from financial stress. The utility and meaning of information concerning fixed investment activity, however, appears to be driven in part by contextual considerations, as detailed herein. Research on precursor conditions, such as stress, and how they impact measurable default risk, is a relatively new extension of a voluminous literature estimating the likelihood of corporate failure using standard statistical accounting-based models. As a result, numerous additional avenues of research are evident. For brevity we mention only two. First, new research is needed to assess the processes and 24

25 conditions that signal an increased likelihood that corporate stress might lead to distress, i.e., a condition whereby a firm does not have the resources to pay claims and obligations that are required and due and is thus likely to fail. Second, additional research is needed to better understand the dynamic processes that govern emergence from conditions of stress and distress, including how financial reporting information can be used to measure and assess such processes. This paper attempts to extend the literature in this area by focusing on one response to stress: growth/contraction of the fixed investment base. Numerous other responses to stress, however, are possible, including changes in marketing activity, changes in inventory, changes in financial structure, and receivables control, business combinations and mergers, just to mention a few. The relatively recent arrival of non-linear and contextual modeling techniques into the research literature opens the door to exciting research questions in all these topical areas. 25

26 TABLE 1 Predictor variable means, candidate growth indicator means, and t-tests of individual differences between stressed and non-stressed groups: one year following initial stress diagnosis Mean Stressed Non-stressed t Predictor Variables: NITA * CATA * CACL * CASHTA CASALE * LTDTA * LSALES * Candidate Growth Indicator: PPECH Number of stressed observations: 9,841. Number of non-stressed observations: 2,810. These are the annual observations that were available on Compustat for the period 1978 to *Significant at the " <.0001 level using a two-tailed test. Significant at the " <.05 level using a two-tailed test. NITA = Net income/total assets CATA = Current assets/total assets CACL = Current assets/current liabilities CASHTA = Cash/total assets/total assets CASALE = Current assets/sales LTDTA = Long-term debt/total assets LSALES = Natural logarithm of sales PPECH = Change in property, plant & equipment (The percentage change from the previous two-year average of PP&E). 26

27 TABLE 2 Correlation matrices for independent variables and surviving growth indicators a NITA CATA CACL CASHTA CASALE LTDTA LSALES NITA CATA (.1803) CACL CASHTA (.9817) CASALE LTDTA LSALES PPECH (.0078) (.1757) (.0107) (.0004) (.0132) (.0079) (.0594) Pearson correlation coefficients (prob > /R/ under H 0 : Rho = 0) appear above the diagonal; Spearman rank correlation coefficients (prob > /R/ under H 0 : Rho = 0) appear below the diagonal. a The first number listed is the correlation; the number in parentheses is the significance level. Variables are defined in Table 1. 27