30 385 J. 1695 COPY 2 STX BEBR FACULTY WORKING PAPER NO. 90-1695 An Empirical Study of the Corporate Choice Among Equity, Convertible Bonds and Straight Debt: A Cash Flow Components Approach Hei Wat Lee James A. Gentry 3 mo College of Commerce and Business Administration Bureau of Economic and Business Research University of Illinois Urbana-Champaign BEBR FACULTY WORKING PAPER NO. 90-1695 College of Commerce and Business Administration University of Illinois at CIrbana-Champaign October 1990 An Empirical Study of the Corporate Choice Among Equity, Convertible Bonds and Straight Debt: A Cash Flow Components Approach Hei Wai Lee Assistant Professor of Finance University of South Florida and James A. Gentry 1BE Distinguished Professor of Finance University of Illinois, Urbana-Champaign ABSTRACT This study uses cash flow information to examine the validity of the informational signalling hypothesis and the optimal capital struc- ture hypothesis concerning the type of securities offered, given that the companies decide to seek external financing. Earlier empirical studies used financial ratios and data to interpret the choice of securities offered. An objective of this study is to extend these earlier studies by using relative cash flow components to classify and predict the type of securities issued in an offering. Several hypotheses are developed to test empirically the cash flow character- istics of companies that issued straight debt, convertible bonds or common stock. The results of the cash flow based tests support the informational signalling hypothesis, but do not support the optimal capital structure hypothesis. Thus the cash flow information supports the hypothesis that companies offering equity securities are less financially healthy than those offering straight debt. Digitized by the Internet Archive in 2011 with funding from University of Illinois Urbana-Champaign http://www.archive.org/details/empiricalstudyof1695leeh I. Introduction The study of security offerings by publicly held companies has made significant contributions to the financial economics literature. A number of empirical studies have examined the valuation impact of different types of security offering announcements on common stocks. These studies found that companies offering common stocks and con- vertible bonds experience significant negative abnormal stock returns around announcement dates. However, the straight debt offering announcements have no significant economic impact on the value of respective common stocks. The informational signalling hypothesis 2 A shows promise of explaining these empirical regularities. According to the informational signalling hypothesis, security offering deci- sions are interpreted as signalling devices used by insiders to convey their superior information to the market concerning the value and 3 earnings prospect of the firms. In particular, signalling models show that the market interprets equity (debt) security offerings as bad (good) signals of the companies. The empirical findings indicate the more junior the security offered, the more negative the perception of the company by investors. These findings are consistent with the general implication of the signalling models discussed above. There are only a few studies that have used financial data to examine the choice of the securities offered. The early works of Baxter and Cragg [1970], Martin and Scott [1974] and Taub [1975] focused on the choice of securities offered. In general, they found that companies offering equities are relatively highly levered and -2- smaller in size. They also found that companies offered equities when the P/E ratios for their stocks were high. Marsh [1982] used logit analysis to examine the equity and debt offering choice of U.K. companies. He found companies offer equity (debt) when there was a favorable stock (bond) market. Additionally, Marsh concluded that companies behave as if they have a target capital structure in mind. Recently, Billingsley, Lamy and Thompson (BLT) [1988] extended Marsh's study with U.S. data to examine the choice among debt, equity and convertible bond offerings. BLT's major finding was that companies offering convertible bond have financial characteristics more in common with companies offering common stock than those offering straight debt. The Marsh and BLT studies used financial and stock return data of companies offering securities and their results were consistent with the optimal capital structure hypothesis. However, these findings were also consistent with the implications of the informational signalling hypothesis. Thus, in explaining security offering decisions of companies, these studies could not discriminate between the informational signalling hypothesis and the optimal capital structure hypothesis. The objective of this study is to extend these earlier empirical studies by using cash flow information to evaluate the choice of securities offered by companies. The cash flow information is used to examine the validity of the informational signalling hypothesis and the optimal capital structure hypothesis concerning the types of securities offered. This study examines the cash flow characteristics of companies offering securities prior to their offering decisions. -3- Relative cash flow components are used to classify and predict the type of securities offered. This study differs from the previous studies by using complete balance sheet and income statement data of the offering companies; it uses cash flow components developed by Gentry, Newbold and Whitford (GNW) [1985a, 1985b], which means the choice of exogenous variables are not sample dependent; finally, the study examines the trend of the cash flow characteristics of companies offering securities. Part II presents the cash flow model and other financial variables and Part III develops the hypotheses. The data and methodologies are found in Part IV and the empirical results are in Part V. The summary and conclusions are in Part VI. II. Empirical Specification T he Modified Cash Flow Model The Helfert [1982] cash flow model was modified by Gentry, Newbold and Whitford (GNW) [1985a, 1985b] and has twelve major components. The twelve cash flow components (CFC) are net operating (NOF), A receiv- ables (AARF), A inventories (AINVF), A other current assets (AOCAF), A payables, (AAPF), A other current liabilities (AOCLF), A net financ- ing (ANFF), fixed coverage expenses (FCE), net investment (NIF), divi- dends (DIV), A other asset and liability flows (AOA&L) and the change in cash and marketable securities (ACash). A cash inflow has a posi- tive sign and a payment has a negative sign. The algebraic sum of the components is equal to the change in cash and marketable securities. The following equation presents a formulation of the modified cash flow model. -4- NOF + AARF + AINVF + AOCAF + AAPF + AOCLF + ANFF + FCE + NIF + DIV + AOA&L - ACash =0 (1) In order to avoid the overidentif ication problem, the ACash component is omitted. The accounting convention underlying the cash flow statement re- sults in total net cash inflow (TCI) being equal to the absolute value of total net cash outflow (TCO). We have simplified the notation by substituting the expression total net cash flow (TCF) for TCI and TCO, i.e., TCF = TCI = TCO. By dividing each absolute cash flow component by TCF, CFC/TCF, a relative cash flow component (CFC*) is created. The CFC* represents the percentage each component contributes to the firm's total net cash inflow or outflow. The CFC*s are useful in evaluating a firm's financial health. By establishing a hierarchy of the CFC*, the availability of free operating cash flow can be used to assess the trend of a firm's financial strengths and weaknesses. The cash flow model has been successful in the prediction of bankruptcy, 4 bond ratings and loan risk. The hypotheses developed for the study are based on the concept of relative free cash flow components (FCF*). An example of the hierarchy of the CFC* and the FCF* is presented in Table 1. Table 1 shows Company A has 100 percent of its cash in- flows originating from operations. After subtracting the basic outflows for capital investment (40%), dividends (10%), fixed coverage expenditures (5%) and working capital (10%), the FCF* from operations after working capital is 35 percent. In contrast Company D has 25 percent of its cash inflows coming from operations. After deducting -5- the cash outflows for investment (20%) and fixed coverage expenditures (30%), Company D has a FCF* after working capital of a -25 percent. Because the financial health of Company A is substantially better than Company D, it has internal cash flow available for retiring any long- or short-term source of financing or investing in marketable securities. In contrast, Company D has negative internal cash flow and must seek external financing or sell marketable securities to offset the short- fall in FCF* after working capital. Table 1 illustrates several basic concepts related to FCF* and financial health of a company. First, as the percentage of cash inflow coming from net operations declines, there is a decline in FCF* before and after working capital. Under these conditions the firm is likely to seek external financing. Second, as the FCF* before and after work- ing capital declines, the financial health of the firm deteriorates and the riskiness of the firm increases. Alternatively, the example shows the higher the FCF* the lower the risk. Third, as the relative cash inflow from operations (NOF*) decreases from Company A to Company D, the relative cash outflow to investment (NIF*) decreases and the relative cash outflow for fixed coverage expenditures (FCE*) increases. In each of these scenarios there is an increase in risk. The GNW cash flow model is further modified in order to examine the implications of the informational signalling hypothesis and the optimal capital structure hypothesis. The modification relates to the change in net financial flow (ANFF) component, where the ANFF is sub- divided into three parts: A short term borrowing (ASTB), A debt financing (AFIND) and A equity financing (AFINE) components. With the -6- modif ication, there are thirteen (13) relative cash flow components (CFC*) used to test the hypotheses. III. Hypotheses Informational Signalling Hypothesis A primary objective of this paper is to use cash flow information to extend the interpretation of the information signalling hypothesis and the optimal capital structure hypothesis on security offering decisions of publicly-held companies. The study determines if the types of securities offered by companies are associated with their cash flow performance. The following hypotheses are used to focus the empirical examination. It is hypothesized that there is a negative relationship between the size of total cash inflow coming from operations (NOF*) and the seniority of the securities offered. That is . . . Hypothesis 1 : Companies offering more junior securities have a smaller and decreasing fraction of their total cash inflows coming from operations (NOF*). Hypothesis 1 indicates that the declining proportion of cash inflows coming from operations (NOF*), such as Companies B, C and D in Table 1 , is a result of a decrease in sales, an increase in oper- ating costs or a combination of declining demand and increasing operating costs. These events indicate potential profitability and inefficient asset management problems of a financially unhealthy company. In an informational asymmetric framework, Myers and Maj luf -7- [1984] showed that firms, such as Company A, using internal financing to cover cash outflow are preferred to firms that use external financing. According to their model, the market interprets the not- to-issue decision as a positive signal about the company and the equity offering decision as a negative signal. Furthermore, when there is a need for external financing, companies offering debt securities are interpreted by the market as better than those offer- ing equity securities. Given this kind of market perception, a financially healthy company, which has a larger share of its cash inflow coming from operations, would distinguish itself from the less healthy ones by offering debt securities when it needs external financing. Besides, firms like Company A are more likely to choose debt when offering securities because they can sustain a higher debt level and justify the increase in bankruptcy risk, Ross [1977]. Thus, it is hypothesized that companies offering convertible bonds and/or common stocks have a smaller proportion of their cash inflows coming from operations than companies offering straight bonds. In turn, a relatively small percentage of total cash inflow coming from NOF* results in a reduction in the availability of internally generated cash for capital investment expenditures (NIF*) and divi- dends (DIV*). As shown in Table 1, relatively low NOF*, NIF* and DIV* results in greater financial risk. Companies that have more value creating investment opportunities are more profitable and so healthier financially. These companies will have a larger share of their cash outflow going to capital investment expenditure (NIF*). Thus, it is hypothesized that there -8- is a positive relationship between the share of total cash outflow going to net investment (NIF*) and the seniority of the securities offered. As dividends are interpreted as positive signals about the value and future earnings prospect of the company, the larger the share of cash outflow going to dividend payments and the more healthy the company. Thus, it is hypothesized that there is a positive relationship between the dividends (DIV*) and the seniority of the securities offered. That is . . . Hypothesis 2 : Companies offering more junior securities have a smaller and decreasing proportion of their total cash outflows going to net investment (NIF*). H ypothes is 3: Companies offering more junior securities have a smaller and decreasing proportion of their total cash outflows going to dividends (DIV*). The thrust of these two hypotheses is that companies offering con- vertible bonds and/or common stocks have a smaller proportion of their total cash outflow going to NIF* and DIV* than companies that offer straight debt. Alternatively, these hypotheses presume the financial markets perceive companies offering straight debt as having the potential to generate the future NOF* that is required to meet their investment, dividend and debt servicing obligations. It is hypothesized that companies with an increasing trend of deferred income taxes and/or a decreasing trend of other assets are -9- experiencing a growth in operating cash inflows or cash outflows going to investment, or vice versa. Thus, it is proposed that . . . Hypothesis 4 ; There is a positive relationship between net other asset and liability flow (NOA&L*) and the senior- ity of the security type. When a company's NOF* is low or has been declining, a larger and increasing function of its cash inflow comes from external financing (ANFF). Another set of hypotheses follow from the preceding observa- tion. That is . . . Hypothesis 5 : Companies that offer more junior securities have a larger and increasing fraction of their cash inflow coming from external financing (ANFF*). It is not only important to determine that the ANFF* is relatively larger for companies offering junior securities, but it is funda- mental to show that the external capital raised was in the form of common stock. Thus . . . Hypothesis 6 : Companies that offer more junior securities have a larger proportion of their external financing (ANFF*) coming from equity securities (AFINE*) and a smaller proportion coming from short term borrowing (ASTB*) and long term debt (AFIND*). -10- The smaller the proportion of net cash inflow coming from outside borrowing, the smaller the financial risk. However, if it is neces- sary to seek external financing, the source of the inflow makes a dif- ference in discriminating between strong and weak companies. As dis- cussed in the previous section, the general implication of the sig- nalling model is that the market interprets, among different alternatives of external financing, riskless debt borrowing as the least negative signal, while common stock offering as the most nega- tive signal of the company. Given this interpretation, when companies have to seek external financing, the healthier ones prefer debt to equity financing. Within the debt financing, they prefer short-terra riskless borrowing to long-terra risky debt offering. The signalling literature indicates that companies do not issue equity securities unless there are no other choices or their common stock is overpriced in the market. It is expected that companies offering debt securities will have a larger proportion of short term borrowing (ASTB*) or long terra debt (AFIND*), and a smaller share in equity (AFINE*). Thus, it is proposed that there is a positive relationship between the seniority of the security to be offered and the short term borrowing flow (ASTB*) and the debt financing flow (AFIND*) components, but a negative relationship with the equity financing flow (AFINE*) com- ponent. The informational signalling hypothesis does not have direct empirical implications on the remaining cash flow components. The predicted signs of the coefficients for the tests of the informational signalling hypothesis are presented in Table 2. -11- Optimal Capital Structure Hypothesis The optimal capital structure hypothesis is based on the concept that there exists a target debt level for each firm. It is hypothe- sized that companies attempt to maintain or retain a target debt level through their security offerings. Because security offerings are lumpy, companies deviate from their target level after each offering. In managing the target debt level, companies tend to reverse the security type vis-a-vis the last offering, which is the basis of the following hypothesis. Thus, companies that recently offered debt securities choose to offer equity securities at the next offering, and vice versa. It is hypothesized that . . . Hypothesis 7 : Companies that offer more junior securities have a smaller share of their external financing coming from equity securities (AFINE*) and a larger share coming from short-terra borrowing (ASTB*) and long- terra debt (AFIND*). The payment of interest and leasing expenses has a direct affect on the target debt level. The optimal capital structure hypothesis assumes the higher the proportion of the total outflow going to fixed coverage expenditures (FCE*) the greater the chance that a company will offer equity, or vice versa. Thus . . . Hypothesis 8 : There is a negative relationship between the FCE* and the seniority of the security type offered. -12- There are three other variables that are directly or indirectly related to the optimal capital structure hypothesis. They are the deviation of the debt ratio from its target level (DEVLTD) variable, a size (SIZE) variable and the asset composition (ASCOMP) variable. A premise of the optimal capital structure hypothesis is that companies operate with a target debt ratio (long term debt/total capital) and they tend to maintain it. Occasionally, the debt ratio deviates from the firm's target level as a result of changes in re- tained earnings and lumpy security offerings. In this study the proxy for the target ratio is the five year historical average of the debt ratio immediately preceding the year of the offering. The DEVLTD variable is the deviation of the debt ratio at the beginning of the offering year from its historical average. For example, when the debt ratio increases above the target level, the optimal capital structure infers the firm issues equity securities in order to return to the target level, or vice versa. Thus, it is hypothesized that a firm experiencing substantial increase above its target debt level will tend to issue equity securities, and vice versa. It is proposed that . . . Hypo thesi s 9: There is a positive relationship between the DEVLTD variable and the security type of offering. Marsh [1982] developed the SIZE and ASCOMP variables as proxies for determining the target capital structure of offering companies. The SIZE variable is defined as the logarithm of the market equity -13- value of the offering company. The ASCOMP variable is defined as Che ratio of fixed-to-total assets (both net of depreciation). Both variables are measured at the beginning of the offering year. Larger companies are more likely to issue debt securities than smaller com- panies because of an economy of scale effect in issuing costs. In addition, the U.S. bond market is more receptive to larger companies. Companies tend to fund permanent assets with long term financial instruments, therefore, it is expected that companies with a larger fixed-to-total assets ratio will utilize more debt. Alternatively, companies offering more junior securities are smaller in size and utilizes less fixed assets. It is proposed that . . . Hyp othesis 10 : There is a negative relationship between the SIZE and ASCOMP variables with the security type of offering. The predicted signs of the coefficients for the two hypotheses are found in Table 2. IV. Data and Methodologies Common stocks, convertible bonds and straight debt offerings by industrial companies over the ten year period from 1977 to 1986 were identified from the Invest me nt Dealers ' Digest and the Wall St reet Journal Index . The offerings were included in the study if they met the following requirements: 1. The companies were listed on the AMEX or NYSE at the time of the offerings. -14- 2. The offerings are publicly underwritten and registered with the Securities Exchange Commission (SEC) under the tradi- tional method. 3. The offerings are not for swapping other securities nor for the explicit purposes of acquisitions and restructuring. 4. The offerings are primary or combinations of primary and secondary offerings. 5. The companies have only one class of voting common stock. 6. The offering announcements are reported in the Wall Street Journal (WSJ). 7. The offering announcements are not contaminated by other company specific announcements such as mergers, takeovers or organizational changes within a calendar week of the announce- ment date. Based on the above criteria, 540 companies were identified as having offered securities over the ten year horizon. Complete finan- cial information was available on the COMPUSTAT Industrial Annual tape for 247 companies during the study period. The time period used in the study started five years prior to the offering year and ended one year after the offering year. Among the 247 companies, 104 offered straight debt, 37 offered convertible bonds and 106 offered common stocks. Table 3 presents the distribution of the offering companies by calendar year and the number of industries represented for each of the three types of security offered. The information in Table 3 sug- gests that industrial companies tend to offer securities during bull markets, e.g., the years of 1983, 1985 and 1986. This observation is -15- consistent with the Myers and Majluf [1984] adverse selection model. In addition, all subsamples were well represented by different industries. A three part study is used to examine the validity of the infor- mational signalling hypothesis and the optimal capital structure hypothesis in security offering decisions. Additionally the effec- tiveness of cash flow components in predicting the type of security offered is examined. First, the Duncan's Multiple Range test and the F test are used to test the equality of each financial variable at the beginning of the offering year across security types. Next, a series of student t tests are used to determine if there was a change in the mean of each financial variable over time. It is achieved by testing the difference between the mean of each variable measured at the beginning of the offering year and the mean of the four year historical average of the variable up to the beginning of the year preceding the offering for each security type. Lastly, the n- chotoraous multivariate probit model developed by McKelvey and Zavonis [1975] is utilized to examine the classification and predictive abil- ity of the financial variables in security offering decisions. In order to test the predictive ability of the model, the master sample in this study is subdivided into an original sample and a holdout sample. In order to have the two subsamples share similar distribu- tion of security type and economic conditions, the holdout sample con- sists of offerings in the years of 1981 and 1985. The offerings in the other eight years are included in the original sample. As a -16- result, the original sample has 80 straight debt offerings, 28 conver- tible bond offerings and 83 common stock offerings. The holdout sample has 24 straight debt offerings, 9 convertible bond offerings and 23 common stock offerings. In the probit analysis, the dependent variable is the dummy variable for the security type in the offerings. The dummy variable takes the value of 1 for straight debt offerings, and the values of 2 and 3 for convertible bond and common stock offerings, respectively. V. Empirical Results The empirical results are divided into two major sections. The first section is composed of the findings based on cross sectional analysis as well as time series data. The results represent a differ- ence between the means test for the financial variables associated with the three types of securities. The cross sectional test includes the F statistic results from the analysis of variance tests and Duncan's Multiple Range test. The time series section presents the student t test results. The second section presents the findings of the multivariate polychotomous probit model. Cross Sectional Analysis Table 4 presents the results of the F statistics generated by the analysis of variance tests. The null hypotheses that companies offer- ing different types of securities have the same financial character- istics are rejected at the 5% level of significance for five variables. These five hypotheses and the respective variables are: 1-NOF, 3-DIV, 6-FINE, 11-SIZE and 1L-ASC0MP. Contrary to Hypothesis 9, that is -17- related to the optimal capital structure hypothesis, there is no significant difference in the DEVLTD variable for companies offering different types of securities. Because the analysis of variance test does not give the direction of the differences among the various groups, the Duncan's Multiple Range test are also presented in Table 4. Four of the five variables that were significant in the ANOVA tests are also significant at the .05 level. For the three CFC*s, the findings are consistent with the implications of the informational signalling hypothesis. In par- ticular, companies that offer equity securities have a smaller propor- tion of their cash inflow coming from operations (NOF*), and a larger proportion coming from A external equity financing (AFINE*). Companies that offer equity securities have a smaller percentage of the total outflow going to Investment (NIF*) and dividend payments (DIV*). In addition, companies offering straight debt have a larger share of their cash inflow coming from short term borrowing (ASTB*). Contrary to Hypothesis 7, which is associated with optimal capital structure hypothesis, companies that offer debt (equity) securities have more external debt (equity) financing in the preceding year rather than reversing their financing pattern in order to regain the target level. In addition, there is no significant difference in terms of the cash flow going to fixed coverage expenses (FCE*), which is Hypothesis 8. Table 4 shows the optimal capital structure hypothesis is weakly supported by the other variables. First, as seen in the F test, companies offering different types of securities experience similar -18- deviations in their debt ratios from the target level. The supportive findings come from the proxy variables for the target debt level. For companies offering debt securities, Table 4 indicates they are larger in size and have more fixed assets in their asset composition. However, these observations do not prove the optimal capital struc- ture hypothesis. The U.S. bond market is more receptive to larger companies, therefore, it is not surprising that the likelihood of issuing debt increases with the size of the company, because the issuing cost is lower for debt issue. Furthermore, permanent assets grow with the size of a company. Thus, it is not surprising that the companies offering debt have more fixed assets in their total asset compositions than companies offering convertible bonds or common stock. Time Series Anal y sis The recent cash flow performance is compared to a four year aver- age for each set of companies that issued straight debt (SD), con- vertible bonds (CB) and common stock (CS). These comparisons are presented in Tables 5, 6 and 7 for SB, CB and CS, respectively. In the convertible bond group, Table 6, none of the variables experienced significant change throughout the four years, therefore, the discus- sion focuses on the straight debt and common stock groups. For the specific implication of the informational signalling hypothesis that debt offerings provide positive signals while equity offerings carry negative signals, the findings of these t tests are mixed. For the general implication of the information signalling hypothesis that debt -19- offerings are not as bad a signal as equity offerings, the findings are supportive. Tables 5 and 7 show companies that offer straight debt and common stock experience a significant decrease in net operating cash flow (NOF*) over time. The magnitude of the NOF* decrease is much larger for companies offering common stock. On the other hand, companies in both groups increased their reliance on external financing (ANFF*) for cash inflow. At the bottom of Table 5, the results show companies offering straight debt increasingly rely on external debt financing (AFIND*) as a source of cash inflow. Besides, the results in Table 7 show that companies offering common stocks increasingly rely on ex- ternal equity financing (AFINE*) as a source of cash inflow. These observations contradict a basic assumption of the optimal capital structure hypothesis that firms attempt to maintain a balanced capital structure by alternating the type of securities offered. Current DIV* were significantly smaller than the four year average DIV* for companies offering straight debt and common stock, Tables 5 and 7, respectively. However, the size of decrease in outflow going to dividends (DIV*) is markedly lower for the companies offering common stock, than for the companies that issued straight debt, which is supportive of the informational signalling Hypothesis 3. Although the difference between the current and the average outflow going to investment (NIF*) is not statistically significant, the size of in- crease in outflow going to investment (NIF*) for companies offering straight debt is substantially larger than for companies offering common stock, which is supportive of signalling Hypothesis 2. The -20- change in cash outflow going to fixed coverage expenses (FCE*) for the two groups of companies contradicts the optimal capital structure hypothesis presented in Hypothesis 9. Finally, the significant increase in short terra borrowing (ASTB*) as a source of cash inflow for companies offering straight debt over time, as shown in Table 5, is consistent with the implications of the informational signalling hypothesis. In summary, the findings of the t tests are not con- sistent with the optimal capital structure hypothesis, but they are supportive of the informational signalling hypothesis. Multivariate N-Probit Model-Combined Test The probit analysis is a combined test of all three security type groups, plus a series of three pair-wise comparisons. The model is constructed such that the more junior security group takes on a larger value for the dependent variable. Each analysis estimates the proba- bility that the more junior security will occur. Tables 8 and 9 pre- sent the results of the modified cash flow model with and without the three financial variables, respectively. The findings presented in Table 8 provide support to the informa- tional signalling hypothesis. The significant coefficient estimates for AFINE* and DIV* indicate that companies offering more junior securities are more heavily reliant on external equity financing, but commit less to dividend payments. These are signs of a less healthy company. The findings presented in Table 9 give even stronger support to the informational signalling hypothesis. The significant coef- ficient estimates for the ASTB*, AFINE*, NIF* and DIV* variables give -21- further insight that companies offering more junior securities are spending less on investment expenditure and have less cash inflows generated from short term borrowing. The shortage of investment op- portunities is also a sign of a less healthy company. The optimal capital structure hypothesis is not supported by the test results in Tables 8 and 9. The DEVLTD, AFIND* and FCE* variables fail to have a significant coefficient estimate. In addition, the AF1NE* variable is significant but with the wrong sign for its coefficient. Although the SIZE and ASCOMP variables are significant with correct sign for their coefficients, they are not necessarily a result of an optimal capital structure. Rather the results might be due to the institutional fac- tors discussed in the previous section. However, the findings on the SIZE and ASCOMP variables are consistent with those in previous studies. The modified cash flow model was modestly accurate in classifying and predicting the corporate choice of straight debt and common stock offerings. For the extended model, 79% of the companies offering straight debt (SD) and 72% offering common stock (CS) were classified correctly, as shown in Table 8. None of the 28 companies offering convertible bonds were properly classified, but 22 of them were mis- classified as common stock offerings. . Table 8 indicates the predic- tion results for companies offering straight debt and common stock were accurate in 79% and 61% of the tests, respectively. Similarly, no convertible bond offerings were correctly predicted, but two-thirds of them were incorrectly predicted to be common stock offerings. The results of the modified cash flow model are presented in Table 9. -22- Both the classification and prediction performance are inferior to the extended model. The classification and prediction accuracies for straight debt offerings were 74% and 75%, respectively, and 64% and 48%, respectively, for common stock offerings. Pair-Wise Tests The findings in a pair-wise comparison test that involved straight debt offerings were similar to those in the preceding combined test. In a pairwise comparison of straight debt to common stock offerings that use the extended model, Table 10 shows the same set of five variables (SIZE, ASCOMP, ANOA&L*, AFINE* AND DIV*) were significant and with the same signs as those in the combined test. The modified cash flow model results in Table 11 show there were three additional significant cash flow variables, ASTB*, NIF* and FCE*. The sign of the coefficient estimate for FCE* is contradictory to the implications of the optimal capital structure hypothesis, but is consistent with the informational signalling hypothesis. Table 11 indicates that com- panies offering straight debt have more cash inflow from short term borrowing and more cash outflows to investment expenditure and fixed coverage expenses than companies offering common stock. The classi- fication and prediction accuracies of the extended model in this comparison were 75% and 66%, respectively. These results are com- parable to the performance of the Marsh and BLT models. For the modified cash flow model, the classification accuracy was 67% and the prediction accuracy was 66%. / -23- The results of the pairwise comparison of straight debt to conver- tible bonds offerings, that utilize the extended model, are found in Table 10. The significant variables are the SIZE, NOF*, AFINE*, DIV* AND ASTB* variables. All except the NOF* variable had the expected signs. The positive coefficient estimate of the NOF* variable implies that companies offering convertible bonds have a larger share of their cash inflows generated internally than those offering straight debt. Table 11 shows the results of the modified cash flow model are similar to the straight debt-common stock offerings comparison, except that the AFINE* variable was not significant. The accuracy of the classi- fication and prediction tests of the extended model were 89% and 82%, respectively. For the modified cash flow model, the accuracy of the corresponding tests were 78% and 73%. The results of the probit analysis were markedly better than the BLT model. The findings in the convertible bond-common stock offerings com- parison were markedly different from the preceding pairwise com- parisons. There were only two significant variables in Tables 10 and 11, the AFINE* and ASTB* variables. The AFINE* variable had the same sign as the other analyses, but the ASTB* variable had the opposite sign. The test indicates that companies offering common stock had larger proportion of their cash inflows coming from short term borrowing than those offering convertible bonds. The same findings were observed for both the extended and modified cash flow models. In addition, many of the variables in the analysis, though not signifi- cant, had signs that were opposite to the other analyses. This observation is different from those in the BLT study which found that -24- corapanies offering convertible bonds behaved in between those offering straight debt and common stock. The classification and prediction performance results associated with the cash flow variables were not promising. Like the results in the combined test, the cash flow model failed to classify or predict companies offering convertible bond correctly. Given the findings in the Duncan's Multiple Range test, it is not a surprise to observe that companies offering convertible bonds have a lot in common with companies offering common stock. In addition, except the pair-wise probit analysis on the con- vertible bond-common stock offerings, the chi square test statistics of the probit analysis were significant at the 1% level for both the extended and modified cash flow models. These findings indicate that cash flow components are important in explaining the type of security offered by corporate issuers. VI. Summary and Conclusions The financial information used in prior studies to interpret and explain the informational signalling and optimal capital structure hypotheses on security offering decisions of publicly-held companies provided the motivation to use relative cash flow components to predict the type of security issued in an offering. Given that the companies decide to seek external financing, the study found the relative cash flow components provide significant information in classifying and predicting the security type to be offered. The results in this study confirm prior empirical studies that were based on rate of return data. That is, the informational signalling -25- hypothesis is strongly supported, but the optimal capital structure hypothesis is not substantiated by the cash flow information. In summary, when compared to companies offering straight debt, companies offering equity securities have a smaller and decreasing portion of their sources of cash inflows generated from operations, but a larger and increasing portion coming from external equity financing. In addition, they have smaller and decreasing shares of their cash outflows going to capital investment and dividend payments. An anchor of this study is that the financial health, of companies offering equity securities is worse than the health of companies offering straight debt. Additionally, companies that offer equity securities are experiencing a deterioration in their financial health. -26- FOOTNOTES See Dann and Mikkelson (1984), Asquith and Mullins (1986), Eckbo (1986), Masulis and Korwar (1986), and Mikkelson and Partch (1986, 1988). 2 The other hypotheses include the implied changes in cash flow hypothsis, the optimal capital structure hypothesis, the price pressure hypothesis, the tax leverage hypothesis and the wealth redistribution hypothesis. These hypotheses also offer interpretations to immediate stock price reactions to security offering announcements. 3 See Leland and Pyle (1977), Ross (1977), Heinkel (1982), Myers and Majluf (1984), Krasker (1986), Blazenko (1987) and Narayanan (1988). See Gentry, Newbold and Whitford (1985b, 1988, 1990). For the hypotheses in this paper, it is assumed for the dependent variable that the most senior securities, straight debt, would be located close to the intercept on the y axis and the least senior securities, common stock, would be located at the upper region of the y axis. -27- BIBLIOGRAPHY Asquith, P. and D. Mullins. "Equity Issues and Offering Dilution." Journal of Financial Economics 15 (January/February 1986), 61-89. Blazenko, G. "Managerial Preference, Asymmetric Information, and Financial Structure." Journal of Finance 42 (September 1987), 839-862. Barclay, M. and R. Litzenberger. "Announcement Effects of New Equity Issues and the Use of Intraday Price Data." Journal of Financial Economics 21 (May 1988), 71-99. Baxter, N. and J. Cragg. "Corporate Choice Among Long-Terra Financing Instruments." Review of Economics and Statistics 52 (August 1970), 225-235. Billingsley, R. , R. Lamy and G. Thompson. "The Choice Among Debt, Equity, and Convertible Bonds." Journal of Financial Research 11 (Spring 1988), 43-55. Dann, L. and W. Mikkelson. "Convertible Debt Issuance, Capital Structure and Financing — Related Information: Some New Evi- dence." Journal of Financial Economics 13 (June 1984), 157-186. Eckbo, B. "Valuation Effects of Corporate Debt Offerings." Journal of Financial Economics 15 (January/February 1986), 119-152. Gentry, J., P. Newbold and D. Whitford. "Classifying Bankrupt Firms with Funds Flow Components." Journal of Accounting Research 23 (Spring 1985a) , 146-160. . "Predicting Bankruptcy: If Cash Flow's Not the Bottom Line, What Is?" Financ ial Analysts Journal 41 (September/ October 1985b), 47-58. . "Predicting Industrial Bond Ratings with a Probit Model and Funds Flow Components." The Financial Review 23 (August 1988), 269-286. . "Using an Expert System with Inductive Learning to Predict Loan Risk." University of Illinois Working Paper (September 1990). Heinkel, R. "A Theory of Capital Structure Relevance Under Imperfect Information." Journal of Financ e 39 (December 1982), 1141-1150. Helfert, E. Techniques in Financial Analysis . 5th ed. Homewood, 111.: Richard D. Irwin, 1982. -28- Krasker, W. "Stock Price Movements in Reponse to Stock Issues Under Asymmetric Information." Journal of Finance 41 (March 1986), 93-105. Leland, H. and D. Pyle. "Informational Asymmetries, Financial Struc- ture and Financial Intermediation." Journal of Finance 22 (May 1977), 371-387. Marsh, P. "The Choice Between Equity and Debt: An Empirical Study." Journal of Finance 37 (March 1982), 121-144. Martin, J. and D. Scott. "Industry Influence on Financial Struc- ture." Financial Management 4 (Spring 1974), 67-73. Masulis, R. and A. Korwar. "Seasoned Equity Offerings: An Empirical Investigation." Journal of Financial Economics 15 (January/ February 1986), 91-118. McKelvey, R. and W. Zavonis. "A Statistical Model for the Analysis of Ordinal Level Development Variables." Journal of Mathematical Sociology 4 (1975), 103-120. Mikkelson, W. and M. Partch. "Valuation Effects of Security Offerings and the Issuance Process." Journal of Financial Economics 15 (January/February 1986), 31-60. . "Withdrawn Security Offerings." Journal of Financial and Quantitative Analysis 23 (June 1988), 119-133. Myers, S. and N. Majluf. "Corporate Financing and Investment Deci- sions When Firms Have Information that Investors Do Not Have." Journal of Financial Economics 13 (June 1984), 187-221. Narayanan, M. "Debt versus Equity Under Asymmetric Information." J ournal of Financial and Quantitative Analysis 23 (March 1988), 39-51. Ross, S. "The Determination of Financial Structure: The Incentive Signalling Approach." Bell Journal of Economics 8 (Spring 1977), 23-40. D/72 TABLE 1 AN EXAMPLE OF THE HIERARCHY OF RELATIVE CASH FLOW COMPONENTS AND RELATIVE FREE CASH FLOW (FCF*) MEASURES UNDER VARIOUS RISK CONDITIONS Company Relative Cash Flow Measures Net Operating (NOF*) Net Investment (NIF*) Dividends (DIV*) Fixed Coverage Exp. (FCE*) FCF* Before Working Capital A Net Working Capital (ANWC*) FCF* After Working Capital 1 Lowest Highest Risk Risk A B C D 100% 75% 50% 25% -40 "11 -30 -20 60 40 20 5 -10 -15 -20 -_5 -10 -L5 -30 45 15 -15 -25 -10 35% 7% - 5 -20% -25% Balancing Components A Net Financing (ANFFF*) A Net Other A & L (ANOA&L*) A Cash (ACash*) FCF* After All Flows 0% 0% 0% 0% 1 ANWC* = AARF* + AINVF* + AOCA* + AAP* + AOCL* TABLE 2 PREDICTED SIGNS OF COEFFICIENT ESTIMATES FOR THE TWO HYPOTHESES Cash Flow Components Net Operating (NOF*) Short Terra Financing (ASTB*) Long Terra Financing (AFIND*) Equity Financing (AFINE*) Fixed Coverage Expenditures (FCE*) Net Investment (NIF*) Dividend (DIV*) Other A&L (AOA&L*) Informational Signalling Hypothesis + + + Optimal Capital Structure Hypothesis + + Other Variables SIZE ASCOMP DEVLTD TABLE 3 DISTRIBUTION OF SECURITY OFFERINGS BY SECURITY TYPE BETWEEN 1977 and 1986 Year Straight Debt Convertible Bonds Common Stocks Total 1977 6 2 8 1978 1 2 1 4 1979 11 3 14 1980 20 8 13 41 1981 9 4 5 18 1982 11 3 7 21 1983 3 3 33 39 1984 3 2 5 10 1985 15 5 18 38 1986 25 10 19 54 NT 104 37 106 247 NI 58 33 73 123 where NT: total number of offerings. NI: total number of industries represented. TABLE 4 ANOVA F TEST RESULTS AND DUNCAN'S MULTIPLE RANGE TEST RESULTS ANOVA F Straight Debt Convertible Common Stock Variable Statistics 17.25*** (SD) (80) * Bonds (CB) (28) 233.7 (CS) (83) SIZE 1768.8 520.2 ASCOMP 6.84** 0.45974 * 0.35330 = 0.36060 DEVLTD 0.07 0.02069 s 0.02631 = 0.01714 TCF/TA 1.12 4.88*** 1 0.26691 = 0.29554 = 0.28772 NOF* 0.54327 = 0.47028 = 0.44314 ANFF* 0.43 0.12874 = 0.11533 = 0. 16068 ASTB* 3.55* 0.04046 * -0.02951 = 0.00213 AFIND* 1.92 0.15798 = 0.13206 = 0.07309 AFINE* 6.71** -0.02924 = -0.01671 * 0.08759 FCE* 0.30 -0. 11345 = -0.10197 = -0. 11444 NIF* 3.51* -0.40530 = -0.33696 = -0.31378 DIV* 8.68*** -0.09230 * -0.05127 = -0.06240 ANOA&L* 0.34 -0.01161 = 0.00654 = 0.00869 AARF* 0.01 -0.07684 = -0.08264 = -0.07750 AINVF* 0.17 -0.07524 = -0.08926 = -0.06711 AOCAF* 0.02 -0.02068 = -0.02312 = -0.02068 AAPF* 0.24 0.04852 = 0.06535 = 0.05398 AOCLF* 1.06 0.05771 = 0.04407 = 0.03464 1 NOF*s for straight debt sample and common stock sample are signif- icantly different from one another at the 5% level. ***signif icant at the 0.01 level **signif icant at the 0.05 level *signif icant at the 0. 10 level TABLE 5 MEANS AND STANDARD DEVIATIONS FOR CFC* AND OTHER VARIABLES FOR TIME PERIOD t-1, 4 YEAR AVERAGE AND DIFFERENCE BETWEEN THE TWO MEANS FOR COMPANIES THAT ISSUED STRAIGHT DEBT (ORIGINAL SAMPLE; n = 80) SIZE ASCOMP DEVLTD TCF/TA NOF* ANOA&L* ANFF* AFIND* AFINE* FCE* NIF* DIV* ACash* AARF* AINVF* AOCAF* AAPF* ASTB* AOCLF* t- -1 4 Year Mean Average S.D. Differe Mean nee Mean S.D. S.D. 1768.77 2123.07 — — — — 0.4597 0.1972 — — — — 0.0207 0.0953 — — — — 0.2669 0. 1057 0.2517 0.0792 0.0152 0.1014 0.5433 0.2136 0.5825 0. 1524 -0.0392** 0.1758 -0.0116 0. 1588 0.0167 0.0553 -0.0283 0. 1642 0. 1287 0.2315 0.0562 0. 1573 0.0725** 0.2827 0. 1580 0.2359 0.0603 0. 1153 0.0977*** 0.2852 -0.0293 0.1845 -0.0040 0.0996 -0.0253 0.1916 -0.1135 0.0792 -0.1055 0.0554 -0.0079 0.0583 -0.4053 0.2064 -0.3752 0. 1725 -0.0301 0. 1870 -0.0923 0.0594 -0.1006 0.0585 0.0083** 0.0366 0.0172 0. 1419 -0.0363 0.0729 0.0535*** 0. 1796 -0.0768 0. 1335 -0.0682 0.0709 -0.0087 0. 1425 -0.0752 0. 1322 -0.0439 0.0927 -0.0313** 0. 1424 -0.0207 0.0549 -0.0075 0.0357 -0.0132* 0.0667 0.0485 0.0923 0.0422 0.0451 0.0063 0. 1094 0.0405 0. 1167 0.0066 0.0378 0.0338** 0. 1336 0.0577 0.0762 0.0397 0.0522 0.0180* 0.0936 ***signif icant at the 0.01 level **signif icant at the 0.05 level *signif icant at the 0.10 level TABLE 6 MEANS AND STANDARD DEVIATIONS FOR CFC* AND OTHER VARIABLES FOR TIME PERIOD t-1 , 4 YEAR AVERAGE AND DIFFERENCE BETWEEN THE TWO MEANS FOR COMPANIES THAT ISSUED CONVERTIBLE BONDS (ORIGINAL SAMPLE; n = 28) t- 1 4 Year Mean Average S.D. Diff Mean erence Mean S.D. S.D. SIZE 233.73 242.17 — — — — — ASCOMP 0.3533 0.2180 — — — — DEVLTD 0.0263 0.1217 — — — — TCF/TA 0.2955 0.0960 0.2747 0.0893 0.0208 0.0772 NOF* 0.4703 0. 1484 0.4893 0.1925 -0.0190 0.1627 ANOA&L* 0.0065 0.1133 -0.0066 0.0644 0.0132 0.1307 ANFF* 0.1153 0.3096 0. 1086 0.2160 0.0067 0.2995 AFIND* 0. 1321 0.2807 0.0908 0. 1648 0.0412 0.3221 AFINE* -0.0167 0.2296 0.0178 0. 1278 -0.0345 0.1851 FCE* -0. 1020 0.0619 -0. 1112 0.0603 0.0092 0.0616 NIF* -0.3370 0.2379 -0.3045 0. 1818 -0.0325 0. 1689 DIV* -0.0513 0.0382 -0.0521 0.0381 0.0009 0.0196 ACash* -0.0164 0.1714 -0.0290 0.0810 0.0126 0.1918 AARF* -0.0826 0. 1894 -0.0840 0.1145 0.0013 0.1899 AINVF* -0.0893 0.2202 -0.0838 0. 1275 -0.0055 0.2298 AOCAF* -0.0231 0.0766 -0.0093 0.0158 -0.0138 0.0785 AAPF* 0.0654 0.0913 0.0414 0.0519 0.0240 0.0815 ASTB* -0.0295 0. 1157 -0.0024 0.0867 -0.0271 0.1620 AOCLF* 0.0441 0. 1258 0.0412 0.0481 0.0029 0.1141 ***signif icant at the 0.01 level **signif icant at the 0.05 level ^significant at the 0. 10 level TABLE 7 MEANS AND STANDARD DEVIATIONS FOR CFC* AND OTHER VARIABLES FOR TIME PERIOD t-1, 4 YEAR AVERAGE AND DIFFERENCE BETWEEN THE TWO MEANS FOR COMPANIES THAT ISSUED COMMON STOCKS (ORIGINAL SAMPLE; n = 83) t- ■1 4 Year Mean Average S.D. Differe Mean nee Mean S.D. S.D. SIZE 520.24 1068.23 — — — — ASCOMP 0.3606 0. 1629 — — — — DEVLTD 0.0171 0. 1228 — — — — TCF/TA 0.2873 0. 1077 0.2632 0.2632 0.0240* 0. 1127 NOF* 0.4431 0.2179 0.5300 0. 1479 -0.0869*** 0.251 ANOA&L* 0.0087 0. 1815 -0.0002 0.0531 0.0089 0. 1993 ANFF* 0. 1607 0.2922 0.0879 0. 1618 0.0727** 0.3383 AF1ND* 0.0731 0.3163 0.0635 0. 1338 0.0096 0.3738 AFINE* 0.0876 0.2322 0.0245 0. 1006 0.0631** 0.2532 FCE* -0. 1144 0.0768 -0. 1180 0.0648 0.0035 0.0669 NIF* -0.3138 0.2352 -0.3099 0. 1408 -0.0039 0.2182 DIV* -0.0624 0.0550 -0.0752 0.0603 0.0128*** 0.044 ACash* -0.0452 0.2076 -0.0291 0.0633 -0.0161 0.2245 AARF* -0.0775 0.1738 -0.0803 0.0834 0.0028 0. 1889 AINVF* -0.0671 0.2007 -0.0974 0. 1062 0.0303 0.2046 AOCAF* -0.0207 0.0505 -0.00635 0.0214 -0.0143** 0.0513 AAPF* 0.0540 0. 1297 0.04672 0.0545 0.0073 0. 1412 ASTB* 0.0021 0. 1463 0.0116 0.0669 -0.0095 0. 1848 AOCLF* 0.0346 0. 1128 0.0518 0.0490 -0.0171 0. 1229 ***signif icant at the 0.01 level **signif icant at the 0.05 level *significant at the 0. 10 level TABLE 8 PROBIT COEFFICIENTS FOR TEST INCLUDING CASH FLOW COMPONENTS AND OTHER FINANCIAL VARIABLES Name Coefficient Intercept SIZE ASCOMP DEVLTD TCF/TA NOF* ASTB* AFIND* AFINE* FCE* NIF* DIV* ANOA&L* AARF* AINVF* AOCAF* AAPF 1.2056* -0.0003*** -1.7640** 0.7312 0.5553 0.8430 -0.7216 0.4385 2.2822*** -0.1503 0.6098 4.0414* 1.7738** 1.2448 1.1830 1.0396 0.3590 R en X 60.67*** 40.96% -162.22 Classification and Prediction Accuracy Security Sample Classification Type Size 80 n (%) SD 63 (79%) CB 28 (0%) CS 83 60 (72%) Total 191 123 (64%) Security Sample Pre diction Type Size n (%) SD 24 19 (79%) CB 9 (0%) CS 23 14 (61%) Total 56 33 (59%) ***signif icant at the 0.01 level **signif icant at the 0.05 level *signif icant at the 0. 10 level TABLE 9 PROBIT COEFFICIENTS FOR TEST INCLUDING CASH FLOW COMPONENTS Name Coefficient Intercept 1.4489** TCF/TA -0.2954 NOF* 0.4236 ASTB* -1.5376* AFIND* 0.8041 AFINE* 1.8979*** FCE* 2.0201 NIF* 1.9311*** DIV* 5.7548*** ANOA&L* 1.9556** AARF* 1.0260 AINVF* 0.7178 AOCAF* 0.1185 AAPF* 0.3119 xl 39.36*** R 26.80% en X -172.88 Classification Prediction Accuracy Sample Security Size T yp e _ _* SD CB CS Total 80 28 83 191 Sample Size n Security T yp e SD 24 CB 9 CS 23 Total 56 Classification Accuracy n (%) 59 (74%) (0%) 53 (64%) 112 (59%) Pre diction Ac curacy n (%) 18 (75%) (0%) 11 (48%) 29 (52%) ***signif icant at the 0.01 level **signif icant at the 0.05 level *signif icant at the 0.10 level TABLE 10 PROBIT COEFFICIENTS FOR PAIRWISE TESTS USING THE EXTENDED MODEL Coefficients SD & CS SD & CB CB & CS Intercept 2.0011** 0.0676 0. 1647 Size -0.0002** -0.0029*** 0. 0009 ASCOMP -2.4835** -3.3529 -0. 8984 DEVLTD 0.8778 0.4580 1. 1833 TCF/TA -0.3465 4.2032 -0. 1109 NOF* 0.9055 4.8789** -0. 2488 ASTB* -1.5689 -7.6492** 2. 8217* AFIND* 0.6053 0.6762 -0. 0339 AFINE* 2.5010*** 3.0613* 2. 3799** FCE* 1.2570 1.4164 -3. 2843 NIF* 0.9288 0.4607 -0. 1474 DIV* 5.2491** 9.7768* -4. 3103 ANOA&L* 2.2388** 2.8306 0. 8210 AARF* 1.7704 2.6914 0. 1836 AINVF* 1.7200* 1.2042 0. 7542 AOCAF* 1.2032 -0.1670 0. 2157 AAPF* 0.7324 -1.9585 0. 6073 2 57.26*** 71. 12*** 13. 34 49.68% 97.03% 52. 37% dn X -84.32 -26.24 -53. 52 Classification and Pre idiction Ac xuracy Sample Classification Security Size SD & CS SD 6 . CB CI 5 & CS Type n n 59 (%) (74%) n 74 (%) (93%) n (%) SD 80 __ CB 28 - 22 (79%) 4 (14%) CS 83 63 (76%) — 78 (94%) Total 191 Sample L22 (75%) 96 (89%) 82 (74%) Security Size Prediction Type n n 19 (%) (79%) n 23 (%) (96%) n (%) SD 24 __ CS 9 - 4 (44%) 1 (11%) CB 23 12 (52%) — 23 (100%) Total 56 31 (66%) 27 (82%) 24 (75%) ***signif icant at the 0.01 level **signif icant at the 0.05 level *significant at the 0. 10 level TABLE 11 PROBIT COEFFICIENTS FOR PAIRWISE TESTS USING THE MODIFIED CASH FLOW MODEL Coefficients SD & CS SD & CB CB & CS Intercept 1.8699** 1.4804 0.0243 TCF/TA -1.1118 -0.0403 0.4156 NOF* 0.5005 1.3603 -0.2491 ASTB* -2.3557** -6.3247*** 3.2580** AFIND* 1.1803 0.9063 -0.0224 AFINE* 2.1612** 1.0262 2.0156** FCE* 3.4612* 7.1116** -3.8156 NIF* 2.5019*** 2.7238** 0.3690 DIV* 6.4449** 13.2768*** -5.7323 ANOA&L* 2.2858** 2.6458* 0.2556 AARF* 0.9361 1.7563 0.0764 AINVF* 1.0032 0.3065 0.7433 AOCAF* -1.0713 -0.7309 -0.4339 AAPF* 0.2871 1.3645 0.5355 2 R x > 37.81*** 34.74*** 14.56 34.07% 49.96% 24.37% en \ -94.05 -44.43 -55.41 Classification and Prediction Accuracy Sample Class: .f ication Security Size SD & CS SD & CB CB & CS Type n n 55 (%) (69%) n 72 (%) (90%) n (%) SD 80 __ CB 28 — 12 (43%) 4 (14%) CS 83 54 (6 5%) — 80 (96%) Total 191 Sample 109 (67%) 84 (78%) 84 (76%) Security Size Prediction Type n n 18 (%) (75%) n 20 (%) (83%) n (%) SD 24 CS 9 — 3 (33%) 1 (11%) CB 23 13 (57%) -- 22 (96%) Total 56 31 (66%) 23 (73%) 23 (72%) ***signif icant at the 0.01 level **signif icant at the 0.05 level *signif icant at the 0.10 level