Financial contracting and managerial flexibility.
In this paper, we examine the efficacy of provisions in warrant agreements that allow managerial discretion in the timing of raising capital. We identify three flexibility provisions in warrant agreements: the right to call the warrant, the right to extend the life of the warrant, and the right to lower the exercise price of the warrant. Our purpose is to explain the cross-sectional presence or absence of these provisions in warrant agreements.
Why might firms include flexibility provisions in warrant agreements? In their seminal article, Smith and Warner (1979) identify two possible explanations. The first of these, the irrelevance hypothesis, states that flexibility provisions are, in essence, neutral mutations--their inclusion or exclusion has no economic consequences. This theory predicts that flexibility provisions will appear randomly across firms, a prediction that is soundly rejected by our data.
The second hypothesis advanced by Smith and Warner is the costly contracting hypothesis. It acknowledges that contracting is a costly undertaking. Consequently, only those provisions in contracts whose benefits exceed their costs will be observed. The costly contracting hypothesis is described in more recent literature as optimal or efficient contracting (cf., Core and Guay, 1999).
The costly contracting hypothesis predicts that the flexibility provisions will be included when the benefits of flexibility are substantial. Kahan and Yermack (1998) show empirically that a firm's investment opportunities are negatively related to the presence of covenants for convertible bonds. They argue that covenants that restrict managerial decisions are costly in the presence of growth opportunities, as they may prevent managers from pursuing profitable projects. The inclusion of flexibility provisions in warrant agreements is analogous to the exclusion of bond covenants. The economic logic of Kahan and Yermack suggests that the flexibility provisions will be positively related to growth opportunities. This relation is the central prediction of the costly contracting hypothesis.
Chalmers, Dann, and Harford (2002) present a third hypothesis, that of managerial opportunism. In this narrative, managers have influence over which provisions are included, and provisions are included when they generate private benefits to the managers. In more recent literature, this has been referred to as an agency explanation (cf., Bebchuk and Fried, 2004, for an application in the context of executive compensation). For example, the provision that allows managers to lower the exercise price of the warrants might allow for an inflow of capital and continuation of the business (and its managers' salaries) even if the firm has no economic viability.
The managerial opportunism hypothesis is difficult to test directly. The ability of managers to engage in self-serving behavior, however, is affected in part by the information environment. If the firm and its managers' actions are transparent, it will be difficult for managers to behave in a manner contrary to the best interests of shareholders. In contrast, if there is a high degree of information asymmetry, such behavior is more likely. Thus, the central prediction of the managerial opportunism hypothesis is a positive relation between measures of information asymmetry and inclusion of the flexibility provisions.
An additional avenue for distinguishing the costly contracting hypothesis from the managerial opportunism hypothesis is to examine the quality of the governance of the sample firms. If a firm has a strong board of directors, the opportunities for significant agency problems are reduced. In this setting, the board is more likely to allow flexibility provisions because the cost of those provisions is mitigated by board monitoring. Consequently, the costly contracting hypothesis predicts a positive relation between the inclusion of flexibility provisions and corporate governance.
In contrast, a weak board is more likely to allow self-serving provisions to be included. Consequently, the managerial opportunism hypothesis predicts a negative relation between the inclusion of flexibility provisions and corporate governance. Our final test examines the long-run performance of firms with and without flexibility provisions in their warrant agreements. If the provisions constitute an agency problem, firms with the provisions will underperform. In contrast, if the provisions are chosen when it is efficient to do so, there will be no cross-sectional difference in performance between the two samples (Demsetz and Lehn, 1985).
We find that the extension provision and the exercise-price-reduction provision are related, and that firms typically choose either both or neither of them. Inclusion of the call provision is independent of decisions about the other two provisions. The evidence shows that the inclusion of the extension and exercise-price-reduction provisions is associated with proxies for growth opportunities. This result supports the costly contracting hypothesis. We find no evidence that including these provisions is associated with measures of information asymmetry, as predicted by the managerial opportunism hypothesis. Additionally, there is a positive association between the inclusion of these two provisions and measures of better corporate governance, suggesting that a firm in which governance structure is stronger is more likely to grant its managers a higher degree of flexibility. We also show that the long-run performance of firms following warrant issuance does not depend on the inclusion or exclusion of the flexibility provisions, consistent with the costly contracting hypothesis.
Sample Selection and Descriptive Statistics
Our sample is formed by looking for the prospectuses of all warrants outstanding in 1995 according to the Irwin Yearbook of Convertible Securities.  We are able to find information on 421 warrant offerings by 368 different firms. For each issuing company, we also read annual reports and press releases to collect additional data about the companies and their warrants.
We classify warrants by their origins. The public offerings sample includes the 369 observations associated with initial public and seasoned equity offerings (328), preferred stock and rights offerings (21), as well as other warrant offerings. The reorganizations sample contains the remaining 52 observations from Chapter 11 proceedings (11), mergers (16), private placements, and lawsuit settlements.
Our sample shows some concentration in the years 1992 (20 percent of the observations), 1993 (31 percent), and 1994 (31 percent). This concentration occurs because our sample selection starts with warrants that were outstanding in 1995. Our sample covers 49 different industries.
We collect data from Compustat and the Center for Research in Security Prices (CRSP) and compare characteristics of the firms in the public offerings and reorganizations samples. We measure all variables in the year prior to warrant issuance. Consistent with Schultz (1993) and How and Howe (2001), the public offerings sample firms are smaller, have less debt than the reorganizations sample firms, and appear to be young, growth-oriented firms. They have a higher ratio of market value of equity to book value of equity and higher R&D expenses by sales than the reorganizations sample firms.
Managerial Flexibility Provisions
Jensen and Meckling (1976), Cheung (1983), and others have described a firm as a nexus of contracts. Debt contracts are the most widely investigated financial contracts of firms. In their seminal article, Smith and Warner (1979) analyze the features of U.S. corporate bond contracts, with emphasis on how those contracts are written to mitigate conflicts between bondholders and stockholders. They propose the irrelevance and costly contracting hypotheses. The latter posits that the use of covenants in financial contracts can increase the value of the firm even though such provisions have opportunity costs.
Table 1 shows the distribution of the three provisions in warrant agreements that allow for managerial flexibility: the right to call the warrant, the right to extend the life of the warrant, and the right to lower the exercise price of the warrant. Callability is the most common provision and appears in 86 percent of the warrant agreements. Forty-one percent of the agreements include an extendibility provision. The option to lower the exercise price appears in 39 percent of the sample warrants. For the public offerings sample, 92 percent of the warrant agreements contain a call provision, 44 percent contain an extendibility provision, and 40 percent contain a provision to lower the exercise price.
We examine each option by first giving an example and then analyzing the economic rationale for its inclusion. We discuss the costs of these managerial options.
Commencing on the Initial Warrant Redemption Date, the Company may, on 30 days prior written notice, redeem all, and not less than all, the outstanding warrants at $0.20 per Warrant, provided, however, that before any such call for redemption of Warrants can take place, the (A) high closing bid price for the Common Stock in the over-the-counter market as reported by the NASD Automated Quotation System or (B) the closing sale price on the primary exchange on which the Common Stock is traded, if the Common Stock is traded on a national securities exchange, shall have for twenty (20) consecutive trading days subsequent to the Initial Warrant Redemption Date equaled or exceeded $9.00 per share (subject to adjustment in the event of any stock splits or other similar events as provided in Section 8 hereof)
Veterinary Centers of America, Inc.,
Warrant Agreement dated October 10, 1991
Analysis: The initial warrant redemption date is listed as April 10, 1992, allowing a call protection period of six months. The exercise price of the warrant is $7.20 per share. Because the stock price must have been greater than $9.00 before the warrant could be called, a redemption announcement would force conversion.
Most warrants are callable (redeemable), often at a nominal price. Of the 338 callable warrants in our public offerings sample, 196 (58 percent) allow for a call protection period during which the warrants cannot be called. In our sample, the mean (median) call protection period is 11.1 (12) months. Warrants are called in order to force exercise in much the same way that in-the-money convertible bonds are called to force conversion. Callability allows managers to finance the exercise of growth options with greater timing flexibility.
Warrant Expiration Date shall mean, unless the warrants are redeemed as provided in Section 9 hereof prior to such date, 5:00 p.m. (New York time) on October 10, 1996 ... subject to the Company's right, prior to the Warrant Expiration Date, in its sole discretion, to extend such Warrant Expiration Date ...
Veterinary Centers of America, Inc.,
Warrant Agreement dated October 10, 1991
Analysis: Warrants can be extended if managers feel that the optimal time to exercise a growth option is after the warrants are scheduled to expire. An extension makes sense especially if, at the time of the announcement of the extension, the warrants are expected to be out of the money at their original expiration date, but have a good chance of later becoming in the money (e.g., after favorable information about the firm's investment opportunity set has been released to the market). A firm might be unwilling to let the warrants expire and seek other sources of financing because the transaction costs faced by a small growth firm are likely to be high. As with callability, extendibility provides managers with greater timing flexibility.
Discretionary Reductions in Warrant Exercise Price
Purchase Price shall mean, subject to modification and adjustment as provided in Section 8, $7.20 and further subject to the Company's right, in its sole discretion, to decrease the Purchase Price for a period of not less than 30 days on not less than 30 days' prior written notice to the Registered Holders.
Veterinary Centers of America, Inc.,
Warrant Agreement dated October 10, 1991
Analysis: The requirement that the reduction in exercise price last at least 30 days is common in warrant agreements. The reduction and notification requirements prevent self-dealing by warrant-owning corporate officials and to allow time for warrant holders to learn of the exercise price reduction and to act on it.
Approximately half of the sample warrants grant the company the right to lower the exercise price of a warrant. Howe and Su (2001) show that these reductions typically involve warrants that are out of the money before the reduction and in the money after the reduction. Further, about half of the reductions are short term and thus can be viewed as exercise-forcing. A short-term reduction in exercise price allows managers to control the timing of financing, much like the call of a warrant. The new shares are issued at the new, lower exercise price, but this disadvantage is balanced by the certainty of receiving the capital infusion.
Costs of Managerial Flexibility
The benefits of managerial discretion are not without costs. Each of the provisions described above permits managers to influence the timing of the raising of capital. One disadvantage of these provisions is that any capital infusion represents cash that could be squandered by managers.
The extension provision allows the warrants to remain alive even though the managers have not shown that the firm has profitable investment opportunities. The reduction provision allows managers to induce an equity offering even if the stock price has languished below the (old) exercise price. A call provision places an upper limit on warrant holder gains, and rational investors will pay less for a warrant when a call provision is present. The inclusion of flexibility provisions creates the possibility for opportunistic behavior by managers. We thus describe them as potentially costly.
Before we begin our statistical analyses, we re-examine Table 1. The data presented there suggest that the warrant provisions in the public offerings sample differ materially from the warrant provisions in the reorganizations sample. For example, 46 percent of the reorganizations have none of the flexibility provisions, but only 5 percent of the public offerings have none. Further, 34 percent of the public offerings include all three provisions, but only 15 percent of the reorganization sample does. These findings are not consistent with the irrelevance hypothesis, which predicts that contract characteristics will be randomly distributed across firms. Consequently, we disregard the irrelevance hypothesis in the analysis that follows.
Associations among the Call, Reduction, and Extension Provisions
We examine the interrelations among the three contract provisions, which we treat as categorical variables. We use C to represent callability, E to denote extendibility, and R to represent the option to reduce exercise price. We wish to investigate whether firms consider the flexibility provisions as complements, substitutes, or independent of one another.
To determine the associations among flexibility provisions in warrant contracts, we fit a loglinear model to our data. Loglinear models are appropriate for our investigation because none of the provisions can be treated as a dependent variable relative to the others. As described in the appendix, we find that the loglinear model that best fits the data requires the inclusion of the interaction term between E and R, but no interaction effects with C, and no three-way interaction. This finding suggests that the call provision and the extension and reduction provisions are independent, which indicates that companies view the decision to include a call provision as separate from the decision to include the extension and reduction provisions. In contrast, the probability of the inclusion of E when R is included compared to cases in which R is not included is 81.5 percent. The probability of not including E when R is not included compared to cases in which R is included is 90.7 percent. Firms appear to consider the extension and reduction provisions as complements, often including either both or neither of them.
Determinants of the Flexibility Provisions
The costly contracting hypothesis suggests that if flexibility provisions are efficient contractual provisions, they should be included only when their benefits outweigh their costs. This would be the case for firms for which flexibility is more valuable, namely, companies with substantial investment (growth) opportunities.
Following Barclay and Smith (1995), we use the ratio of the market value of equity to the book value of equity (MTBE) as our primary proxy for growth opportunities. In robustness checks reported later, we use five other measures of growth opportunities.
Alternatively, the flexibility provisions might create an agency conflict in which managers' interests are not well aligned with shareholders' interests--the managerial opportunism hypothesis. For example, managers could exercise the call provision even though there are no profitable projects in which the firm might invest. The extension provision allows the warrants to remain outstanding even though the managers have not been able to raise the stock price through value-revealing investments. The exercise-price-reduction provision allows managers to create an inflow of capital to the firm even if the stock price remains below the original exercise price.
The managerial opportunism hypothesis suggests that the flexibility provisions will be related to the degree of information asymmetry about the future actions of the managers. (See, for example, the commitment hypothesis of Brav and Gompers, 2003.) Under the managerial opportunism hypothesis, higher information asymmetry will be associated with a higher likelihood of inclusion. Following Krishnaswami, Spindt, and Subramaniam (1999) and others, we use the residual standard deviation (RSD) from the market model as our primary proxy for information asymmetry.
In our sample of public offerings, 257 (71 percent) observations are associated with IPOs. For these companies, we compute the ratio of the market value of equity to the book value of equity using the market value 20 days after the first trading day for these firms. To compute the residual standard deviation for IPO firms, we estimate their market model residuals using days +20 to +270 relative to the first trading day. For non-IPO firms, we measure the market value of equity 20 days prior to the offering and residual standard deviation using days -270 to -20 relative to the offering day. In our results, an indicator variable for IPO firms is never significant and thus is not shown.
We control for size by using the market value of equity and control for leverage by using total debt (long-term debt plus short-term debt) as a percentage of assets the year prior to the offering. Results are similar when we use total assets at the end of the fiscal year prior to the offering year, rather than the market value of equity, to control for size.
Determinants of the Extension and Reduction Provisions
Given the evidence that the call provision is independent of the extension and reduction provisions, we collapse the analysis over C and further explore the determinants of E and R.
Table 2 displays descriptive statistics by groups based on whether the extension and/or reduction provisions are included in the warrant agreement. The median RSD for firms that include both E and R is 5.71 percent, 5.05 percent for firms that only include E, 5.83 percent for firms that only include R, and 5.65 percent for firms that include neither E nor R. The Kruskal-Wallis test for differences among the four groups does not reject the null that the group medians are equal. (The p-value is 0.6709.) When we compare only firms that include either both provisions (ER) or none (NN), the Wilcoxon rank sum test does not reject the null that the two groups' medians are equal over the alternative that RSD is larger for firms that include both provisions. The p-value for the one-sided test is 0.3526. The finding that there are no differences in information asymmetry between firms that include the flexibility provisions and those that do not does not support the managerial opportunism hypothesis.
The median MTBE for firms that include both provisions is 3.85, 3.70 for firms that include only E, 3.09 for firms that include only R, and 3.38 for firms that include neither E nor R. There is no evidence that the four groups are different (the p-value for the Kruskal-Wallis test is 0.1914). There is evidence that firms that include both provisions have more investment opportunities (higher market-to-book ratios) than do firms that include none (the p-value for the one-sided Wilcoxon rank sum test is 0.0698), which supports the costly contracting hypothesis.
Table 2 also shows values for the market value of equity, free cash flow, and leverage. Firms that include both provisions have a smaller median market value, less negative free cash flow relative to assets, and lower financial leverage than do firms that include neither. None of the differences, however, is significant at conventional levels.
To further distinguish between the costly contracting and managerial opportunism hypotheses, we run a logistic regression in which the response variable can take four levels, corresponding to the decisions to include E and R. The base case does not include either provision.
Table 3 reports results using our primary proxies for growth opportunities and information asymmetry, MTBE and RSD. Because MTBE is skewed, we use its natural logarithm (lnMTBE) in the analysis. We note that none of our results materially change if we use MTBE (without taking natural logarithm) instead. We include the natural logarithm of market value of equity (lnMVE) as a control variable. Although not reported in the table, the likelihood ratio statistic for the overall model is 17.5333 with nine degrees of freedom. A chi-square test rejects the null that all parameters except the intercept are equal to zero (the p-value is 0.0410).
In Table 3, Panel A, we show that the effects of lnMTBE on the log odds of including E and R compared to not including them is significant and positive at the 5 percent level. (The parameter estimate is 0.51 and the p-value is 0.022.) The effects of lnMTBE on the log odds of including E but not R compared to not including them are significant and positive at the 1 percent level. (The parameter estimate is 0.71 and the p-value is 0.01.) These results are consistent with the costly contracting hypothesis. The effect of lnMTBE on the log odds of including R is not significant at the 10 percent level. (The estimate is -0.30 and the p-value is 0.568.) On the other hand, RSD is never significant. This finding suggests that information asymmetry does not play a role in the decision to include either E or R or both, a finding that is not consistent with the managerial opportunism hypothesis.
Table 3, Panel B, presents information similar to that of Panel A, but stated in terms of probabilities. We compute predicted probabilities for the inclusion of the flexibility provisions at mean values of lnMVE and RSD, but at different values of lnMTBE. In this way, we can examine the effect of given changes in growth opportunities on the decision to include E and/or R. The second column of Panel B shows predicted probabilities at mean values of lnMVE, RSD, and lnMTBE. The predicted probability of including both E and R for an average company is 36.96 percent, 9.88 percent for the inclusion of E only, 5.92 percent for the inclusion of R only, and 47.24 percent if we include neither E nor R. These probabilities are similar to sample frequencies.
The third column in Panel B shows the predicted probabilities for a one-standard-deviation change in the mean value of the lnMTBE. With a one-standard-deviation increase in lnMTBE, the probability of including both E and R increases to 44.35 percent, and the probability of including E alone increases to 13.86 percent. In contrast, the probability of including R decreases to 3.76 percent, and the probability of including neither of them decreases to 38.03 percent.
The last two columns of Table 3, Panel B, display predicted probabilities at the median and the third-quartile value of lnMTBE. When lnMTBE is set at median value, the predicted probabilities are 35.33 percent for the inclusion of E and R, 9.14 percent for the inclusion of E alone, 6.45 percent for R, and 49.07 percent for neither. When lnMTBE increases to its third-quartile value, predicted probabilities change to 39.22 percent, 10.98 percent, 5.21 percent, and 44.59 percent. Growth opportunities as measured by lnMTBE have a significant effect on the decision to include E and R and E alone.
In Table 4, we run a logistic regression in which the response variable takes the value of one if both E and R are present or if only E is present and zero if either R only is present or neither of them is present. We obtain similar results when we partition the sample to compare the inclusion of any flexibility provision against the inclusion of neither of them. The results are again similar when we compare the inclusion of both E and R against the inclusion of neither of them, although this last comparison is not a chi-squared partition.
The first specification in Table 4 shows that the coefficient on our proxy for growth opportunities (lnMTBE) is positive and significant at the 1 percent level. The second specification includes our proxy for information asymmetry (RSD). Its coefficient is not significant. The third specification includes both our proxies for growth opportunities and information asymmetry. Again, the coefficient on growth opportunities is significant at the 1 percent level, and the coefficient on information asymmetry is not significant. The fourth specification includes a leverage control variable, the ratio of debt to total assets. The growth opportunities proxy remains positive and significant at the 1 percent level, a finding that supports the costly contracting hypothesis. The information asymmetry proxy remains insignificant, lending no support to the managerial opportunism hypothesis. Because leverage is never significant and does not affect the results, we do not show it elsewhere in the table.
The table also reports several measures of goodness of fit. First, the panel shows the likelihood ratio test statistic and p-value for the null hypothesis that all parameters except the intercept are equal to zero. Second, the panel shows both a generalized r-square and a pseudo r-square. We compute the generalized r-square as 1 - [[L(0)/L(M)].sup.2/n] and the pseudo r-square as -2lnL(M)/-2lnL(M) + n, where n is the sample size and L(M) and L(0) are the maximized likelihoods for the fitted model and the intercept-only model. Third, the panel shows the percentage of correct predictions under the assumption that the outcomes are equally probable. It also shows the gamma measure of association between the predictions of the model and actual outcomes at each level of the independent variable. The goodness-of-fit measures reinforce our conclusion that inclusion of the flexibility provisions is related to growth opportunities, as measured by MTBE, but that there is no association between the inclusion of the provisions and information asymmetry, as measured by RSD.
For alternative measures of investment opportunities, we follow Pilotte (1992) and Denis (1994) and use ex-post growth measures: The average ratio of capital expenditures to total assets, annual sales growth, and annual asset growth for the three years following the warrant offering. Following Goyal, Lehn, and Racic (2002), we also use the ratio of capital expenditures to the book value of assets at the end of year 0 to measure growth opportunities. We use the ratio of capital expenditures to sales at the end of the offering year as an alternative to normalizing investment by total assets.
Three of the five alternative proxies for growth opportunities are significantly associated with the flexibility provisions.  Asset growth,' the ratio of capital expenditures to book value of assets at the end of year 0, and the ratio of capital expenditures to sales are positively related to the flexibility provisions and significant at the 5 percent level. Capital expenditures over years 0, 1, and 2 and sales growth are not significant. We interpret these findings as supporting the costly contracting hypothesis.
We use four alternative proxies for information asymmetry. The presence of free cash flow creates uncertainty about whether firm managers will disgorge it to shareholders or spend it on unprofitable projects (Jensen, 1986). Although proceeds and offer price might be better proxies for investor uncertainty about the real value of the firm than they are for future managerial actions, we use them to assess the sensitivity of our results to the use of alternative variables. The managerial opportunism hypothesis predicts a negative relation between proceeds from the offering and the inclusion of the flexibility provisions. Similarly, higher offer prices are associated with more seasoned firms and thus less information asymmetry. Following Brav and Gompers (2003), we use a dummy variable that equals one if the securities include a lock-up provision and equals zero otherwise. The managerial opportunism hypothesis predicts a negative relation between the presence of a lock up provision and the flexibility provisions.
None of the alternative proxies for information asymmetry affects the decision to include the flexibility provisions. In regression results, we find that none of our information asymmetry proxies is significantly related to the inclusion of the flexibility provisions.
In this section, we examine two board characteristics of the sample firms: board size and proportion of independent directors. We conjecture that a firm with a stronger governance structure is more likely to grant its managers a higher degree of flexibility. Thus, we expect the inclusion of the extension and reduction provisions to be negatively correlated with board size and positively correlated with the proportion of independent directors. We also include data on insider ownership as a control variable, but the influence of insider ownership is not obvious ex ante. We gather data on governance from Compact Disclosure and proxy statements.
Firms that include both provisions (ER) have significantly smaller board sizes than firms that include neither (NN). (The p-value for the Wilcoxon rank sum test is 0.0111.) There are no differences between the categories of firms in terms of the proportion of independent directors. The median insider ownership percent is higher for ER firms (39 percent) than for NN firms (28 percent), but the difference is not statistically significant.
Table 5 reports the results of logistic regressions that predict the inclusion of the extension and reduction provisions. In each of the five models, board size is negatively related to the inclusion of the provisions--firms with smaller boards are more likely to include them. The coefficient on board size is significant at the 5 percent level or better in each case. Additionally, the proportion of independent directors is positively related to the inclusion of the provisions (at the 90 percent confidence level) in two of the specifications.
A plausible interpretation of these findings is that the benefits of flexibility are more likely to exceed the costs of flexibility for a firm with stronger governance characteristics (and hence lower agency costs).
Long-run Return Performance
In this section, we compare the long-run stock return performance of firms that include and firms that do not include the flexibility provisions in their warrant agreements. If the costly contracting hypothesis is valid, firms choose provisions optimally, and their performance should not depend on whether or not they include the flexibility provisions. If the managerial opportunism hypothesis is valid, firms that choose the flexibility provisions might underperform firms that do not.
Firms that include the flexibility provisions experience an average (median) cumulative three-year return of 17.26 percent (20.27 percent), while firms that do not include the flexibility provisions experience an average (median) -2.04 percent (12.87 percent) cumulative three-year return. A t-test for the difference in means cannot reject the null hypothesis that the average three-year return is the same for the two samples. (The t-statistic is 1.34.) Similarly, the Wilcoxon rank sum test cannot reject the null hypothesis that the two distributions of three-year returns are equal. (The z-statistic is 1.24.)
Results are similar when we examine firms for which we have 36 months of data (three-year survivors). Average (median) three-year returns for firms that include and firms that do not include flexibility provisions are 40.62 percent (41.08 percent) and 23.83 percent (30.19 percent), respectively. Again, both a parametric and a nonparametric test cannot reject the null hypothesis of no difference between the means and distributions of three-year returns for the two samples.
If the two samples comprise firms of different levels of risk, comparing unadjusted returns is not appropriate. One commonly-used approach to adjust for possible risk differences is to run a calendar-time regression of portfolio returns on the three factors identified by Fama-French (1993) as relevant to asset pricing. Securities are grouped into a portfolio and the following regression is estimated on portfolio returns:
[R.sub.pt]-[R.sub.ft] = [alpha] + [beta]([R.sub.mt] - [R.sub.ft])+ [sSMB.sub.t] + [hHML.sub.t] + [u.sub.pt] (1)
[R.sub.pt] = The average (equally or value-weighted) daily return on the calendar-time portfolio;
[R.sub.ft] = The daily return on the three-month Treasury bills;
[R.sub.mt] = The return on a value-weighted market index;
[SMB.sub.t] = The difference in the returns of a value-weighted portfolio of small stock and big stocks; and
[HML.sub.t] = The difference in the returns to a value-weighted portfolio of high book-to-market equity stocks and low book-to-market equity stocks.
The estimate of the average abnormal return is [alpha]. We run the regression over the three years following the month of the offering date.
When portfolio returns are equally weighted, the intercept estimate (t-statistic) is 0.0001 (0.28) for the sample of firms that include the flexibility provisions and 0.0015 (2.44) for the sample of firms that do not include the provisions. When portfolio returns are value weighted, however, the intercept estimate (t-statistic) is -0.0004 (0.75) and 0.0010 (1.76) for firms that include and firms that do not include the provisions, respectively. Thus, the abnormal returns of companies that do not use the flexibility provisions appear to be driven by very small firms.
We conclude that the inclusion of flexibility provisions does not produce significant return differentials. We interpret this evidence as supportive of the notion that firms adopt the flexibility provisions necessary to produce an optimal financial contract.
Summary and Conclusions
In this paper, we examine a unique set of financial contracts: warrant agreements. We identify three provisions of these contracts that allow for managerial discretion in the timing of the raising of capital: the right to call the warrants, the right to extend the life of the warrants, and the right to lower the exercise price of the warrants. We examine the benefits and costs of these flexibility provisions. The frequency with which these provisions appear in warrant agreements underscores the value of managerial flexibility in growth companies.
We find a robust positive association between the inclusion of the extension and reduction provisions but no association between them and the call provision. We document a significant and positive association between the inclusion of the extension and reduction provisions and the ratio of market value to book value of equity, our proxy for investment opportunities. This finding is robust to alternative measures of investment opportunities. There is no significant association between the inclusion of the provisions and proxies for information asymmetry. These results support the costly contracting hypothesis and fail to support the managerial opportunism hypothesis. We also find that a firm with a smaller board and a higher proportion of independent directors is more likely to include the extension and reduction provisions. Controlling the agency costs associated with the flexibility allows the benefits of flexibility to outweigh its agency costs. Thus, a firm with a stronger governance structure is more likely to include the provisions. Last, we show that the long-run performance of firms following warrant issuance does not depend on the inclusion or exclusion of the flexibility provisions, consistent with the costly contracting hypothesis.
We conclude that timely access to capital is more important than the potential agency costs associated with the increase in managerial flexibility. [See Bradley and Roberts (2003) for a similar point in a study of bond covenants.]
We begin by examining the interrelations among the three contract provisions, which we treat as categorical variables. Most of the methods in this section follow Agresti (1990).
The motivation for our analysis comes from the statistical anomaly that when a third variable is ignored, the association between a pair of variables (marginal association) can be in a different direction from the direction it takes when we include the third variable (partial association). This result is known as Simpson's paradox. We control for the third variable by examining the association between the two variables at different cross-sectional levels of the third. Sufficient conditions that allow for collapsing over the third variable depend on whether the marginal and partial odds ratios are the same. If they are, we can simplify the analysis of the two-variable association by examining the marginal table.
To determine the collapsibility conditions for our sample of flexibility provisions in warrant contracts, we fit a loglinear model to our data. Loglinear model parameters are functions of conditional odds ratios (i.e., odds ratios within fixed levels of the third variable). Hence, model parameters provide information about the appropriateness of collapsing over any given variable. Loglinear models estimate the log of the expected cell frequency.
Table A1 shows fitted values for several loglinear models. The last column repeats the observed cell counts. Fitted values are close to the data for all models except the three-way independence (C, E, R) model. Fitted values not shown correspond to models that fit the data poorly, because the fitted values are different from the observed values. The table indicates independence between the call provision and the other provisions, although two conditional models also seem to fit the data.
We report goodness-of-fit tests in Table A2, Panel A. For given degrees of freedom, larger values of the likelihood-ratio test statistic [G.sup.2] represent poorer fits. The table indicates that models (C, E, R), (CE, R), (CR, E), and (CE, CR) fit the data poorly. The common feature of these models is the omission of the association between the extension and reduction provisions. This omission suggests that such an association is important.
We also can use Panel A to show the results of the backward elimination process for model selection. At each stage, we delete the term for which the resulting increase in [G.sup.2] is smallest, beginning with the saturated model (bottom of table). As the model (CE, CR, ER) adequately fits the data, we consider the three models obtained by elimination of one two-variable interaction term. Model (CR, ER) provides a good fit, as does (CE, ER).
To find the most parsimonious model that fits the data, we examine the next three models obtained by elimination of one two-variable interaction term. Model (ER, C) provides an adequate fit, but a further simplification to model (C, E, R) is not appropriate. Therefore, the most parsimonious model that best seems to fit the data indicates that the call provision is independent of the other two provisions but that an important association exists between the extension and reduction provisions. Model (CR, ER), however, also renders a good fit. This model indicates that the call and extension provisions are independent, once we control for the exercise-price-reduction provision.
Additional criteria for model selection include the Akaike information criterion (AIC) and the Schwarz Bayesian criterion (SBC). When we use them to compare models (CR, ER) with (C, ER), they both favor the more parsimonious specification (AIC = -2.77, SBC = -2.15).
Panel B, Table A2, presents the results of chi-squared partitioning to compare models. Here, we compare models that differ only by the inclusion of a certain association term. For example, the value of 0.08 for the difference column represents the value for the likelihood-ratio statistics for model (CR, ER) conditional on model (CE, CR, ER). The results from the conditional test are the same as those from unconditional tests. These results persist even after we adjust critical values so that the overall Type-I error probability for the four comparisons does not exceed 10 percent (4.96 critical value for a chi-squared statistic with one degree of freedom).
Based on the analysis of three-way interactions, we believe the type of association that exists among the flexibility provisions in warrant agreements is such that the inclusion of the call provision is independent of the inclusion of the other two. There is also evidence of a strong relation between the extension and reduction provisions.
Because we conclude that model (C, ER) adequately describes variable interactions, we collapse our analysis over the call provision to obtain the marginal contingency table for the decision to include the extension provision and the decision to include the reduction provision. Fisher's exact test, Pearson's chi-square, and the likelihood ratio test all reject the null hypothesis of independence between the extension and reduction provisions, with p-values of less than 0.001. The sample odds ratio is 43.1, which implies that the odds of observing the exercise-price reduction provision in a warrant agreement are 43 times higher when we include the extension provision than if we do not. Put differently, the probability of the inclusion of E when R is included compared to when R is not included is 81.5 percent. The probability of not including E when R is not included compared to when R is included is 90.7 percent. Thus, firms appear to consider the extension and reduction provisions as complements, including either both or neither of them.
(1.) Agresti, A., Categorical Data Analysis (New York, John Wiley & Sons, 1990).
Table A1-Fitted Values for Loglinear Models Table Al reports fitted values of several loglinear models fitted to the call, extension, and exercise-price reduction provisions in 369 warrant contracts. In the models, C means the call provision is present in warrant contract, E means the extension provision is present, and R means the exercise-price-reduction provision is present. The last column shows cell counts. CER refers to the saturated model, which includes a three-way interaction term. Warrant Contract Provision Model Callability Extension Reduction C, E, R C, ER CR, ER Callable Extendable Yes 59.5 119.1 123.9 No 87.9 28.4 27.6 Non Yes 76.9 17.4 18.1 Extendable No 113.6 173.1 168.4 Non Extendable Yes 5.46 10.9 6.1 Callable No 8.1 2.6 3.4 Non Yes 7.0 1.6 0.9 Extendable No 10.4 15.9 20.6 Warrant Contract Provision Model Callability Extension Reduction CE, ER CE, CR, ER CER Callable Extendable Yes 121.9 123.8 124 No 29.1 27.2 27 Non Yes 17.1 18.2 18 Extendable No 169.9 168.8 169 Non Extendable Yes 8.1 6.2 6 Callable No 1.9 3.8 4 Non Yes 1.9 0.8 1 Extendable No 19.1 20.2 20 Table A2--Goodness-of-Fit Tests for Loglinear Models The table reports unconditional (Panel A) and conditional (Panel B) goodness-of-fit tests for several loglinear models fitted to the call, extension, and exercise-price-reduction provisions in 369 warrant contracts. The table also shows the results of the backward elimination process for model selection, with the model selected at each step indicated by (*). In the models, C means the call provision is present in warrant contract, E means the extension provision is present, and R means the exercise-price-reduction provision is present. CER refers to the saturated model, which includes a three-way interaction term. Panel A: Unconditional Tests Model [G.sup.2] Df P-value (C, E, R) 217.83 4 0.0001 (ER, C) 4.93 3 0.1769 * (CE, R) 216.01 3 0.0001 (CR, E) 213.06 3 0.0001 (CR, ER) 0.16 2 0.9247 * (CE, ER) 3.10 2 0.2119 (CE, CR) 211.23 2 0.0001 (CE, CR, ER) 0.08 1 0.7780 * (CER) 0 0 -- Panel B: Conditional Tests Model [G.sup.2] Difference Df (C, E, R) 217.83 4 212.90 1 (ER, C) 4.93 3 2.27 1 (CR, ER) 0.16 2 0.08 1 (CE, CR, ER) 0.08 1 0.08 1 (CER) 0.00 0
(1.) Barclay, M., and C. Smith, "The Maturity Structure of Corporate Debt," Journal of Finance (1995), pp. 609-631.
(2.) Bebchuk, L., and J. Fried, Pay without Performance: The Unfilled Promise of Executive Compensation (Harvard University Press, 2004).
(3.) Bradley, M., and M.R. Roberts, "Are Bond Covenants Priced?" working paper, Duke University (2003).
(4.) Brav, A., and P. Gompers, "The Role of Lock-Ups in Initial Public Offerings," Review of Financial Studies (2003), pp. 1-29.
(5.) Chalmers, J.M.R., L.Y. Dann, and J. Harford, "Managerial Opportunism: Evidence from Directors' and Officers' Insurance Purchases," Journal of Finance (2002), pp. 609-636.
(6.) Cheung, S., "The Contractual Nature of the Firm," Journal of Law and Economics (1983), pp. 1-21.
(7.) Core, J., and W. Guay, "The Use of Equity Grants to Manage Optimal Equity Incentive Levels," Journal of Accounting and Economics (1999), pp. 151-184.
(8.) Demsetz, H., and K. Lehn, "The Structure of Corporate Ownership: Causes and Consequences," Journal of Political Economy (1985), pp. 1155-1177.
(9.) Denis, D.J, "Investment Opportunities and the Market Reaction to Equity Offerings," Journal of Financial and Quantitative Analysis (1994), pp. 159-177.
(10.) Fama, E.F., and K.R. French, "Common Risk Factors in the Returns on Stocks and Bonds," Journal of Financial Economics (1993), pp. 3-56.
(11.) Goyal, V.K., K. Lehn, and S. Racic, "Growth Opportunities and Corporate Debt Policy: The Case of the U.S. Defense Industry," Journal of Financial Economics (2002), pp. 35-59.
(12.) How, J.C.Y., and J.S. Howe, "Warrants in Initial Public Offerings: Empirical Evidence," Journal of Business (2001), pp. 433-457.
(13.) Howe, J.S., and T. Su, "Discretionary Reductions in Warrant Exercise Prices," Journal of Financial Economics (2001), pp. 227-252.
(14.) Jensen, M, "Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers," American Economic Review (1986), pp. 323-329.
(15.) Jensen, M., and W. Meckling, "Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure," Journal of Financial Economics (1976), pp. 305-360.
(16.) Kahan, M., and D. Yermack, "Investment Opportunities and the Design of Debt Securities," Journal of Law, Economics and Organization (1998), pp. 136-151.
(17.) Krishnaswami, S., P. Spindt, and V. Subramaniam, "Information Asymmetry, Monitoring, and the Placement Structure of Corporate Debt," Journal of Financial Economics (1999), pp. 407-434.
(18.) Pilotte, E., "Growth Opportunities and the Stock Price Response to New Financing," Journal of Business (1992), pp. 371-394.
(19.) Schultz, P., "Unit Initial Public Offerings: A Form of Staged Financing," Journal of Financial Economics (1993), pp. 199-229.
(20.) Smith, Jr., C.W., and J. Warner, "On Financial Contracting: An Analysis of Bond Covenants," Journal of Financial Economics (1979), pp. 117-161.
Florida Atlantic University
John S. Howe
University of Missouri
 We requested warrant agreements from the 476 firms with warrants outstanding in 1995. Using the 34 cases for which we had both the warrant agreement and the prospectus, we determine that the prospectus contains all of the information in the warrant agreement.
 All results not reported in a table are available from the authors upon request.
Table 1--Frequency of Inclusion of the Call, Extension, and Exercise-Price-Reduction Provisions in Warrant Agreements The table reports the frequency of inclusion of the call (Call), extension (Extension), and exercise-price reduction (Reduction) provisions in 421 warrant contracts. We divide the observations into two groups. The public offerings sample consists of warrant issues associated with initial public offerings, seasoned equity offerings, preferred stock offerings, rights offerings, and warrant dividend, exercise, and exchange offers. The reorganizations (Reorgs.) sample consists of warrant issues related to Chapter 11 proceedings, mergers, private placements, and lawsuit settlements. All Public Reorgs Warrant Contract Offerings Provision (N=421) (N=369) (N=52) Call Extension Reduction Freq % Freq % Freq % Y Y Y 132 31.4 124 33.6 8 15.4 Y Y N 28 6.7 27 7.3 1 1.9 Y N Y 19 4.5 18 4.9 1 1.9 Y N N 181 43.0 169 45.8 12 23.1 N Y Y 9 2.1 6 1.6 3 5.8 N Y N 4 1.0 4 1.1 0 0.0 N N Y 4 1.0 1 0.3 3 5.8 N N N 44 10.5 20 5.4 24 46.2 Table 2--Descriptive Statistics Mean, (median), and [sample size] for firm characteristics. Firms are classified by whether they include an extension provision (E), a reduction provision (R), both provisions (ER), or none of them (NN). Residual standard deviation (RSD) is computed using the market model for days +20 to +270 relative to the offering day for IPO firms and for days -270 to -20 for non-IPO firms. Market value of equity (MVE) is measured 20 days after the first trading day for IPO firms and 20 days prior to the offering day for non IPO firms. All other variables are as of the year prior to the warrant offering. Free cash flow (FCF) is operating cash flow plus cash dividend minus capital expenditures divided by book assets. Extension and Reduction Provisions Variable ER E R NN RSD (%) 6.43 5.79 6.11 6.14 (5.71) (5.05) (5.83) (5.65)     MTBE 7.06 35.96 4.08 5.50 (3.85) (3.70) (3.09) (3.38)     MVE ($ millions) 323.8 30.66 47.46 26.16 (15.11) (21.98) (18.51) (34.88) [991    FCF -0.35 -0.15 -0.34 -0.58 (-0.09) (-0.09) (-0.06) (-0.17)     DEBT/ASSETS 0.63 0.38 0.47 0.58 (0.36) (0.41) (0.25) (0.40)     Board Size 5.12 5.74 5.55 5.90 (5.00) (5.00) (5.00) (6.00)     Independent Directors 0.52 0.51 0.59 0.49 (Percent) (0.50) (0.50) (0.60) (0.50)     Insider Ownership 0.39 0.34 0.40 0.37 (Percent) (0.39) (0.33) (0.42) (0.28)     Table 3--Multinomial Logistic Regression Predicting the Inclusion of the Extension and Reduction Provisions Panel A displays results of a logistic regression in which the response variable can take four levels, corresponding to the inclusion of both the extension and reduction provisions (ER), only the extension provision (E), or only the exercise-price reduction provision (R). The base case is the inclusion of neither provision (NN). Panel A reports coefficient estimates (Coeff), standard errors (Se), and p-values (P-val). Panel B reports predicted probabilities at mean values of the (log of) market value of equity (lnMVE) and residual standard deviation (RSD), but at different values of the (log of) ratio of the market-to-book value of equity (lnMTBE). Significance indicators: 1 percent (***), 5 percent (**). The number of observations is 248. Panel A: Logistic Regression Logit (ER/NN) Logit (E/NN) Variable Coeff Se P-val Coeff Se P-val Intercept 0.31 0.72 0.669 -2.65 1.14 0.021 lnMVE -0.51 ** 0.22 0.019 0.07 0.30 0.818 lnMBTE 0.51 ** 5.25 0.022 0.71 *** 0.28 0.010 RSD 0.01 0.05 0.906 -0.05 0.09 0.606 Logit (R/NN) Variable Coeff Se P-val Intercept -1.43 1.40 0.307 lnMVE -0.06 0.40 0.881 lnMBTE -0.30 0.53 0.568 RSD 0.00 0.11 0.978 Panel B: Predicted Probabilities at Mean Values of lnMVE and RSD and at Different Values of lnMTBE Contract One Standard Provision Mean Deviation Median Third Quartile Included lnMTBE Increase lnMTBE lnMTBE ER 36.96 44.35 35.33 39.22 E 9.88 13.86 9.14 10.98 R 5.92 3.76 6.45 5.21 NN 47.24 38.03 49.07 44.59 Table 4--Logistic Regressions Predicting the Inclusion of the Extension and Reduction Provisions The response variable equals one when both the extension and reduction provisions are included or when the extension provision is included. The response variable equals zero when we include only the reduction provision or when neither of the provisions is included. For IPO firms, we compute the residual standard deviation (RSD) using the market model (equally-weighted index) for days +20 to +270 relative to the offering day. For non-IPO firms, we compute the residual standard deviation using the market model for days -270 to -20 relative to the offering day. For IPO firms, we measure the market value of equity (MVE) 20 days after the first trading day. For non-IPO firms, we measure the market value of equity 20 days prior to the offering day. We use only firms with positive book value of equity to compute the (log of) market-to-book equity ratio (lnMTBE). DEBT/ASSETS is the ratio of long-term plus short-term debt the year prior to warrant offering. Significance indicators: 1 percent (***), 5 percent (**), 10 percent (*). All regression specifications use 248 observations except the specification containing the DEBT/ASSETS variable, which uses 216. Parameter estimates (p-values) of logistic regressions using primary proxies for growth opportunities and information asymmetry DEBT/ Likelihood Constant lnMVE lnMTBE RSD ASSETS Ratio (p-val) -0.06 -0.35 * 0.58 *** 10.478 (0.904) (0.055) (0.004) (0.0053) 0.12 -0.13 0.02 1.1821 (0.852) (0.431) (0.601) (0.5537) -0.04 -0.36 * 0.58 *** 0.00 10.482 (0.954) (0.064) (0.004) (0.952) (0.0149) 0.13 -0.42 ** 0.69 *** -0.04 0.11 10.3990 (0.825) (0.046) (0.004) (0.474) (0.523) (0.0342) K-square % Correct Constant (pseudo) (gamma) -0.06 0.0419 56.7 (0.904) (0.5727) (0.19) 0.12 0.0048 50.6 (0.852) (0.5795) (0.05) -0.04 0.0419 55.1 (0.954) (0.5728) (0.19) 0.13 0.0470 56.0 (0.825) (0.5715) (0.21) Table 5--Logistic Regressions Predicting the Inclusion of the Extension and Reduction Provisions. The response variable equals one when both the extension and reduction provisions are included or when the extension provision is included and zero when only the reduction provision or when neither of the provisions is included. Board size is the number of members of the board of directors. Indep. dtors. is the percentage of outside directors in the board of directors. Insider own. is the percentage of shares outstanding owned by insiders. Board and ownership information are for the fiscal year prior to the offering year when available or for the fiscal year corresponding to the offering year otherwise. Significance indicators: 1 percent (***), 5 percent (**), 10 percent (*). All regression specifications use 196 observations except the specification containing the DEBT/ASSETS variable, which uses 175. Parameter Estimates (P-Values) of Logistic Regressions Predicting the Inclusion of the Extension and Reduction Provisions Indep. Insider DEBT/ Constant lmMVE Board Size Dtors. Own. ASSETS 0.94 0.00 -0.20 ** (0.162) (0.992) (0.021) -0.22 -0.05 0.45 (0.735) (0.767) (0.529) 0.56 0.02 -0.26 *** 1.29 (0.444) (0.894) (0.008) (0.108) 0.33 0.00 -0.23 ** 1.40 * 0.16 (0.678) (0.990) (0.025) (0.096) (0.752) 0.10 -0.20 -0.21 ** 1.17 (0.912) (0.383) (0.041) (0.150) 0.29 0.01 -0.25 ** 1.43 * 0.47 (0.731) (0.951) (0.011) (0.085) (0.569) Likelihood Constant lnMBTE SD Ratio 0.94 5.763 (0.162) (0.0056) -0.22 0.494 (0.735) (0.7811) 0.56 8.406 (0.444) (0.0383) 0.33 7.114 (0.678) (0.1300) 0.10 0.58 ** -0.01 14.033 (0.912) (0.026) (0.855) (0.0154) 0.29 8.1749 (0.731) (0.0854)
|Printer friendly Cite/link Email Feedback|
|Author:||Garcia-Feijoo, Luis; Howe, John S.|
|Publication:||Quarterly Journal of Finance and Accounting|
|Date:||Mar 22, 2010|
|Previous Article:||Antitakeover amendments and managerial entrenchment: new evidence from investment policy and CEO compensation.|
|Next Article:||The importance of being known: relationship banking and credit limits.|