The Pecking Order Theory and Time-Varying Adverse Selection Costs Don Autore * Tunde Kovacs Department of Finance Pamplin College of Business Virginia Tech Blacksburg, VA 24061 First version: May 2003 This version: November 2004 JEL Classification: G32 Keywords : capital structure, pecking order theory, time-varying adverse selection costs Abstract The extensive use of equity financing in the 1990s is in sharp contrast to the prediction of Myers and Majlufs (1984) pecking order theory that debt issues strictly dominate equity issues. We provide evidence in favor of a multi-period pecking order [Viswanath (1993)], in which time- varying adverse selection costs can make equity issues optimal (even for firms with sufficient debt capacity). We find that time-varying adverse selection costs are directly related to firms preference for internal over external funds and for debt over equity financing. We verify our results using a sample of SEOs and public debt offers. Additionally, we find that two increasingly common methods of raising equity shelf and PIPE offers are less subject to an adverse selection problem than SEOs. Our results support the underlying prediction of the pecking order that adverse selection costs give rise to a hierarchy of financing choices. * We are grateful to Greg Kadlec for his guidance and for many helpful discussions. We also thank Senay Agca, Don Chance, Jennifer Conrad, Dave Denis, John Easterwood, Ken French, Jeff Hobbs, Raman Kumar, Abon Mozumdar, Alexei Ovtchinnikov, Barbara Remmers, Dilip Shome, and seminar participants at the 2004 Financial Management Association meeting and Virginia Tech for helpful comments and suggestions. The authors gratefully acknowledge the contribution of Thomson Financial for providing earnings per share forecast data, available through the Institutional Brokers Estimate System. This data has been provided as part of a broad academic program to encourage earnings expectations research.
54
Embed
The Pecking Order Theory and Time-Varying Adverse Selection Costs
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
The Pecking Order Theory and Time-Varying Adverse Selection Costs
Don Autore*
Tunde Kovacs
Department of Finance Pamplin College of Business
Virginia Tech Blacksburg, VA 24061
First version: May 2003
This version: November 2004
JEL Classification: G32
Keywords :
capital structure, pecking order theory, time-varying adverse selection costs
Abstract
The extensive use of equity financing in the 1990s is in sharp contrast to the prediction of Myers and Majluf�s (1984) pecking order theory that debt issues strictly dominate equity issues. We provide evidence in favor of a multi-period pecking order [Viswanath (1993)], in which time-varying adverse selection costs can make equity issues optimal (even for firms with sufficient debt capacity). We find that time-varying adverse selection costs are directly related to firms� preference for internal over external funds and for debt over equity financing. We verify our results using a sample of SEOs and public debt offers. Additionally, we find that two increasingly common methods of raising equity � shelf and PIPE offers � are less subject to an adverse selection problem than SEOs. Our results support the underlying prediction of the pecking order that adverse selection costs give rise to a hierarchy of financing choices.
* We are grateful to Greg Kadlec for his guidance and for many helpful discussions. We also thank Senay Agca, Don Chance, Jennifer Conrad, Dave Denis, John Easterwood, Ken French, Jeff Hobbs, Raman Kumar, Abon Mozumdar, Alexei Ovtchinnikov, Barbara Remmers, Dilip Shome, and seminar participants at the 2004 Financial Management Association meeting and Virginia Tech for helpful comments and suggestions. The authors gratefully acknowledge the contribution of Thomson Financial for providing earnings per share forecast data, available through the Institutional Brokers Estimate System. This data has been provided as part of a broad academic program to encourage earnings expectations research.
1
The Pecking Order Theory and Time-Varying Adverse Selection Costs
First version: May 2003
This version: November 2004
Abstract
The extensive use of equity financing in the 1990s is in sharp contrast to the prediction of Myers and Majluf�s (1984) pecking order theory that debt issues strictly dominate equity issues. We provide evidence in favor of a multi-period pecking order [Viswanath (1993)], in which time-varying adverse selection costs can make equity issues optimal (even for firms with sufficient debt capacity). We find that time-varying adverse selection costs are directly related to firms� preference for internal over external funds and for debt over equity financing. We verify our results using a sample of SEOs and public debt offers. Additionally, we find that two increasingly common methods of raising equity � shelf and PIPE offers � are less subject to an adverse selection problem than SEOs. Our results support the underlying prediction of the pecking order that adverse selection costs give rise to a hierarchy of financing choices.
2
In this paper we study the link between time-variation in adverse selection costs
and the pecking order of financing choices. This link is important. We use it to (i)
reconcile conflicting prior evidence in favor of a multi-period pecking order theory, and
(ii) provide evidence that firms are increasingly raising equity capital via methods that
are less prone to adverse selection costs than seasoned equity offerings (SEOs). Our
results are consistent with the underlying prediction of the pecking order that adverse
selection costs give rise to a hierarchy of financing choices.
According to the pecking order theory, formalized by Myers and Majluf (1984)
and Myers (1984), firms seeking to finance new investments prefer to use funds
according to a hierarchy: first internal funds, then debt issuance, and finally equity
issuance. This �pecking order� arises because managers, not wanting to dilute existing
shareholders� claim, will issue only overvalued securities. Aware of this, market
participants discount firm value to reflect adverse selection costs. Myers and Majluf
(1984) show that because adverse selection costs are always larger for equity issues than
for debt issues, issuing equity is never optimal. Viswanath (1993) extends the single-
period framework of Myers and Majluf to a multi-period setting in which adverse
selection costs vary over time. In this setting, equity issuance can be optimal, even if
cash or debt capacity is available.1
Specifically, Viswanath considers a two-period world, in which a firm has the
option to take on a positive NPV project in each period, but has enough cash available to
finance only one of them. Financing both projects requires an equity issue. In this world,
the manager has to consider the timing of an equity issue, in addition to accepting or
1 Myers and Majluf (1984) allude to this possibility in their conclusion, stating that the �[�] way to build slack is by issuing stock in periods when managers� information advantage is small; firms with insufficient
3
rejecting a project [Myers and Majluf (1984)]. Stock is issued when two conditions are
satisfied: 1) the expected level of information asymmetry in the future is high, and 2) the
dilution of the equity issue is small compared to the cost of passing up the projects. In
general, issuing equity when adverse selection costs are low allows the manager to save
available cash to ensure financing for future positive NPV projects. Thus, Viswanath�s
(1993) model predicts temporal variation in the pecking order hierarchy as a result of
time-variation in adverse selection costs.2
Although adverse selection costs distinguish the pecking order theory from other
capital structure theories, empirical studies fail to effectively control for this key feature
of the theory. We argue that this is why the empirical evidence concerning the pecking-
order theory is mixed. Shyam-Sunder and Myers (1999) find support for firms� reliance
on debt financing for a sample of large firms drawn from the 1970s and 1980s. By
contrast, Frank and Goyal (2003) find a significant increase in firms� reliance on equity
financing for a sample of firms drawn from the 1990s. Lemmon and Zender (2003) offer
an explanation for these conflicting findings based on debt capacity. They argue that
firms close to their debt capacity may be forced to issue equity. Consistent with this,
Agca and Mozumdar (2003) estimate a piecewise linear model and find that firms prefer
debt to equity before reaching their debt capacity. While the issue of debt capacity is
important, it cannot explain the findings of Fama and French (2004) that equity issues are
common even for large firms that are not under duress. Thus it seems that the debt
capacity argument is not sufficient to explain all equity issues within the pecking order
theory. However, in the multi-period pecking order, equity issuance can be optimal even
slack to cover possible future investment opportunities would issue in periods when managers have no information advantage.� However, Myers and Majluf do not formalize a multi-period model.
4
when firms have internal cash or sufficient debt capacity. We provide evidence that the
conflicting empirical results can be reconciled, in favor of the pecking order theory, when
the empirical specification explicitly allows for time-variation in adverse selection costs.
Several empirical studies provide evidence that adverse selection costs are
important in security issues. First, equity issue announcements are met with a negative
market reaction [Asquith and Mullins (1986), Mikkelson and Partch (1986), and Masulis
and Korwar (1986)]. Second, the market�s reaction to security issue announcements is
more negative for issues of riskier securities [Hadlock and James (2002)]. Third, equity
issues cluster immediately after information release events to minimize adverse selection
costs [Korajczyk, Lucas, and McDonald (1991, 1992) and Dierkens (1991)]. Fourth,
equity issues tend to cluster in certain �windows of opportunity,� which are associated
with low aggregate adverse selection costs [Bayless and Chaplinsky (1996)]. Finally,
Choe, Masulis, and Nanda (1993) argue that adverse selection costs are low in business
expansion periods when equity financing is more common. Overall, these studies suggest
that adverse selection costs vary over time and that firms adjust their issuance policies
accordingly. Therefore, tests of the pecking order theory should consider a multi-period
setting in which time-varying adverse selection costs play a crucial role. To our
knowledge, there is no study that explicitly controls for time-varying adverse selection
costs in tests of the pecking order theory.
We provide statistically significant and economically meaningful evidence in
support of Viswanath�s (1993) multi-period pecking order theory. We consider several
potential proxies for time-series variation in adverse selection costs and show that the
2 Cooney and Kalay (1993) provide a different extension to the Myers and Majluf (1984) model. They allow for negative NPV projects and show that the reaction to stock issues can be positive. However, the Cooney and Kalay model is a single-period model.
5
dispersion in analysts� earnings forecasts is the most appropriate proxy to test the multi-
period pecking order theory. Using this proxy, we find that firms rely more on the
external markets for financing needs when adverse selection costs are low. Additionally,
although Viswanath�s (1993) model does not explicitly compare debt and equity issues,
predictions can be formulated by noting that the difference in adverse selection costs
between debt and equity issues is smaller when informational discrepancies are smaller
[Myers and Majluf (1984)]. By issuing equity when information asymmetry is low, a firm
can reduce the incremental adverse selection costs imposed from using equity instead of
debt.3 Given that a firm chooses external financing, we find that the choice of debt
financing is positively related to adverse selection costs. Similarly, firms are more likely
to deviate from the strict pecking order when adverse selection costs are low and growth
options are high. In our tests we emphasize time-series relations (and control for
heterogeneity) by including firm fixed effects.
Furthermore, we conduct tests on additional samples to address the findings of
Fama and French (2004) that firms issue equity in ways that are potentially less prone to
adverse selection costs. We verify our results using a restricted sample that is similar to
using a sample of new debt and equity issues. Additionally, we provide support for the
multi-period pecking order using a sample of SEOs and public debt offers. Specifically,
we illustrate that adverse selection costs significantly decline in the three months prior to
SEO announcements, and significantly increase in the three to six month period after
SEOs. We find a small (insignificant) decrease in adverse selection costs prior to public
debt offers.
3 We assume that debt capacity is limited, so firms need to issue equity at some point.
6
Finally, we extend the findings of Fama and French (2004) that firms raise more
capital via methods that potentially suffer less from an adverse selection problem [e.g.
mergers and options] than they raise through SEOs. In particular, we examine shelf-
registered and PIPE offers - methods of raising equity that are most similar to SEOs - and
find that the timing of these offers has little relation to time-varying adverse selection
costs. Moreover, the market imposes a significantly larger discount on the stock of SEO
issuers than on the stock of shelf or PIPE issuers. These findings are consistent with the
relative financing mix of SEOs, shelf offers, and PIPEs during 1990-2003, as issuers are
increasingly avoiding SEOs in favor of shelf offers and PIPEs.
Overall, our evidence supports the premise of the pecking order that adverse
selection costs affect financing choices. Using various samples, we find that firms take
adverse selection costs into account by timing their use of internal versus external and
debt versus equity financing, as well as by moving to equity issues that are less prone to
adverse selection problems. These findings suggest the pecking order as a theory of
financing choices lives.
The remainder of the paper is organized as follows. Section 1 describes our data
and variables. Section 2 provides tests of the multi-period pecking order theory. In
section 3 we provide additional tests using various samples. We conclude in section 4.
1. Data and Variables
Our primary sample consists of firms in the intersection of CRSP, quarterly
COMPUSTAT, and IBES files. CRSP and COMPUSTAT provide security return and
quarterly accounting data, respectively. We exclude all financial firms and utilities, firms
7
that are likely to be financially distressed (average assets under $ 50 million, average
sales under $ 5 million, or share price under $ 5), non-U.S. companies, and companies
that are not traded on the NYSE, AMEX, or NASDAQ. We further exclude individual
firm quarters with missing values for total book assets, sales, financing deficit, net change
in debt, net change in equity, long term debt, property, plant, and equipment (PPE),
market-to-book ratio, or profitability. We also exclude firm quarters that are
contaminated by significant (greater than 50% of sales) merger and acquisition activity.
In addition, we exclude a small number of companies with extreme values that are likely
due to data errors.4 We use IBES to obtain analysts� earnings forecast data. To be
included in our sample, firms must have a minimum of four analyst earnings forecasts
available (in at least one month of a given fiscal quarter). We choose this cutoff point to
improve the reliability of the standard deviation of analysts� earnings forecasts. Since the
quarterly IBES coverage starts in 1984, our sample period is 1984-2002. Our sample
consists of 69,944 firm quarters over this 19-year period.
The restriction on the number of analysts following a firm introduces a selection
bias towards large, widely followed firms. Several studies have documented that large
firms tend to use more debt, have smaller growth opportunities, and are more transparent
to investors. Although our sample tends to exclude many small, high growth firms by
construction, it still exhibits substantial variation in size, growth opportunities, adverse
selection costs, and collateral.
4 We exclude the following extreme values (all variables are scaled by total book assets): profitability<-1; net change in debt, net change in equity, or total financing deficit < -0.5; market-to-book>20; dispersion in analysts� earnings forecasts scaled by the stock price>0.1. These exclusions decrease our sample by less than 0.5%. As an alternative to excluding extreme values, we winsorize our variables at the 1% level. The results using this alternative dataset do not differ from those reported in the paper.
8
1.1. Measuring adverse selection costs
The pecking order theory is distinguished from other capital structure theories by
its prediction about the relation between adverse selection costs and firms� incremental
financing decisions. Inferences about the relevance of the pecking order theory thus
depend critically on choosing an appropriate proxy variable for adverse selection costs.
Several studies reject the pecking order by interpreting firm characteristics such as
size, growth, or age as proxies for adverse selection costs. For instance, Frank and Goyal
(2003) and Fama and French (2002) argue that the pecking order theory should perform
best among small, high-growth firms because these firms are more likely to have
information asymmetry problems, and therefore incur higher adverse selection costs
when issuing securities. Both studies reject the single-period pecking order theory based
on the finding that small, high-growth companies issue more equity and are less levered
than large and old firms. These variables, however, capture cross-sectional differences
rather than time-series variations. Thus, they are inappropriate to test the multi-period
pecking order theory, which predicts a time-series relation between adverse selection
costs and firms� financing decisions.
Previous corporate finance studies [e.g. Krishnaswami and Subramaniam (1999)]
have identified several proxies to measure the level of adverse selection costs (dispersion
in analysts� earnings forecasts, mean analyst forecast error, residual standard deviation,
and trading volume). These variables capture time-series variation, but they assess
different aspects of asymmetric information. While one of them may be useful in one
context, another may be more suitable in another context.
In the context of the pecking order, it is desirable to use a proxy that reflects
dispersion of opinion among investors, because this is arguably the best approximation of
9
the market�s reaction to an equity issue announcement. Since managers issue equity only
when the market overvalues the firm, an issue decision signals bad news about the firm�s
value and truncates from above the distribution of investors� expectations about the future
states of the world. The new consensus (mean) value of the firm, conditional on the
equity issue decision, is thus lower than the previous unconditional price. This discount
to the security price reflects adverse selection costs and depends on the dispersion of
investors� beliefs about the firm�s prospects. This is what we would like to capture. A
more formal justification of this argument is presented in Appendix A.
To approximate the dispersion of investor beliefs, we use dispersion in analysts�
earnings forecasts (DISPERSION). This variable is our proxy for adverse selection costs.
D�Mello and Ferris (2000) identify a negative relation between forecast dispersion and
equity issue announcement reactions, providing empirical justification for the link we
make between DISPERSION and adverse selection costs. In a different context, Diether,
Malloy, and Scherbina (2002) also argue that dispersion in analysts� earnings forecasts
captures dispersion in investor beliefs, and not risk. Other proxies such as analysts�
earnings forecast error5, trading volume6, and volatility7 are noisier estimates of
dispersion of investors� opinion.
5 Analysts� forecast error is the difference between the mean analysts� earnings forecast and the actual realized earnings in a period. Forecast error is not observable to the market until after the earnings release and therefore does not necessarily predict the adverse selection costs the market imposes on the firm at a security issue announcement. An issue announcement does not fully reveal the information the manager has, it only signals the quality of it. The market�s new consensus value of the firm�s price, therefore, is not directly related to the value known by the manager, as shown in the appendix. Thus, forecast error is not a suitable proxy to measure adverse selections costs imposed by the market. 6 Higher trading volume could represent more dispersion in beliefs among investors. However, higher volume can also result from increased liquidity trading, portfolio rebalancing, tax-driven trading [Harris and Raviv (1993)], and the influence of other corporate events, so there is not necessarily a direct connection between trading volume and the dispersion of opinions. 7 Volatility (residual standard deviation) is used in the literature as a proxy for both adverse selection costs and risk. In some situations adverse selection costs and risk have similar effects. However, in tests of debt versus equity financing, theories based on adverse selection costs and theories based on risk yield opposite predictions. On the one hand, higher adverse selection costs lead to more debt financing [Myers and Majluf (1984)]. On the other hand, higher risk is associated with more equity issues [Fluck (1998)]. Volatility is
10
We should note that this proxy is likely to understate the level of dispersion in
investor beliefs.8 First, analysts tend to be optimistic and publish positive forecasts. When
analysts have bad news for investors they often choose not to cover the firm at all.
Second, analysts tend to herd and publish similar forecasts. This leads again to a smaller
dispersion.
We compute dispersion of analysts� earnings forecasts in the following way. We
calculate the standard deviation of quarterly forecasts for each month in which at least
four analysts covered the firm. We average these standard deviations quarterly and scale
this number by the end-of-quarter stock price.9 This gives us a measure of DISPERSION
for each quarter for each firm. This variable exhibits a large amount of skewness, which
is an undesirable property in the estimation of normal linear regressions, and more
importantly, in assessing the economic significance of adverse selection costs. Therefore,
we transform this variable by taking its natural logarithm.10 Since DISPERSION is zero
when all analysts agree, we need to introduce a new binary variable (DISP_DUMMY)
that takes the value of one when DISPERSION is zero. Figure B.1 in Appendix B
illustrates the skewness in DISPERSION and Figure B.2 in Appendix B depicts how
taking the natural logarithm eliminates this problem. This has implications for any study
that uses dispersion of analysts� earnings forecasts to make inferences about corporate or
investor behavior.
more likely to be associated with risk since it measures time-series fluctuations in the market�s consensus of the firm�s value, in other words, fluctuations in the mean of the distribution of opinions. Accordingly, we use volatility as a control variable for risk in our tests. 8 Diether, Malloy, and Scherbina (2002) document a data handling procedure at IBES that alters the dispersion of analysts� earnings forecasts. When IBES adjusts for stock splits they round off earnings per share forecasts to the nearest cent. However, this should not create a systematic bias, since some forecasts will be rounded down, others will be rounded up. 9 Alternatively, we scale the standard deviation by the mean analyst forecast. The results (unreported) are consistent with those reported in the paper. 10 Our results (not reported) are qualitatively similar when we use DISPERSION without a log transformation.
11
1.2. Control variables
Several factors influence firms� choices between debt and equity financing. Rajan
and Zingales (1995) identify four variables that are closely related to leverage: the
market-to-book ratio, asset tangibility, firm size, and profitability. In our empirical tests
we control for these variables. Asset tangibility is defined as the ratio of fixed assets
(plant, property, and equipment) to total assets (PPE). Tangible assets can serve as
collateral in debt contracts. Firms with more tangible assets can, in general, take on debt
easier. The market-to-book ratio (MTB) is defined as market value of equity plus book
value of total assets minus book value of equity, divided by total book assets. Market-to-
book can proxy for growth opportunities or overvaluation. High growth opportunities
mitigate dilution costs [Myers and Majluf (1984)] and increase debt overhang problems
[Myers (1977)]. Overvaluation induces managers to issue equity [Baker and Wurgler
(2002)]. In each case, high market-to-book values are associated more with equity than
with debt issuance. Similarly, the multi-period pecking order theory implies that firms
with high growth opportunities will issue equity to build financial slack for future
financing, and also that firms will prefer to issue equity when their stock is overvalued to
reduce dilution to existing shareholders. We use the natural logarithm of net sales to
proxy for firm size (SIZE). Larger firms are generally better recognized and thus have
easier access to debt. We define profitability as income before extraordinary items
(Compustat data item number 237) divided by total book assets (PROFITABILITY).
More profitable firms are less reliant on external financing, and upon accessing the
external markets, they are more able to issue debt.
12
Our tests also include a young firm dummy variable (YOUNG), which equals one
for firms less than ten years old, measured from the time they first appeared on
Compustat, and zero otherwise. Younger companies are generally not as well known and
have less access to debt markets. Additionally, we include the residual standard deviation
of stock returns (VOLATILITY) to capture idiosyncratic risk and cash flow volatility.
We calculate VOLATILITY as the residual standard deviation from a modified market
model in which we regress daily returns in a given quarter on the lead, contemporaneous,
and lagged daily value-weighted market return. This approach is similar to that of
Dimson (1979) and Scholes and Williams (1977) and helps to minimize the effect of
nonsynchronous trading. Firms with higher volatility are likely to have higher costs of
financial distress, because volatile cash flows make it difficult to support promised
payments to bondholders [Fluck (1998)]. Therefore, higher VOLATILITY should be
associated with relatively less debt and more equity financing. We also include an interest
rate variable (INTRATE), which is the 10-year AAA-rated corporate bond rate. Firms�
choice of debt over equity financing should be inversely related to INTRATE. Finally, in
our tests of internal versus external financing, we include the slope of the yield curve
(YIELDSLOPE), defined as the 20-year Treasury bond rate minus the 3-month Treasury
bill rate, as a control variable for business conditions. Interest rate data is obtained from
the St. Louis Federal Reserve.
[INSERT TABLE 1 ABOUT HERE]
Table 1 presents descriptive statistics and correlation coefficients between our
variables. There is substantial variation in financing source variables (net change in debt
[∆D], net change in equity [∆E], and total financing deficit [FINDEF]) as well as in firm
13
characteristics. All correlation coefficients in Panel B are significant. To address potential
multicollinearity, we provide robustness tests using orthogonalized variables.
2. Empirical Tests of the Multi-Period Pecking Order
To motivate our study of the relation between time-varying adverse selection
costs and firms� financing choices, we first illustrate that adverse selection costs and
other firm characteristics vary considerably over time. Figures 1-8 illustrate trends in
financing patterns and characteristics of our sample firms. Figure 1 shows an increasing
reliance on external financing in the mid- to late-1990s, which contradicts the strict
pecking order. Figure 2 illustrates substantial time variation in the average portion of the
current financing deficit covered by debt, with lower values occurring in 1999-2000.
Figures 3-8 plot cross-sectional averages of dispersion in analysts� earnings forecasts,
market-to-book ratios, asset tangibility, size (natural logarithm of sales), profitability, and
return volatility (residual standard deviation) for each quarter in our sample. The figures
reveal that several factors exhibit time-series variation (decrease in analysts� earnings
forecasts dispersion, increase in market-to-book ratios and return volatility, and decrease
in asset tangibility, size, and profitability) that could have contributed to firms� increased
reliance on external financing and on equity issues in the late 1990s. In the following
sections we provide formal tests of how these variables affect firms� financing decisions.
[INSERT FIGURES 1-8 ABOUT HERE]
2.1. Internal versus external financing
Previous tests of capital structure theory are often organized as a �horse race�
between various theories, such as tradeoff, pecking order, and market timing. It is often
14
difficult to devise tests that are capable of distinguishing between the predictions of
competing theories. For example, Baker and Wurgler (2002) interpret their findings as
support for the market timing theory, although they concede that they cannot rule out the
dynamic pecking order.
One of our primary contributions is to identify and test a prediction that is unique
to the pecking order: adverse selection costs give rise to a financing hierarchy. We
establish two key hypotheses to test the multi-period pecking order: (i) a firm�s
preference for internal over external financing is stronger when adverse selection costs
are higher, and (ii) given that a firm accesses the external markets, it will prefer debt to
equity when adverse selection costs are higher.
According to the multi-period pecking order, firms should time the use of external
financing to periods of low adverse selection costs. We test the relation between external
financing and adverse selection costs by estimating the pooled cross-sectional time-series
regression of financing deficit (FINDEF) on dispersion in analysts� forecasts
(DISPERSION) and our control variables:11
FINDEFit = a + b1 DISPERSIONit+ b2 DISP_DUMMY it + b3 MTB it + b4 PPEit + b5 SIZEit + b6 YOUNGit+
Note that by estimating Eq. [1] we do not assume that adverse selection costs are
capable of influencing firms� investment policies, rather, we examine how adverse
selection costs influence whether the firm uses internal or external (FINDEF) financing to
11 Our definition of FINDEF follows Frank and Goyal (2003): FINDEFt = DIVt + It + ∆Wt � Ct = ∆Dt + ∆Et. The financing deficit (FINDEF), defined as the sum of dividends (DIV), investment (I), and the change in working capital (∆W), less cash flow after interest and taxes (C), equals the sum of net equity (∆E) and net long-term debt (∆D) issued in the given quarter (all variables are scaled by total book assets).
15
cover its total financing needs. We hypothesize that DISPERSION is positively related to
the use of internal over external funds to cover total financing needs.
The estimates for Eq. (1) are reported in Table 2. We report estimates for both
pooled and firm-fixed effect regressions. The fixed effect estimates emphasize our main
focus, time-series relations. There is a statistically significant negative relation between
DISPERSION and FINDEF, indicating that firms rely more heavily on external funds for
financing needs when adverse selection costs are low. Similarly, DISP_DUMMY has a
positive and significant coefficient in both estimations, also suggesting that firms rely
more on external financing when adverse selection costs are low. The economic
magnitude of these effects is substantial. The coefficient on DISPERSION is -0.0029,
and from Table 1, the standard deviation of DISPERSION is 2.49. A one standard
deviation increase in adverse selection costs decreases external financing by 0.65% of
total assets. Comparing this figure to the average external financing (0.9% of total assets)
implies that the effect of adverse selection costs is both statistically and economically
significant. These findings are consistent with the prediction that low adverse selection
costs provide incentive for firms to 1) issue risky securities [Viswanath (1993)], and 2)
increase the issue size [Krasker (1986)].
[INSERT TABLE 2 ABOUT HERE]
Higher values of MTB are associated with using more external funds, possibly
because firms attempt to issue overvalued equity, or because high growth firms have
limited internal funds available. As expected, larger, older, and more profitable firms rely
less on external funds for financing. Our interest rate macro variable (INTRATE) is
positively associated with the use of external financing, while the slope of the yield curve
(YIELDSLOPE) is not significant. Accounting for firm fixed effects changes the impact
16
of some variables that capture mainly cross-sectional differences (PPE and
VOLATILITY), but the results pertaining to the role of adverse selection costs get
somewhat stronger. A possible reason for the low R-squared in our Table 2 regressions is
that there is more variation in FINDEF due to variation in investments than due to
variations in the internal versus external financing choice.
2.2. Debt versus equity issues for external financing
To test our second hypothesis, that firms prefer debt to equity when adverse
selection costs are higher, we implement several modifications to the methodology used
in Shyam-Sunder and Myers (1999), Frank and Goyal (2003), and Lemmon and Zender
(2003). These studies use the following regression framework:
∆D it = a + bPO FINDEFit + eit , (2)
where ∆D it is the net change in long-term debt of firm i in quarter t.
We extend Eq. (2) to explicitly control for cross-sectional and time-series
variation in adverse selection costs. Specifically, we add analyst forecast dispersion
(DISPERSION) and our control variables, while conditioning on the amount of financing
This conditional approach avoids the critique of Chirinko and Singha (2000) that
tests of Eq. [2] can be misleading since it is difficult to determine how close the
coefficient on financing deficit must be to one to provide support for the pecking order.
Lemmon and Zender (2003) also use a conditional model, but focus on debt capacity
17
rather than time-varying adverse selection costs. They also discuss unreported results that
use analysts� earnings forecast error and residual standard deviation as information
asymmetry proxies, both of which we find improper to test the multi-period pecking
order. See our Section 1.1 and Appendix A for a discussion.
We report results for two samples, our full sample and a subsample that includes
only firms that have a financing deficit, not a surplus. The choice of this subsample is
motivated by the fact that the pecking order theory�s predictions are derived for
investment driven external financing. From our full sample of 69,944 firm quarters, only
38,177 (or about 55%) observations contain firms that have a net financing deficit, i.e.
accessed the external markets on net for financing.12 While our subsample is intended to
ensure that our results are not driven by repurchases, we note that Myers and Majluf�s
model implies that equity repurchases will occur when adverse selection costs are low. In
our regression framework, this corresponds to the same prediction for the coefficient of
adverse selection costs when FINDEF is negative. Shyam-Sunder and Myers (1999) state
that �The Myers-Majluf reasoning works in reverse when the company has a surplus
([FIN]DEF<0) and wants to return cash to investors. [�] Thus the simple pecking
order�s predictions do not depend on the sign of [FIN]DEF.� The intuition is that
repurchase announcements drive the stock price up, more so when adverse selection costs
are higher. This makes repurchases expensive to continuing shareholders. For a more
complete discussion, see Shyam-Sunder and Myers (1999), page 225.
12 The use of an even more restrictive data set, requiring both net change in debt and net change in equity to be positive, could also be justified. When ∆Dt and ∆Et have different signs but FINDEF>0, a firm issues equity (debt) sufficient to pay down debt (repurchase equity) and to finance the firm�s deficit. In this case, there is no relationship between the firm�s net change in debt (equity) and its financial deficit. However, using this more restrictive data set would compromise the size of the sample. Also, we might lose valuable information by disregarding companies that rebalance their portfolios by issuing equity and paying down debt to build debt capacity. The multi-period pecking order predicts that firms will pay down debt when
18
Column 1a and 1b of Table 3 report estimates for the pooled cross-sectional time-
series regression [Eq. (3)] for the full sample and for the financing deficit only sample,
respectively. The coefficient on the interaction term of DISPERSION * FINDEF
(DISP_DUMMY * FINDEF) is positive (negative) and significant in both specifications.
Once a firm has decided to access the external markets, higher adverse selection costs are
associated with more debt issuance and lower adverse selection costs are associated with
more equity issuance. These results are economically significant: a one standard
deviation increase in DISPERSION increases the percentage of FINDEF covered with
debt by 10.83%.
[INSERT TABLE 3 ABOUT HERE]
Both profitability and market-to-book are negatively related to debt issues conditional on
external financing. The negative coefficient on MTB is consistent with the multi-period
pecking order as firms with more growth options are more likely to preserve debt
capacity. The negative effect of PROFITABILITY is somewhat surprising considering
that both the pecking order and the trade-off theory predict a positive sign given that a
firm accesses the external markets (conditional model). The variables PPE, YOUNG,
SIZE, VOLATILITY, and INTRATE all have the expected signs in our regressions.
Firms with more tangible assets, as well as older and larger firms, use more debt
compared to equity, consistent with the empirical findings of Rajan and Zingales (1995).
Riskier firms use relatively more equity then debt. This is consistent with the prediction
of Fluck (1998) that riskier firms have less access to the debt markets, and with the
findings of Jung, Kim, and Stulz (1996) that greater stock return volatility correspond to a
adverse selection costs are low. Thus, for our subsample, we require only that change in debt and change in equity together represent a net financing deficit.
19
greater likelihood of equity issues. Finally, firms� reliance on debt compared to equity
increases when interest rates are low.
Column 2a and 2b of Table 3 report firm fixed effect estimates for each sample.
The coefficients of DISPERSION and DISP_DUMMY remain significant at the 1%
level, indicating a positive (negative) time-series relation between debt (equity) financing
and adverse selection costs.
As a robustness check, we re-estimate our results in Table 2 and Table 3 using
orthogonalized variables to address potential multicollinearity between our variables. We
regress DISPERSION on our set of control variables, and use the orthogonal part of
DISPERSION in our tests. In addition to this, we use a second orthogonalization
technique in which we orthogonalize all of our variables in the following order:
VOLATILITY, PROFITABILITY, SIZE, PPE, MTB, and finally DISPERSION.
Specifically, we remove the effects of YOUNG and DISP_DUMMY from
VOLATILITY, and then we remove the effects of these three variables from
PROFITABILITY. We continue this procedure until the effects of all the variables are
removed from DISPERSION. The results using orthogonalized variables (unreported) are
qualitatively similar to those we report, indicating that multicollinearity does not affect
our inferences.
In Eq. (1) FINDEF is the dependent variable and in Eq. (3) FINDEF is an
independent variable. This motivates us to estimate Eq. (1) and Eq. (3) simultaneously
using a seemingly unrelated regression (SUR) system. The results (untabulated) are
qualitatively similar to those we report.
Frank and Goyal (2003) advocate the use of first differences in tests of
incremental financing choices. Accordingly, we estimate a conditional model similar to
20
Eq. (3), but we replace the levels of independent variables with first differences. Table 4
reports the regression estimates. The importance of adverse selection costs is unchanged
from our earlier results: the effect of ∆DISPERSION is significant at the 1% level.
[INSERT TABLE 4 ABOUT HERE]
Our findings extend the prior literature on the role of time-varying adverse
selection costs. We use a direct and firm-specific proxy (analyst forecast dispersion) for
adverse selection costs, whereas previous studies use aggregate measures such as time
varying �hot� and �cold� markets (Bayless and Chaplinsky (1996)) and the business
cycle (Choe, Masulis, and Nanda (1993)). We are the first to our knowledge to explicitly
control for time-variation in adverse selection costs in tests of the pecking order�s
predictions about internal versus external and debt versus equity financing.
2.3. Predictive logistic regressions
A potential problem with models based on Eq. (2) is that the size of FINDEF and
∆D are not related when ∆D and ∆E have opposite signs. If such capital structure
rebalancing occurs, the magnitude and sign of FINDEF is not directly linked to ∆D. For
example, if a firm issues $100 worth of equity and pays down $80 worth of debt,
FINDEF equals $20 and regressing ∆D on FINDEF can give misleading inferences.
In this section we implement logistic regressions that do not suffer from this
problem. In particular, we test how our explanatory variables affect the probability of
firms deviating from the strict pecking order. We define a �mild deviation� from the
strict (single-period) pecking order as accessing the market on net (FINDEF>0), and a
�strong deviation� as issuing equity on net (∆E>0).
21
Table 5 reports the results of logistic regressions for each definition. Since we are
estimating probabilities, we lag all our explanatory variables to comply with the
predictive nature of the estimation. The specifications in columns 1a and 1b show the
effect of DISPERSION and our set of control variables on the probability of deviation
from the pecking order. In Viswanath�s (1993) model the optimal time to deviate from
the strict pecking order is when growth options (MTB) are high and the dilution to
existing shareholders (DISPERSION) from issuing risky securities is low. The sensitivity
of FINDEF to DISPERSION is thus affected by the value of a firm�s growth options. In
particular, when a firm has high growth options that it does not want to pass up, it will
issue risky securities regardless of DISPERSION. Therefore, we include specifications
with an interaction term of DISPERSION * MTB (columns 2a and 2b).
In Table 5, the coefficient on DISPERSION (DISP_DUMMY) is significantly
negative (positive) and the coefficient on MTB is significantly positive in every
specification, all at the 1% level. Consistent with the multi-period pecking order theory,
firms are more likely to deviate from the strict pecking order when adverse selection
costs are low and / or the firm�s growth options are high. When we interact MTB and
DISPERSION, we find that MTB decreases the effect of DISPERSION on the
probability of deviation. When growth options are higher, adverse selection costs play a
lesser role in the type of financing used. This is consistent with the multi-period pecking
order, since high NPV projects can outweigh the costs of dilution in issuing risky
securities.
[INSERT TABLE 5 ABOUT HERE]
The coefficients on our control variables indicate that larger and more profitable
firms with more collateral are less likely to issue equity, suggesting that these firms have
22
enough financial slack and / or debt capacity and thus are less likely to deviate from the
strict pecking order. Younger firms with greater stock return volatility are more likely to
make financing decisions that deviate from the strict pecking order. We note that the
coefficients on both VOLATILITY and PPE are much more significant in our second
specification (∆E>0), indicating that firms with more volatile returns and less tangible
assets rely more on equity financing.
Overall our logistic regressions indicate that, consistent with predictions in
Viswanath (1993), when firms have low adverse selection costs and high growth options
they are more likely to deviate from the strict financing hierarchy. This suggests that
issuing equity can be a strategic decision to maintain / build financial slack or debt
capacity, to avoid deviations from the strict pecking order in the future, when adverse
selection costs are expected to be higher. Thus, it is not surprising that empirical studies
often find support for single-period pecking order behavior in samples drawn from
periods when aggregate adverse selection costs are high. This finding can in fact be a
result of firms� multi-period financing strategy.
Although we find support for the multi-period pecking order theory, we cannot
rule out other capital structure explanations. Similarly, Fama and French (2002) and
Howakimian, Opler, and Titman (2001) find mixed evidence and argue that competing
theories are complementary rather than mutually exclusive. In our tests, several variables
identified in the trade-off theory contribute significantly to firms� financing decisions.
Additionally, we cannot rule out the market timing argument of Baker and Wurgler
(2002) since we find that high market-to-book corresponds to more external financing,
and in particular, to more equity financing. However, the multi-period pecking order not
only explains the positive relation between market-to-book and equity issues, but also
23
makes additional predictions about the importance of adverse selection costs, which we
also confirm. In our tests adverse selection costs have explanatory power beyond the
effect of market-to-book.
In view of our empirical results, the pecking order theory is not a universal, all-
encompassing capital structure theory. Rather, Myers and Majluf�s (1984) and
Visvanath�s (1993) pecking order theories are descriptions of an important time-series
aspect of firms� financing decisions � with the implicit assumption of �everything else
equal�. This leaves room for other considerations, such as debt tax shields or debt
capacity concerns, to enter firms� financing structure decisions. Even though the pecking
order theory may not fully explain firms� capital structure decisions on its own right, its
underlying assumption, that adverse selection costs matter, is supported by our data.
3. The Method of Raising Equity
In a recent study, Fama and French (2004) point out that firms can raise equity
capital in ways that are potentially less subject to adverse selection costs. This, they
argue, is extremely damaging to the pecking order since equity financing is no longer a
last resort. In this section we provide tests that address the above concerns.
3.1. The Fama-French critique
Our tests to this point have used the Compustat variables change in equity and
change in debt to approximate equity and debt issues. A drawback to using this approach
in tests of the pecking order is that Compustat variables group together methods of
raising equity (debt) that are affected differently by adverse selection costs. Fama and
French (2004) report that option exercises and especially mergers and acquisitions have
24
recently accounted for most of the equity raised by the average firm. These alternative
methods of raising equity presumably have lower adverse selection costs than the less
frequent seasoned equity offerings. Fama and French (2004) interpret the large frequency
and the significant amount of capital raised via these methods as a �deadly blow� to the
pecking order since debt does not necessarily precede equity in the financing hierarchy.
So how do our results stand up in light of the Fama-French critique? In general,
grouping methods of raising equity that are more prone to adverse selection costs [e.g.
SEOs] with methods that are less prone to adverse selection costs [e.g. options and
mergers] should bias against us finding support for the multi-period pecking order. We
note that our Compustat variables are calculated from cash flow statements, and not from
balance sheets, thus our ∆E does not incorporate the effect of stock financed acquisitions.
∆E is affected, though, by option exercises and employee stock purchase plans. It is
possible that option exercises occur during periods of relatively high market valuation,
which may correspond to low adverse selection costs.
To address this concern, we re-estimate our specifications in Tables 2 and 3 using
a restricted sample which includes only firm-quarter observations in which ∆D and ∆E
are greater than 5% of book assets. Hovakimian, Opler, and Titman (2001) find that using
this restricted sample is similar to using a sample of new debt and equity issues from
Securities Data Company (SDC). We report results using the restriction that the absolute
value of ∆D (∆E) is greater than 5%, and the narrower restriction that the value of ∆D
(∆E) is greater than +5%. If ∆D (∆E) is less than the absolute value of 5% or is less than
+5%, we set it equal to zero.
Table 6 reports OLS and firm fixed-effect estimates for Eq. (1) with the above
restrictions. The effect of adverse selection costs (DISPERSION and DISP_DUMMY)
25
remains strong: higher DISPERSION decreases firms� reliance on external funds for
financing needs. The only notable change from Table 2 is that the slope of the yield curve
(YIELDSLOPE) is now positively related to firms� choice of external over internal funds.
This is possibly because expansionary periods are associated with more investment
opportunities, causing firms to save internal cash for future financing. It is also possible
that business cycle expansions coincide with reduced aggregate adverse selection costs
[Choe, Masulis, and Nanda (1993)]. Our firm-specific measure of adverse selection costs
(DISPERSION), however, has additional explanatory power beyond the effect of
YIELDSLOPE. Table 7 reports estimates for Eq. (3) with the 5% restrictions. In each
regression, the explanatory power (R2) is improved compared to the Table 3 estimations.
The results are consistent with our earlier findings that firms use relatively less debt and
more equity to cover their investment needs when adverse selection costs are low.
[INSERT TABLE 6 ABOUT HERE]
[INSERT TABLE 7 ABOUT HERE]
To provide specific evidence that seasoned equity offers (SEOs) are timed to
periods of low adverse selection costs, we examine variations in adverse selection costs
in event time around SEOs. Our SEO sample is drawn from SDC and contains 1428
primary common stock issues of U.S. non-financial and non-utility corporations, during
the period 1990-2003. We exclude unit, rights, and shelf offers. We match this SEO
sample with monthly analysts� forecasts dispersion, and identify the event month as the
month of the SEO filing date. We require that companies have dispersion data available
(according to our previous requirement of a minimum of 4 analysts) in every month
during the interval [-9, 9], which results in a sample of 187 SEOs. We also examine a
sample of 814 public debt offers during 1990-2003 that satisfy our analyst coverage
26
requirements. Shortening the data requirement period increases the sample sizes but does
not change our results.
Figure 9 plots mean values of DISPERSION around SEOs in event time. The
multi-period pecking order predicts that equity issues occur when adverse selection costs
are currently low and are expected to increase in the future. Consistent with this, the level
of adverse selection costs declines sharply prior to SEOs, and increases three to six
months after SEOs. To check for statistical significance, we report tri-monthly changes in
DISPERSION in Panel A of Table 8. There is a significant drop in DISPERSION in the
interval starting three months prior to and ending at the month of the SEO. The largest
decline occurs in the intervals [-3,-2] and [-2,-1]. Additionally, there is a significant
increase in DISPERSION in the three to six month period after SEOs. In contrast, Panel
B of Table 8 reports a small and insignificant decline in DISPERSION prior to public
debt offers and virtually no change after public debt offers. The larger decline (increase)
in DISPERSION prior to (after) seasoned equity offers compared to public debt offers is
consistent with our earlier tests and provides support for the multi-period pecking order.
Note that we compare changes in DISPERSION over time rather than the level of
DISPERSION across securities to focus on the time-series predictions of the pecking
order.
[INSERT FIGURE 9 ABOUT HERE]
[INSERT TABLE 8 ABOUT HERE]
Our finding that SEOs are timed to periods of low adverse selection costs
complements and extends the prior literature on the role of time-varying adverse selection
costs in firm financing. We use a direct and firm-specific proxy (analyst forecast
dispersion) for adverse selection costs, whereas previous studies use aggregate measures
27
such as time varying �hot� and �cold� markets (Bayless and Chaplinsky (1996)) and the
business cycle (Choe, Masulis, and Nanda (1993)). Additionally, we provide evidence
that SEOs are filed after the largest drop in adverse selection costs in a 1.5-year window
around the SEO, whereas Korajczyk, Lucas, and McDonald (1991, 1992) find that equity
issues occur shortly after information releases such as earnings announcements, thereby
reducing adverse selection costs. Earnings announcements are periodic events that
happen multiple times during the period we plot, but we cannot observe a cycle in
adverse selection costs corresponding to these.
Still, we agree with Fama and French (2004) that the pecking order in its most
general form � internal equity, external debt, external equity - breaks down when certain
types of equity issues suffer less (or not at all) from an adverse selection problem. At the
time when Myers and Majluf (1984) and Myers (1984) first described the pecking order
theory, stock financed acquisitions and option exercises were rare, thus the pecking order
was a financing hierarchy that considered public debt issues and SEOs as the full extent
of external financing. Over time, academic researchers (for example, Hadlock and James
(2002) and Hertzel and Smith (1993)) added more layers to this two-tier hierarchy,
including bank loans as the debt with the least adverse selection costs, convertible bonds
as the debt with the most adverse selection costs, and private placements of stock as the
equity with the least adverse selection costs. Similarly, the new low-adverse-selection-
cost ways to issue equity reported by Fama and French (2004) give rise to a more
sophisticated, multi-layered pecking order in which equity issues can have lower and
higher ranks depending on the method of their issuance. After all, the �pecking order� is
an outcome of the Myers and Majluf model in which information asymmetry gives rise to
adverse selection costs. The theory was derived to explain incremental financing choices
28
rather than to explain overall capital structure. Instead of discarding the pecking order as
no longer relevant, we believe future research should focus on updating the pecking order
to include today�s new methods of raising capital. In the next section we provide
preliminary evidence that non-specific methods of raising equity capital � in addition to
the specific purpose equity issues (e.g. mergers and option exercises) mentioned by Fama
and French (2004) � can be less subject to adverse selection costs.
3.2. Methods to raise equity and the pecking order
Motivated by Fama and French (2004), we test the hypothesis that different
methods of raising equity are affected by adverse selection costs to different degrees. We
examine ways to issue equity that are the most similar to SEOs and are used for financing
investments or for general corporate purposes. In particular, we compare our sample of
1428 primary SEOs to a sample of primary shelf-registered equity offerings and Private
Investment in Public Equity (PIPE) offers during 1990-2003. Our sample contains 386
primary shelf offers, identical to the sample studied by Autore, Kumar, and Shome
(2004), and 929 PIPE offers collected from the SDC New Issues Database. Although
PIPEs are privately placed, they are similar to SEOs because the issuer is required to file
a registration statement that covers the resale of the shares, thus eliminating the illiquidity
that is associated with Rule 144 private placements.13
We expect offerings of equity securities that are more prone to adverse selection
costs to occur after significant declines in DISPERSION, and offerings of equity
securities that are less prone to adverse selection costs to have less of a relation to time-
variation in DISPERSION. Panels C-E of Table 8 provide statistics on changes in
29
DISPERSION around the filing and offer dates14 of shelf-registered offerings and around
PIPE offers, respectively. Our restriction that companies have dispersion data available (a
minimum of 4 analysts) in every month during the interval [-9, 9] reduces the sample of
shelf filings to 97, shelf offers to 128, and PIPEs to 47. Statistics in Panels C and D
indicate no significant change in DISPERSION prior to shelf filings or shelf offers. Panel
E reports no significant change in DISPERSION prior to PIPE offers. These findings are
in contrast to the significant decline in DISPERSION prior to SEO announcements.
Furthermore, we expect offerings of equity securities that are more subject to
adverse selection costs to be associated with a more severe market reaction, and offerings
of equity securities that are less subject to adverse selection costs to generate a smaller
market reaction. We calculate the market-adjusted mean and median cumulative
abnormal returns (CARs) for the three-day window centered on the announcement date
for our sample of 1428 SEOs and 386 shelf offerings, and on the offer date for our
sample of 929 PIPE offers. The mean (median) market reaction is -1.68% (-2.07%) at
SEO announcements, -0.67% (-0.82%) at shelf announcements, and 1.34% (-0.16%) at
PIPE offers. Differences between means (t-test) and medians (Wilcoxon sign-rank test)
are all significant at the 1% level, with the exception that the median shelf announcement
reaction is less than the median PIPE offer reaction at the 5% level.
The evidence suggests that shelf equity and PIPEs are less affected by adverse
selection costs than are SEOs. Issuers appear to have little or no concern about time-
varying adverse selection costs in the issuance of shelf-registered equity or PIPEs,
although they conduct SEOs after declines in adverse selection costs. In addition, the
13 Ordinary private placements are structurally different from SEOs, due to the resale restriction and the resulting illiquidity. Accordingly, we do not include them in our equity sample. 14 Shelf offer dates are included because shelf issuers can delay offerings for several months or even years after filing a registration statement.
30
market punishes SEO issuers significantly more than it punishes shelf issuers at the offer
announcement, and it does not penalize PIPE issuers.15
Our findings are consistent with the relative mix of SEOs, shelf offers, and PIPEs
in recent years. Figures 10 and 11 illustrate the relative proportions of each type of equity
offer in terms of the number of offers and total proceeds, respectively. Both figures
illustrate that the relative use of traditional SEOs has declined substantially in the past
decade, while the use of shelf registration and PIPEs has grown. For example, during
2000-2003 SEOs account for only about 15% (258 offers) of the sample equity offers,
while PIPEs account for about two-thirds (627 offers) of our sample equity offers.
Additionally, during 2000-2003 the proceeds from SEOs account for only about 30% of
our total sample equity proceeds, while the proceeds from shelf offers account for more
than half of the sample equity proceeds. These findings are consistent with the results of
Fama and French (2004) that issuers are increasingly favoring ways to issue equity that
have low adverse selection costs. Our evidence on shelf and PIPE issues is important
because these are methods of raising equity similar to an SEO. In contrast, options do not
accommodate investment-driven needs and mergers are specific transactions.
[INSERT FIGURES 10-11 ABOUT HERE]
Considering our evidence, it is unlikely that the pecking order can completely
explain firms� overall capital structure, especially if certain types of equity issues become
intertwined with debt issues in a multi-layered hierarchy. However, the pecking order
remains important as a theory of incremental financing choices. The fact that firms seem
15 We note that, in the cross-section, DISPERSION is substantially higher for PIPE issuers than for SEO or shelf issuers. PIPE issuers are, on average, smaller and less recognized firms and thus are likely to have greater informational discrepancies than SEO or shelf issuers. The fact that PIPE issuers are not penalized by the market implies that PIPEs can resolve the adverse selection problems that equity issuers with high information asymmetry are expected to have. Thus PIPE offers can, in part, explain small firms� reliance on equity even within a single-period adverse selection framework.
31
to avoid SEOs and favor instead issuing equity through channels that bear lower adverse
selection costs, is in fact support for the underlying principle of the pecking order:
adverse selection costs play an important role in firms� financing decisions and, holding
all else constant, create a hierarchy of financing choices.
4. Conclusion
In this paper, we find support for a multi-period pecking order [Viswanath
(1993)] that can accommodate equity issues of firms that are not financially constrained.
The key difference between our study and previous studies is that we explicitly control
for time-series variation in adverse selection costs by using dispersion in analysts�
earnings forecasts as a proxy. We find that firms rely more on external markets for
financing needs when adverse selection costs are low, and given that a firm chooses
external financing, we report that the choice of debt over equity is positively related to
adverse selection costs. Furthermore, we find that firms are more likely to deviate from
the strict pecking order when adverse selection costs are low and growth options are high,
consistent with the predictions of the multi-period pecking order.
We conduct tests on additional samples to address the findings of Fama and
French (2004) that firms issue equity in ways that are less prone to adverse selection
costs. We verify our results using a restricted sample that is similar to using a sample of
new debt and equity issues. Additionally, we provide support for the multi-period
pecking order using a sample of SEOs and public debt offers. Specifically, we illustrate
that adverse selection costs significantly decline in the three months prior to SEO
announcements, and significantly increase in the three to six month period after SEOs.
32
We find a small (insignificant) decrease in adverse selection costs prior to public debt
offers.
We also examine methods of raising equity similar to SEOs that appear to suffer
less from an adverse selection problem. In particular, we find that the timing of shelf-
registered and PIPE offers has little relation to time-varying adverse selection costs.
Moreover, the market places a significantly larger discount on the stock of SEO issuers
than on the stock of shelf or PIPE issuers. The evidence is consistent with the relative
financing mix of SEOs, shelf offers, and PIPEs during 1990-2003, as issuers are
increasingly avoiding SEOs in favor of shelf offers and PIPEs.
In sum, our evidence supports the premise of the pecking order that adverse
selection costs affect financing choices. Using various samples, we find that firms take
adverse selection costs into account by timing their use of internal versus external and
debt versus equity financing, as well as by moving to equity issues that are less prone to
adverse selection problems. These findings suggest the pecking order as a theory of
financing choices lives, although a new multi-layered hierarchy is necessary to reflect
current methods of raising capital. We leave the formal development of a new pecking
order that includes various types of equity securities to future research.
33
Appendix A
Myers and Majluf (1984) and Myers (1984) develop the pecking order theory by
assuming that managers have an information advantage over investors. In their model, the
manager knows both the value of assets in place and the value of a project (growth
option), but investors know only the joint distribution of these two random variables. The
manager, acting in the interest of existing shareholders, issues securities to take on a
positive NPV project if the dilution caused by the issue decision is smaller than the value
gain from the project. The dilution costs arise because the market�s expectation of the
firm value changes conditional on whether or not the manager announces a security issue.
To illustrate how adverse selection and dilution are related, we focus on the
asymmetric information between managers and investors about the value of assets in
place and keep the value of growth options (denoted as b) as given. Similar to Myers and
Majluf (1984), let �a� denote the value of assets in place known only by the manager.
Investors know the distribution of the value of assets in place, the distribution of the
random variable �A�. When the manager announces a security issue, investors will
discount the value of assets in place according to their knowledge of A�s distribution,
conditional on the issue decision, E( A | issue). The manager issues equity if the value of
the existing shareholders� stake is greater with an issue than without it. His decision rule
is thus the following, with the simplification that the firm has no cash, and has to collect I
from the issue proceeds to invest:
aIP
PIba ≥+
++'
')( , (A1)
34
where P�= E( A | issue)+b. Let the value of assets in place, for which the manager
is indifferent between issuing and not issuing, be a*. If the market is rational it can apply
the above decision rule and calculate a level for a*. If a > a* the manager forgoes the
investment, and if a < a* the manager will issue and invest. The existing shareholders�
claim, given the issue decision, is: bdaaafPa
a
+= ∫*
min
)('
Intuitively, the issue decision signals bad news about the value of assets in place
and truncates the distribution of A from above. However, the issue decision does not fully
reveal the manager�s information, thus the discounted share price reflects the market�s
conditional expectation.
We consider an example to illustrate how different potential proxies for adverse
selection are associated with the equity issue decision. Let I=1, b=0.2, and A be
distributed uniformly over [amin, amax]. It is not necessary to assume a uniform
distribution; we use it to simplify the discussion. The value of a* can be calculated using
equation (A1) and the distributional assumption; E( A | issue)=(a*+amin)/2 .
When A~U[1, 3], the value of a* is 2.1. Managers will only issue equity if the
value of assets in place is less than 2.1, thus the market�s expected value of assets in
place conditional on an issue falls to 1.55. If we increase the dispersion of A�s
distribution to U[0, 4], a* becomes 0.6 and the conditional value of assets in place is 0.3.
This illustrates that adverse selection costs of an issue are higher when the dispersion in
the market�s assessment of assets in place is larger, all else constant.
Note, however, that the amount of asymmetric information between managers and
investors is not directly related to adverse selection costs, since the issue announcement
does not fully reveal the realized value known by the manager. When A is uniformly
35
distributed [1, 3], the market�s initial expectation is E(A) equals 2. Conditional on equity
issuance, A is uniformly distributed [1, 2.1]. Asymmetric information is the difference
between the market�s initial expectation and the realization of the value of assets in place
(between 1 and 2.1). Thus the level of asymmetric information between managers and
investors can be anywhere between �0.1 and 1, given an equity issue. This illustrates that
there is no direct relation between the amount of asymmetric information and the adverse
selection costs imposed by the market.
The discussion above assumes that the value of growth options is known with
certainty. This assumption facilitates the focus on the origins of adverse selection costs in
the model. In the more realistic case where the market knows only the joint probability
distribution of the value of assets in place and of the growth option, the issue decision
reveals information about possible pairs of (a, b) in a, b space. However, numerical
simulations in Myers and Majluf (1984) show that when the distribution of assets in place
has a wider dispersion, ceteris paribus, adverse selection costs become greater.
36
Appendix B
Histograms of the dispersion in analysts� earnings forecasts (Figure B.1), and its natural logarithm (Figure B.2) Dispersion is calculated as the standard deviation of analysts� earnings forecasts scaled by the end-of-quarter stock price. For the computation of this variable we require that at least 4 analysts follow the stock.
Agca, S., and A. Mozumdar, 2003, �Firm size, debt capacity, and the pecking order
theory of financing choices,� working paper, Virginia Tech.
Asquith, P., and D.W. Mullins, 1986, �Equity issues and offering dilution,� Journal of
Financial Economics, 15, 61-89.
Autore, D., R. Kumar, and D. Shome, 2004, �The revival of shelf-registered corporate
equity offerings,� working paper, Virginia Tech.
Baker, M., and J. Wurgler, 2002, �Market timing and capital structure,� Journal of
Finance, 57, 1-32.
Bayless, M., and S. Chaplinsky, 1996, �Is there a window of opportunity for seasoned
equity issuance?,� Journal of Finance, 51, 253-278.
Chirinko, R.S., and A.R. Singha, 2000, �Testing static tradeoff against pecking order
models of capital structure: a critical comment,� Journal of Financial Economics,
58, 417-425.
Choe, H., R.W. Masulis, and V. Nanda, 1993, �Common stock offerings across the
business cycle: theory and evidence,� Journal of Empirical Finance, 1, 3-31.
Cooney, J.W., and A. Kalay, 1993, �Positive information from equity issue
announcements,� Journal of Financial Economics, 33, 149-172.
D�Mello, R., and S.P. Ferris, 2000, �The information effects of analyst activity at the
announcement of new equity issues,� Financial Management, 29, 78-95.
38
Dierkens, N., 1991, �Information asymmetry and equity issues,� Journal of Financial
and Quantitative Analysis, 26, 181-199.
Diether, K.B., C.J. Malloy, and A. Scherbina, 2002, �Differences of opinion and the cross
section of stock returns,� Journal of Finance, 57, 2113-2141.
Dimson, E., 1979, �Risk measurement when shares are subject to infrequent trading,�
Journal of Financial Economics, 7, 197-226.
Fama, E.F., and J.D. MacBeth, 1973, �Risk, return, and equilibrium: empirical tests,�
Journal of Political Economy, 81, 607-636.
Fama, E.F., and K.R. French, 2002, �Testing trade-off and pecking order predictions
about dividends and debt,� Review of Financial Studies, 15, 1-33.
Fama, E.F., and K.R. French, 2004, �Financing decisions: Who issues stock?,� Journal of
Financial Economics, forthcoming.
Fluck, Z., 1998, �Optimal Financial Contracting: Debt versus Outside Equity,� Review of
Financial Studies, 11, 383-419.
Frank, M.Z., and V.K. Goyal, 2003, �Testing the pecking order theory of capital
structure,� Journal of Financial Economics, 67, 217-248.
Hadlock, C.J., and C.M. James, 2002, �Do banks provide financial slack?,� Journal of
Finance, 57, 1383-1420.
Harris, M., and A. Raviv, 1993, �Differences of opinion make a horse race,� Review of
Financial Studies, 6, 473-506.
Hertzel, M., and R.L. Smith, 1993, �Market discounts and shareholder gains for placing
equity privately,� Journal of Finance, 48, 459-485.
39
Howakimian, A., T. Opler, and S. Titman, 2001, �The debt-equity choice,� Journal of
Financial and Quantitative Analysis, 36, 1-24.
Jung, K., Y. Kim, and R.M. Stulz, 1996, �Timing, investment opportunities, managerial
discretion, and the security issue decision,� Journal of Financial Economics, 42,
159-185.
Korajczyk, R.A., D.J. Lucas, and R.L. McDonald, 1991, �The effect of information
releases on the pricing and timing of equity issues,� Review of Financial Studies,
4, 685-708.
Korajczyk, R.A., D.J. Lucas, and R.L. McDonald, 1992, �Equity issues with time-varying
asymmetric information,� Journal of Financial and Quantitative Analysis, 27,
397-417.
Krasker, W.S., 1986, �Stock price movements in response to stock issues under
asymmetric information,� Journal of Finance, 41, 93-105.
Krishnaswami, S., and V. Subramaniam, 1999, �Information asymmetry, valuation, and
the corporate spin-off decision,� Journal of Financial Economics, 53, 73-112.
Lemmon, M.L., and J.F. Zender, 2003, �Debt capacity and tests of capital structure
theories,� working paper, University of Utah.
Masulis, R.W., and A.N. Korwar, 1986, �Seasoned equity offerings: An empirical
investigation,� Journal of Financial Economics, 15, 91-118.
Mikkelson, W.H., and M. Partch, 1986, �Valuation effects of security offerings and the
issuance process,� Journal of Financial Economics, 15, 31-60.
40
Myers, S.C., 1977, �Determinants of corporate borrowing,� Journal of Financial
Economics, 5, 147-175.
Myers, S.C., 1984, �The capital structure puzzle,� Journal of Finance, 39, 575-592.
Myers, S.C., and N.S. Majluf, 1984, �Corporate financing and investment decisions when
firms have information investors do not have,� Journal of Financial Economics,
13, 187-221.
Rajan, R.G., and L.Z., 1995, �What do we know about capital structure? Some Evidence
from International Data,� Journal of Finance, 50, 1421-1460.
Scholes, M., and J. Williams, 1977, �Estimating betas from nonsynchronous data,�
Journal of Financial Economics, 5, 309-327.
Shyam-Sunder, L., and S.C. Myers, 1999, �Testing static tradeoff against pecking order
models of capital structure,� Journal of Financial Economics, 51, 219-244.
Viswanath, P. V., 1993, �Strategic considerations, the pecking order hypothesis, and
market reactions to equity financing,� Journal of Financial and Quantitative
Analysis, 28, 213-234.
41
Table 1 Descriptive statistics and correlations between explanatory variables
Panel A displays the mean, median, and standard deviation of variables computed from the full data sample. Panel B exhibits Pearson correlation coefficients computed from the full sample. All correlation coefficients are significant at the 1% level. ∆D, ∆E, and FINDEF are the net change in debt, net change in equity, and total financing deficit in a quarter. The financing deficit (FINDEF) is defined as the sum of net equity and debt issues for a firm in a given quarter. DISPERSION is the natural logarithm of the average standard deviation of analysts� quarterly earnings forecasts scaled by end of quarter stock price. The dispersion dummy (DISP_DUMMY) takes the value of one when DISPERSION is zero and thus the log operation is invalid. Market-to-book (MTB) is the end of quarter market value of equity plus book value of debt divided by total book assets. YOUNG is a dummy variable equal to one for firms under ten years old and zero otherwise. SIZE is defined as the natural logarithm of sales. Tangibility (PPE) is defined as property, plant and equipment to total book assets. PROFITABILITY is income before extraordinary items scaled by total book assets. Residual standard deviation (VOLATILITY) is calculated from daily returns using a modified market model with the contemporaneous, lead, and lagged market return as explanatory variables in the spirit of Scholes and Williams (1977) and Dimson (1979). The sample period is 1984-2003. There are 69,944 firm-quarter observations in our sample.
Panel A: Descriptive statistics
Variable Mean Median Standard deviation
∆D 0.0046 0 0.0492
∆E 0.0043 0.0002 0.0499
FINDEF 0.0090 0.0005 0.0689
DISPERSION -5.8118 -6.5505 2.4868
DISP_DUMMY 0.1311 0 0.3375
MTB 2.2693 1.6932 1.8025
SIZE 19.2346 19.1427 1.6268
PPE 0.3287 0.2761 0.2265
YOUNG 0.1505 0 0.3576
Profitability 0.0111 0.0146 0.0417
VOLATILITY 0.0274 0.0236 0.0156
Panel B: Correlation coefficients Dispersion MTB SIZE PPE YOUNG Profitability
OLS and fixed effects regressions of financing deficit regressed on firm characteristics. Financing deficit (FINDEF) is defined as the sum of net equity and debt issues for a firm in a given quarter. DISPERSION is the natural logarithm of the average standard deviation of analysts� quarterly earnings forecasts scaled by end of quarter stock price. The dispersion dummy (DISP_DUMMY) takes the value of one when DISPERSION is zero and thus the log operation is invalid. Market-to-book (MTB) is the end of quarter market value of equity plus book value of debt divided by total book assets. YOUNG is a dummy variable equal to one for firms under ten years old and zero otherwise. SIZE is defined as the natural logarithm of sales. Tangibility (PPE) is defined as property, plant and equipment to total book assets. PROFITABILITY is income before extraordinary items scaled by total book assets. Residual standard deviation (VOLATILITY) is calculated from daily returns using a modified market model with the contemporaneous, lead, and lagged market return as explanatory variables in the spirit of Scholes and Williams (1977) and Dimson (1979). INTRATE equals the rate on 10-year AAA-rated corporate bonds and YIELDSLOPE is the slope of the yield curve (YIELDSLOPE), defined as the 20-year Treasury bond rate minus the 3-month Treasury bill rate. Interest rate data is obtained from the St. Louis Federal Reserve. Heteroskedasticity adjusted standard errors are in parentheses. The sample period is 1984-2002. There are 69,944 firm-quarter observations in our sample.
OLS Fixed effects Intercept
0.0413*** (0.0058) -
DISPERSION
-0.0028*** (0.0003)
-0.0042*** (0.0004)
DISP_DUMMY
0.0201*** (0.0024)
0.0325*** (0.0032)
MTB
0.0027*** (0.0003)
0.0048*** (0.0004)
PPE
0.0034*** (0.0011)
-0.0827*** (0.0055)
SIZE
-0.0036*** (0.0001)
-0.0102*** (0.0007)
YOUNG (dummy)
0.0093*** (0.0009)
0.0084*** (0.0014)
PROFITABILITY
-0.1046*** (0.0119)
-0.0314*** (0.0116)
VOLATILITY
0.1558*** (0.0321)
0.0047 (0.0405)
INTRATE
0.0010*** (0.0002)
0.0011*** (0.0002)
YIELDSLOPE
0.017 (0.0547)
0.026 (0.0533)
Adjusted R2 0.0322
0.0859
*, **, *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.
43
Table 3 Determinants of the source of external financing
Conditional model estimates to examine the source of external financing. The dependent variable is the change in debt (∆D). Financing deficit (FINDEF) is defined as the sum of net equity and debt issues for a firm in a given quarter. DISPERSION is the natural logarithm of the average standard deviation of analysts� quarterly earnings forecasts scaled by end of quarter stock price. The dispersion dummy (DISP_DUMMY) takes the value of one when DISPERSION is zero and thus the log operation is invalid. Market-to-book (MTB) is the end of quarter market value of equity plus book value of debt divided by total book assets. YOUNG is a dummy variable equal to one for firms under ten years old and zero otherwise. SIZE is defined as the natural logarithm of sales. Tangibility (PPE) is defined as property, plant and equipment to total book assets. PROFITABILITY is income before extraordinary items scaled by total book assets. Residual standard deviation (VOLATILITY) is calculated from daily returns using a modified market model with the contemporaneous, lead, and lagged market return as explanatory variables in the spirit of Scholes and Williams (1977) and Dimson (1979). INTRATE equals the rate on 10-year AAA-rated corporate bonds and is obtained from the St. Louis Federal Reserve. Heteroskedasticity adjusted standard errors are in parentheses. The sample period is 1984-2002. There are 69,944 firm-quarter observations in our full sample, and 38,177 firm-quarter observations in our financing deficit only sample. Panel A: Full sample Panel B: Findef>0 subsample OLS Fixed effects OLS Fixed effects 1a 2a 1b 2b Intercept
0.0009*** (0.0001) - -0.0019***
(0.0002) -
FINDEF
0.0209 (0.1761)
0.0153 (0.1629)
-0.6839*** (0.2100)
-0.9373*** (0.2010)
Dispersion*FINDEF
0.0435*** (0.0095)
0.0423*** (0.0088)
0.0332*** (0.0118)
0.034*** (0.0108)
DISP_DUMMY *FINDEF
-0.3724*** (0.0727)
-0.3635*** (0.0681)
-0.3124*** (0.0913)
-0.314*** (0.0847)
MTB*FINDEF
-0.0201*** (0.0035)
-0.0202*** (0.0033)
-0.0162*** (0.0036)
-0.0136*** (0.0034)
PPE*FINDEF
0.2031*** (0.0373)
0.1854*** (0.0364)
0.2782*** (0.0443)
0.2835*** (0.0491)
SIZE*FINDEF
0.0755*** (0.0060)
0.0769*** (0.0056)
0.103*** (0.0069)
0.1156*** (0.0072)
YOUNG *FINDEF
-0.0375* (0.0227)
-0.0394* (0.0221)
-0.0731*** (0.0257)
-0.0857*** (0.0267)
PROFITABILITY *FINDEF
-0.743*** (0.1764)
-0.7941*** (0.1712)
-0.8196*** (0.2458)
-0.9633*** (0.2553)
VOLATILITY *FINDEF
-2.3388*** (0.5787)
-2.2246*** (0.5649)
-2.4413*** (0.6624)
-1.777*** (0.6684)
INTRATE*FINDEF
-0.066*** (0.0073)
-0.0685*** (0.0067)
-0.0442*** (0.0098)
-0.0444*** (0.0093)
Adjusted R2 0.5094
0.5225 0.4951
0.5360
*, **, *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.
44
Table 4 Robustness checks: determinants of the source of external financing
This table presents a conditional model, in which the independent variables are first differenced. The dependent variable is change in debt, and the conditioning variable is FINDEF. The financing deficit (FINDEF) is defined as the sum of net equity and debt issues for a firm in a given quarter. DISPERSION is the average standard deviation of analysts� quarterly earnings forecasts scaled by end of quarter stock price. Market-to-book (MTB) is the end of quarter market value of equity plus book value of debt divided by total book assets. YOUNG is a dummy variable equal to one for firms under ten years old and zero otherwise. SIZE is defined as the natural logarithm of sales. Tangibility (PPE) is defined as property, plant and equipment to total book assets. PROFITABILITY is income before extraordinary items scaled by total book assets. Residual standard deviation (VOLATILITY) is calculated from daily returns using a modified market model with the contemporaneous, lead, and lagged market return as explanatory variables in the spirit of Scholes and Williams (1977) and Dimson (1979). INTRATE equals the rate on 10-year AAA-rated corporate bonds and is obtained from the St. Louis Federal Reserve. Heteroskedasticity adjusted standard errors are in parentheses. The sample period is 1984-2002. There are 61,251 firm-quarter observations in our sample.
Conditional model with first differences of independent variables
Dependent variable: change in debt (∆D)
Panel A: Full sample Panel B: Findef >0 subsample
OLS Fixed effects OLS Fixed effects
1a 2a 1b 2b
Intercept
0.001*** (0.0001) - 0.001***
(0.0003) -
FINDEF
0.5465*** (0.0101)
0.5565*** (0.0096)
0.5726*** (0.0156)
0.5721*** (0.0149)
∆DISPERSION*FINDEF
4.1976*** (1.3242)
4.2148*** (1.2364)
5.3811*** (1.8086)
5.2818*** (1.6366)
∆MTB*FINDEF
0.013** (0.0064)
0.013** (0.0060)
0.015** (0.0067)
0.0132** (0.0062)
∆PPE*FINDEF
0.501*** (0.0916)
0.5098*** (0.0866)
0.6965*** (0.1004)
0.6923*** (0.0927)
∆SIZE*FINDEF
0.0556*** (0.0151)
0.0588*** (0.0141)
0.0691*** (0.0159)
0.0672*** (0.0144)
∆PROFITABILITY*FINDEF
-0.7095*** (0.2092)
-0.6763*** (0.2120)
-0.9639*** (0.3037)
-0.8603*** (0.3193)
∆VOLATILITY*FINDEF
-0.5925 (0.7974)
-0.4654 (0.7617)
-0.7042 (0.9063)
-0.2243 (0.8992)
∆INTRATE*FINDEF
-0.0164 (0.0119)
-0.0186* (0.0111)
0.0048 (0.0160)
0.0036 (0.0142)
Adjusted R2 0.4866 0.4900 0.4565
0.5013
*, **, *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.
45
Table 5 Determinants of the probability that a firm deviates from the strict pecking order
Logit regressions to assess how our variables affect the probability of a firm deviating from the strict pecking order. In Panel A, deviating from the strict pecking order is defined as FINDEF>0. In Panel B, deviating from the strict pecking order is defined as ∆E>0. All independent variables except YOUNG are lagged to ensure compliance with the predictive nature of the regression. Financing deficit (FINDEF) is defined as the sum of net equity and debt issues for a firm in a given quarter. DISPERSION is the natural logarithm of the average standard deviation of analysts� quarterly earnings forecasts scaled by end of quarter stock price. The dispersion dummy (DISP_DUMMY) takes the value of one when DISPERSION is zero and thus the log operation is invalid. Market-to-book (MTB) is the end of quarter market value of equity plus book value of debt divided by total book assets. YOUNG is a dummy variable equal to one for firms under ten years old and zero otherwise. SIZE is defined as the natural logarithm of sales. Tangibility (PPE) is defined as property, plant and equipment to total book assets. PROFITABILITY is income before extraordinary items scaled by total book assets. Residual standard deviation (VOLATILITY) is calculated from daily returns using a modified market model with the contemporaneous, lead, and lagged market return as explanatory variables in the spirit of Scholes and Williams (1977) and Dimson (1979). INTRATE equals the rate on 10-year AAA-rated corporate bonds and is obtained from the St. Louis Federal Reserve. Standard errors are in parentheses. The sample period is 1984-2002. There are 61,251 firm-quarter observations.
Panel A
Dependent variable:
FINDEF>0
Panel B
Dependent variable:
∆E>0
1a 2a 1b 2b
Intercept
1.147*** (0.1640)
0.9411*** (0.1677)
3.2514*** (0.1733)
3.005*** (0.1772)
Lag(DISPERSION)
-0.0504*** (0.0095)
-0.0814*** (0.0109)
-0.0873*** (0.0100)
-0.1234*** (0.0113)
Lag(DISP_DUMMY)
0.4741*** (0.0723)
0.4752*** (0.0723)
0.9003*** (0.0756)
0.8978*** (0.0757)
Lag(MTB)
0.1017*** (0.0064)
0.1581*** (0.0118)
0.0806*** (0.0068)
0.1495*** (0.0129)
Lag(DISPERSION)* Lag(MTB)
-
0.0101*** (0.0017)
-
0.0122*** (0.0019)
Lag(PPE)
0.1275*** (0.0392)
0.1283*** (0.0392)
-0.3371*** (0.0402)
-0.3357*** (0.0402)
Lag(SIZE)
-0.1187*** (0.0059)
-0.1191*** (0.0059)
-0.2114*** (0.0062)
-0.2115*** (0.0062)
YOUNG
0.1825*** (0.0266)
0.1807*** (0.0266)
0.0343 (0.0282)
0.0327 (0.0282)
Lag(PROFITABILITY)
-1.0046*** (0.2494)
-1.0555*** (0.2498)
-3.2097*** (0.3218)
-3.2818*** (0.3225)
Lag(VOLATILITY)
9.3012*** (0.7488)
9.633*** (0.7513)
20.9684*** (0.8360)
21.3747*** (0.8388)
Lag(INTRATE)
0.0574*** (0.0077)
0.0608*** (0.0077)
-0.0124 (0.0080)
-0.0084 (0.0080)
*, **, *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.
46
Table 6 Determinants of financing deficit using restricted samples
OLS and fixed effects regressions of financing deficit regressed on firm characteristics. Financing deficit (FINDEF) is defined as the sum of net equity (∆E) and debt (∆D) issues for a firm in a given quarter. In Panel A, if ∆E (∆D) is less in absolute value than 5% in a given quarter then ∆E (∆D) is set equal to zero. In Panel B, if ∆E (∆D) is less than +5% in a given quarter then ∆E (∆D) is set equal to zero. DISPERSION is the natural logarithm of the average standard deviation of analysts� quarterly earnings forecasts scaled by end of quarter stock price. The dispersion dummy (DISP_DUMMY) takes the value of one when DISPERSION is zero and thus the log operation is invalid. Market-to-book (MTB) is the end of quarter market value of equity plus book value of debt divided by total book assets. YOUNG is a dummy variable equal to one for firms under ten years old and zero otherwise. SIZE is defined as the natural logarithm of sales. Tangibility (PPE) is defined as property, plant and equipment to total book assets. PROFITABILITY is income before extraordinary items scaled by total book assets. Residual standard deviation (VOLATILITY) is calculated from daily returns using a modified market model with the contemporaneous, lead, and lagged market return as explanatory variables in the spirit of Scholes and Williams (1977) and Dimson (1979). INTRATE equals the rate on 10-year AAA-rated corporate bonds and YIELDSLOPE is the slope of the yield curve (YIELDSLOPE), defined as the 20-year Treasury bond rate minus the 3-month Treasury bill rate. Interest rate data is obtained from the St. Louis Federal Reserve. Heteroskedasticity adjusted standard errors are in parentheses. The sample period is 1984-2002. There are 69,944 firm-quarter observations in our sample.
*, **, *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.
47
Table 7 Determinants of the source of external financing using restricted samples
Conditional model estimates to examine the source of external financing. The dependent variable is the change in debt (∆D). Financing deficit (FINDEF) is defined as the sum of net equity (∆E) and debt (∆D) issues for a firm in a given quarter. For regressions in Panel A, if ∆E (∆D) is less in absolute value than 5% in a given quarter then ∆E (∆D) is set equal to zero. For regressions in Panel B, if ∆E (∆D) is less than +5% in a given quarter then ∆E (∆D) is set equal to zero. DISPERSION is the natural logarithm of the average standard deviation of analysts� quarterly earnings forecasts scaled by end of quarter stock price. The dispersion dummy (DISP_DUMMY) takes the value of one when DISPERSION is zero and thus the log operation is invalid. Market-to-book (MTB) is the end of quarter market value of equity plus book value of debt divided by total book assets. YOUNG is a dummy variable equal to one for firms under ten years old and zero otherwise. SIZE is defined as the natural logarithm of sales. Tangibility (PPE) is defined as property, plant and equipment to total book assets. PROFITABILITY is income before extraordinary items scaled by total book assets. Residual standard deviation (VOLATILITY) is calculated from daily returns using a modified market model with the contemporaneous, lead, and lagged market return as explanatory variables in the spirit of Scholes and Williams (1977) and Dimson (1979). INTRATE equals the rate on 10-year AAA-rated corporate bonds and is obtained from the St. Louis Federal Reserve. Heteroskedasticity adjusted standard errors are in parentheses. The sample period is 1984-2002. There are 69,944 firm-quarter observations. Panel A:
Adjusted R2 0.5749 0.5880 0.6080 0.6230 *, **, *** indicate statistical significance at the 10%, 5%, and 1% level, respectively.
48
Table 8 Change in dispersion around security offerings
Changes in DISPERSION measured in three-month increments prior to and after SEO filings, public debt filings, shelf equity filings and offers, and private investment in public equity (PIPE) offers conducted during 1990-2003 by U.S. firms. The sample of SEOs and shelf-registered offers contains primary offerings of common stock. The sample of SEOs, public debt offers, shelf offers, and PIPEs are each constrained to those offers for which dispersion data is available in every month during the interval [-9, 9] around the event date. Change in dispersion is calculated for each firm for the given interval, and then it is averaged across firms.
Panel A: Change in dispersion around SEO filings
Interval in eventmonth Mean ∆DISPERSION T-value Number of observations
-9 to -6 0.00020 0.77 187
-6 to -3 0.00012 0.40 187
-3 to 0 -0.00087 -2.71 187
0 to 3 -0.00020 -1.14 187
3 to 6 0.00089 1.94 187
6 to 9 0.00054 0.66 187
Panel B: Change in dispersion around public debt filings
Interval in eventmonth Mean ∆DISPERSION T-value Number of observations
-9 to -6 0.00002 0.12 814 -6 to -3 -0.00022 -1.25 814 -3 to 0 -0.00003 -0.20 814 0 to 3 -0.00002 -0.14 814 3 to 6 0.00003 0.26 814 6 to 9 -0.00003 -0.25 814
Panel C: Change in dispersion around shelf filing dates
Interval in eventmonth Mean ∆DISPERSION T-value Number of observations
-9 to -6 -0.00127 -1.26 97 -6 to -3 0.00006 0.28 97 -3 to 0 0.00009 0.25 97 0 to 3 0.00025 0.60 97 3 to 6 0.00059 0.68 97 6 to 9 -0.00090 -1.39 97
49
Table 8 (cont.)
Change in dispersion around security offerings Changes in DISPERSION measured in three-month increments prior to and after SEO filings, public debt filings, shelf equity filings and offers, and private investment in public equity (PIPE) offers conducted during 1990-2003 by U.S. firms. The sample of SEOs and shelf-registered offers contains primary offerings of common stock. The sample of SEOs, public debt offers, shelf offers, and PIPEs are each constrained to those offers for which dispersion data is available in every month during the interval [-9, 9] around the event date. Change in dispersion is calculated for each firm for the given interval, and then it is averaged across firms.
Panel D: Change in dispersion around shelf offer dates
Interval in eventmonth Mean ∆DISPERSION T-value Number of observations
-9 to -6 0.000032 0.07 128 -6 to -3 0.00047 0.99 128 -3 to 0 -0.00022 -0.41 128 0 to 3 0.00077 0.92 128 3 to 6 -0.00055 -0.75 128 6 to 9 0.00057 0.89 128
Panel E: Change in dispersion around PIPEs
Interval in eventmonth Mean ∆DISPERSION T-value Number of observations
-9 to -6 0.00081 0.39 47 -6 to -3 -0.00174 -0.89 47 -3 to 0 0.00002 0.02 47 0 to 3 -0.00032 -0.25 47 3 to 6 0.00013 0.17 47 6 to 9 -0.00105 -1.37 47
50
Figures 1-8 The time-series of average firm characteristics
Figure 1, and Figures 3-8 plot the sample mean of the firm characteristic labeled for each quarter in the sample. Figure 2 plots the coefficient of financing deficit from the cross-sectional regression of net change in debt on financing deficit. The financing deficit is defined as the sum of net equity and debt issues for a firm in a given quarter. Dispersion is the average standard deviation of analysts� quarterly earnings forecasts scaled by the end of quarter stock price. Market-to-book is the end of quarter market value of equity plus book value of debt divided by total book assets. The natural logarithm of sales variable is our SIZE proxy throughout the paper. Property, plant and equipment is scaled by total book assets. Profitability is income before extraordinary items scaled by total book assets. Return volatility is the residual standard deviation calculated from daily returns using a modified market model with the contemporaneous, lead, and lagged market return as explanatory variables in the spirit of Scholes and Williams (1977) and Dimson (1979). The sample period is 1984-2002. There are 69,944 firm-quarter observations.
51
Figure 1: Average financing deficit
-0.02-0.01
00.010.020.030.04
1984 1987 1990 1993 1996 1999 2002
Figure 3: Average Dispersion of Analysts' Earnings Forecasts
-7.0-6.5-6.0
-5.5-5.0
1984 1987 1990 1993 1996 1999 2002
Figure 4: Average Market-to-Book
1.5
2.0
2.5
3.0
3.5
1984 1987 1990 1993 1996 1999 2002
Figure 5: Average Ratio of Property, Plant, end Equipment
0.200.240.280.320.360.40
1984 1987 1990 1993 1996 1999 2002
Figure 6: Average Logarithm of Sales
19.019.119.219.319.419.519.6
1984 1987 1990 1993 1996 1999 2002
Figure 7: Average Profitability
-0.01
0.00
0.01
0.02
1984 1987 1990 1993 1996 1999 2002
Figure 8: Average Return Volatility
0.010
0.020
0.030
0.040
0.050
1984 1987 1990 1993 1996 1999 2002
Figure 2: Coefficient on Financing Deficit
0.10
0.30
0.50
0.70
0.90
1984 1987 1990 1993 1996 1999 2002
52
Figure 9 Dispersion around SEOs in event time
The figure plots the mean monthly dispersion around SEOs in event time. Our sample of SEOs contains primary SEOs of common stock by non-financial and non-utility U.S. firms during 1990-2003. Unit offerings, right offerings and shelf offers are excluded. The sample is constrained to SEOs for which dispersion data is available in every month [-9, 9] around the SEO. There are 187 sample SEOs that meet this requirement.
Figure 9: Average monthly dispersion around SEOs in event time
0.00%
0.05%
0.10%
0.15%
0.20%
0.25%
0.30%
0.35%
0.40%
-9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9
eventmonth
disp
ersi
on
53
Figures 10-11 Relative mix of SEOs, shelf offers, and PIPEs
Figure 10 illustrates the relative number of equity offers (in terms of annual percentages) during our sample period 1990-2003 that are SEOs (middle portion), shelf offers (bottom portion), and PIPEs (top portion). Figure 11 illustrates the relative proceeds of equity offers (in terms of annual percentages) during our sample period 1990-2003 that are SEOs (middle portion), shelf offers (bottom portion), and PIPEs (top portion).