Does tradeoff theory explain high-frequency debt issuers?
∗
B. Espen Eckbo
†
Michael Kisser
‡
July 2014
Current draft: July 2015
Abstract
Over the past forty years, one-third of the publicly listed industrial firms in the U.S. raised two-
thirds of all public and private debts (net of debt rollovers). We use these high-frequency debt issuers
(HFIs)—large and highly leveraged, investment-intensive firms with low Tobin’s Q—to test tradeoff
theory of debt financing. Relative to low-frequency net-debt issuers (LFIs)—small, low-leveraged,
R&D-intensive firms with high Q—HFIs appear to face low total and fixed issue costs. Under dynamic
tradeoff theory, HFIs should therefore exhibit smaller issue sizes, lower leverage ratio volatility, and
higher speed-of-adjustment to deviations from target leverage ratios than LFIs, which our evidence
fails to support. However, consistent with dynamic financing and investment models, over-leveraged
firms occasionally issue debt followed by equity issues and leverage ratio reductions. Finally, we show
that CEO equity ownership and stock-based compensation are both higher for HFIs than for other
sample firms.
∗
Earlier versions of this research were circulated under the titles “Corporate funding: who finances externally?” and
“Corporate debt issues and leverage dynamics”. We have benefitted from the comments and suggestions of Andras Danis,
Harry DeAngelo, Pierre Chaigneau, Michael Hertzel, Michael Roberts, Karin Thorburn, Toni Whited, and seminar partici-
pants at Concordia University, the Norwegian School of Economics, the Norwegian School of Management, Tuck School of
Business at Dartmouth, Tulane University, University of Adelaide, University of Bristol, University of St. Gallen, and the
Vienna Graduate School of Finance. This research has also been presented at the University of Stavanger Corporate Fi-
nance Conference, and at the annual meetings of the Financial Management Association, the European Finance Association,
the Society for Financial Studies (SFS) Cavalcade, and the Southern Finance Association. Financial support from Tuck’s
Lindenauer Center for Corporate Governance is gratefully acknowledged.
†
‡
1 Introduction
It is well established that the distribution of leverage ratios among publicly traded nonfinancial U.S. firms
has a persistent and sizable left tail consisting of firms with zero or near-zero debt (Strebulaev and Yang,
2013). These are firms that never or almost never issue debt during their lifetime as public companies.
What is much less documented, however, is the right tail of the cumulative debt-issue frequency distribu-
tion: listed firms that persistently fund themselves by issuing new debt (beyond debt rollovers) and which,
as it turns out, raise the bulk of all public and private debts. These firms evidently view debt-financing
as uniquely beneficial, and they face sufficiently low debt issue costs to issue frequently. We argue that
the dual combination of high debt benefits and low issue costs makes these firms ideally suited to test
dynamic tradeoff theory: if high-frequency debt issuers do not dynamically rebalance capital structure
towards a leverage target, who does?
The paper has three distinct objectives. The first is to document little-known differences in the firm
characteristics of high- and low-frequency net-debt issuers, beginning with the year of public listing.
Second, we provide evidence on the fundamental notion that low-frequency issuers have greater fixed
debt-issue costs than high-frequency issuers. According to the dynamic tradeoff theory pioneered by
Fischer, Heinkel, and Zechner (1989), which adds issue costs and optimal leverage conditions to the
continuous-time asset pricing framework of Merton (1974), fixed issue costs contribute to long periods
of issue inactivity. This theory also makes predictions about the relative issue behavior of high and low-
frequency issuers when they do issue debt, and our third objective is to test these additional theoretical
implications. In the process, we also address a key prediction of the more recent, discrete-time financing
model with endogenous investment developed by DeAngelo, DeAngelo, and Whited (2011).
We sample debt issues and retirements (both private and public debts) using Compustat cash-flow
statements for public industrial companies over the period from 1971 to 2012. As Lemmon, Roberts, and
Zender (2008) and DeAngelo and Roll (2015), we sample at the annual frequency.
1
The annual debt issue
frequency treats multiple debt issues within one year as being part of the same strategic time horizon, as
may be the case for corporate investment decisions (that require external financing) as well. Importantly,
to avoid pooling older firms with newly listed companies, we require firms to go public during the sample
1
Our life cycle analysis (beginning with public listing) requires contiguous data, and there are more gaps for quarterly
data. However, when requiring at least four contiguous quarters in the data, redoing our main tests based on quarterly
observations yields similar inferences.
1
period and condition the analysis on firm age since going public.
Throughout the paper, high-frequency net-debt issuers (henceforth HFIs) are firms in the upper quar-
tile of the annual cumulative net-debt issue frequency distribution (net of debt retirements) conditional
on listing age. Moreover, low-frequency net-debt issuers (LFIs) are firms in the lower quartile of this
conditional distribution. We focus on the net-debt issue frequency as we are primarily interested in the
frequency of issue decisions that change the leverage ratio, rather than on debt-rollovers designed to ex-
tent the maturity of existing debt. Over the sample period, HFIs receive as much as 70% of all net-debt
issue proceeds, while LFIs only receive 7%. Importantly, there is only minimal migration over time of
firms between the HFI and LFI categories.
Our issue-frequency classification creates a substantial spread between HFIs and LFIs in terms of
leverage stability, debt-issue dynamics and firm characteristics. Interestingly, high-frequency issuers stand
out as relatively large investment-intensive firms with high leverage and low Tobin’s Q. In contrast, low-
frequency issuers are relatively small R&D-intensive firms with low leverage and high Q. Some of these
differences are apparent already soon after public listing (thus the minimal migration between firms in
the HFI and LFI classifications). Market leverage ratios quickly rise to an average of 35% for HFIs versus
an average of 10% for LFIs. Moreover, while as much as 32% of the LFIs have zero leverage in a typical
firm-year, only 1% of HFIs ever reduce leverage to zero during their lifetime as public companies. Overall,
HFIs tend to finance a relatively capital-intensive investment (capex) program externally through debt
issues, while LFIs tend to finance a relatively intensive R&D program internally.
In general, large fixed issue costs may slow issue frequency whether or not the firm manages capital
structure towards a target. Thus, it is natural to assume at the outset that HFIs face smaller and less
fixed issue costs than LFIs. As extant evidence on issue costs is sparse (Altinkilic and Hansen, 2000;
Eckbo, Masulis, and Norli, 2007), and in order to corroborate this issue-cost assumption, we provide our
own empirical issue-cost analysis, in two steps. First, we estimate dynamic issue hazard shapes using
our full sample of public and private net-debt issues, and compare them to the simulated hazard shapes
reported by Leary and Roberts (2005) for firms with fixed versus proportional issues costs, respectively.
This simple comparison indeed suggests that LFIs have largely proportional and LFIs largely fixed issue
costs.
Second, we directly identify a fixed-cost component in underwriter spreads using the Altinkilic and
Hansen (2000) issue cost function for 3,773 public bond issues over the sample period. This estimation
2
also suggests that low-frequency public debt issuers have significantly higher fixed issue costs than high-
frequency issuers. Finally, other than these cost differences, extant quantitative indices (Kaplan and
Zingales, 1997; Whited and Wu, 2006; Hadlock and Pierce, 2010) do not suggest that LFIs are fundamen-
tally more financially constrained than HFIs. It appears that the spread between HFIs and LFIs may be
a spread on direct issue costs but not on more fundamental sources of financial constraints that motivate
such quantitative indices.
We then turn to predictions of dynamic tradeoff theory for differences in issue behavior of HFIs and
LFIs. Under the assumption that LFIs have higher total and fixed issue costs than HFIs, we focus on
three hitherto unexplored predictions: relative to LFIs, HFIs should exhibit (1) smaller issue sizes scaled
by total assets, (2) lower leverage ratio volatility, and (3) higher speed-of-adjustment (SOA) to deviations
from leverage targets.
Our evidence fails to support any of these predictions. First, net-debt issue sizes are roughly equal
across HFIs and LFIs—and not greater for LFIs as predicted. Second, leverage ratio volatilities are
substantially higher for HFIs than for LFIs (not lower as predicted). Medium to low-frequency net-debt
issuers tend to have relatively stable leverage ratios. Thus, we also suspect that our HFIs drive much of
the firm-level leverage instability by highly leveraged firms shown by DeAngelo and Roll (2015).
Third, notwithstanding the fact that the debt-issue frequency is typically ten times higher for HFIs
than for LFIs, the two groups of firms receive statistically indistinguishable SOA coefficient estimates.
In the SOA estimation, we use empirical leverage targets found elsewhere in the literature (Fama and
French, 2002; Flannery and Rangan, 2006; Hovakimian and Li, 2012; Faulkender, Flannery, Hankins, and
Smith, 2012). Apparently, net-debt changes (the denominator of the leverage ratio change) and total
asset changes (the numerator) work in such a way as to produce similar SOA coefficient estimates for
HFIs and LFIs. This finding, which fails to support tradeoff theory, also adds to the concern raised by
Welch (2004) and others about the use of leverage ratio changes to identify the true speed with which
firms manage leverage towards a target.
The data also indicate strongly that net-debt issues are driven by ongoing investments. In fact, we
show that capex is the main determinant of the probability of becoming an HFI. Moreover, net-debt issue
size and the size of capex are highly correlated for HFIs, which is also reflected in so-called financing-deficit
regressions.
2
Since the ongoing debt funding of investments can mask true tradeoff behavior in the data,
2
The financing deficit captures the shortfall in internally available funds for investment (Shyam-Sunder and Myers, 1999;
3
it may be an important reason why classical dynamic tradeoff theories that abstract from investment
appear inadequate in terms of explaining actual debt issue behavior by HFIs.
In order to account for investment financing, we turn to the discrete-time, structural financing and
investment model of DeAngelo, DeAngelo, and Whited (2011). In this model, firms dynamically maintain
a long-term target leverage ratio while at the same time responding to persistent investment shocks. We
are particularly interested in this model’s prediction that over-leveraged firms may issue “transitory” debt
to help finance new investments. In theory, as the investment shocks fade, these transitory debt issues
will be followed by debt repurchases to restore leverage targets. Interestingly, when using the empirical
leverage targets from the speed-of-adjustment analysis, we identify net-debt issues by firms that are both
over-leveraged and invest in excess of the industry median capex. Moreover, these net-debt issues signif-
icantly increase the likelihood of subsequent leverage ratio reductions. On the other hand, this leverage
ratio reduction is financed through a combination of retained earnings and additional (costly) equity issues
rather than by debt repurchases—as ongoing funding needs dominate debt-retirement considerations.
We end the paper with a check on the stock ownership and stock-based compensation of Chief Ex-
ecutive Offer (CEO) of HFIs relative to all other firms in our sample. Eckbo, Thorburn, and Wang
(2014) show that corporate bankruptcy imposes significant personal costs on the median CEO. Thus,
some managers may choose to hedge against bankruptcy risk ex ante by reducing the firm’s exposure
to leverage. Interestingly, in their study of zero and near-zero leverage firms—a substantial fraction of
which end up being classified as LFIs here—Strebulaev and Yang (2013) find that firms with large CEO
equity ownership and more CEO-friendly boards are more likely to pursue low-leverage policies. In our
data, HFIs have both greater equity ownership and a greater proportion of their total compensation paid
in stock than is the case for other firms, suggesting that the impact of CEO ownership is more complex
and perhaps nonlinear in firm leverage policy.
The rest of the paper is organized as follows. Section 2 identifies HFIs and LFIs and displays their
firm characteristics, and documents the stability of the HFI classification over time. Section 3 shows
the results of our dynamic hazard estimation and our estimates of the direct issue cost function using
public bond issues. In Section 4, we explore predictions of the classical tradeoff model without corporate
investment, followed by an examination of dynamic financing and investment theory in Section 5. We
Frank and Goyal, 2003). See also Leary and Roberts (2010) and Lemmon and Zender (2010) for empirical tests involving
financing deficits, and DeAngelo, DeAngelo, and Stulz (2010) for evidence that firms issue equity after “running out of cash”.
4
discuss CEO stock compensation in Section 6, while Section 7 concludes the paper.
2 Who are high-frequency debt issuers?
2.1 Sample selection
As summarized in Panel A of Table 1, our main sample consists of 12,131 nonfinancial, non-government,
U.S. domiciled corporations and a (unbalanced) panel of 93,031 firm-years from the period 1971-2012. To
arrive at this sample, we start with the merged CRSP/Compustat database and impose a largely standard
set of sample requirements listed in Panel A. These restrictions include dropping financial firms and
regulated utilities, foreign companies, and firms with missing entries of key balance sheet characteristics.
The definition of the balance sheet and cash flow variables used in this paper are summarized in Appendix
Tables 1 and 2 using Compustat mnemonics.
Our requirement that a sample firm went public during the sample period is unusual relative to the
extant capital structure literature (Graham and Leary, 2011). This restriction excludes a total of 2,411
firms and 39,449 firm-years, primarily from the early sample years. Conditioning on the age of the firm is
important for our lifecycle issue frequency analysis, and it allows us to properly document issue-frequency
persistence. Moreover, the very notion of “issue frequency” is related to asset growth which likely differs
with firm age. We define listing age as the difference between the reporting date of the financial statement
and the date of the first month a company is reported in the CRSP/Compustat merged database. The
average annual number of listed firms in the sample is 6,471 (median 6,848), and the average number of
years a firm is listed is 8 (median 5). A total of 5,921 firms are listed for more than five years, 2,931 for
more than ten years, and 782 firms for twenty years or more.
Panel B of Table 1 lists a sample of 3,773 public bond issues by 1,075 public companies drawn from
the security issue section of SDC, 1980-2011. This sample is used in Section 3 below exclusively for the
purpose of estimating a fixed-cost component in underwriter fees across high and low-frequency bond
issuers. We return to a description of the selection criteria for this sample at that point.
5
2.2 Annual cumulative issue frequencies
2.2.1 Net-debt issues
Table 2 shows the annual cumulative frequencies of net-debt issues (Panel A), net-debt retirements
(Panel B), and equity issues (Panel C). The frequency begins with the year of listing (event year 0),
and it continues over the subsequent twenty years (the empirical analysis uses all years on record, also
if longer than twenty). Since the issue frequency will depend on the issue size threshold used to define
it, we report for illustrative purposes two frequencies in Table 2: the first based on a minimum issue size
of 2.5% of total assets (Panel A), and the second based on the 5% threshold (Panel B) more commonly
found in the extant literature on security issuances. To maximize sample size, the subsequent empirical
analysis is based on the 2.5% issue-size threshold.
In each year following public listing, firms are classified as belonging to the group of high-frequency
issuers (henceforth HFIs), low-frequency issuers (LFIs) or the group in between. HFIs are firms in the
top quartile (above the 75
th
percentile) of the annual cumulative net-debt issue distribution, while LFIs
are firms in the bottom quartile (below the 25
th
percentile). Of the total sample of 93,031 firm-year
observations, 34% are represented by HFIs and 41% by LFIs. Moreover, the average annual number of
HFIs (LFIs) is 2,474 (4,040). HFIs receive 70% of the total dollar value of all net-debt issues over the
sample period, while LFIs receive only 7%. In sum, the HFIs not only issue frequently: they also receive
a highly disproportional share of total issue proceeds.
Beginning with the year of public listing (year 0), Table 2 shows the annual distribution of the number
of LFIs and HFIs, the annual average fraction of total net-debt issue proceeds received by each of the two
categories, and the average cumulative number of net-debt issues. As shown, the cumulative net-debt
issue frequency is skewed towards HFIs throughout the public firm lifecycle, both in terms of volume and
number of issues.
For example, in Panel A.1, of the 5,921 firms that have been listed for five years, 2,739 or 46% fall in
the LFI category and 1,823 or 31% are HFIs. Over these first five years of public listing, HFIs undertake
on average 3.59 net-debt issues above the 2.5% size threshold, while LFIs undertake on average only 0.49
such issues. Moreover, in year five, HFIs raised 73% and LFIs 10% of aggregate issue proceeds. After
ten years of listing, HFIs have on average made 5.98 issues, while LFIs have still made only 0.49 issues.
Also, in year ten, HFIs raised 59% and LFIs only 4% of aggregate issue proceeds. Looking at the overall
6
sample, the median cumulative number of net-debt issues is two after five years and three after ten years.
Raising the issue-size threshold to 5% in Panel A.2 of Table 2 shows that, after five years of public
listing, HFIs have issued on average 2.73 net-debt issues and LFIs close to zero such issues. Moreover, in
year five, HFIs raised 79% and LFIs 3% of aggregate proceeds. After ten years of listing, the cumulative
number of issues averages 4.94 for HFIs (one issue every two years), and 0.51 by LFIs (one issue every
twenty years). HFIs (LFIs) raised 56% (7%) of total issue proceeds, respectively, in year ten. The
sample median cumulative number of net-debt issues at the 5% threshold is one after five years, two after
ten years, and four after twenty years. Thus, the cumulative net-debt issue distribution is highly skewed
towards HFIs also with the 5% threshold—both in terms of the number of issues and issue proceeds—with
the median listed firm issuing net-debt at roughly half the pace of HFIs.
Quarterly sampling increases the issue frequency somewhat as net-debt issues occasionally cluster
within the year. Also, it is interesting to compare the annual frequency in Panel A.2 of Table 2 to that
reported by Leary and Roberts (2005). They find that firms on average issue net debt of 5% or more
once every eight quarters (sample period 1984-2001). Assuming there are no years with multiple issues,
this extrapolates into one debt issue every two years on average. In comparison, Panel A.2 of Table 2
shows that the actual annual net-debt issue frequency is less than half that: once every five years.
2.2.2 Net-debt retirements and equity issues
Our examination of dynamic financing and investment models also involves searching for debt retirements
following what DeAngelo, DeAngelo, and Whited (2011) call “transitory” debt issues (positive net debt
issues in a period where the firm is over-leveraged). While we return to an analysis of such retirements
below (Section 5), Panel B of Table 2 shows the cumulative frequency and volume of net-debt retirements
in the total sample. Interestingly, net debt retirements are much less skewed towards HFIs than is the
case for net-debt issues.
For example, at the 2.5% size threshold in Panel B.1 and after five years of listing, HFIs and LFIs
have almost identical average number of net-debt retirements: 1.22 and 1.30, respectively. Moreover, in
year five, the percentage of total retirement volume is on average 37% for both categories of firms. The
difference between HFIs and LFIs increases after ten years, with HFIs making 2.49 net-debt retirements
on average and LFIs 1.54 such issues. The corresponding fractions of total retirement volume in year
ten are 36% and 10%, respectively. In the overall sample, the median cumulative number of net-debt
7
retirements is one after five years and two after ten years of public listing. A similar picture emerges
when using the 5% size threshold in Panel B.2. The fact that HFIs issue net-debt more often and in much
greater dollar volumes than LFIs, while at the same time retiring net-debt at a fairly similar rate to that
of LFIs, suggest that HFIs on average develop a greater leverage ratio over time (confirmed below).
Finally, Panel C of Table 2 shows that, when maintaining the definition of LFI and HFI based on
net-debt issue activity, the average cumulative number of equity issues and percent of annual total issue
volume is similar across LFIs and HFIs. This is true whether using a 2.5% or a 5% equity-issue size
threshold. In the overall sample and using the 2.5% threshold, the average cumulative number of equity
issues is 2.87 after ten years and 3.81 after twenty years, with a median of two and three for both years.
3
In sum, our issue-frequency classification of firms into LFIs and HFIs identifies firms that differ
significantly in terms of net-debt issue activity, but that are quite similar in terms of net-debt retirements
and equity issue activity. We next show that the HFI and LFI classifications also result in a relatively
stable set of firms within each category.
2.3 Firm-level stability of the issue-frequency classification
As Table 2 above demonstrates, the annual differences in cumulative issue frequencies between HFIs and
LFIs persist across listing age and issue-size thresholds. Using the 2.5% issue size threshold, Table 3
further shows that individual companies, once classified as either HFI or LFI, tend to persist in that
classification. Table 3 uses two measures of persistence, one backward-looking and one forward-looking.
To illustrate, focus first on the backward-looking measures in columns (1) - (3) for the HFIs in Panel
A. Column (1) says that 100% of the firms that are classified as HFI in year five were also classified as
HFI one year earlier. Moreover, 88% were also classified as HFIs two years earlier, and 65% three years
earlier. After ten years of listing, the corresponding backward-looking percentages are even higher: of
the firms classified as HFIs in year ten, 100% were so classified also one year earlier, 92% two year earlier,
and 82% three years earlier, respectively.
Turning to the forward looking measures of persistence, column (6) of Table 3 shows that only 1% of
all firms classified as HFI after five years of listing change this classification to LFI in any future sample
year. Thus, there is almost no migration from HFI to LFI. Moreover, 73% of the HFIs in year five retain
3
While our breakdown of the listing age is new to the literature, the overall equity-issue frequencies in Panel C of Table
2 are in line with that reported elsewhere. See, e.g., Eckbo and Masulis (1995), Fama and French (2005), Eckbo, Masulis,
and Norli (2007) and Leary and Roberts (2010).
8
the HFI classification for the future, with the balance of 26% migrating to the medium frequency issuers
sometime in future sample years. After ten years of listing, these percentages are even more striking: 0%
move from HFI to LFI while 82% remain HFI, respectively, with 18% migrating from HFI to become a
“medium” frequency issuer (somewhere between the 25
th
and 75
th
percentiles).
Finally, Panel B of Table 3 further shows that there is almost no migration from LFI to HFI over the
public lifecycle. In most of the years, 100% of the firms that are classified as LFI were also classified as
LFI in each of the previous three years. The only major exception from the high level of persistence is
due to rebalancing: in year five following public listing the threshold for being classified as a LFI increases
from 0 to 1 (Table 2), which changes the portfolio composition of LFIs such that, in this year, only 59%
of LFIs were also classified as low frequency in the previous year. However, looking forward, only 4% of
these LFIs migrate to become HFIs in future sample years. After ten years of public listing, 100% of the
LFIs were LFIs in each of the preceding three years, 89% remain LFI in future sample years, and 0%
migrate to become HFIs. This, of course, is consistent with the evidence in Table 2 that a large majority
of firms classified as LFI remain “dormant” in terms of net-debt issues for most of their lifecycle as public
companies.
Having established the lack of firm migration between HFIs and LFIs, we next characterize the
difference between HFIs and LFIs in terms of key firm characteristics.
2.4 Comparing average firm characteristics of HFIs and LFIs
The average annual firm characteristics listed in Table 4 show clear differences between HFIs and LFIs.
First, LFIs are much less leveraged and have higher cash balances than HFIs. Using the overall average
values at the bottom of each panel, the leverage ratio (both market and book leverage) is 35% for HFIs
and 10% for LFIs. This difference in average leverage ratios is also reflected in column (3) which shows
the fraction of the sample firms that are all-equity financed (AE): it is 32% for LFIs and only 1% for
HFIs.
Moreover, the cash ratio C in column (4) is interesting. It is 31% for LFIs and 10% for HFIs.
This means that much of the build-up of cash balances reported by Bates, Kahle, and Stulz (2009) is
concentrated among our low-frequency net-debt issuers, causing these firms to have negative net leverage
(debt minus cash) on average. High-frequency debt issuers have relatively high leverage ratios whether
measured using gross debt or debt net of liquid assets such as cash balances.
9
Notice also the significant differences in asset structure and growth rates. Relative to LFIs, HFIs
are large on average ($804 million versus $345 million for LFIs), have high degree of asset tangibility
defined as PPE/Assets (0.36 versus 0.22 for LFIs), and exhibit low Q (1.80 versus 2.56 for LFIs). The
relatively low Q for HFIs is reflected in low R&D spending (RD in column (11)), which is 3% for HFIs
and 8% for LFIs. The dividend rate (relative to total assets) is also somewhat lower for HFIs than LFIs
(on average 0.5% versus 0.9%, respectively). However, HFIs have greater capital expenditures (Capex
in column (12)): 10% of total assets versus 6% for LFIs. Perhaps a reflection of the greater investment
intensity, HFIs have relatively high growth rates of both total assets (g
A
) and total sales (g
S
), as shown
in the last two columns.
4
In sum, Table 4 shows that HFIs tend to be large, highly leveraged, low-Q companies with a par-
ticularly active investment program that generates high asset and sales growth. We next show that
investment and R&D expenditure rates are significant drives of issue frequencies.
2.5 Probability of becoming HFI in T years
Panel A of Table 5 presents estimates of the determinants of the probability of becoming a HFI after T
years of listing. This probit estimation employs firm characteristics in the year of public listing (year 0)
in the following model:
Y
∗
iT
= α + βX
i0
+
iT
(1)
Y
iT
= 1 if Y
∗
iT
≥ 75
th
percentile and 0 otherwise
where Y
∗
iT
is the latent variable for the probability of firm i being a HFI after T years of listing, and Y
iT
is the dummy variable for Y
∗
iT
. The vector X
i0
of firm characteristics includes most variables listed in
Table 4 but now measured in year 0.
5
In addition, the estimation of Eq. (1) includes industry dummies
4
To some degree, differences in firm characteristics relate to industry effects. Using the Fama-French 12-industry clas-
sification, we find the following distribution of HFIs and LFIs across different industries: consumer non-durables (HFIs:
8%; LFIs: 6%), consumer durables (HFIs: 3%, LFIs: 3%), manufacturing (HFIs: 13%, LFIs: 11%), energy (HFIs: 8%,
LFIs: 4%), chemicals (HFIs: 2%, LFIs: 2%), business equipment (HFIs: 16%, LFIs: 31%), shops (HFIs: 18%, LFIs: 11%),
healthcare (HFIs: 10%, LFIs: 16%) and other (HFIs: 21%, LFIs: 15%). The subsequent regression analysis will therefore
account for industry (or firm-fixed) effects in addition to the individual firm characteristics.
5
Asset and sale growth rates are not included as they are not available in the year of public listing. In addition, we include
depreciation expenditures (the ratio of depreciation expenditures to assets) to account for non-debt related tax shields.
10
for eight of the 12 Fama-French (FF12) industries (excluding financial firms and regulated utilities). All
covariates are winsorized at the 1(99) percent level.
Panel A shows parameter estimates for forecasting periods of 3, 6, 9, 12 and 15 years following
public listing. These estimates strongly suggest that the issue frequency classification far out in time
is predictable based on year-zero values of several of the characteristics. The initial investment (Capex)
and leverage ratio (L) are the two characteristics most strongly associated with a greater probability of
becoming a HFI. Moreover, initial R&D and cash balance C are associated with a significantly lower
probability of becoming a HFI.
6
We next turn to whether the differential net-debt issue activity for HFIs and LFIs may be attributed
to HFIs facing lower total and fixed issue costs than LFIs.
3 Do fixed issue costs drive issue frequencies?
Dynamic capital structure theory predicts funding activity in part based on the level and form of security
issuance costs (Fischer, Heinkel, and Zechner, 1989). Since high fixed issue costs slow the firm’s optimal
issue frequency, it is natural to view LFIs as high fixed-cost issuers relative to HFIs. In this section,
we present new empirical evidence that, as it turns out, supports this view. This helps motivate our
subsequent tests of dynamic capital structure theory exploiting the different issue behaviors of LFIs and
HFIs.
Extant evidence on issue cost structure (fixed versus variable costs) is sparse (Eckbo, Masulis, and
Norli, 2007). Below, we approach the possible presence of a fixed cost component in debt issue costs from
two complementary angels. The first is an informal comparison of the shapes of estimated dynamic debt
issue hazards (time between net-debt issues) with the simulated shapes in Leary and Roberts (2005).
The second directly estimates fixed-cost components in underwriter fees paid by high- and low-frequency
public bond issuers (a subset of our full Compustat sample of debt issues). We end the section with
a brief check on whether LFIs are also more financially constrained than HFIs—perhaps beyond direct
issue costs—as per quantitative indices suggested by the extant literature.
6
The average marginal effects on the probability are as follows: a 10 percentage point increase in Capex (L) increases the
probability of being classified as a HFI nine years following public listing by 7.9 points (2.1 points). A similar increase in
R&D decreases the probability by 5.3 percentage points.
11
3.1 Dynamic issue-hazard estimation
Panels B and C of Table 5 show the results of estimating the determinants of the time between successive
net-debt issues—the issue hazard or financing spell. In Panel B, the exponential hazard model is of the
following form:
h
i
= h
0
exp(β
0
+ βx
i
)α
i
, (2)
where h
0
is the baseline hazard (when all covariates are equal to zero and assumed constant in Panel B),
and α
i
captures unobserved heterogeneity analogous to a regression error term.
7
The firm characteristics
x
i
enter after subtracting the median value across all firms each year, which means that the baseline
hazard is
ˆ
h
0
= exp(
ˆ
β
0
) for the median firm (Leary and Roberts, 2005). We perform the hazard rate
estimation separately for LFIs and HFIs.
In Panel B, a hazard ratio which is statistically indistinguishable from unity means that the control
variable does not change the likelihood of the financing event taking place the following year. The
tabulated results are consistent with the probit estimation in Panel A in that Capex has a strong and
significant impact on the net-debt issue decision. Specifically, a 10 percentage point increase in capital
expenditures (relative to the median firm) raises the issue hazard by a factor of 1.5 for HFIs and a factor
of 1.4 for LFIs. In other words, it appears that net-debt issues tend to be driven by the need to fund
investment. Panel B also shows that the availability of internal funds—either C or P rof—reduces issue
hazards with similar marginal effects for LFIs and HFIs.
In Panel C, we expand Eq. (2) using a cubic function of the number of years t between issues:
h
i
(t) = h
0
exp(β
0
+ βx
i
+ γf(t))α
i
where f(t) = t + t
2
+ t
3
. (3)
As shown in Panel C, estimation of Eq. (3) exacerbates the difference in the impact of Capex in the issue
hazards of LFIs and HFIs. This effect likely reflects a combination of the persistently lower investment
activity for LFIs (shown earlier in Table 4) and the use of net-debt issues to fund new investment projects.
7
If T is a random variable measuring the time between net-debt issues, the issue hazard function is defined as
h(t) = lim
m→0
Pr(t ≤ T < t + m|T ≥ t)
m
.
Here, h(t) is the instantaneous rate at which a firm issues net debt conditional on not having done so for t periods. Intuitively,
h(t)m is the probability that a firm will issue over the next m periods, conditional on not having issued up to time t, For
example, the hazard function for net-debt issuances at t=5 years tells us the probability that the firm will issue net debt
over the next year (m=1).
12
Below, we exploit this cubic function estimation more specifically in the context of issue-cost differences
between HFIs and LFIs.
3.2 Inferring fixed issue costs from dynamic issue hazard shapes
Panels A and B of Figure 1 plot the shapes of the dynamic net-debt issue hazards that follow from
estimating Eq. (3) for LFIs and HFIs, respectively. The horizontal axis is years since last issue, which
occur in year 0. For example, at year five, the dynamic hazard function gives the estimated probability
of a debt issue in year six conditional on not having issued debt over the previous five years. The plots
of the estimated hazard shapes have steps because time has been discretized to the annual frequency.
To explain the intuition behind these shapes, it is useful to refer to Figure V in Leary and Roberts
(2005), which depicts dynamic hazard shapes using the above cubic function of time between issues on
simulated data (thus without our firm characteristics x
i
). For expositional simplicity, we copy their
simulated hazard shapes directly into panels C and D of Figure 1. In these simulations, firms adjust the
leverage ratio L towards a target L
∗
. As firm value drifts upwards, L falls to a lower boundary L, which
triggers a debt issue. The distance L
∗
− L depends on the magnitude and form of issue costs, with the
distance being greatest for firms with high and largely fixed costs.
8
The estimated shape for LFIs in Panel A is strikingly similar to the simulated shape in Panel C
generated by Leary and Roberts under the assumption of fixed issue costs. Similarly, the estimated
shape for HFIs in Panel B is similar to the simulated shape in Panel D assuming proportional issue costs
only. To better understand the different intercepts and slopes of these dynamic issue hazard shapes,
consider first the fixed-cost issuer in Panel A and C. Large fixed issue cost create a large wedge L
∗
− L.
Since, in year 0, the firm has just issued debt, the expected time to reach L again is relatively low, and
so the intercept is low. Furthermore, the slope of the issue hazard is positive since the longer the time
since the last issue, the closer (in expectation) L is to L and so the greater the probability that the firm
will hit L again and issue in the next period.
9
8
As explained by Leary and Roberts (2005), the initial L
∗
is set equal to the midpoint between an upper and lower
recapitalization boundary of 0.60 and 0.15, respectively, selected to match the maximum and minimum leverage ratios in a
sample of industrial Compustat firms with at least four years of contiguous data, 1984-2001. Stock returns are assumed to
follow a geometric Brownian motion with positive drift: mean 12% and standard deviation 46% (also selected to match the
sample Compustat firms). The leverage process starts at the midpoint of the spread between the upper and lower boundaries
(the assumed initial L
∗
) and are subsequently updated each period using the simulated equity returns.
9
The hazard function in Panel A is hump-shaped, sloping downwards after ten years. This is driven by the fact that the
number of observations available in the estimation diminishing rapidly with time t. Since t is reset to zero when the firm
issues debt, firms with two or more issues do not influence the estimation beyond year 10. Leary and Roberts (2005) make
13
Turning to the relatively small proportional cost case in Panel B and D, note first that the equity
value function (the Brownian motion) need not drift upwards much before the optimality condition for
an issue is again met (where marginal issue cost equals marginal issue benefit). Thus, following an issue
in year 0, the probability of another issue next year is high, and so the intercept is also high. However,
this probability declines with t: the longer the time since the last debt issue, the more likely the stock
return process has drifted downwards to create over-leverage, and the smaller the probability of hitting
the lower recapitalization boundary over the next period.
As a caveat, the similarity of the empirical and simulated issue hazard shapes in Figure 1, while
consistent with tradeoff behavior under fixed and proportional issue costs, may also be driven by other
potential motivations for the observed issue activity of HFIs and LFIs. That is, while the shapes may
be necessary, they are not sufficient to conclude in favor of tradeoff theory. For example, we know from
Table 4 above that HFIs invest intensively, which may drive the bulk of their debt issues rather than
rebalancing efforts. Also, we know from Table 5 that investment is a highly significant driver of the issue
hazard estimates themselves, which the simulations do not address. We return to these issues in the
empirical analysis of tradeoff theory predictions below. However, we first present direct evidence of a
fixed-cost component in underwriter fees for public bond issues which, as it turns out, also depends on
issuer frequency.
3.3 Estimating fixed issue costs in public debt offerings
The net-debt issues identified through our Compustat cash flow statements cover both public and private
debt. While we do not have data on the issue cost of private debt issues, SDC’s Global Issue database
provide issue cost information for public bond issues by a subsample of our firms (1980–2012). In this
section, we use this subsample to provide new and direct evidence on whether relatively high-frequency
public debt issuers have lower fixed costs than low-frequency public debt issuers.
As summarized in Panel B of Table 1, we sample all available public bonds and medium term notes,
and then restrict the issuer to be present in CRSP/Compustat. We follow Altinkilic and Hansen (2000)
and impose several screens for data availability. Our final sample consists of 3,773 public debt issues
made by 1,075 different firms. The sample average underwriter spread is 1.2% of offering proceeds, and
firms raise on average $334 million with an average total equity market value of $11,523 million. Thus,
a similar observation when explaining the hump-shape in Panel C.
14
the average firm in the public bond issue sample used in this section is substantially larger than the
average firm size of $804 million for our HFIs in the broader CRSP/Compustat sample (discussed earlier
in Table 4 above).
Note that, in this section, we use the sample distribution of the public debt issues—not the Compustat
sample classification—to classify high- and low-frequency public debt issuers, here labeled LFPIs and
HFPIs, respectively. The reason for this is that, of the total of 9,283 LFIs in our sample, only 52 are
recorded in SDCs Global Issue database as ever issuing public debt. The public debt issue frequency
distribution based on this particular classification is also skewed: the mean and median are 3.5 and 2
public debt issues, 48% of the firms undertake a single public debt issue, and 66% undertake at most two
such issues. In the right tail of the public debt issue frequency distribution, 27% of the firms undertake
at least four issues and 6% do more than ten public debt issuances over the 23-year sample period. LFPIs
are firms with at most two issues in the sample, while HFPIs are firms with at least four such debt issues.
As Altinkilic and Hansen (2000), we estimate the following type of issue cost function:
s = β
0
+ β
1
1
x
1
+ β
2
x
1
x
2
+ γX + , (4)
where s is the bond underwriter spread,
10
x
1
is total offering proceeds, and x
2
is the issuer’s market value
of total equity on the day preceding the offering day. The vector X contains a set of control variables
that includes five ratings dummies for debt issues (AAA, A, BBB, BB, and B-CCC), the lagged ratio of
operating cash flow to total assets, the lagged book leverage ratio and a lagged market-to-book ratio. In
addition, we construct an issue wave variable which compares the yearly aggregate net debt issue (equity
issue) volume to the aggregate book value of assets.
11
In Eq. (4), the parameter β
0
is the variable issue cost component of the spread, β
1
is the fixed
issue cost component, while β
2
captures issue cost convexity. Table 6 shows the estimates of these three
parameters. Panel A provides estimates based on the total sample of 3,773 as well as on the subsamples
of 896 LFPIs and 2,622 HFPIs, respectively. The dollar value of the estimated fixed cost is obtained by
multiplying the estimated parameter value by $10,000.
As shown in the total-sample column, the fixed-cost estimate β
1
is a highly significant $249,000 for
10
The spread s is the ratio of the SDC variable “gross spread paid” to total issue proceeds. SDC defines ”Gross Spread”
as follows: ”Total manager’s fee. The fee is shared among lead managers, co-managers and syndicate group. Includes
management fee, underwriting fee and selling concession.”
11
The issue wave variable is standardized relative to its mean and standard deviation.
15
the average public bond issue. Splitting the sample into low- and high-frequency issuers shows that the
fixed-cost estimate β
1
is substantially lower both in magnitude and significance for HFPIs than for LFPIs.
For LFPIs, the fixed-cost estimate β
1
range from $448,000 to $487,000, whereas the estimate drops to
$106,000 for firms issuing at least four times. Moreover, for the most active public debt issuers, the
estimated fixed cost is a statistically insignificant $43,000.
It is possible that younger firms have greater fixed bond issue costs, reflecting greater valuation
uncertainty. However, the results in Panel B of Table 6 reject that the fixed-cost estimate for the LFPIs
in Panel A are driven by firm age. The estimate of β
1
is large and highly significant across all the age
categories shown, including firms that have been publicly traded for at least ten years. Across age groups,
the fixed cost estimates ranges from $356,000 to $450,000.
The analysis so far suggests that low-frequency debt issuers may face greater fixed financing costs
than high-frequency debt issuers. As a final check on issue costs, we next examine whether quantitative
indices of financial constraints classify our sample LFIs as fundamentally more financially constrained
than the HFIs. As these indices work with firm characteristics—and not issue frequency or direct cost
estimates—the index scores may be viewed as a check on whether LFIs are constrained beyond what our
dynamic issue hazard and direct issue cost analyses suggests.
3.4 Are LFIs fundamentally more financially constrained?
As discussed further below, the tradeoff theory concentrates on differences in debt-issue behavior driven
by differences in issue cost structures alone, holding other firm-specific factors constant. Importantly,
the evidence above suggests that sorting firms into groups of LFIs and HFIs also results in a spread
between the type of high fixed and low proportional issue costs that the theory addresses. However,
other firm-specific differences between LFIs and HFIs—not addressed by this theory—may also affect
issue frequencies. For example, firms may differ in the nature of the benefits from debt financing, in asset
composition and investment opportunities, and in agency costs (discussed further below). Moreover, as
addressed in this section, LFIs may be fundamentally more financially constrained than HFIs—beyond
the theoretical notion of relative fixed and variable issue costs.
It widely recognized that, absent debt collateral, the presence of moral hazard and informational
asymmetries may significantly reduce a firm’s ability to raise debt financing (Stiglitz and Weiss, 1981;
Bernanke and Gertler, 1989; Innes, 1990). As these are fundamentally unobservable characteristics, the
16
empirical capital structure literature suggests quantitative indices designed to indirectly identify the
resulting financial constraints. In the construction of indices, researchers have used the sensitivity of
investment to cash flow (Fazzari, Hubbard, and Petersen, 1988), univariate sorts based on individual
firm characteristics (Almeida, Campello, and Weisbach, 2004), a classification based on firm size and
age (Hadlock and Pierce, 2010), and various indices based on accounting and stock market information
(Kaplan and Zingales, 1997; Lamont, Polk, and Saa-Requejo, 2001; Whited and Wu, 2006).
Debating the construction of these indices goes beyond our purpose (Farre-Mensa and Ljungqvist,
2014). Rather, we compute these simply as another descriptive check on whether LFIs on average appear
to be more financially constrained than HFIs. Also, a lack of difference in index scores arguably increases
the likelihood that the documented differences in net-debt issue activities between LFIs and HFIs is rooted
in the type of direct issue cost differences discussed above and which lie at the core of our subsequent
tests of tradeoff theory.
In Table 7, we classify, starting in the year of public listing, LFIs and HFIs using the three indices
proposed by Kaplan and Zingales (1997) (KZ), Whited and Wu (2006) (WW) and Hadlock and Pierce
(2010) (SA), respectively, as follows:
KZ = −1.001909 × P rof + 3.139193 × L − 39.36780 × Div − 1.314759 × C + 0.2826389 × Q
W W = −0.091 × P rof − 0.062 × Divpos + 0.021 × T ldt − 0.044 × Size + 0.102 × ISG − 0.035 × g
S
SA = −0.737 × Size + 0.043 × Size
2
− 0.04 × Age. (5)
In KZ, Div is the ratio of dividends to assets. In WW , Divpos is a dummy variable equal to one in
case the firm paid dividends, T ldt is the ratio of long-term debt to assets, ISG is industry sales growth
(defined using 3-digit SIC codes) and g
S
is sales growth.
The W W -index and the SA-index in Table 7 give similar scores to LFIs and HFIs, while the KZ-index
scores HFIs as more financially constrained than LFIs. The anatomy of the index scores is as follows:
First, the KZ-index attaches a relatively large weight to leverage, and HFIs are indeed highly leveraged
(Table 4). Second, the W W -index also accounts for sales growth and firm size, and we have shown that
HFIs grow substantially and are relatively large. In addition, it loads substantially less on leverage. As it
turns out, the negative effect of firm size and sales growth on the financial constraint score dominates the
positive effect of leverage, to the point where the difference between HFIs and LFIs becomes small under
17
the W W index. Third, under the SA-index, the effect of firm size is non-linear and gives older (longer
listed) firms a lower score. However, the variable Age dominates in the construction of the SA index as
dollar differences in the book value of assets are mitigated by the logarithmic transformation of the size
variable. The net effect of Size and Age is such that the SA-index is similar for HFIs and LFIs. Based
on these index scores, there is no evidence that LFIs are fundamentally more financially constrained than
HFIs beyond, perhaps, the effect of firm characteristics on fixed issue costs.
In sum, we have shown in this section that HFIs exhibit an issue hazard shape consistent with low and
largely proportional issues debt issue costs. Moreover, LFIs, who issue net-debt rarely over their lifetime,
exhibit hazard shapes consistent with large and fixed debt issue costs. Also, direct fixed cost estimation
indicates that underwriter fees in bond issues have a statistically significant fixed-cost components only
in the subgroup of infrequent bond issuers. There is no evidence that LFIs and HFIs also on average
differ in terms of quantitative financial constraints scores.
4 Does tradeoff theory explain HFIs?
4.1 Predictions for HFIs relative to LFIs
Abstracting from corporate investments, classical dynamic tradeoff theory combines assumptions about
issue-cost structure with the corporate objective of managing leverage ratios toward an optimal target
(Fischer, Heinkel, and Zechner, 1989; Goldstein, Ju, and Leland, 2001; Strebulaev, 2007).
12
Whether
or not firms have leverage targets, the presence of large fixed issuance costs can cause long financing
spells (periods of issue inactivity). Thus, as discussed above, our issue cost evidence for HFIs and LFIs
is necessary but not sufficient to conclude in favor of the tradeoff theory. In this section, we examine
three additional, testable implications of this class of capital structure theories for the differences in issue
behavior of HFIs and LFIs.
The three additional implications address the relative values across LFIs and HFIs of issue size (stan-
dardized by the market value of assets), leverage ratio volatility, and speed-of-adjustment to deviations
from a target leverage ratio:
Proposition 1 (dynamic tradeoff without investment):
12
Brennan and Schwartz (1984) and Kane, Marcus, and McDonald (1984, 1985) provide dynamic capital structure models
without issue costs. Early examinations of the impact of adjustment costs on optimal financing behavior include the cash
management model of Miller and Orr (1966) and the portfolio selection theory of Constantinides (1979).
18
Suppose HFIs face lower fixed and total debt issuance costs than LFIs. Abstracting from
corporate investments, classical dynamic tradeoff theory then implies the following:
(1) Net-debt issues by HFIs are smaller than those of LFIs.
(2) Leverage ratios of HFIs are less volatile than those of LFIs.
(3) HFIs exhibit greater speed-of-adjustment to target leverage ratio deviations than do LFIs.
The basic intuition behind these three predictions follows readily from the model simulations in Table V
of Fischer, Heinkel, and Zechner (1989) and from the discussion of the dynamic hazard shapes in Section
3.2 above. Again, because the theory abstracts from the need to finance new investment, debt issues are
a response to exogenous increases in equity value causing the leverage ratio to fall below the target ratio
L
∗
. The (presumed) fixed issue costs of LFIs drive a greater wedge between L
∗
and the recapitalization
boundary L than do the (presumed) proportional issue costs of HFIs. This implies both a period of
inactivity (to reach the boundary) and, when the firm does issue, a debt issue size that is sufficiently
large to increase the leverage ratio all the way back up the associated (concave) firm value function to its
maximum at L
∗
. With proportional issue costs, the distance L
∗
− L is smaller, and so both the length of
the financing spell and the resulting issue size are also smaller.
13
Moreover, since greater fixed issuance
costs imply a wider spread L
∗
− L, they lead to more volatile leverage ratios for LFIs and slower speed
of adjustment back to L
∗
than for HFIs.
4.2 Relative issue size
Since the tradeoff theory models the behavior of a single firm, cross-sectional tests of Proposition 1 must
control for cross-sectional differences between HFIs and LFIs that are not modelled. Specifically, the issue
size prediction of Proposition 1 requires a firm size standardization (later, we also control for other firm
characteristics entering the empirical model for L
∗
). We use two different methods of standardization.
The first method scales the net-debt issue size with the lagged value of total assets (using market value for
total equity), as is common in the literature. With this scaling, annual net-debt issue sizes average 16%
for HFIs and 17% for LFIs, with medians of 10% and 9%, respectively, which fails to support prediction
(1) of Proposition 1.
13
Issue size is smaller for proportional costs because the first order condition for an optimal issue size (where marginal
issue cost equals marginal issue benefit) is more restrictive relative to the case with fixed issue costs, where the marginal
cost is zero.
19
The second method follows Eckbo and Kisser (2015) and scales issue size by the firm’s total funding
sources. The tradeoff theory does not distinguish between internal and external sources of equity funding,
and it abstracts from financing costs other than for debt issues. However, if these other financing costs
are empirically important, they may affect the overall funding mix and therefore debt issue size relative
to total funding sources. We therefore scale the net-debt issues by the total funding sources obtained
from the cash flow statement, as summarized in Appendix Tables 1 and 2.
Define a funding ratio for the (non-negative) funding source S
j
as R
j
≡ S
j
/
P
7
i
S
i
, where the denom-
inator sums over the seven mutually exclusive sources that make up the total funding reported in the
firm’s cash flow statement:
7
X
i
S
i
≡ CF
+
+ EI + NDI
+
+ ∆C
−
+ I
−
+ ∆W
−
+ O
+
, (6)
where CF
+
is the positive portion of operating cash flow, EI is proceeds from equity issues, NDI
+
is
positive net debt issues (debt issues exceeding debt retirements), ∆C
−
is draw-down of cash balances, I
−
is sale of investments, sale of property, plant and equipment (PPE) and cash flows from other investment
activities, ∆W
−
is reduction in net working capital, and O
+
is a small residual that maintains the cash
flow identity.
14
Table 8 and Figure 2 show the annual values of each of the funding ratios and their components.
To simplify the exposition, Figure 2 aggregates the contribution of liquid and illiquid asset sales into
a single Asset Sales ratio: R
AS
≡ (∆C
−
+ ∆W
−
+ O
+
+ I
−
)/
P
7
i
S
i
. The other three ratios shown
in the figure are, respectively, the Net-Debt Issue ratio R
NDI
+
≡ NDI
+
/
P
7
i
S
i
, the Equity Issue ratio
R
EI
≡ EI/
P
7
i
S
i
, and the positive Operating Cash Flow ratio R
CF
+
≡ CF
+
/
P
7
i
S
i
. By construction,
these four ratios sum vertically to one in Figure 2.
Interestingly, HFIs exhibit substantially greater net-debt funding ratios than LFIs: the annual value
of R
NDI
+
averages 26% for HFIs and only 2% for LFIs, respectively. Thus, net-debt issues by HFIs are
14
In 1988, Statement of Financial Accounting Standards (SFAS) instituted a new and uniform reporting system for
working capital, including its component assets and liabilities. We work with net working capital over the entire sample
period. Separate analysis on the post-1988 period shows that splitting net working capital into assets and liabilities does
not affect our main conclusions below. Also, debt and equity issues extracted from cash flow statements may differ from
balance sheet induced changes of debt and equity when a transaction impacting the balance sheet does not also involve a
cash flow. Examples include stock swaps, balance sheet consolidation following acquisitions, convertible debt conversions,
and stock issued under employee stock option plans. Computing the difference between net debt issues and balance-sheet
implied positive changes in debt, shows that this effect is small in our data: the mean (median) difference (scaled by assets)
is 0% (0%). For equity issues, the distribution is slightly more skewed with a mean (median) of 2% (-1%).
20
substantially larger than those of LFIs in terms of their importance for the annual funding mix. In sum,
HFIs raise at least as much net-debt as LFIs per issue, issue net-debt more often, and rely to a much
greater extent than LFIs on net-debt as an overall funding source, all of which reject prediction (1) of
Proposition 1.
4.3 Relative leverage instability and volatility
Prediction (2) of Proposition 1 holds that the leverage ratio volatility of LFIs should exceed that of HFIs.
To test this, Table 9 provides evidence on annual leverage ratio “instability” and volatility for HFIs and
LFIs following public listing. In the table, L
0
is the leverage ratio in the listing year (year 0). The first
four columns report the fraction of the sample firms in year t > 0 with an absolute change in the leverage
ratio of at least 20 percentage points (pp.): |L
t
− L
0
| > 0.2.
15
The last four columns show average annual
leverage volatility, computed as the average of the standard deviation of the individual annual leverage
ratios, beginning with five annual observations in year five.
Panel A of Table 9 shows that HFIs exhibit substantially higher leverage ratio volatilities than LFIs.
This is true in virtually every year since public listing, and it holds whether we use the fraction of leverage
instability or average leverage volatility. For example, after five years of listing, 50% of the individual
HFIs have experienced a leverage ratio change exceeding +/-20 pp., while only 16% of the leverage ratios
among the LFIs were similarly unstable. Moreover, after five years, the average leverage ratio volatility
is 17% for HFIs—double that of the 8% for LFIs. The failure to support prediction (2) of Proposition
1 also holds if we use net rather than gross leverage ratios. Subtracting cash balances, the net-leverage
volatility is 19% for both HFIs and LFIs. Moreover, net-leverage ratios of HFIs are relatively unstable as
43% of all HFIs experience significant changes in net-leverage ratios relative to the year of public listing,
whereas only 26% of the LFIs do so.
Panel B of Table 9 repeats the analysis using the estimated target leverage ratio: L
∗
i,t
(βX
i,t−1
), where
the determinants X
i,t−1
are the lagged values of size, profitability, Q, cash ratio, tangibility, depreciation,
R&D expenses, capital expenditures and the median industry leverage ratio (all winsorized at the 1(99)
percent level). Here, instability is the exception as only 6% of all HFIs experience a target leverage ratio
change in excess of 20 percentage points. More to the point, the target leverage ratio volatility is similar
across LFIs and HFIs, indicating that the greater leverage ratio instability and volatility shown in Panel
15
DeAngelo and Roll (2015) refer to a leverage ratio change exceeding this threshold as “unstable”.
21
A of Table 9 for HFIs is not reflecting changes in the target L
∗
.
Since our HFIs are highly leveraged both absolutely and relative to LFIs (Table 4 above), the evidence
in Table 9 corroborates the conclusion of DeAngelo and Roll (2015) that leverage ratio volatility of
relatively high-leveraged companies is substantial relative to low-leveraged firms. In addition, given the
relatively stable leverage ratios of LFIs evidenced in Table 9, our results suggest that much of the leverage
ratio instability reported by DeAngelo and Roll (2015) is also driven by high-frequency net-debt issuers.
4.4 Relative speed-of-adjustment to deviations from L
∗
Prediction (3) of Proposition 1 holds that HFIs, through their high-frequency debt issue activity, will
exhibit greater speed-of-adjustment (SOA) than LFIs to deviations from the putative leverage target L
∗
.
To examine this prediction, we follow the capital structure literature and estimate the SOA parameter
φ using the following dynamic regression with the change in the market leverage ratio L
i,t
− L
i,t−1
as
dependent variable:
L
i,t
− L
i,t−1
= α + η
i
+ φ
L
∗
i,t
(βX
i,t−1
) − L
i,t−1
+
i,t
. (7)
here, the determinants X
i,t−1
of the target leverage ratio L
∗
i,t
are the same as in the previous section. The
estimation of Eq. (7) also accounts for firm fixed effects (η
i
), that the regressor L
∗
i,t
is itself estimated,
and that the lagged dependent variable L
i,t−1
also features as a regressor.
16
As recommended by the
literature, we use system GMM estimation.
17
The first column of Table 10 shows the estimated value of φ for the total sample of 10,783 firms (80,900
firm-years) that are listed for at least two years. This estimate is statistically significant with a value
of 0.259 (p-value of 0.01), which is close to GMM estimates of 0.25 reported by Lemmon, Roberts, and
Zender (2008). It suggests that it takes the average firm about three years to recover half of the target
leverage deviation (the half-life implied by φ is ln(0.5)/ln(1 + φ)). More surprising, as shown in columns
16
To see this, note that Eq. (7) is equivalent to: L
i,t
= α + η
i
+ φL
∗
i,t
(βX
i,t−1
) + (1 − φ)L
i,t−1
+
i,t
.
17
See, e.g., Blundell and Bond (1998), Lemmon, Roberts, and Zender (2008), and Flannery and Hankins (2013). Alter-
natives to the dynamic panel estimation used here are long difference estimation (Hahn, Hausman, and Kuersteiner, 2007;
Huang and Ritter, 2009) and bias correction methods (Kiviet, 1995; Bruno, 2005). The application of these methods is often
complicated by the fact that corporate finance panels are unbalanced and suffer from non-contiguous data due to missing
observations. Flannery and Hankins (2013) simulate data with similar properties and compare the performance of these
estimates. Their simulations suggest that bias correction methods and system GMM estimates emerge as the most accurate
methodologies. We implement system GMM in Stata using the command xtabond2 and treat lagged leverage and the vector
X
i,t−1
as predetermined and use a maximum of 3 (lagged leverage) and 5 (X
it,−1
) lags when constructing instruments.
Changing the specification and modelling X
i,t
as endogenous does not change our results.
22
two and three, the SOA coefficient estimate is large and highly significant for both HFIs (φ = 0.316,
p-value of 0.01) and LFIs (φ = 0.272, p-value of 0.02). As shown in column four, the difference between
the SOA coefficient estimates for HFIs and LFIs is 0.044, which is different from zero only at the 10%
level of significance. Thus, prediction (3) of Proposition 1 is rejected at conventional levels of statistical
significance.
Note also that the high SOA coefficient estimate for the LFIs is interesting in of itself as the a priori
expected size of this coefficient is low. Recall from Panel A of Table 2 that LFIs on average undertake
only one-half net-debt issue (with the 2.5% threshold) during the first ten years of listing, and only 1.6
issues over the first twenty years. With this in mind, the SOA coefficient estimate for LFIs in Table
10 simply cannot be driven by net-debt issues. Rather, for LFIs, the dynamic behavior of the market
leverage ratio must be driven by changes in the denominator of the leverage ratio. However, this raises
concerns about the very informativeness of the SOA estimation for LFIs: it appears to be confounded by
the dynamics of the asset side of the balance sheet (Welch, 2004, 2010).
The empirical analysis in this section fails to support the three predictions of dynamic tradeoff theory
summarized in Proposition 1. While there may be multiple reasons for this rejection, we are particularly
interested in the role of the firm’s ongoing need to finance investment, measured here by capital expen-
ditures (Capex). The need to finance investment shocks may mask true tradeoff behavior in the data as
it gives rise to transitory or temporary debt issues. We address this issue next.
5 Tradeoff with investment financing and transitory debt
Dynamic financing and investment models allow firms to issue debt to finance investment projects in
addition to responding to deviations from a target leverage ratio (Hennessy and Whited, 2005, 2007;
DeAngelo, DeAngelo, and Whited, 2011). We focus in this section on the implication that firms may
optimally issue debt to finance new investments even if current leverage exceeds a long-run target. In
the vernacular of DeAngelo, DeAngelo, and Whited (2011), such debt issues are “transitory” in the sense
that the excess debt will optimally be retired as the investment shocks recede. Proposition 2 summarizes
this prediction:
Proposition 2 (dynamic tradeoff with investment):
Suppose firms jointly determine financing and investment in a dynamic tradeoff setting. Dy-
23
namic tradeoff theory with investment implies that an over-leveraged firm may issue debt to
finance investments. Such debt issues are transitory as they are subsequently repurchased to
restore the leverage target.
Below, we examines Proposition 2 empirically. We focus on the HFIs as, of the total sample of transitory
net debt issues (defined below), 74% are undertaken by HFIs.
18
We begin by identifying a close association
between net-debt issue activity and investments for the HFIs, followed by a search for transitory debt
issuances and retirements in the data.
5.1 Net-debt issue size and investment
Recall from Table 5 that Capex is perhaps the strongest predictor of the net-debt issue frequency and
hazard. In this section we further show that net-debt issue size and Capex are also positively correlated
for HFIs. This positive correlation is apparent both from Figure 3 and from the regression results in Table
11. The horizontal axis in Figure 3 sorts all net-debt issues (relative to book assets) into ten deciles,
while the vertical axis plots the associated average Capex. For HFIs, Capex averages 10% with a median
of 6%, and its 95
th
percentile is 32% of book assets. As shown, average Capex increases monotonically
with the net-debt issue decile from low to high.
Table 11 shows a similar effect in a multivariate regression framework using the following modified
“financing deficit” regression (Shyam-Sunder and Myers, 1999; Frank and Goyal, 2003; Lemmon and
Zender, 2010):
NDI
i,t
A
i,t
= α + β
1
Capex
i,t
+ β
2
NetDeficit
i,t
+ β
3
Capex
2
i,t
+ β
4
NetDeficit
2
i,t
+
i,t
. (8)
NDI/A is net-debt issues (retirements when negative) scaled by total assets, and NetDeficit is the
financing deficit net of Capex.
19
The inclusion of squared terms addresses the possibility that, in a
dynamic financing pecking order setting, firms financing larger projects may optimally choose to issue
both debt and equity (and not just debt) in order to preserve future debt capacity.
Column (1) of Table 11 shows a positive and significant (univariate) correlation between invest-
18
LFIs only account for 5% of all transitory net debt issues, implying that medium frequency issuers undertake the
remaining 21%.
19
Using the Compustat mnemonics in Appendix Table 2: NetDefecit ≡ dv + aqc + ivch − siv − ivstch − sppe − ivaco −
oancf + chech. All variables are scaled by total assets.
24
ment (Capex) and issue size. This slope coefficient on Capex increases, from 0.34 to 0.52, after adding
NetDeficit in column (2), and it increases further to 0.57 in the full model in column (4). Moreover,
Capex
2
also receives a positive and significant coefficient, suggesting that larger capital expenditures are
financed with a greater fraction of net debt issues. In sum, net-debt issues are strongly related to Capex,
and the more so the greater the capital expenditure program in relation to total assets.
5.2 Cumulative debt and equity issues by over-leveraged firms
Within the issue-cost structure assumed by DeAngelo, DeAngelo, and Whited (2011), transitory debt
issues are followed by net-debt repurchases as the investment shocks recede. Below, we provide a simple
visual impression of the existence of net debt issues by over-leveraged firms, followed in the next section
by estimates of the probability that such debt issues trigger the predicted debt repurchases and leverage
ratio reductions.
Let Dev
t
= L
t−1
− L
∗
t
(βX
t−1
) denote the leverage ratio deviation from the target going into period
t, where L
∗
t
(βX
t−1
) is estimated as in Table 10 above. Moreover, let Ecapex denote the firm’s capital
expenditure in excess of the industry median. The dummy variable TN DI
i,t−1
indicates whether the firm
made a transitory net-debt issue in year t − 1. Specifically, T NDI
t−1
equals one if, in period t − 1, the
following three conditions are met: the firm (1) was over-leveraged going into the period (Dev
t−1
> 0),
(2) issued net-debt in excess of 2.5% of total assets (NDI
t−1
> 0), and (3) invested in excess of the
industry median (Ecapex
t−1
> 0). For the HFIs, T NDI equals one in 15% of the sample firm-years
where N DI > 0. Moreover, when T N DI = 1, firms are on average over-leveraged by 8 percentage
points.
Figure 4 shows in event time what happens on average after a transitory versus a non-transitory
net-debt issue in year zero (N DI
0
> 0). The figure plots, over the next three years, the annual average
leverage ratio L (“Debt”), cumulative net-debt issue (“Cumulative NDI”), cumulative net equity issue
(“Cumulative NEI”, net of stock repurchases and dividends), and capital expenditure (“Capex”). Panel
A conditions on T N DI = 0 (a non-transitory net debt issue), while Panel B investigates the case when
T NDI = 1 (a transitory net-debt issue).
The leverage ratio L is increasing somewhat in Panel A (where NDI
0
occurs when T NDI = 0), while
it is decreasing in Panel B. This supports the notion that the net-debt issues in Panel B are transitory.
On the other hand, Cumulative NDI and Cumulative NEI are increasing in both panels, while Capex is
25
largely flat following the event year 0. It appears that the leverage decrease in Panel B is driven by net
equity issues rather than by net debt retirements. Thus, notwithstanding that HFIs are relatively active
in terms of undertaking net-debt retirements (Table 2 above), this retirement activity does not appear
in Figure 4 to coincide with the predicted repurchases following transitory debt issues.
While Figure 4 plots the average cumulative net-debt and net-equity issuances following a transitory
net-debt issue, Proposition 2 may also be addressed by estimating the marginal effect of T N DI on the
propensity for a net-debt repurchase in the next period. We turn to this estimation next.
5.3 Probability of debt retirements following transitory debt issues
Table 12 presents coefficient estimates of the determinants of the probability that a transitory net debt
issue in period t−1, as indicated by the dummy variable TNDI
i,t−1
, is followed in year t by, alternatively,
a decrease in leverage (Panel A), a net-debt retirement exceeding 2.5% of total assets (Panel B), or a net
equity issue exceeding 2.5% (Panel C). The model is:
Y
i,t
= α + βT NDI
i,t−1
+ γEcapex
i,t
+ δDev
i,t
+
i,t
, (9)
where Y
i,t
= 1 is the binary dependent variable for firm i in year t. The estimation also includes industry
dummies for eight of the 12 Fama-French (FF12) industries (excluding financial firms and regulated
utilities).
Consistent with the evidence on cumulative issuances in Figure 4 above, the results in Panel A of Table
12 suggest that a transitory debt issue significantly increases the probability that a firm decreases leverage
in the subsequent period. Also, current Ecapex is significantly negatively correlated with reductions in
leverage. This is also consistent with our earlier findings that HFIs tend to finance investments with
debt, and so do not lever down in periods when Ecapex is relatively high.
Panel B estimates the probability of a net-debt retirement directly. The results here are mixed.
In the first of the two regression specifications, the coefficient estimates on T NDI
t−1
and Ecapex are
highly significant and of the sign expected under Proposition 2: having issued transitory net debt last year
increases the probability of a net-debt retirement this year, while greater Ecapex reduces this probability.
However, the coefficient on T NDI
t−1
becomes small and statistically insignificant when we add Dev
t
to
the regression. That is, in the full regression specification, the probability of a net-debt retirement
26
is increasing with the deviation from current-period optimal leverage ratio, while it is independent of
whether or not the firm undertook a “transitory” net-debt issue T N DI in the last period.
Finally, consistent with Figure 4, Panel C of Table 12 shows that net-debt issues T NDI
t−1
and capital
expenditures Ecapex
t
combine to trigger subsequent equity issues in period t. The estimated coefficients
on both T N DI
t−1
and Ecapex
t
are positive and highly significant. Overall, these results suggest that
some of the leverage increase resulting from a debt issue of type T DNI is reduced over the following
year through a combination of retained earnings and (relatively costly) equity issues to finance continued
investments. While the reduction in leverage following T NDI > 0 supports Proposition 2, the lack of
follow-on net-debt repurchases does not. Rather, the leverage ratio reduction reflects a combination of
retained earnings and new (costly) equity issues rather than by debt repurchases—as if ongoing funding
needs dominate debt-retirement considerations.
6 Issue frequency and CEO compensation
The analysis so far has considered tradeoff behavior in terms of optimal managerial adjustments to
investment shocks and issue costs. Leverage policies (and therefore, indirectly, net-debt issue frequencies)
may be driven by considerations other than shareholder value maximization, such as agency cost created
by the personal bankruptcy costs of the Chief Executive Officer (CEO) (Ross, 1977; Eckbo, Thorburn,
and Wang, 2014). Also, cash-strapped young firms with few pledgable assets have low leverage and
routinely compensate their executives in restricted stock. These additional considerations suggest that
CEO stock ownership and stock-based compensation may differ in complex ways across HFIs and LFIs,
which we examine briefly below.
Table 13 shows interesting differences between HFIs and LFIs in terms of both CEO stock ownership
and stock-based compensation. CEO ownership and compensation information is from ExecuComp over
the period 1992-2012. Since ExecuComp samples relatively large companies, and we require firms to go
public during our sample period, ExecuComp data is available for only about ten percent of our total
sample: 1,659 firms and 13,907 firm-year observations. Moreover, since HFIs are significantly larger
than LFIs on average (Table 4), the sample in Table 13 is skewed towards HFIs relative to LFIs (it also
includes, of course, firms that are classified as neither).
While HFIs and LFIs on average hold similar percentage of shares outstanding (approximately 5%),
27
the first row of Table 13 indicates that, after controlling for firm-fixed effects and firm characteristics,
CEOs of HFIs hold a significantly greater percentage equity than all other firms (LFIs and those in
between LFIs and HFIs). Moreover, CEOs of LFIs hold significantly less stock than all other firms in the
regression. Furthermore, the second regression shows that HFIs are associated with significantly greater
stock-based compensation than all other firms.
20
We emphasize that the evidence of a significant association between CEO stock ownership/stock-
based compensation and HFI does not mean that a similar association exists with firm leverage ratios.
Indeed, Strebulaev and Yang (2013) sample all firms on ExecuComp over the period 1992 and 2009 (6,476
firm-years) and find that CEO stock ownership is associated with greater probability of having near-zero
(less than 5%) leverage. We have verified that a similar effect is present in our data, suggesting that
the impact of stock ownership and compensation on both net-debt issue frequency and leverage ratios is
likely non-linear in nature, an interesting topic for future research.
7 Conclusion
Over the past forty years, a group of highly leveraged, high-frequency debt issuers (our HFIs) raised the
bulk of all public and private debts issued by public non-financial U.S. firms. While much attention has
been given to zero or near-zero-leverage firms, little is known about the HFIs. Since these firms evidently
view debt-financing as uniquely beneficial, and face sufficiently low debt issue costs to issue frequently,
they are ideally suited to test tradeoff theory. In this paper, we contrast the firm characteristics of HFIs
versus low-frequency debt issuers (our LFIs), and we exploit the differences in debt issue behavior across
HFIs and LFIs in new tests of capital structure theory.
HFIs stand out already at the time of public listing as relatively large, investment-intensive firms with
high leverage and low Tobin’s Q. On the other hand, LFIs are relatively small R&D-intensive firms with
low leverage and high Q. There is only minimal migration between firms in the HFI and LFI categories.
Market leverage ratios quickly rise to an average of 35% for HFIs versus an average of 10% for LFIs.
Moreover, while as much as 32% of the LFIs have zero leverage in a typical firm-year, only 1% of HFIs
ever reduce leverage to zero during their lifetime as public companies. Overall, HFIs tend to finance a
relatively capital-intensive investment (capex) program externally through debt issues, while LFIs tend
20
Separate estimation (not shown in Table 13) also shows that CEO stock ownership in year t−1 is associated with greater
probability of being classified as HFI in year t.
28
to finance a relatively intensive R&D program internally.
Large fixed issue costs slow issue frequency whether or not a firm manages capital structure towards
a target. It is therefore natural to assume that HFIs face smaller and less fixed issue costs than LFIs.
In preparation for testing dynamic tradeoff theory, we corroborate this issue-cost assumption with new
evidence. The evidence is both indirect (using estimated dynamic issue hazard shapes) and direct (using
underwriter fees of public bond issues). This estimation suggests that LFIs in general, and low-frequency
public bond issuers in particular, face significant fixed issue costs. There is no other obvious difference in
external finance costs as LFIs and HFIs have similar scores on quantitative indices of financial constraints.
Dynamic tradeoff theory, which abstracts from investment financing, predicts that, relative to LFIs,
debt issues by HFIs are smaller, leverage ratio volatility is lower, and the speed-of-adjustment to deviations
from leverage targets is higher. Our evidence fails to support these three predictions: issue size is no
smaller for HFIs than for LFIs, leverage ratio volatilities are substantially higher, and LFIs and HFIs
receive statistically indistinguishable speed-of-adjustment coefficient estimates. Given the extremely low
issue frequency of LFIs (on average one tenth of that of HFIs), the latter finding raises concern about
the use of leverage ratio changes to identify the true speed with which firms manage leverage towards a
target.
True tradeoff behavior by HFIs may be masked by the ongoing funding of their intensive investment
program. We show that not only are capital expenditures a main determinant of the probability of
becoming an HFI but issue size of HFIs is also highly correlated with capital expenditures. Recent
investment and financing models predict that temporarily over-leveraged firms may issue debt in response
to persistent investment shocks. As the investment shocks fade, these “transitory” debt issues are followed
by debt repurchases to restore leverage targets. Using the empirical leverage targets from the speed-of-
adjustment analysis, there is evidence of the existence of net-debt issues by over-leveraged firms. This
extra debt is transitory in the sense that it is followed by a reduction in leverage—however, the reduction
is achieved using mostly equity issues (not debt repurchases), which suggests that ongoing and immediate
funding needs tend to dominate debt-retirement considerations also for temporarily over-leveraged firms.
Finally, we show briefly that, as perhaps expected, CEOs of HFIs have high stock ownership and
receive a greater portion of total pay in terms of stocks and options relative to all other sample firms
(with available compensation data). In other words, CEOs of HFIs appear to have relatively strong
incentives to maximize shareholder value. However, we leave it to future research to identify a possible
29
causal effects of CEO stock ownership and incentive-based pay on firms’ net-debt issue frequency and
leverage policy.
30
References
Almeida, H., M. Campello, and M. S. Weisbach, 2004, The cash flow sensitivity of cash, Journal of
Finance 59, 1777–17804.
Altinkilic, Oya, and Robert S. Hansen, 2000, Are there economies of scale in underwriting fees? Evidence
of rising external financing costs, Review of Financial Studies 13, 191–218.
Bates, Thomas W., Kathleen M. Kahle, and Rene M. Stulz, 2009, Why do U.S. firms hold so much more
cash than they used to?, Journal of Finance 64, 1985–2021.
Bernanke, Ben, and Mark Gertler, 1989, Agency costs, net worth, and business fluctuations, American
Economic Review 79, 14–31.
Blundell, Richard, and Steve Bond, 1998, Initial conditions and moment restrictions in dynamic panel
data models, Journal of Econometrics 87, 115–143.
Brennan, Michael J., and Eduardo S. Schwartz, 1984, Optimal financial policy and firm valuation, Journal
of Finance 39, 593–607.
Bruno, Giovanni, 2005, Approximating the bias of the lsdv estimator for dynamic unbalanced panel data
models, Economic Letters 87, 361–366.
Constantinides, George M., 1979, Multiperiod consumption and investment behavior with convex trans-
action costs, Management Science 25, 1127–1137.
DeAngelo, Harry, Linda DeAngelo, and Rene M. Stulz, 2010, Seasoned equity offerings, market timing,
and the corporate lifecycle, Journal of Financial Economics 95, 275–295.
DeAngelo, Harry, Linda DeAngelo, and Toni M. Whited, 2011, Capital structure dynamics and transitory
debt, Journal of Financial Economics 99, 235–261.
DeAngelo, Harry, and Richard Roll, 2015, How stable are corporate capital structures?, Journal of
Finance 70, 373–418.
Eckbo, B. Espen, and Michael Kisser, 2015, Financing decisions: Who sells assets?, Working Paper,
Dartmouth and Norwegian School of Economics.
Eckbo, B. Espen, and Ronald W. Masulis, 1995, Seasoned equity offerings: A survey, in Robert A. Jarrow,
Vojislav Maksimovic, and William T. Ziemba, ed.: Finance . chap. 31, pp. 1017–1072 (Elsevier/North-
Holland, Handbooks in Operations Research and Management Science Series).
, and Øyvind Norli, 2007, Security offerings, in B. E. Eckbo, ed.: Handbook of Corporate Finance:
Empirical Corporate Finance, vol. 1 . chap. 6, pp. 233–374 (Elsevier/North-Holland, Handbooks in
Finance Series).
Eckbo, B. Espen, Karin S. Thorburn, and Wei Wang, 2014, How costly is corporate bankruptcy for top
executives?, Working Paper, Dartmouth College, the Norwegian School of Economics, and Queen’s
University.
Fama, Eugene F., and Kenneth R. French, 2002, Testing trade-off and pecking order predictions about
dividends and debt, Review of Financial Studies 15, 1–33.
, 2005, Financing decisions: Who issues stock?, Journal of Financial Economics 76, 549–582.
Farre-Mensa, Joan, and Alexander Ljungqvist, 2014, Do measures of financial constraints measure finan-
cial constraints?, Working Paper, Harvard and New York University.
31
Faulkender, Michael, Mark J. Flannery, Kristine Watson Hankins, and Jason M. Smith, 2012, Cash flows
and leverage adjustments, Journal of Financial Economics 103, 632–646.
Fazzari, Steven M., R. Glenn Hubbard, and Bruce C. Petersen, 1988, Financing Constraints and Corpo-
rate Investment . pp. 141–195 (Brookings Papers on Economic Activity).
Fischer, Edwin O., Robert Heinkel, and Josef Zechner, 1989, Dynamic capital structure choice: Theory
and tests, Journal of Finance 44, 19–40.
Flannery, Mark J., and Kristine W. Hankins, 2013, Estimating dynamic panels models in corporate
finance, Journal of Corporate Finance 19, 1–19.
Flannery, Mark J., and Kasturi P. Rangan, 2006, Partial adjustment toward target capital structures,
Journal of Financial Economics 79, 469–506.
Frank, Murray Z., and Vidhan K. Goyal, 2003, Testing the pecking order theory of capital structure,
Journal of Financial Economics 67, 217–248.
Goldstein, Robert, Nengjiu Ju, and Hayne Leland, 2001, An EBIT-based model of dynamic capital
structure, Journal of Business 74, 483–512.
Graham, John R., and Markt T. Leary, 2011, A review of empirical capital structure research and
directions for the future, Annual Review of Financial Economics 3, 309–345.
Hadlock, Charles J., and Joshua R. Pierce, 2010, New evidence on measuring financial constraints: moving
beyond the KZ index, Review of Financial Studies 23, 1909–1940.
Hahn, Jinyong, Jerry Hausman, and Guido Kuersteiner, 2007, Long difference instrumental variables
estimation for dynamic panel models with fixed effects, Journal of Econometrics 140, 574–617.
Hennessy, Christopher A., and Toni M. Whited, 2005, Debt dynamics, Journal of Finance 60, 1129–1165.
, 2007, How costly is external financing? Evidence from a structural estimation, Journal of
Finance 62, 1705–1745.
Hovakimian, Armen, and Guangzhong Li, 2012, Is the partial adjustment model a useful tool for capital
structure research, Review of Finance 16, 733–754.
Huang, Rongbing, and Jay R. Ritter, 2009, Testing theories of capital structure and estimating the speed
of adjustment, Journal of Financial and Quantitative Analysis 44, 237–271.
Innes, 1990, Limited liability and incentive contracting with ex-ante action choices, Journal of Economic
Theory 52, 45–67.
Kane, Alex, Alan J. Marcus, and Robert L. McDonald, 1984, How big is the tax advantage to debt?,
Journal of Finance 39, 841–853.
, 1985, Debt policy and the rate of return to leverage, Journal of Financial and Quantitative
Analysis 20, 479–499.
Kaplan, Steven N., and Luigi Zingales, 1997, Do investment cash-flow sensitivities provide useful measures
of financing constraints?, Quarterly Journal of Economics 112, 169–215.
Kiviet, Jan F., 1995, On bias, inconsistency, and efficiency of various estimators in dynamic panel data
models, Journal of Econometrics 68, 53–78.
Lamont, Owen, Christopher Polk, and Jesus Saa-Requejo, 2001, Financial constraints and stock returns,
Review of Financial Studies 14, 529–554.
32
Leary, Mark T., and Michael Roberts, 2005, Do firms rebalance their capital structures?, Journal of
Finance 60, 2575–2619.
, 2010, The pecking order, debt capacity, and information asymmetry, Journal of Financial
Economics 95, 332–355.
Lemmon, Michael, Michael R. Roberts, and Jamie F. Zender, 2008, Back to the beginning: Persistence
and the cross-section of corporate capital structure, Journal of Finance 63, 1575–1608.
Lemmon, Michael L., and Jaimie F. Zender, 2010, Debt capacity and tests of capital structure theories,
Journal of Financial and Quantitative Analysis 45, 1161–1187.
Merton, Robert C., 1974, On the pricing of corporate debt: the risk structure of interest rates, Journal
of Finance 29, 449–470.
Miller, Merton H., and Daniel Orr, 1966, A model of the demand for money by firms, Quarterly Journal
of Economics 80, 413–435.
Ross, Stephen, 1977, The determination of financial structure: The incentive signalling approach, Bell
Journal of Economics 8, 23–40.
Shyam-Sunder, Laskshmi, and Stewart C. Myers, 1999, Testing static trade-off against pecking order
models of capital structure, Journal of Financial Economics 51, 219–244.
Stiglitz, Joseph E., and Andrew Weiss, 1981, Credit rationing in markets with imperfect information,
American Economic Review 71, 393–410.
Strebulaev, Ilya A., 2007, Do tests of capital structure theory mean what they say?, Journal of Finance
62, 1747–1787.
, and Baozhong Yang, 2013, The mystery of zero-leverage firms, Journal of Financial Economics
109, 1–23.
Welch, Ivo, 2004, Capital structure and stock returns, Journal of Political Economy 112, 106–131.
, 2010, A critique of recent quantitative and deep-structure modeling in capital structure research
and beyond, Working Paper, UCLA.
Whited, Toni M., and Guojun Wu, 2006, Financial constraints risk, Review of Financial Studies 19,
531–559.
33
Figure 1: Estimated and simulated hazards curves
The estimated hazard curves in Panel A and B use our sample of 12,131 firms and 93,031 firm-years and the following
exponential shared frailty hazard model:
h
i
(t) = h
0
exp(β
0
+ βx
i
+ γf(t))α
i
and f (t) = t + t
2
+ t
3
,
where t is time (years) since last issue, h
0
is the baseline hazard (when all covariates are equal to zero), and α
i
captures
unobserved heterogeneity analogous to a regression error term. The covariates are investment (Capex), R&D expenditures
(R&D), market leverage ratio (L), cash ratio (C), logarithm of assets (Size), operating cash flow (P rof), tangibility (T an),
Tobin’s Q (Q) and depreciation expenditures (Depr). All covariates are winsorized at the 1(99) percent level or must lie
between zero and one (cash ratio and leverage). They enter after subtracting their median values. The high and low-
frequency classification (HFI and LFI) is performed annually and based on a 2.5% issue size threshold. Variable definitions
are in Appendix Tables 1 and 2.
The simulated hazard curves in Panel C and D are from Figure 5 in Leary and Roberts (2005), which are generated using
the cubic polynomial f (t)) but not the covariates x
i
. The simulations assume stock returns follow a geometric Brownian
motion, with a mean (12%) and standard deviation (46%) of equity returns matching a sample of industrial Compustat
firms with at least four years of contiguous data, 1984-2001. Firms are assumed to recapitalize when they hit an upper or a
lower leverage boundary, matched to the sample median maximum (0.60) and minimum (0.15), respectively. The leverage
process starts at the midpoint of the spread between the upper and lower boundaries and are updated each period using the
simulated equity returns. Firms in Panel C face fixed issues costs and rebalance all the way back to the midpoint. Firms in
Panel D face proportional issue costs and optimally rebalances to the nearest boundary only.
A: Eckbo-Kisser estimated hazard for LFIs
.006 .008 .01 .012 .014 .016
Baseline Hazard
0 5 10 15 20
Time Since Last Issue
B: Eckbo-Kisser estimated hazard for HFIs
.1 .2 .3 .4 .5 .6
Baseline Hazard
1 2 3 4 5 6
Time Since Last Issue
C: Leary-Roberts simulated hazard
with fixed issue costs
D: Leary-Roberts simulated hazard
with proportional issue costs
34
Figure 2: Lifecycle financing ratios for high- and low-frequency net-debt issuers
The figure plots four financing ratios R
j
≡ S
+
j
/
P
7
i
S
+
i
, where
P
7
i
S
+
i
is the firm’s total cash contribution from each of its
seven (non-negative) sources of funds:
P
7
i
S
+
i
= EI + N DI
+
+ CF
+
+ ∆C
−
+ I
−
+ ∆W
−
+ O
+
. EI is proceeds from
equity issues, N DI
+
is positive net debt issues (net of debt retirements), CF
+
is positive operating cash flow, ∆C
−
is
cash drawdowns, I
−
is sale of illiquid assets (sale of investments, PPE and other investments), ∆W
−
is reduction in net
working capital, and O
+
is “other” sources of funds (a small residual closing the cash flow identity). By construction,
the four ratios sum vertically to one: R
EI
= EI/
P
7
i
S
+
i
, R
NDI
+
= N DI
+
/
P
7
i
S
+
i
, R
CF
+
= CF
+
/
P
7
i
S
+
i
, and
R
AS
= (∆C
−
+ I
−
+ ∆W
−
+ O
+
)/
P
7
i
S
+
i
. Year 0 is the year of public listing. Variable definitions are in Appendix Tables
1 and 2. Sample of 12,131 U.S. public industrial firms and 93,101 firm-years, 1971-2012.
Panel A: Average funding ratios for HFI
0 .1 .2 .3 .4 .5 .6
Financing Ratio
0 5 10 15 20
Event Year
R(NDI
+
) R(EI)
R(CF
+
) R(AS)
Panel B: Average funding ratios for LFI
0 .1 .2 .3 .4 .5 .6
Financing Ratio
0 5 10 15 20
Event Year
R(NDI
+
) R(EI)
R(CF
+
) R(AS)
35
Figure 3: Capital expenditures and size of net debt issues for HFIs
The figure plots average capital expenditures (Capex) for ten different portfolios of HFIs. The decile sort is based on the
size of net debt issues (N DI), expressed relative to the book value of assets. Decile 1 corresponds to the smallest net debt
issues, decile 10 denotes the largest net debt issues. Variable definitions are in Appendix Tables 1 and 2. Total sample of
6,684 U.S. public firms and 31,449 firm-years, 1971-2012.
0 .05 .1 .15
Average Capex
1 2 3 4 5 6 7 8 9 10
Net Debt Issue Deciles (1 = low, ..., 10 = high)
36
Figure 4: Financing and investment following non-transitory and transitory debt issues by HFIs
The figure plots the evolution of leverage, cumulative net debt issues, cumulative net equity issues and average investment
over three years following a non-transitory (T NDI = 0) and transitory net-debt issue (T NDI = 1), respectively, in year 0.
A net-debt issue is defined as transitory if the following three conditions are met: (1) N DI exceeds 2.5% of total assets, (2)
the firm had excess leverage in the year prior to the issue (Dev > 0), and (3) capital expenditures exceed the industry median
capital expenditures (in percent of assets). Dev
t
= L
t−1
−L
∗
t
(βX
t−1
) and the leverage target (L
∗
) is estimated as in Table 10.
Variable definitions are in Appendix Tables 1 and 2. Total sample of 6,684 U.S. public firms and 31,449 firm-years, 1971-2012.
A: Non-transitory net debt issues
0 .1 .2 .3 .4
Percent of firm market value
0 1 2 3
Year since non-transitory debt issue
Debt Cumulative NDI
Cumulative NEI Capex
B: Transitory net-debt issues by
0 .1 .2 .3 .4
Percent of firm market value
0 1 2 3
Year since transitory debt issue
Debt Cumulative NDI
Cumulative NEI Capex
37
Table 1: Selection of CRSP/Compustat firm-years and SDC public debt offerings
Sample restriction Firm-years Firms
A: CRSP/Compustat sample, 1971-2012
Sample after merging CRSP and Compustat 250,053 23,081
U.S. domiciled firms only -20,473 -2,158
Nongovernmental, industrial firms only
a
-65,011 -5,593
No multiple annual observations -1,817 0
No missing information on book value of assets -449 -16
Firm age positive
b
-5,463 -382
Consistent cash-flow statement data
c
-4,888 -333
Consistent balance sheet data
d
-2,528 -57
Went public during sample period -39,449 -2,411
Contiguous annual observations
e
-16,944 0
Final CRSP/Compustat sample 93,031 12,131
B: SDC public bond issue sample, 1980-2012
Sample after merging SDC with CRSP and Compustat
f
12,481 1,845
Total asset data available -230 -2
No deals within 30 days -115 -28
Underwriter fee data available -5,683 -125
Merge SDC with CRSP -2,213 -458
Firm on CRSP for 260 trading days before issue -165 -101
Data on last fiscal year available -72 -16
Issue size at least $10mn (in 1990 dollars) -33 -11
Issue size at most $10bn (in 1990 dollars) -35 -2
Underwriter fee at least 5 bp -80 -5
Final SDC/CRSP/Compustat public bond issue sample 3,773 1,075
a
Eliminates utilities (SIC codes 4899-5000), financial firms (SIC codes 5999-7000), and government entities (SIC codes
greater than 8999).
b
Firm age is the difference between the reporting date of the financial statement and the date of the first month a
company is reported in the CCM monthly stock price database, rounded to the next smaller integer.
c
For cash-flow data consistency, we first drop observations with negative values for the following Compustat variable
names (see Appendix Table 1 and 2 for variable definitions): dltis, dltr, sstk, prstkc, dv, capx, aqc, ivch, sppe and
siv (eliminates 3,493 obs.). Moreover, we set missing entries for items in the cash flow statement to zero and drop
observations in case total sources or uses of funds equal zero (eliminates 1,165 obs.) or deviate by more than 1% from
each other (eliminates 230 obs.).
d
For balance sheet data consistency, we set missing entries for deferred taxes and investment tax credit (txditc) and
preferred stock liquidation value (pstkl) equal to zero and subsequently require non-missing data for the market value
of the firm’s equity (prcc f × csho), Tobin’s Q (lt + pstkl - txditc + prcc f × csho)/at), total debt (dltt + dlc), cash
holdings (che), property plant and equipment (ppent) and further drop observations in case the book leverage ratio
is outside the unit interval or cash holdings are negative.
e
We eliminate observations once the underlying annual data become non-contiguous. For example, suppose a firm has
a total 11 annual observations on Compustat but only the first 8 years are contiguously available. In this case, only
the first 8 annual observations are included in the final sample.
f
The SDC sample is merged with CRSP/Compustat after having accounted for rows (1) to (8) in Panel A.
38
Table 2: Annual cumulative net-debt issues, retirements and equity issues following public listing
In Panel A, a net-debt issue (N DI
+
) is the positive portion of total debt issue minus debt retirement as given
by annual Compustat cash flow statements. Event year 0 is the year of public listing. In each event year, firms
in the bottom quartile of the cumulative net-debt issue distribution are classified as low-frequency issuers (LFIs),
while high-frequency issuers (HFIs) are in the top quartile. This LFI and HFI classification is based on net-debt
issues and is maintained also in Panel B and C that report on net-debt retirements (N DI
−
) and equity issues
(EI), respectively. Each panel first shows the issue frequency assuming a 2.5% issue size threshold (relative to total
assets) and, second, assuming a 5% size threshold. In the columns reporting the cumulative number of issues, the
reported number is the mean for the LFI and the HFI columns, while the Mean and Median columns are for the
total sample of 12,131 U.S. public firms, 1971-2012. Variable definitions are in Appendix Tables 1 and 2.
Fraction of aggregate
Number of Firms issue proceeds Cumulative number of issues
Year All LFI HFI LFI HFI LFI Mean Median HFI
A.1 Net-debt issues with a 2.5% threshold
0 12131 9028 3103 0.02 0.98 0.00 0.26 0 1.00
1 10783 5500 5283 0.02 0.98 0.00 0.61 0 1.25
2 9264 3652 2495 0.02 0.73 0.00 0.94 1 2.22
3 7918 2509 3029 0.01 0.74 0.00 1.24 1 2.45
4 6848 1850 3236 0.01 0.83 0.00 1.52 1 2.68
5 5921 2739 1823 0.10 0.73 0.49 1.79 2 3.59
6 5046 2058 1900 0.28 0.59 0.50 2.07 2 3.80
7 4344 1574 1107 0.15 0.56 0.51 2.33 2 4.72
8 3815 1219 1161 0.08 0.65 0.51 2.60 2 4.92
9 3345 954 1206 0.04 0.67 0.50 2.87 3 5.10
10 2931 754 763 0.04 0.59 0.49 3.11 3 5.98
15 1549 419 552 0.11 0.63 1.14 4.50 4 7.53
20 782 215 243 0.04 0.45 1.64 5.78 6 9.77
A.2 Net-debt issues with a 5% threshold
0 12131 n.A. n.A. n.A. n.A. n.A. 0.21 0 n.A.
1 10783 6275 4508 0.04 0.96 0.00 0.51 0 1.21
2 9264 4378 4886 0.03 0.97 0.00 0.76 1 1.45
3 7918 3137 2284 0.04 0.65 0.00 1.00 1 2.38
4 6848 2351 2482 0.05 0.75 0.00 1.22 1 2.56
5 5921 1788 2492 0.03 0.79 0.00 1.43 1 2.73
6 5046 1325 1297 0.06 0.49 0.00 1.64 1 3.65
7 4344 2065 1309 0.21 0.63 0.51 1.83 2 3.79
8 3815 1654 1322 0.16 0.66 0.52 2.03 2 3.94
9 3345 1310 1301 0.14 0.63 0.52 2.23 2 4.07
10 2931 1047 767 0.07 0.56 0.51 2.41 2 4.94
15 1549 592 482 0.20 0.54 1.10 3.43 3 6.22
20 782 223 237 0.04 0.43 1.07 4.26 4 7.58
Continued on next page
39
Table 2 – continued from previous page
Fraction of aggregate
Number of Firms issue proceeds Cumulative number of issues
Year All LFI HFI LFI HFI LFI Mean Median HFI
B.1 Net-debt retirements with a 2.5% threshold
0 12131 9028 3103 1.00 0.00 0.46 0.34 0 0.00
1 10783 5500 5283 0.64 0.36 0.62 0.51 0 0.40
2 9264 3652 2495 0.38 0.20 0.74 0.70 1 0.44
3 7918 2509 3029 0.30 0.39 0.82 0.92 1 0.80
4 6848 1850 3236 0.16 0.60 0.91 1.15 1 1.12
5 5921 2739 1823 0.37 0.37 1.30 1.37 1 1.22
6 5046 2058 1900 0.26 0.57 1.39 1.59 1 1.55
7 4344 1574 1107 0.18 0.33 1.47 1.81 2 1.62
8 3815 1219 1161 0.20 0.42 1.52 2.05 2 2.01
9 3345 954 1206 0.12 0.56 1.52 2.28 2 2.37
10 2931 754 763 0.10 0.36 1.54 2.50 2 2.49
15 1549 419 552 0.07 0.49 2.28 3.68 4 4.02
20 782 215 243 0.06 0.33 2.74 4.61 4 5.12
B.2 Net-debt retirements with a 5% threshold
0 12131 n.A. n.A. n.A. n.A. n.A. 0.27 0 n.A.
1 10783 6275 4508 0.68 0.32 0.41 0.37 0 0.31
2 9264 4378 4886 0.43 0.57 0.49 0.49 0 0.50
3 7918 3137 2284 0.35 0.28 0.54 0.63 0 0.59
4 6848 2351 2482 0.21 0.47 0.60 0.78 1 0.81
5 5921 1788 2492 0.15 0.50 0.62 0.92 1 1.02
6 5046 1325 1297 0.10 0.45 0.62 1.05 1 1.11
7 4344 2065 1309 0.30 0.50 0.96 1.19 1 1.32
8 3815 1654 1322 0.31 0.47 1.01 1.34 1 1.56
9 3345 1310 1301 0.19 0.59 1.04 1.49 1 1.74
10 2931 1047 767 0.17 0.37 1.04 1.63 1 1.91
15 1549 592 482 0.16 0.51 1.52 2.33 2 2.85
20 782 223 237 0.36 0.30 1.42 2.84 3 3.68
Continued on next page
40
Table 2 – continued from previous page
Fraction of aggregate
Number of Firms issue proceeds Cumulative number of issues
Year All LFI HFI LFI HFI LFI Mean Median HFI
C.1 Equity issues with a 2.5% threshold
0 12131 9028 3103 0.76 0.24 0.79 0.75 1 0.66
1 10783 5500 5283 0.43 0.57 1.07 1.04 1 1.01
2 9264 3652 2495 0.31 0.44 1.34 1.30 1 1.30
3 7918 2509 3029 0.26 0.47 1.61 1.54 1 1.52
4 6848 1850 3236 0.20 0.57 1.86 1.76 1 1.73
5 5921 2739 1823 0.37 0.42 2.04 1.99 2 1.99
6 5046 2058 1900 0.37 0.47 2.33 2.19 2 2.13
7 4344 1574 1107 0.29 0.42 2.61 2.38 2 2.31
8 3815 1219 1161 0.33 0.39 2.86 2.56 2 2.46
9 3345 954 1206 0.23 0.51 3.07 2.70 2 2.55
10 2931 754 763 0.23 0.43 3.30 2.87 2 2.81
15 1549 419 552 0.15 0.53 3.97 3.56 3 3.36
20 782 215 243 0.11 0.38 4.08 3.81 3 3.79
C.2 Equity issues with a 5% threshold
0 12131 n.A. n.A. n.A. n.A. n.A. 0.73 1 n.A.
1 10783 6275 4508 0.48 0.52 0.99 0.97 1 0.95
2 9264 4378 4886 0.36 0.64 1.19 1.18 1 1.18
3 7918 3137 2284 0.32 0.39 1.35 1.37 1 1.43
4 6848 2351 2482 0.28 0.48 1.50 1.53 1 1.62
5 5921 1788 2492 0.26 0.50 1.68 1.70 1 1.79
6 5046 1325 1297 0.24 0.40 1.85 1.83 1 1.99
7 4344 2065 1309 0.38 0.48 1.91 1.95 2 2.08
8 3815 1654 1322 0.41 0.44 2.04 2.06 2 2.20
9 3345 1310 1301 0.31 0.54 2.12 2.16 2 2.25
10 2931 1047 767 0.29 0.39 2.25 2.27 2 2.54
15 1549 592 482 0.24 0.48 2.68 2.71 2 2.93
20 782 223 237 0.24 0.43 2.58 2.70 2 3.23
41
Table 3: Persistence of the net-debt issue-frequency classification
In each event year, firms in the bottom quartile of the cumulative net-debt issue frequency distribution shown in Table
2 are classified as low-frequency issuers or LFIs, while high-frequency issuers (HFIs) are in the top quartile. The issue
frequency classification uses the 2.5% issue-size threshold. Year 0 is the year of public listing. Columns (1) through (3) are
backward looking and show the fraction of firms classified as HFI and LFI in a given year that are also classified as HFI
and LFI in the previous year (n=1), in the previous two years (n=2), and in the previous three years (n=3), respectively.
Columns (4) through (6) are forward looking and show the fraction of HFIs and LFIs in a given year that are also classified
as LFI, medium frequency issuer, or HFI in any of the firm’s remaining sample years. Sample of 12,131 U.S. public firms,
1971-2012.
Percent of sample firms Percent of sample firms
with no classification change with no classification change
Years over past n sample years over all future sample years
since n=1 n=2 n=3 LFI Median HFI
listing (1) (2) (3) (4) (5) (6)
A: High frequency issuers (HFIs)
0 n.A. n.A. n.A. 0.06 0.27 0.67
1 0.51 n.A. n.A. 0.08 0.36 0.56
2 1.00 0.59 n.A. 0.02 0.24 0.74
3 0.69 0.69 0.41 0.02 0.31 0.66
4 0.80 0.55 0.55 0.03 0.39 0.58
5 1.00 0.88 0.65 0.01 0.26 0.73
6 0.81 0.81 0.71 0.01 0.33 0.66
7 1.00 0.88 0.88 0.00 0.19 0.81
8 0.84 0.84 0.75 0.00 0.23 0.77
9 0.84 0.70 0.70 0.00 0.30 0.70
10 1.00 0.92 0.82 0.00 0.18 0.82
15 0.88 0.76 0.76 0.00 0.30 0.70
20 0.92 0.92 0.88 0.00 0.25 0.73
Mean 0.81 0.76 0.69 0.01 0.21 0.78
B: Low frequency issuers (LFIs)
0 n.A. n.A. n.A. 0.46 0.29 0.25
1 1.00 n.A. n.A. 0.58 0.30 0.12
2 1.00 1.00 n.A. 0.68 0.26 0.07
3 1.00 1.00 1.00 0.76 0.20 0.03
4 1.00 1.00 1.00 0.82 0.17 0.02
5 0.59 0.59 0.59 0.67 0.29 0.04
6 1.00 0.65 0.65 0.72 0.26 0.03
7 1.00 1.00 0.69 0.77 0.22 0.02
8 1.00 1.00 1.00 0.82 0.17 0.01
9 1.00 1.00 1.00 0.87 0.13 0.01
10 1.00 1.00 1.00 0.89 0.10 0.00
15 1.00 1.00 1.00 0.91 0.09 0.00
20 1.00 1.00 1.00 0.94 0.18 0.03
Mean 0.94 0.89 0.83 0.88 0.11 0.02
42
Table 4: Average annual firm characteristics following public listing
In each event year, firms in the bottom quartile of the cumulative net-debt issue frequency distribution shown in Table
2 are classified as low-frequency issuers or LFIs, while high-frequency issuers (HFIs) are in the top quartile. The issue
frequency classification uses the 2.5% issue-size threshold. Year 0 is the year of public listing. The table lists the average
annual values, scaled by total book value of assets (when relevant), of the market leverage ratio (L), the book leverage
ratio (BL), the fraction of firms that are all-equity financed firms (AE, defined as zero-leverage firms), the cash ratio
(C), the book asset value (Assets) itself, the operating profitability (P rof), asset tangibility (T an), Tobin’s Q (Q), R&D
expenditures, dividend payouts (Div), capital expenditures (Capex), asset growth rate (g
A
) and sales growth rate (g
S
).
Variable definitions are in Appendix Tables 1 and 2. Sample of 12,131 U.S. public firms, 1971-2012.
Year L BL AE C Assets Prof T an Q R&D Div Capex g
A
g
S
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
A: High frequency issuers (HFIs)
0 0.26 0.34 0.01 0.19 294 -0.01 0.33 2.56 0.04 0.01 0.14 n.A. n.A.
1 0.30 0.33 0.01 0.13 276 -0.02 0.34 2.16 0.05 0.00 0.12 0.44 1.92
2 0.39 0.40 0.00 0.09 391 0.02 0.37 1.79 0.03 0.00 0.11 0.38 0.68
3 0.38 0.37 0.01 0.09 392 0.03 0.36 1.72 0.03 0.00 0.09 0.30 1.62
4 0.37 0.35 0.02 0.10 404 0.04 0.35 1.69 0.03 0.00 0.08 0.23 0.36
5 0.41 0.39 0.01 0.08 500 0.05 0.37 1.55 0.03 0.00 0.09 0.22 0.32
6 0.39 0.37 0.02 0.08 512 0.05 0.37 1.57 0.03 0.01 0.08 0.18 0.18
7 0.41 0.40 0.02 0.07 581 0.06 0.38 1.56 0.02 0.01 0.09 0.79 0.20
8 0.39 0.37 0.02 0.08 581 0.06 0.38 1.63 0.02 0.01 0.09 0.15 0.35
9 0.39 0.36 0.02 0.08 877 0.06 0.37 1.52 0.02 0.01 0.08 0.17 0.20
10 0.38 0.37 0.01 0.08 960 0.06 0.38 1.60 0.02 0.01 0.08 0.15 0.14
15 0.35 0.32 0.01 0.08 1,540 0.08 0.35 1.47 0.02 0.01 0.07 0.13 0.12
20 0.31 0.31 0.02 0.06 2,469 0.09 0.36 1.50 0.02 0.01 0.07 0.08 0.07
Mean 0.35 0.35 0.01 0.10 804 0.03 0.36 1.80 0.03 0.01 0.10 0.27 0.72
B: Low frequency issuers (LFIs)
0 0.11 0.14 0.21 0.33 181 0.01 0.23 3.17 0.06 0.01 0.08 n.A. n.A.
1 0.11 0.11 0.28 0.31 206 -0.02 0.22 2.70 0.08 0.00 0.08 0.25 0.94
2 0.10 0.09 0.33 0.31 236 -0.02 0.22 2.55 0.09 0.01 0.06 0.20 1.02
3 0.09 0.08 0.39 0.33 256 0.00 0.21 2.54 0.10 0.01 0.05 0.18 0.60
4 0.07 0.06 0.45 0.34 281 0.01 0.20 2.51 0.10 0.01 0.05 0.16 0.33
5 0.12 0.11 0.30 0.28 311 0.02 0.22 2.27 0.08 0.01 0.06 0.18 0.28
6 0.11 0.10 0.34 0.28 368 0.03 0.22 2.28 0.08 0.01 0.06 0.17 0.35
7 0.10 0.09 0.37 0.29 428 0.04 0.21 2.26 0.09 0.01 0.05 0.17 0.39
8 0.08 0.08 0.39 0.31 468 0.04 0.21 2.23 0.09 0.01 0.05 0.19 0.20
9 0.08 0.07 0.42 0.31 387 0.04 0.20 2.19 0.09 0.01 0.05 0.09 0.13
10 0.06 0.06 0.44 0.32 406 0.05 0.20 2.27 0.08 0.01 0.05 0.10 0.99
15 0.06 0.06 0.44 0.29 451 0.06 0.22 2.14 0.08 0.01 0.05 0.09 0.11
20 0.06 0.07 0.42 0.28 459 0.08 0.23 2.17 0.07 0.02 0.05 0.09 0.08
Mean 0.10 0.10 0.32 0.31 345 0.02 0.22 2.56 0.08 0.01 0.06 0.17 0.53
43
Table 5: Net-debt issue hazards and the probability of becoming a high-frequency issuer
Panel A presents coefficient-estimates determining the probability of becoming a HFI T years following public listing (year
0) conditional on the year-0 covariates:
Y
iT
= α + βX
i,0
+
iT
,
where Y
iT
= 1 if firm i is classified as a HFI in T years, T = 3, 6, 9, 12, 15. Panel B presents coefficient estimates determining
the number of years between successive net-debt issues (financing spells) for LFIs and HFIs, respectively, using the following
exponential shared frailty hazard model:
h
i
= h
0
exp(β
0
+ βx
i,t−1
)α
i
,
where h
0
denotes a constant baseline hazard, and α
i
is a frailty term capturing unobserved observation-specific effects. Panel
C allows a time-varying baseline hazard by extending the hazard model
h
i
= h
0
exp(β
0
+ βx
i,t−1
+ γf(t))α
i
where f (t) = t + t
2
+ t
3
. The covariates are investment (Capex), R&D expenditures (R&D), market leverage ratio (L),
cash ratio (C), logarithm of assets (Size), operating cash flow (P rof), tangibility (T an), Tobin’s Q (Q) and depreciation
expenditures (Depr). In Panel A, the estimation includes industry dummies, based on the Fama-French 12 Industry
Classification. In Panel B and C, the covariates x
i,t−1
enter subtracting their median values. All covariates are winsorized at
the 1(99) percent level or must lie between zero and one (cash ratio and leverage). The high and low-frequency classification
(HFI and LFI) is performed annually and based on a 2.5% issue size threshold. Variable definitions are in Appendix
Tables 1 and 2. *, ** indicate significance at the 5% and 1% level, respectively. Sample of 12,131 U.S. public firms, 1971-2012.
Firm-specific explanatory variables (X)
N Capex R&D L C Size P rof T an Q Depr
A: Probability of becoming HFI T years from public listing (X measured in year 0): Y
iT
= α + βX
i,0
+
iT
T = 3 years 7,918 2.658** -1.180** 0.321** -1.245** -0.013 -0.813** -0.056 -0.022** -2.345**
T = 6 years 5,046 2.509** -1.359** 0.489** -1.248** -0.038** -0.854** -0.176 -0.031** -1.157
T = 9 years 3,345 2.426** -1.632** 0.653** -1.375** -0.036* -0.961** -0.245 -0.027* -0.475
T = 12 years 2,307 2.215** -2.324** 0.842** -1.280** -0.050** -1.133** -0.176 -0.029 -1.946
T = 15 years 1,549 2.103** -2.310** 1.037** -1.352** -0.048 -0.653* -0.375 -0.009 -1.768
B: Time between net-debt issues,with constant baseline hazard: h
i
= h
0
exp(β
0
+ βx
i,t−1
)α
i
LFI only 29,421 13.652** 1.412 0.338** 0.055** 1.039* 0.375** 0.702 1.02 0.183
HFI only 28,346 14.765** 1.113 0.567** 0.446** 1.007 0.705** 0.767** 0.997 0.033**
C: Time between net-debt issues, with time-varying baseline hazard: h
i
= h
0
exp(β
0
+ βx
i,t−1
+ γf(t))α
i
LFI only 29,421 15.159** 1.243 0.486** 0.056** 1.034 0.363** 0.72 1.027 0.160*
HFI only 28,162 3.081** 0.914 0.410** 0.619** 0.999 0.887* 0.962 0.984** 0.159**
44
Table 6: Fixed-cost estimation using underwriter fees in public debt issues
The table displays parameter estimates for the direct issue cost function
s = β
0
+ β
1
1
x
1
+ β
2
x
1
x
2
+ γX + ,
where s is underwriter spread, x
1
is offering proceeds, x
2
is issuer’s total equity value. The coefficient β
1
represents the
fixed issue-cost component, β
0
is the variable cost component, while β
2
represents a quadratic cost component. The vector
X of firm-specific control variables contains five ratings dummies for debt issues (AAA, A, BBB, BB, and B-CCC), the
lagged ratio of operating cash flow to total assets, the lagged book leverage ratio, a lagged market-to-book ratio and an
issue wave variable. Issue wave is constructed the difference between the annual aggregate net debt issue volume and
the aggregate book value of assets, standardized by its total sample mean and standard deviation. Firms with at most
two public debt issues in the sample are labeled low-frequency public debt issuers (LFPI), while firms with at least four
sample debt issues are labeled high-frequency (HFPI). Standard errors in parentheses. Variable definitions are in Appendix
Tables 1 and 2. *, ** indicate significance at the 5% and 1% level, respectively. Sample of 3,773 public debt issues, 1980-2012.
A: Issue cost components by absolute public debt issue frequency (LFPIs and HFPIs)
Total LFPI HFPI
Issue cost parameter sample
P
DI = 1
P
DI ≤ 2
P
DI ≥ 4
P
DI ≥ 11
Fixed-cost component (β
1
) 24.915** 48.722** 44.827** 10.644** 4.256
(3.42) (7.03) (5.95) (3.48) (3.72)
Variable cost component (β
0
) 2.157** 2.290** 2.449** 1.704** 1.633**
(0.10) (0.26) (0.20) (0.10) (0.12)
Quadratic-cost component (β
2
) 0.210** 0.103* 0.124* 0.734** 0.708*
(0.08) (0.05) (0.05) (0.11) (0.33)
Firm controls yes yes yes yes yes
R2 0.672 0.774 0.767 0.457 0.241
Number of debt issues 3774 513 897 2622 1331
Number of issuing firms 1075 513 705 285 66
B: Issue cost components by age of low-frequency public debt issuers (LFPIs)
P
DI = 1
P
DI ≤ 2
Issue cost parameter Age > 3 Age > 6 Age > 9 Age > 3 Age > 6 Age > 9
Fixed-cost component (β
1
) 45.043** 42.873** 37.535** 41.224** 39.899** 35.637**
(7.30) (8.66) (9.12) (6.15) (6.93) (7.28)
Variable cost component (β
0
) 2.220** 2.368** 2.397** 2.310** 2.341** 2.396**
(0.30) (0.33) (0.34) (0.24) (0.26) (0.26)
Quadratic-cost component (β
2
) 0.081 0.059 0.265* 0.105* 0.093 0.275**
(0.05) (0.07) (0.11) (0.05) (0.08) (0.09)
Firm controls yes yes yes yes yes yes
R2 0.798 0.804 0.834 0.784 0.784 0.808
N 406 330 275 738 620 527
45
Table 7: Financial constraint scores for high- and low-frequency debt issuers
In each event year, firms in the bottom quartile of the cumulative net-debt issue frequency distribution shown in Table 2 are
classified as low-frequency issuers or LFIs, while high-frequency issuers (HFIs) are in the top quartile. The issue frequency
classification uses the 2.5% issue-size threshold. Year 0 is the year of public listing. The financial constraint scores are the
KZ-index (Kaplan and Zingales, 1997), the WW-index (Whited and Wu, 2006), and the SA-index (Hadlock and Pierce,
2010). Variable definitions are in Appendix Tables 1 and 2. Sample of 12,131 U.S. public firms, 1971-2012.
KZ-index WW-index SA-index
Year LFI HFI LFI HFI LFI HFI
0 0.54 1.24 n.a. n.a. -1.92 -1.95
1 0.56 1.33 -0.19 -0.23 -2.02 -2.07
2 0.42 1.47 -0.21 -0.21 -2.11 -2.26
3 0.30 1.33 -0.20 -0.25 -2.21 -2.33
4 0.18 1.22 -0.20 -0.21 -2.29 -2.41
5 0.32 1.32 -0.20 -0.23 -2.37 -2.53
6 0.24 1.24 -0.22 -0.23 -2.45 -2.60
7 0.13 1.30 -0.22 -0.25 -2.53 -2.75
8 0.02 1.25 -0.22 -0.25 -2.60 -2.79
9 -0.05 1.19 -0.22 -0.25 -2.66 -2.84
10 -0.13 1.25 -0.25 -0.25 -2.73 -2.94
15 -0.21 1.02 -0.25 -0.29 -3.00 -3.27
20 -0.42 0.96 -0.27 -0.30 -3.32 -3.56
Avg. 0.28 1.24 -0.22 -0.25 -2.34 -2.56
46
Table 8: Average annual corporate funding mix following public listing
In each event year, firms in the bottom quartile of the cumulative net-debt issue frequency distribution shown in Table
2 are classified as low-frequency issuers or LFIs, while high-frequency issuers (HFIs) are in the top quartile. The issue
frequency classification uses the 2.5% issue-size threshold. Year 0 is the year of public listing. The annual (non-negative)
cash contribution of the i’th funding source is the ratio R
j
≡ S
j
/
P
7
i
S
i
, where the denominator is the sum of the seven
individual funding sources in the firm’s total cash flow statement:
7
X
i
S
i
= NDI
+
+ EI + CF
+
+ ∆C
−
+ ∆W
−
+ I
−
+ O
+
The four columns are: R
NDI
+
is the net debt issue ratio (NDI
+
in the numerator), R
EI
is the equity issue ratio, R
CF
+
is
the operating cash flow contribution, R
∆C
−
is the contribution from cash draw-downs, R
∆W
−
is contribution of reductions
in net working capital and R
I
−
is the fraction of funds provided by illiquid asset sales. Variable definitions are in Appendix
Tables 1 and 2. Sample of 12,131 U.S. public firms, 1971-2012.
R
NDI
+
R
EI
R
CF
+
R
∆C
−
R
∆W
−
R
I
−
R
O
+
Year HFI LFI HFI LFI HFI LFI HFI LFI HFI LFI HFI LFI HFI LFI
0 0.33 0.00 0.36 0.60 0.21 0.27 0.02 0.03 0.04 0.04 0.03 0.04 0.01 0.01
1 0.31 0.01 0.13 0.17 0.29 0.40 0.12 0.20 0.07 0.10 0.07 0.12 0.01 0.01
2 0.35 0.01 0.12 0.16 0.31 0.42 0.06 0.14 0.08 0.11 0.07 0.15 0.01 0.01
3 0.27 0.01 0.12 0.14 0.36 0.44 0.06 0.14 0.09 0.10 0.09 0.16 0.01 0.01
4 0.23 0.00 0.11 0.14 0.40 0.44 0.06 0.14 0.10 0.09 0.09 0.17 0.01 0.01
5 0.25 0.04 0.10 0.13 0.40 0.46 0.05 0.12 0.09 0.10 0.09 0.14 0.01 0.01
6 0.23 0.04 0.09 0.13 0.42 0.46 0.06 0.12 0.09 0.09 0.10 0.15 0.01 0.01
7 0.25 0.03 0.09 0.12 0.42 0.47 0.04 0.11 0.08 0.09 0.10 0.15 0.01 0.01
8 0.21 0.03 0.10 0.12 0.44 0.47 0.05 0.11 0.10 0.08 0.10 0.17 0.01 0.01
9 0.21 0.03 0.08 0.10 0.44 0.47 0.05 0.11 0.11 0.10 0.09 0.18 0.01 0.01
10 0.21 0.01 0.11 0.11 0.46 0.48 0.04 0.14 0.08 0.08 0.10 0.17 0.01 0.01
15 0.18 0.04 0.08 0.09 0.47 0.45 0.05 0.13 0.11 0.08 0.10 0.20 0.01 0.01
20 0.17 0.05 0.05 0.08 0.50 0.49 0.06 0.11 0.11 0.09 0.11 0.17 0.01 0.01
Mean 0.26 0.02 0.13 0.24 0.38 0.41 0.06 0.11 0.08 0.08 0.09 0.13 0.01 0.01
Median 0.19 0.00 0.01 0.04 0.36 0.35 0.00 0.00 0.00 0.00 0.01 0.00 0.00 0.00
Funding source standardization by lagged total assets
Mean 0.11 0.01 0.11 0.22 0.11 0.12 0.02 0.04 0.03 0.03 0.03 0.07 0.00 0.01
Median 0.03 0.00 0.00 0.01 0.09 0.10 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
47
Table 9: Leverage ratio instability and volatility following public listing
In each event year, firms in the bottom quartile of the cumulative net-debt issue frequency distribution shown in Table 2 are
classified as low-frequency issuers or LFIs, while high-frequency issuers (HFIs) are in the top quartile. The issue frequency
classification uses the 2.5% issue-size threshold. Year 0 is the year of public listing. Leverage instability is fraction of firms
for which the leverage ratio change from year 0 to year t exceeds +/− 20 percentage points (|L
t
− L
0
| > 0.2) and leverage
volatility is the average standard deviation of the firm-level leverage ratio up to year t using a minimum of five yearly
observations. Net leverage is gross leverage net of cash balances. Panel A works with actual leverage ratios, while Panel
B works with estimated target leverage ratios: L
∗
i,t
(βX
i,t−1
), where the determinants X
i,t−1
are the lagged values of size,
profitability, Q, cash ratio, tangibility, depreciation, R&D expenses, capital expenditures and the median industry leverage
ratio (all winsorized at the 1(99) percent level). Variable definitions are in Appendix Tables 1 and 2. Sample of 12,131 U.S.
public firms, 1971-2012.
Fraction of Average
leverage instability leverage volatility
|(L
t
− L
0
)| > 0.2
1
T
P
τ
(L
τ
−
¯
L)
2
Gross leverage Net leverage Gross leverage Net leverage
Year LFI HFI LFI HFI LFI HFI LFI HFI
A: Actual leverage ratios
0 n.A. n.A. n.A. n.A. n.A. n.A. n.A. n.A.
1 0.06 0.24 0.20 0.31 n.A. n.A. n.A. n.A.
2 0.09 0.45 0.30 0.49 n.A. n.A. n.A. n.A.
3 0.09 0.45 0.35 0.51 n.A. n.A. n.A. n.A.
4 0.09 0.46 0.35 0.52 n.A. n.A. n.A. n.A.
5 0.16 0.50 0.38 0.55 0.08 0.17 0.20 0.20
6 0.16 0.49 0.39 0.56 0.07 0.17 0.20 0.20
7 0.15 0.50 0.39 0.57 0.07 0.17 0.19 0.19
8 0.14 0.48 0.41 0.55 0.07 0.17 0.20 0.19
9 0.13 0.47 0.41 0.53 0.06 0.16 0.20 0.19
10 0.12 0.44 0.36 0.52 0.06 0.16 0.19 0.19
15 0.16 0.46 0.43 0.52 0.07 0.16 0.19 0.18
20 0.20 0.44 0.37 0.47 0.08 0.17 0.18 0.19
Avg. 0.10 0.38 0.26 0.43 0.08 0.16 0.19 0.19
B: Target leverage ratios
0 n.A. n.A. n.A. n.A. n.A. n.A. n.A. n.A.
1 0.00 0.00 0.00 0.00 n.A. n.A. n.A. n.A.
2 0.01 0.01 0.00 0.01 n.A. n.A. n.A. n.A.
3 0.02 0.04 0.01 0.03 n.A. n.A. n.A. n.A.
4 0.03 0.06 0.02 0.05 n.A. n.A. n.A. n.A.
5 0.04 0.09 0.03 0.09 0.06 0.06 0.05 0.06
6 0.05 0.10 0.03 0.09 0.06 0.06 0.05 0.06
7 0.05 0.12 0.04 0.11 0.06 0.06 0.05 0.06
8 0.05 0.10 0.03 0.12 0.05 0.06 0.05 0.06
9 0.05 0.12 0.04 0.10 0.05 0.06 0.05 0.06
10 0.05 0.10 0.04 0.10 0.05 0.06 0.05 0.06
15 0.06 0.11 0.07 0.11 0.06 0.06 0.05 0.06
20 0.06 0.11 0.07 0.14 0.05 0.06 0.06 0.07
Avg. 0.03 0.06 0.02 0.06 0.06 0.06 0.05 0.06
48
Table 10: Speed of leverage ratio adjustments to target leverage deviations
The table reports estimates of the speed-of-adjustment coefficient φ in the regression:
L
i,t
− L
i,t−1
= α + η
i
+ φ
L
∗
i,t
(βX
i,t−1
) − L
i,t−1
+
i,t
.
where the dependent variable is firm i’s change in market leverage ratio L from time t − 1 to t,
i,t
is the regression error
term, α is the constant, η
i
is firm fixed effect, L
∗
i,t
(βX
i,t−1
) is the (estimated) target leverage ratio where the determinants
X
i,t−1
are the lagged values of size, profitability, Q, cash ratio, tangibility, depreciation, R&D expenses, capital expenditures
and the median industry leverage ratio (all winsorized at the 1(99) percent level). Coefficients are estimated using system
GMM, implemented using the stata command xtabond2, assuming that all regressors are predetermined. We use a
maximum number of lags of 3 (5) for market leverage (target leverage ratio regressors). Variable definitions are in Appendix
Tables 1 and 2. *, ** indicate significance at the 5% and 1% level, respectively. Total sample of 10,783 U.S. public firms
and 80,900 firm-years, 1971-2012.
GMM estimates of φ
Dependent
variable All HFI LFI HFI - LFI
Market leverage change 0.259** 0.316** 0.272** 0.044
(0.009) (0.014) (0.020) (z-score: 1.793)
49
Table 11: Association between net-debt issue size, capex, and financing deficit for HFIs
The table reports coefficient estimates using the following regression:
NDI
i,t
A
i,t
= α + β
1
Capex
i,t
+ β
2
NetDeficit
i,t
+ β
3
Capex
2
i,t
+ β
4
NetDeficit
2
i,t
+
i,t
.
where NDI/A is net-debt issues (or retirements) scaled by total assets, and Capex and NetDeficit are capi-
tal expenditures and the financing deficit net of Capex, respectively, also scaled by total assets. NetDeficit ≡
dv + aqc + ivch − siv − ivstch − sppe − ivaco − oancf + chech, where all Compustat mnemonics are scaled by total assets.
Variable definitions are in Appendix Tables 1 and 2. Standard errors in parentheses, *, ** indicate significance at the 5%
and 1% level, respectively. Total sample of 6,684 U.S. public firms and 31,449 firm-years, 1971-2012.
(1) (2) (3) (4)
Capex 0.334** 0.518** 0.589** 0.567**
(0.01) (0.01) (0.02) (0.02)
NetDeficit 0.354** 0.599** 0.598**
(0.01) (0.01) (0.01)
Capex
2
0.149** 0.164**
(0.04) (0.04)
NetDeficit
2
-0.485** -0.474**
(0.01) (0.01)
α 0.032** 0.004** 0.014** 0.022*
(0.00) (0.00) (0.00) (0.01)
Industry no no no yes
Age no no no yes
N 31449 31449 31449 31449
R
2
0.06 0.40 0.53 0.54
50
Table 12: Transitory debt issues as a determinant of leverage reductions for HFIs
Panel A presents coefficient-estimates determining the probability that a transitory net debt issue in period t−1, T N DI
i,t−1
,
is followed in year t by, alternatively, a decrease in leverage (Panel A), a net-debt retirement exceeding 2.5% of total assets
(Panel B), or a net equity issue exceeding 2.5% (Panel C). The logit model is:
Y
i,t
= α + βT NDI
i,t−1
+ γEcapex
i,t
+ δDev
i,t
+
i,t
,
where Y
i,t
= 1 is the binary dependent variable for firm i in year t. Moreover, Dev
i,t
= L
i,t−1
− L
∗
i,t
(βX
i,t−1
) is the
deviation of the optimal leverage ratio at the beginning of year t from last period’s actual leverage ratio, where L
∗
i,t
(βX
i,t−1
)
is estimated as in Table 10 above. The variable Ecapex
i,t−1
is the capital expenditure in period t − 1 in excess of firm i’s
industry median capex. Finally, T N DI
i,t−1
is a dummy variable for transitory debt issues: it is equal to one if, in period
t− 1, the following three conditions are fulfilled: the firm (1) issued net-debt in excess of 2.5% of total assets (N DI
i,t−1
> 0),
(2) invested heavily (Ecapex
i,t−1
> 0), and (3) was over-leveraged (Dev
i,t−1
> 0). The estimation also includes industry
dummies, based on the Fama-French 12 Industry Classification. The HFI classification is performed annually and based
on a 2.5% issue size threshold. Variable definitions are in Appendix Tables 1 and 2. Standard errors in parentheses, *, **
indicate significance at the 5% and 1% level, respectively. Total sample of 6,684 U.S. public firms and 31,449 firm-years,
1971-2012.
Sample Explanatory variables
with Y=1 T N DI
t−1
Ecapex
t
Dev
t
A: Y=1 if leverage reduction
10,843 0.738** -2.910**
(0.05) (0.21)
10,843 0.211** -1.214** 6.113**
(0.05) (0.23) (0.13)
B: Y=1 if net-debt retirement
6,291 0.438*** -6.063***
(0.05) (0.29)
6,291 -0.013 -4.610*** 5.024***
(0.06) (0.29) (0.13)
C: Y=1 if net equity issue
7,633 0.346** 1.634**
(0.06) (0.24)
7,633 0.423** 1.413** -0.766**
(0.06) (0.24) (0.14)
51
Table 13: CEO stock ownership and stock-based pay at HFIs and LFIs
The table estimates the impact of issue frequency on compensation policy
Y
i,t
= α + β
1
HF I
i,t
+ β
2
LF I
i,t
+ γX
i,t−1
+ γ
i
+
i,t
,
where γ
i
is firm-fixed effect, and the control variables X (lagged by one period) include investment Capex, R&D expenditures
(R&D), the cash ratio (C), the logarithm of assets (Size), operating cash flow (P rof), tangibility (T an), Tobin’s Q (Q),
and depreciation expenditures (Depr). In the first row, the dependent variable Y is the percentage of firm’s outstanding
share owned by the executive. In the second row, the dependent variable is the proportion of total CEO compensation that
is paid in (restricted) stocks and option. Compensation data are from ExecutComp, 1992-2012. Variable definitions are in
Appendix Tables 1 and 2. Standard errors in parentheses, *, ** indicate significance at the 5% and 1% level, respectively.
Total sample of 1,659 U.S. public firms and 13,907 firm-years, 1971-2012.
Dependent var. (Y) HF I LF I Capex R&D C Size P rof T an Q Depr R
2
Firm-years
Shares owned 0.009* -0.012** 0.035 -0.029 -0.008 -0.017** 0.009 0.005 0.001 -0.11 0.073 9484
(0.00) (0.00) (0.02) (0.02) (0.01) (0.00) (0.01) (0.02) (0.00)
Fraction stock 0.034** -0.009 0.170* -0.056 -0.068 0.029** 0.081 -0.078 0.018** -0.025 0.012 13626
(0.01) (0.01) (0.08) (0.11) (0.04) (0.01) (0.05) (0.06) (0.00)
52
Appendix Table 1: Variable construction using Compustat mnemonics
Variable Description (Compustat mnemonics) Mean Median St.Dev.
I: Selected firm characteristics (All Financial Statements)
ML Market leverage: (dlcc + dlt)/(prcc f*csho + dlcc + dlt) 0.22 0.13 0.24
BL Book leverage: (dlcc + dlt)/at 0.22 0.17 0.21
C Cash ratio: che/at 0.20 0.10 0.23
Size log(at) 4.24 4.15 1.89
Prof Profitability: (oancf + nwc
inv)/at 0.03 0.09 0.23
Tan Tangibility: ppent/at 0.28 0.22 0.23
Q Tobin’s Q: (lt + pstkl - txditc + prcc f*csho)/at 2.13 1.45 1.99
R&D xrd/at 0.05 0.00 0.11
Div dv/at 0.01 0.00 0.02
Capex capx/at 0.07 0.05 0.08
Depr dp/at 0.05 0.04 0.04
Tldt dltt/at 0.16 0.09 0.18
KZ -1.001909 × Prof + 3.139193 × L - 39.36780 × Div - 1.314759 × C + 0.2826389 × Q 0.71 0.71 1.20
WW
a
-0.091 × Prof - 0.062 × Divpos + 0.021 × Tldt - 0.044 × Size + 0.102 × ISG - 0.035 × g
S
-0.23 -0.22 0.59
SA -0.737 × Size + 0.043 × Size
2
- 0.04 × Age -2.50 -2.58 0.83
II: Sources of funds (Cash Flow Statement)
EI Equity Issues: sstk 14.6 0.4 95.5
DI Debt Issues: dltis + max[dlcch,0] 68.9 0.4 590.6
CF
+
Positive operating Cash Flow: max[oancf + nwc inv,0] 69.3 4.4 600.5
∆C
−
Draw-down of Cash balance: max[chech*(-1),0] 7.1 0.0 78.4
I
−
Asset sales: siv + min[ivstch,0] + min[ivaco,0] + sppe 29.6 0.1 577.8
∆W
−
Decrease in net Working capital: max[nwc inv*(-1),0] 5.5 0.0 52.5
O
+
Other sources: max[fincf oth,0] 2.1 0.0 31.3
III: Uses of funds (Cash Flow Statement)
ER Distributions to equity-holders: dv + prstkc 22.1 0.0 246.7
DR Debt Retirements: dltr + min[dlcch,0]*(-1) 56.7 0.9 533.8
CF
−
Negative operating Cash Flow: max[(oancf + nwc inv)*(-1),0] 2.8 0.0 31.9
∆C
+
Build-up of Cash balance: max[chech,0] 12.9 0.0 113.3
I
+
Investments: ivch + aqc + min[ivstch*(-1),0] + min[ivaco*(-1),0] + capx 88.6 6.1 919.7
∆W
+
Increase in net Working capital: max[nwc inv,0] 10.8 0.5 103.4
O
−
Other uses: max[fincf oth*(-1),0] 3.2 0.0 85.6
IV: Composite Variables (Cash Flow Statement)
NDI Debt issue minus debt retirement: DI - DR 12.2 0.0 229.1
NDI
+
Positive portion of debt issue minus debt retirement: max[DI - DR,0] 22.1 0.0 200.0
NDI
−
Negative portion of debt issue minus debt retirement: max[DR - DI,0] 10.0 0.0 109.6
NetDeficit (dv + aqc + ivch - siv - ivstch - sppe - ivaco - oancf + chech)/at 0.04 -0.03 0.41
V: Compensation Variables (ExecuComp)
SO CEO stock ownership: max[shrown tot pct,shrown excl opts pct] 0.05 0.02 0.08
CEO cash Salary + bonus + ltip + othcomp 1166.7 786.9 2027.7
CEO stock rstkgrnt 218.4 0.0 5645.7
CEO opt option awards blk value 1710.4 390.9 8163.3
FS Fraction stock: (CEO stock + CEO opt)/(CEO cash + CEO stock + CEO opt) 0.36 0.37 0.31
a
Divpos is a dummy variable equal to one in case the firm paid dividends, ISG is industry sales growth (defined using 3-digit SIC codes) and g
S
is
sales growth.
53
Appendix Table 2: Compustat mnemonics used for variable construction
Variable Description (Compustat mnemonics) Mean Median St.Dev.
I: Compustat balance sheet items
che Cash and cash equivalents 63.6 5.7 511.9
ppent Property, plant and equipment (net of depreciation) 209.8 12.1 1805.8
at Total assets 593.0 63.7 4281.7
dlc Debt in current liabilities 19.6 0.8 195.5
dltt Long-term debt 142.1 3.0 1010.7
lt Tota liabilities 334.4 23.7 2494.0
pstkl Preferred stock liquidation value 3.4 0.0 62.1
txditc Deferred taxes and investment tax credit 30.5 0.0 465.6
prcc f Stock price 13.7 8.5 19.7
csho Common shares outstanding 30.4 9.1 153.8
II: Compustat income statement items
sale Revenues 601.6 66.6 5166.3
xrd Research and development expenditures 11.0 0.0 112.5
dp Depreciation expenses (Income statement) 28.9 2.4 249.1
III: Compustat cash flow statement statement items
ibc Income Before Extraordinary Items 22.9 1.2 376.9
dpc Depreciation and Amortization 31.3 2.5 275.5
ocf oth
a
Other Operating Cash Flow ( = xidoc + txdc +esubc + sppiv + fopo + fsrco + exre) 12.2 0.4 225.3
nwc inv
b
Investment into Net Working Capital [ = (recch + invch + apalch + txach + aoloch)*(-1)] 5.3 0.5 116.5
oancf ibc + dpc + ocf oth + nwc inv 61.2 2.6 577.5
capx Capital Expenditures 40.9 3.0 327.1
aqc Acquisitions 16.0 0.0 173.5
ivch Increase in Investments 23.1 0.0 728.7
siv Sale of Investments 20.4 0.0 567.5
sppe Sale of Property, Plant and Equipment 2.2 0.0 40.8
ivstch Short-term Investments - Change -1.5 0.0 76.8
ivaco
c
Investing Activities - Other -0.1 0.0 132.5
inv total capx + aqc + ivch - siv - sppe - ivstch - ivaco 58.9 4.1 482.5
sstk Sale of Common and Preferred Stock 14.6 0.4 95.5
prstkc Purchase of Common and Preferred Stock 13.6 0.0 175.4
dv Cash Dividends 8.5 0.0 110.0
dltis Long-Term Debt - Issuance 66.2 0.0 580.9
dltr Long-Term Debt - Reduction 54.2 0.7 524.1
dlcch Changes in Current Debt 0.2 0.0 81.0
fincf oth
d
Other Financing Cash Flow [ = (txbcof + fiao) ] -1.1 0.0 90.5
fin total sstk + prstkc + dv + dltis + dltr + dlcch + fincf oth 3.5 0.4 295.1
chech Change in cash and cash equivalents 5.8 0.0 138.5
IV: ExecuComp items
shrown excl opts pct Percentage shares owned excluding options 4.7 1.3 8.3
shrown tot pct Percentages shares owned (total) 4.5 1.7 8.6
salary Salary 577.1 511.5 326.7
bonus Bonus 359.4 90.0 993.2
ltip Long-term incentive plan 44.4 0.0 481.2
othcomp Other compensation 185.8 26.6 1342.2
rstkgrnt Restricted stock grants 336.7 0.0 7007.1
option awards blk value
e
Options granted (Black Scholes) 2059.3 473.9 9826.6
a
ocf
oth
is the sum of extraordinary items and discontinued operations (xidoc), deferred taxes (txdc), equity in net loss (esubc), loss from sale of
PPE and investments (sppiv), funds from operations–other (fopo), other sources of funds (fsrco) and exchange rate effects (exre). The item fsrco
is 0 if the company reports according to format code 7 (scf=7), exre is zero in case of format codes scf=1, 2 or 3.
b
nwc
inv
is constructed as follows: For format code 7, it is the sum of (multiplied by minus 1) accounts receivable-decrease (recch), inventory-decrease
(invch), accounts payable and accrued liabilities-increase (apalch), income taxes-accrued-increase (txach), assets and liabilities-other (aoloch). For
format code 1, it is the variable wcapc. In case of format codes 2 and 3, it is wcapc ∗ (−1).
c
ivaco is replaced by fuseo*(-1) in case of format codes 1, 2 or 3.
d
fincf
oth
is the sum of excess tax benefits of stock options (txbcof) and other financing activities (fiao).
e
option awards blk value is replaced with option awards rpt value and then with option awards fv in case the information is missing.
54
